CUMULATIVERECORD

DefinitiveEdition

2

CUMULATIVERECORD

DefinitiveEdition

BY

B.F.SKINNER

3

Copyright©1999byB.F.SkinnerFoundationwww.bfskinner.org

OriginalversionspublishedbyMeredithCorporationin1959,1961,and1972.

Allrightsreserved.Nopartofthisbookmaybereproducedinanyformorinanymeanswithoutpermission.Forinformation,contacttheB.F.SkinnerFoundation.

B.F.SkinnerReprintSeries

EditorsofthisDefinitiveEdition:VictorG.LatiesandA.CharlesCatania

ISBN:978-0-9899839-7-6ePubISBN:978-0-9899839-9-0Mobi

4

ToE.G.Boring

5

ACKNOWLEDGMENT

Completing thisneweditionofCumulativeRecordhasbeena laborof love.CharlieCatania and Victor Laties undertook that labor. Working from the three originaleditionsofthebook,theywereconfrontedwithmoreproblemsthananyoneimagined—certainly more than they themselves anticipated. Many hours were devoted tofiguringoutthecompositionoftheDefinitiveEdition,andmanydiscussionswereheldaboutwhatother articlesmightbe included.Vicmade several trips toCambridge todelve into Skinner’s papers in the Harvard Archives to find out why the originalselectionsweremade,tomakesurethatrestoringmaterialSkinnerhadomittedinlatereditionswouldnotviolatehiswisheswerehealivetoday.Theforewordinthiseditionrepresents some of what they discovered as well as information from their longassociationwithB.F.Skinner.Wordsofappreciationfallfarshortofwhatisdue.IcanonlyofferthanksonbehalfoftheFoundationandallfuturereaders.TheFoundationisalsoindebtedtoRobertEpsteinandJuliaBecerrawhotookonthe

considerabletaskofcreatinganindex—somethingthatnoneofthethreeprioreditionscontained.Becauseof thediverse contentsof thevolume, theparticularwords tobeindexed were not obvious, but required expertise in the field as well as an unusualbreadth of intellectual background tomatch the range of topics discussed. Thus theFoundationwas fortunate that theywerewilling to undertake the job. The resultingindexhasconsiderablyincreasedtheresearchvalueofthisDefinitiveEdition.FortheFoundationandallusersofthisbook,Iwouldliketoexpressmyappreciation.

JulieS.Vargas,PresidentB.F.SkinnerFoundationCambridge,Massachusetts

6

Acknowledgmentsforthise-bookversion

TheB.F.SkinnerFoundationwouldliketo thanktheBehaviorAnalystCertificationBoard (www.bacb.com) for the encouragement and financial support that made thiselectronicformatpossible.

7

CONTENTS

FOREWORD:AMATTEROFRECORDVictorG.Laties&A.CharlesCatania

PREFACESTOEARLIEREDITIONS

PartI:TheImplicationsofaScienceofBehaviorforHumanAffairs,EspeciallyfortheConceptofFreedom

FreedomandtheControlofMenTheControlofHumanBehavior(Abstract)SomeIssuesConcerningtheControlofHumanBehaviorTheDesignofCultures“Man”TheDesignofExperimentalCommunities

PartII:AMethodfortheExperimentalAnalysisofBehavior—ItsTheoryandPractice,ItsHistory,andaGlimpseofItsFuture

AreTheoriesofLearningNecessary?TheAnalysisofBehaviorACaseHistoryinScientificMethodTheExperimentalAnalysisofBehaviorReinforcementToday

PartIII:TheTechnologyofEducationTheScienceofLearningandtheArtofTeachingTeachingMachinesWhyWeNeedTeachingMachinesReflectionsonaDecadeofTeachingMachinesTeachingScienceinHighSchool—WhatIsWrong?ContingencyManagementintheClassroom

PartIV:TheAnalysisandManagementofNeurotic,Psychotic,andRetardedBehavior

ACritiqueofPsychoanalyticConceptsandTheoriesPsychologyintheUnderstandingofMentalDiseaseWhatisPsychoticBehavior?SomeRelationsBetweenBehaviorModificationandBasicResearchCompassionandEthicsintheCareoftheRetardate

8

PartV:ForExperimentalPsychologistsOnlyCurrentTrendsinExperimentalPsychologyTheFlightfromtheLaboratory

PartVI:CreativeBehaviorCreatingtheCreativeArtistALectureon“Having”aPoem

PartVII:LiteraryandVerbalBehaviorHasGertrudeSteinaSecret?TheOperationalAnalysisofPsychologicalTermsTheAlliterationinShakespeare’sSonnets:AStudyinLiteraryBehaviorAQuantitativeEstimateofCertainTypesofSound-PatterninginPoetryTheProcessesInvolvedintheRepeatedGuessingofAlternatives

PartVIII:TheoreticalConsiderationsWhyAretheBehavioralSciencesNotMoreEffective?TheConceptoftheReflexintheDescriptionofBehaviorTheGenericNatureoftheConceptsofStimulusandResponseTwoTypesofConditionedReflexandaPseudo-typeTwoTypesofConditionedReflex:AReplytoKonorskiandMillerAReviewofHull’sPrinciplesofBehaviorAReviewofBushandMosteller’sStochasticModelsforLearning

PartIX:AMiscellanyAParadoxicalColorEffectSomeQuantitativePropertiesofAnxiety(withW.K.Estes)“Superstition”inthePigeonASecondTypeof“Superstition”inthePigeon(withW.H.Morse)Two“SyntheticSocialRelations”ConcurrentActivityUnderFixed-IntervalReinforcement(withW.H.Morse)SustainedPerformanceDuringVeryLongExperimentalSessions(withW.H.Morse)JohnBroadusWatson,BehavioristHowtoTeachAnimalsBabyinaBoxAWordAboutBoxesThePsychologyofDesignPigeonsinaPelicanSomeResponsestotheStimulus“Pavlov”SquirrelintheYard

PartX:CodaCanPsychologyBeaScienceofMind?

9

APPENDIX:INTRODUCTIONSREVISEDFORTHE1961EDITION

ACKNOWLEDGMENTS

INDEX

INDEXTOTHEPAGINATIONOFEARLIEREDITIONS

10

FOREWORD:

AMatterofRecord

Intheanalysisofbehavior,acumulativerecordisagraphthatconvenientlyshowshowbehaviorchangesovertime.Itisproducedinthelaboratorybyamarkingdevicethatmoves horizontally at a constant rate over time and verticallywith each response insteps of constant size. Thus, the record becomes steeper with faster responding andshallower with slower responding, so that moment-to-moment changes in rate ofrespondingcaneasilybeseenaschangesintheslopeoftherecord.AsthetitleforacollectionofpapersbyB.F.Skinner,thepersonwhofirstusedthecumulativerecorderinstudyingbehavior,CumulativeRecordisapunofsorts,astheauthorhimselfnotedintheFirstEdition’sPreface,whichdirectlyfollowsthisForeword.

TheFirstEdition.AccordingtoalettertoSkinnerfromDanaFerrin,thepresidentofAppleton-Century-Crofts, this bookwas first published onApril 27, 1959 [4/5/1960letterfromDHFtoBFS].Itincluded427pageswiththirtypapers,fewerthanhalfofthoseSkinnerhadpublishedbythen.Most of the papers were preceded by introductions. These were usually brief,

sometimesnomorethananacknowledgmentofpermissiontoreprint.Butonaboutadozenoccasionstheintroductionseitherplacedanarticleinitsintellectualcontextordescribed the circumstances of its creation. Upon reading them after publication,however, Skinner had strong second thoughts about their appropriateness. In a 1960lettertoDanaFerrin,hispublisher,hehaddescribedhisconcernsthisway:

…Iamwonderinghowthesaleshavegoneandhownearyouare toa reprinting. Iamunhappy about the introductorymaterials. I don’t believe the light touchwhich I intendedalwayscomesoff,andIhopeeventuallytoseeasobererpresentation,possiblywithallofthecommentariesomitted.During thenext twoor threeyears,otherpaperswillbe turningupwhichmightbeincludedandwhichshouldaddtoitssales.Forexample,arecentpaperofmineonourwartimeworkonpigeon-directedmissiles isattractingattentionandwouldbeworthincludinginanewedition….Whatwouldyouthinkofaneventualsecondedition,…addingafewpublishedpapersbutfourorfiveunpublishedpapers?[BFStoDHF,3/30/1960]

Years later Skinner described his doubts about the introductions, as well as theoriginsofthebookitself,inAMatterofConsequences(1983),thefinalvolumeofhisautobiography:

ForyearsIhadhadasublimefaiththatthetruthwouldprevail.Iwasquitecontenttogetmypapersintoprintsomewhere;thosewhoneededthemwouldfindthem.(Itwasausefulprinciple,foritpermittedmetocontinueworkinginisolationwhenisolationwasprobablymorevaluablethanbeinginfluential.)Imayhavebeenrightaboutafuturehistorianofideas

11

grubbing about in a library, but I was wrong about my contemporaries. They were notreadingallmywork,and in takingstockduringmyPutneysabbatical Iconsideredputtingtogether a book of collected papers. I could think of eighteen that could be included, butwhen the manuscript went off in October 1958, it contained thirty. I called the bookCumulativeRecord.…(p.163)

Perhapsbecause Iwasgettingabook“fornothing,” Iwent aboutpreparingCumulativeRecord in a curiously debonair way. The very title was a pun. My preface contained acumulative curve showing the number of words in the papers plotted against the year ofpublication (and the publishers put it on the cover above my name). I wrote a briefintroductionforeachpaperandadded“AWordaboutBoxes.”

WhenIsawmyfirstcopy,Iknewthatsomethingwaswrong:Ihadbeenboasting!Ihadtoldthereaderthat“FreedomandtheControlofMen”hadbeenreprintedintheFrenchandItalianeditionsofPerspectivesU.S.A.; thatGertrudeSteinhadsaidofmeasapsychologistthat“…whenheisnottooseriousheisaprettygoodone”;thatalthoughmychapterinTheBehavior of Organisms on “The Conceptual Nervous System” had been “interpreted asshowingananti-physiologicaloranti-neurologicalbias,”Ibelievedthebookwas“apositivecontributiontophysiology”;thatwhenaquestionnairewassenttoseventy-threecoupleswhohadusedbaby-tendersforonehundredthirtybabies,allbutthreehaddescribedthedeviceas“wonderful…withphysicalandpsychologicalbenefits[which]seemedtowarrantextensiveresearch….”(pp.163–164)

Iwasdismayed. Iwasadvertising the“SkinnerBox,”droppingnames,andshowingoffmyLatinandmywitbypunninginLatin.WhatwouldthemanyfriendstowhomIhadsentcopies think of me? Dick Herrnstein was one of them, and my anxiety was somewhatassuagedwhenhewassurprisedtohearthatIwasconcerned.Hadanyonecriticizedme?Hethoughtthematerialwascharming.ButIsufferedacutely,andwithinayearIhadpersuadedthepublisherstobringoutanenlargededition.Itwas“enlarged”bytheadditionofonlythreepapers,addedwithoutrepaginating,butwithmostof theoffendingmaterial removed.AsIwroteinanote:“Gonearethepersonaltouches…goneismyacuteshameinthinkingaboutthem.”(p.165)

DespiteSkinner’suneasewithsomeaspectsofhisbook,hisfourthinsevenyears,hecouldhardlyhavedonebetterinitsreception.Themajorreview,byHaroldSchlosbergintheJanuary1960issueoftheAmericanPsychologicalAssociation’sContemporaryPsychology,beganwiththeflatteringwords“LonghaveIregardedFredSkinnerasapotentialthirdmemberofatrinity,alongwithFreudandPavlov.”AlthoughSchlosbergquite properly concludedhis reviewby noting that any final evaluation ofSkinner’scontributionswouldrestwithfuturegenerations,hemadeclearhisownadmirationfortheir importance and consistency. He made no comment on the suitability of theintroductorymaterial.

TheEnlargedEdition.InearlyApril1960,DanaFerrinreportedtoSkinnerthatabout1,200 copies of Cumulative Record had been sold during the 11 months sincepublication.“Sinceweprinted2,500copies, thereislittlelikelihoodofourrunninga

12

newprintinginlessthantwoyears.However,onecannevertell.”[4/5/1960letterfromDHFtoBFS]Salesmusthavecontinuedstrongbecauseanotherprintingfollowedin1961, affording Skinner the opportunity to deal with his misgivings about theintroductionsaswellastoaddafewarticles.Thethreepapersthatheaddedwere“TheDesignofCultures,”“WhyWeNeedTeachingMachines,”and“PigeonsinaPelican.”Hedealtwithhis“acuteshame” inseveralways.First,hechanged the toneofhis

prose. Compare the vigorous first paragraph of the original introduction to “TheConceptof theReflex in theDescriptionofBehavior” tohis revision.“Thispaper…stillseemstomeimportantforthreereasons.Inthefirstplace,itwasanearlyexampleof….”wascompressedto“Thispaper…wasanearlyexampleof….”Henowavoidedwritinginfirstperson:“IbelievethecluetothedefinitionofreflexcamefromBertrandRussell”became“Theclue to thedefinitionof reflexmayhavecome fromBertrandRussell.”Afewsentenceslater,“IsupportedtheargumentwithaMachiananalysis…”became“TheargumentcouldbesupportedbyaMachiananalysis….”Hesometimesdiminished the informality of his prose, discarding a sentence in the introduction to“PsychologyintheUnderstandingofMentalDisease”thatstated:“Myownscatteredcommentsrevealaconcernwiththeoperationaldefinitionoftermsinthefield….”Notallofthe“offendingmaterial”thatcausedhimsuchdistresswasmodified.For

example,hedidnotbackofffromassertingthat“…astatementofbehavioralfactsinaform which most readily makes contact with physiological concepts and methods”constitutes“apositivecontributiontophysiology.”Hecontinuedtonotethat“Freedomand the Control of Men” had been reprinted in the French and Italian editions ofPerspectivesU.S.A.And,intheintroductionto“ACaseHistoryinScientificMethod,”whichwashis response toa request foraquite formalpaper, themild“Myreactionwas the present paper” became the stronger and more accurate “The present paperscarcelyfollowsthisplan.”Theintroductionto“HasGertrudeSteinaSecret?”causedSkinnerperhapsthemost

discomfort.Heshortened itbyhalf, removing theparagraphonHarvard’sSocietyofFellows,where he had called attention to the fact that his paperwas the first to bepublished by a Junior Prize Fellow and listed some of the Harvard luminaries withwhomthefellowsdinedweekly,agroupthatincludedAlfredNorthWhiteheadaswellas the university’s president and immediate past-president.He also excised the finalparagraph,whichwasdevotedtoMissStein’sreactiontothearticleitselfwhenitfirstappearedintheAtlanticMonthlyandwhichcontainedthewordsofpraisehecitedinthequotationgivenabovefromhisautobiography.ThelastarticleintheFirstEditionwasSkinner’scontributiontotheOctober1945

issueoftheLadiesHomeJournal,“BabyinaBox.”FortheEnlargedEdition,heleftunchanged his introduction, where he reported on the overwhelmingly favorablereviews his “Air-Crib” had received from users, but removed the free-standingafterword, “A Word about Boxes,” in which he first recounted the origin of themagazinearticle’stitle—theeditorshadsuppliedit—andthenumerouswaysinwhichboxes had entered his life (baby box, Skinner box, even the teachingmachine), andthentoldoftheoccasionwherehewaspromptedtomakethedelightful“boxpopuli”Latinpunthatlatercausedhimgriefwhenhesawitinprintandconcludedthathehadbeen“showingoffmyLatinandmywit.”

TheThirdEdition. InNovember 1969, Skinnerwrote to JackBurton, his editor atAppleton-Century-Crofts,asfollows:

13

Iamcurrentlyatworkonthreepapersgivenduringthepastyear,eachofwhichistobepublishedinanappropriateplace.Ishallthenbeturningmyattentionentirelytobooklengthmanuscripts. This seems, therefore, a good time to think of a third edition ofCumulativeRecord. IverymuchdoubtwhetherIwillbewritinganyotherpapersduringthenextfiveyears.A thirdeditionofCumulativeRecordwould, therefore, serve formanyyears as theoneconvenientsourceofallmypapersnotreprintedelsewhere.

Afterdescribingthearticlesthathewantedtoaddordelete,Skinnerconcluded:

InsummaryIamproposingtoaddfivesubstantialpapersandseveralshortarticles.Thisseemstometojustifyaneweditionparticularlyinviewofthefactthattherewillbenootherarticlestoaddforatleastmanyyearstocome.[BFStoJKB,11/10/1969]

(Note that Skinner was unduly pessimistic about his subsequent productivity; inaddition to his books, he published about four papers per year for the next twodecades.)Afewweekslater,JackBurtonagreedthataneweditionwasappropriatebut,noting

thattheystillhad“alittleoveroneyear’ssupplyonhand,”suggestedthattheyaimforpublication“earlyin1971.”[JKBtoBFS,11/25/1969]The 604-pageThirdEdition actually appeared in 1972. It contained 18 additional

papers, but two articles on teaching machines were deleted because they had beenrevised and included inThe Technology of Teaching (1968). Also deleted were theexcerpts from “Some Contributions of an Experimental Analysis of Behavior toPsychologyasaWhole,”whichhadappearedunderthetitle“TheAnalysisofBehavior(excerpts)” in the earlier editions.Thus this edition contained48papers.Among theadditional 18 paperswere five thatwere publishedprior to 1961when theEnlargedEditionhadappeared.Theinclusionofthismaterialplustheadditionofasubtitle,“ASelection of Papers,” made the nature of the collection more explicit. (Skinner’soriginalselectioncriteriaweresetforthintheFirstEdition’sintroductiontohispaperonalliterationinShakespeare’ssonnets;seepage431ofthisvolume.)Almost all introductory material was omitted from the Third Edition; Skinner

includednocommentonanypaperbefore“HasGertrudeSteinaSecret?,”morethanhalfwaythroughthebook.TherehemovedtheshortenedintroductionoftheEnlargedEditiontotheend,offeringitasanepilogue(seetheAppendix).Onlyfourpapersnowcontainedintroductions,allfromthemoretechnicalsectionsofthebook.Hekepttheintroductionsto“TwoTypesofConditionedReflex:AReplytoKonorskiandMiller”and“TheProcessesInvolvedintheRepeatedGuessingofAlternatives,”omittingonlya single sentence from the latter (see the Note in the Appendix). The alreadythoroughlymassagedintroductionto“TheConceptoftheReflexintheDescriptionofBehavior”was retainedand is included in theAppendix so that itmaybe comparedwiththeFirstEditionversion,whichstartsonpage475ofthepresentvolume.IntheintroductiontotheEstesandSkinnerpaper,“SomeQuantitativePropertiesof

Anxiety,”Skinnerhadoriginallywritten that thebehavioral technique theydescribed“…hasprovedtobeausefulbaselineinstudyingmeasureswhich‘relieveanxiety’inhuman subjects.” This was no more than a statement of fact; by that time manypharmaceuticalcompanieswereusingvariantsoftheprocedureaspartoftheireffortsin the rapidly growing field of psychopharmacology. Nevertheless, he dropped thissentencefromtheThirdEdition.

14

Apartfromhis1960statementinthefirstlettertoDanaFerrincitedabove—“Idon’tbelievethelighttouchthatIintendedalwayscomesoff,andIhopeeventuallytoseeasoberer presentation, possiblywith all the commentaries omitted”—nothingwe havefound in Skinner’s correspondence with his publisher alluded to dropping theintroductory commentaries. Since itwas hewho prepared a paste-upmanuscript forthis revisionfromcutsheetsprovidedbyhiseditor,Skinnerprobablysimplydeletedmost of the introductions at that time. [JKB to BFS, 11/25/1969; BFS to JKB,12/1/1969]Wedoknowthatsavingspacewasnotaconsiderationinthedeletions.Actingwith

remarkableforesightshortlyafterpublication,oneofus,neartheendofaletterdealingwithothermatters,askedSkinnerwhyhehadmadethesechanges.

…IwasdisturbedtoseethattheneweditionofCumulativeRecorddoesnotcontaintheintroductionstothearticlesthatappearedintheearliereditions.Imissedthemandthinkthatthereaderswillmissthem.Ihopeyouwriteintroductionsforsomeofthearticlesthatyouaddtothenextedition.Thesearelittlefootnotestohistorywhicharevaluable;Ihatetoseethemjettisonedtosaveafewpenniesonthepriceofabook.[VGLtoBFS,9/1/1972]

Skinnerreplied:

Theintroductorymaterialwasnotleftoutofthethirdeditionforreasonsofeconomy.ItwasmyowndecisionandIthoughtithadtheadvantageofcementingthebookassuchratherthan keeping it disjointed as a collection of papers. Maybe I was wrong. [BFS to VGL,9/14/1972]

A decade later, in the poignant “Epilogue” toAMatter of Consequences (1983),somepagesafterrecalling:“Ihavesaidthat‘IwastaughttofearGod,thepolice,andwhatpeoplewill think,’” this ishowSkinneraddressedhisemotional reaction to thefirstappearanceoftheintroductorymaterial:

I now see that I greatly exaggerated the extent towhich the first edition ofCumulativeRecordwasboastfulandthatIsufferedunnecessarily.[p.410]

Four features of previous editions deserve brief mention before we describe thepresentvolume:therevisionsmadebySkinnerinthesuccessiveprefaces,hisrevisionswithinarticles,hischoiceofatitle,andhisuseofquotations.

The threeprefaces.Theprefacesalso reflected their author’sattempt to switch toamoreimpersonalvoice.AllthreeimmediatelyfollowthisForeword.InthePrefacetotheFirstEdition,Skinner includeda figure showing the totalwordscontained in thepaperschosenforthebook,plottedcumulatively,fortheyearsbetween1930and1958,and devoted most of the Preface to discussing the figure. He had publishedapproximately160,000words,averagingabout3,600wordsperyearthrough1955and23,000peryearforthenextthreeyears,“aperiodofheightenedactivity”thatreflecteda sharp increase in invited papers. This did not include his books, which he notedconstitutedcompetingbehavior forwritingotherpapers.Heused labeledarrowheadson the cumulative record to indicate when the five books had been written. Alsoexcludedfromthegraphwere“19paperswhichwouldgreatlyhaveincreasedtheslope

15

between1931and1937”buthadbeenusedinTheBehaviorofOrganisms (1938).Healso pointed out that Schedules of Reinforcement (with Charles B. Ferster, 1957)contained material that “would normally have appeared in papers during the early50’s.”For the much shorter Preface in the 1961 Enlarged Edition, Skinner updated the

cumulative record but removed the book titles and arrowheads. As in the revisedintroductions,hemovedto thepassivevoiceandexcisedalmostalldiscussionofhisownbehavior.Hementionedneitherhiseditorialchangestotheintroductionsnorhisexcisionof“AWordAboutBoxes,” theafterword thathadaccompanied“Baby inaBox.”ThePrefacetothe1972ThirdEditionwasonlyninelineslongandthecumulative

record itself disappeared, the author noting that “there seems to be no point inextendingthecumulativerecordwhichappearedinearliereditionstoexplainmytitle.”Wecontemplatedpreparinganupdatedcumulativerecordforthisnewedition,butit

was not feasible for us to undertakeword counts across themany articles publishedwith different fonts in different formats, nor couldwe resolve the problem of units.Publication counts that weigh a single-paragraph book review as heavily as entirebooksareinadequate,andevenwordcountsfailtoconsiderthecontributionoffiguresrichindata.Ifonepictureisworthathousandwords,thenSchedulesofReinforcementoutweighs all of his other work; it includesmore than 900 figures.We leave it forothers to explore such quantification when all of Skinner’s writings are eventuallyavailableinanelectronicformatthatwouldeasethelaborofsuchatask.(Oneattempt,displayingcumulativeplotsofbooksandotherpublicationsseparately,appearsontheB.F. SkinnerFoundation Internet site, in connectionwith an extensive bibliographyoriginallycompiledbyEpsteinforTodd&Morris(1995).

Skinner’s revisionswithin articles. Skinner’s copies of the first two editions havenotesinhishandthatperhapswereplansforchangestobeincorporatedintotheThirdEdition. For example, the third paragraph of “The Operational Analysis ofPsychologicalTerms,”whichbeginswith“Theoperationist…,”hadbeenrewrittenintheplural,“Operationists…,”therebyallowinghimtoreplacethelater“Hehasnot…”withtheungendered“Theyhavenot….”Buttheseandotherchangeswerenevermade.PerhapsSkinnerhadreservationsaboutarchivalproblemsifthereprintedversionsnolongeragreedwiththeoriginals.Heevenidentifiedbutletstandsomesubstantiveandgrammatical errors. For example, the seventh paragraph in the section “ComplexLearning” in“AreTheoriesofLearningNecessary?”endswithanexample inwhichthe50percent success ratebychanceofapigeon’smatching-to-sample responses isreferred to as fixed ratio reinforcement. Although Skinner corrected his copy to“variableratio,”thetextremainedas“fixedratio”inalleditions.Inothercontexts,Skinnerdidrevisesomeofhisearlierwork.Forexample,whenhe

laterpreparedsomeofhisclassicpapersforcommentaryinthespecialDecember1984issue of the journal Behavioral and Brain Sciences (revised as Catania & Harnad,1988),heeliminatedmuchofthesexistlanguageandmodernizedsomeofthetechnicalvocabulary (e.g., replacing periodic with intermittent, and induction withgeneralization). He also revised some passages in which he had written about thereinforcementoforganisms,changingthemsothatresponseswereinsteadreinforced.

Cumulative Record as a title. Cumulative Record, first appearing in 1959, was

16

Skinner’s sixth book, and its punning title was totally original, at least withinpsychology.1 So were the titles of his fourth and fifth books, Schedules ofReinforcementandVerbalBehavior,bothpublishedtwoyearsearlierin1957.Forhisfirstthreebooks,Skinnerhadusednameswhichresonatedwithtitlesusedby

others.ThechoiceofTheBehaviorofOrganismsmayhavebeeninfluencedbyJacquesLoeb’sTheOrganismasaWhole (1916),whichhehad readwhileanundergraduate(Skinner, 1976, p. 296; cf.Catania, 1992, for a discussionof the significanceof theterm“organism”inSkinner’swritings),andperhapsbyHerbertS.Jennings’Behaviorof the Lower Organisms (1906). Walden Two (1948), of course, pays tribute toThoreau’sWalden. Science and Human Behavior (1953) resemblesScience and theModernWorld(1925),thebookthatattractedhimtoAlfredNorthWhitehead’scourseat Harvard (Skinner, 1979, p. 30) and which, incidentally, contains the statement:“Scienceistakingonanewaspectthatisneitherpurelyphysicalnorpurelybiological.Itisbecomingthestudyoforganisms”(Whitehead,1925,p.103).InchoosingTheAnalysisofBehavior(1961)asthetitleoftheprogrammedtexthe

co-authored with J. G. Holland only two years after Cumulative Record appeared,SkinnermayhavebeeninfluencedbythenameofacoursehetookwithW.J.Crozierduring his first year as a graduate student at Harvard, “The Analysis of Conduct.”(“Crozierused‘conduct’becauseWatsonandthepsychologistshadsullied‘behavior.’“—Skinner,1979,p.44)

Theuseofquotations.Skinneralwayshadaneyeforfelicitousstatementsbyothersand frequently relied upon an apt quotation to help make a point. The very firstparagraph of Science andHumanBehavior contains a long direct quotation from aseventeenthcenturyscientist,FrancescoLana,uponwhichSkinnerbasesadiscussionof theuseandmisuseofscience. InParticularsofMyLife,hecreditshis friendAlfEvers’ remark that “Science is the art of the twentieth century”with influencinghisturn from literature to science (Skinner, 1976, p. 291). Incidentally, his title is fromShakespeare’sHenryIV,PartI:“Dothoustandformyfatherandexaminemeontheparticularsofmylife”andheplacedthequotationonapageamongthefrontmatter.Near the end ofWaldenTwo, he quotes a passage from Thoreau’sWalden, whichconcludeswiththedelightfulsentence,“Thesunisbutamorningstar.”Thiswashisoriginal title for the book, changed at the lastminute because another star titlewasalreadyonthemarket(Skinner,1979,p.330).Many of Skinner’s articles, especially those aimed at general audiences, quote

liberally, usually from literary rather than scientific sources. The first paper in thiscollection,“Freedomand theControlofMen,” isespeciallynotable,withquotationsfromMarcus Aurelius, Dostoevsky, Lord Acton, Ralph Barton Perry, JosephWoodKrutch,T.H.Huxley,T.S.Eliot,AldousHuxley,andDeanAcheson.TheearliereditionsofCumulativeRecordcontainedasinglefreestandingquotation

onapageotherwiseblank, fromShakespeare’sTimonofAthens. It can be found onpage68ofthisvolumeandisoneoftheonlytwoinstanceswhereShakespeareusedawordthatSkinnermuchlaterclaimedforhisown:“…saucehispalatewiththymostoperantpoison.”Theother,fromHamlet,nowappearsonpage660ofthisedition.

This“Definitive”Edition.CumulativeRecord was not Skinner’s only collection ofpapers.JustbeforepublishingtheThirdEdition,hecombinedninepaperswithsome

17

other material to produce Contingencies of Reinforcement: A Theoretical Analysis(1969). Six years after the Third Edition, Reflections on Behaviorism and Society(1978) appeared, containing 18 articles published from 1972 through 1978. UponFurtherReflection (1987) included 14 more papers, all but one of which had beenpublishedduringthe1980s.Finally,RecentIssuesintheAnalysisofBehavior (1989),publishedtheyearbeforehedied,contained12morefrom1986orlater.InthePrefacetothatfinalbook,helistedonlythesefivecollections,apparentlynotconsideringTheTechnology of Teaching (1968) to be a collection in the same sense as the othersbecauseonlyfourofits11chaptershadpreviouslybeenpublished.MerelyreprintingtheThirdEditionofCumulativeRecord,whichwentoutofprint

someyearsago,wasacourseofactionthatwasentertainedbrieflybutrejected.Giventhe history as outlined in this Foreword, amore appropriate choice for this volumeseemedtobeaneditioncontainingsignificantfeaturesfromalltheprioreditions.First,all 48 articles in the Third Edition have been retained and all of those items thatSkinnerusedearlierbutthenomittedfromthateditionareagainincluded.Second,two“new” articles (see below) have been added. Third, all of the First Edition’sintroductionshavebeenrestoredtotheiroriginalpositions,inlightofSkinner’schangeof heart concerning this material. The two introductions from the 1961 “EnlargedEdition”thatSkinnermodifiedmostextensivelyarepresentedinanAppendix.Lastly,RobertEpsteinandJuliaBecerrahaveprovidedahighlyusefulIndextothisvolume.We have also included an index to the pagination of each article in each edition ofCumulativeRecordonpages698through700.Because any editorial tinkeringwith the text itselfwould have produced a flawed

archival document, none has been attempted. As it stands, the book exhibits theevolution of Skinner’s thinking and writing style and is therefore a cultural andhistoricaldocumentaswellasascientificone.

Reinstated and newly added articles. Three papers that appeared in the first twoeditionsbutweredeletedfromthethirdhavebeenreinstated:“TheScienceofLearningand the Art of Teaching,” “Teaching Machines,” and “The Analysis of Behavior(excerpts).”Theshortcommentary,“AWordAboutBoxes,”whichonlyappeared intheFirstEdition,nowappearsonpage620,following“BabyinaBox.”Twoarticleswithoutahistoryofinclusioninanyearliereditionhavebeenaddedto

thisedition.Theywerepublishedroughlyhalfacenturyapart.Thefirst,“ThePsychologyofDesign,”appeared ina1941 issueofArtEducation

Today.ItrepresentsanearlyexercisebySkinnerinapplyingpsychologicalprinciplesto the understanding of visual art. Skinner omittedmost of his experimental papersfromCumulativeRecord on the grounds that they had been incorporated in revisedformintootherbooks(especiallyTheBehaviorofOrganisims).Healsoomittedrepliesthat would not stand alone and short technical items of specialized interest. But heincludedallotherpaperson topicsofgeneral interest suchasart and literature.Thiswastheonlyoneweidentifiedfromtherelevanttimeperiodforwhichthereseemednorationaleforexclusion.WefoundnothingtosuggestthatSkinnerwasdissatisfiedwiththe paper (he evenbriefly discussed it in the secondvolumeof his autobiography—Skinner,1979,p.238).Thoseclosetohimregarditasunlikelythathelosttrackofthepaper,evengiventheinterveningeventsofWorldWarIIandhissubsequentacademicmoves. One guess is that some technical problem accounts for its omission. Forexample,ifhehadincludedhisonlycopiesoftheartforthefigureswhenhesubmitted

18

thearticleforpublication,copiessuitableforprintingwouldnothavebeenavailableatthetimewhenthisbookwasfirstpublished.The second newly added paper is “Can Psychology be a Science ofMind?” The

American Psychological Association presented Skinner with its first Citation forOutstandingLifetimeContributiontoPsychologyatits1990annualmeetinginBoston,andthispaperwashisresponsetothataward.HedeliveredhistalkonAugust10.OnAugust17,inthehospitalonthedaybeforehediedofleukemia,heputthefinishingtouchestothewrittenversion.ThiswashismostimportantpapersincethosecollectedinRecentIssuesintheAnalysisofBehavior,whichhadappearedthepreviousyear,anditdeservestoconcludethisversionofCumulativeRecord.

Thecurrentcontents.Therangeoftopicsincludedinthiscollectiondefiessummaryand our temptation is to let the articles speak for themselves. Shortly after itspublication, theThirdEditionofCumulativeRecordbecamea main selection of theBehavioralScienceBookService.TheServiceprescientlyinvitedoneofus(ACC)toprepareareviewfortheflierthatannouncedit.Thereviewdescribesthateditionas“acollectionof 48papers byoneof themost controversial figures of our time.To saywhythiscollectionisworthreading,areviewerhaslittlemoretodothanlistsomeofthe topics: scientificmethod; teachingmachines;ESP; ethics; thedesignof cultures;babies; poetry and creativity; pigeons; anxiety; Pavlov; psychoanalysis.” Later itcontinues:“Varietyisoneremarkablefeatureofthiscollection…herearesuchclassicsas ‘TheConceptof theReflex in theDescriptionofBehavior’ and ‘AreTheoriesofLearning Necessary?’… But poetry and literary analysis are represented also.” Byincludingasamplelistoftopics,thereviewsharesafeatureofmanyofthereviewsofCumulativeRecord thatappearedatthetimeoftheearliereditions(Knapp,1974,pp.37–38).Thereviewargues that thevariety is importantnotsomuchbecause thebookwill

appealtodifferentreaders,butalsobecause“somanyhaveassumedthatthebehavioralpoint of view is necessarily narrow and intellectually confining…. The cumulativeeffect of the papers in this collectionwill be to set the record straight.” The reviewcontinues:“Itwouldbemisleadingtoemphasizethevarietyinthesewritingswithoutalsomentioningtheircoherence.”Thatcoherencecanhardlybeunderstated,especiallygiven that thepapers in this selectionwerewrittenover a periodofmore than fortyyears—andaboutsixtyifweincludethelastpaperinthisedition;cf.Epstein,1995.Thebodyofworkrepresentedherevariesstylisticallyaswellassubstantively.Asit

ranges from the relatively formal prose of “The Generic Nature of the Concepts ofStimulusandResponse”tothemoreinformalproseof“HowtoTeachAnimals,”andfrom the implications for literature of “A Lecture on ‘Having’ a Poem” to theimplications for education of “The Science of Learning and the Art of Teaching,”Skinnerremindsusagainandagainthatbehaviorisasubjectmatterinitsownright.Theanalysisofbehaviormustjustifyitselfinitsownterms.It isofcourserelatedtoother sciences and to other levels of analysis but it need not appeal to them tolegitimizeitsbasicunitsortoverifyitstaxonomyofbehavioralprocesses.Thisstanceon the primacy of behavior is evenmore fundamental to the enterprise of behavioranalysis than is ourunderstandingof suchphenomena as the three-termcontingencyandtheshapingofbehaviorbyitsconsequences.ThissingularcontributionbySkinnermadealloftheotherspossible.Whether he dealt with pressing levers, behaving ethically, looking at paintings,

19

pecking keys, writing poems, talking, teaching, designing cultures, pushing buttons,doing experiments, or knowing, Skinner offered explanations of behavior in its ownterms. Consistent with the opening paragraph of “Are Theories of LearningNecessary?”hedidnotappealto“eventstakingplacesomewhereelse,atsomeotherlevelofobservation,described indifferent terms,andmeasured, ifatall, indifferentdimensions.”WheneverSkinnerdealtwiththedetailsoflookingandotherexamplesofdiscriminating,orwithtalkingandthinkinginthecontextsofeducationandliterature,or with social interactions as instances of cultural practices, or with the role of theenvironment in mental disease, he extended the boundaries of what we count asbehavior.Andwhenheexaminedhisownbehaviorandscientificbehavioringeneral,asin“ACaseHistoryinScientificMethod”or“TheFlightfromtheLaboratory,”heeventurnedbehavioranalysisuponitself.

The record shows that the earliest paper in this collection, “The Concept of theReflex in the Description of Behavior,” was written during the summer of 1930(Skinner, 1979, p. 67). The latest, “Can Psychology Be a Science of Mind?” wasfinishedalmostexactly60yearslater.Thesesixdecadescontainacumulativerecordofaccomplishmentthatwillnotsoonbematched.

VICTORG.LATIESUniversityofRochester

A.CHARLESCATANIAUniversityofMarylandBaltimoreCounty

December,1998

We extend our thanks to theHarvardUniversity Archives for courteously providingaccesstoB.F.Skinner’spapers;toJohnGachforahelpfuldiscussionofsignificantfeaturesofhistoricbooksinpsychology;and,aboveall,toJulieVargasandtheB.F.SkinnerFoundationforgivingustheopportunitytoundertakethislaboroflove.

References

CATANIA,A.C. (1992).B.F.Skinner,organism,AmericanPsychologist,47,1521–1530.

CATANIA, A. C. and HARNAD, S. (Eds.). (1988). The selection of behavior: TheoperantbehaviorismofB.F.Skinner.Cambridge:CambridgeUniversityPress.

EPSTEIN,R.(1995).AnupdatedbibliographyofB.F.Skinner’sworks.InTodd,J.T.&Morris, E. K. (Eds.)Modern perspectives on B. F. Skinnerand contemporarybehaviorism(pp.217–226).Westport,CT:GreenwoodPress.

FERSTER,C.B.andSKINNER,B.F.(1957).Schedulesofreinforcement.NewYork:Appleton-Century-Crofts.

HOLLAND,J.G.ANDSKINNER,B.F. (1961).Theanalysisofbehavior.NewYork:McGraw-Hill.

KNAPP,T.J.(1974).ReferenceguidetoB.F.Skinner:Acomprehensivebibliography

20

ofpublishedworksbyandconcerningB.F.Skinnerfrom1929through1973.Reno,NV:CliffordPress.

JENNINGS,H.S.(1906).Behaviorofthelowerorganisms.NewYork:Macmillan.LOEB,J.(1916).Theorganismasawhole.NewYork:Putnam’sSons.SCHLOSBERG,H.(1960).IsthisanewMessiah?ContemporaryPsychology,5,35–36.SEGEL, D. (1938).Nature and use of the cumulative record. Bulletin 1938, No. 3.Washington,DC:U.S.GovernmentPrintingOffice.

SKINNER, B. F. (1938).The behavior of organisms. New York: Appleton-Century-Crofts.

SKINNER,B.F.(1948).Waldentwo.NewYork:Macmillan.SKINNER,B.F.(1953).Scienceandhumanbehavior.NewYork:Macmillan.SKINNER,B.F.(1957).Verbalbehavior.NewYork:Appleton-Century-Crofts.SKINNER, B. F. (1968).The technology of teaching. New York: Appleton-Century-Crofts.

SKINNER, B. F. (1969). Contingencies of reinforcement. New York: Appleton-Century-Crofts.

SKINNER,B.F.(1976).Particularsofmylife.NewYork:Knopf.SKINNER,B.F.(1978).Reflectionsonbehaviorismandsociety.EnglewoodCliffs,NJ:Prentice-Hall.

SKINNER,B.F.(1979).Theshapingofabehaviorist:Parttwoofanautobiography.NewYork:Knopf.

SKINNER, B. F. (1983).Amatter of consequences: Part three of an autobiography.NewYork:Knopf.

SKINNER,B.F.(1987).Uponfurtherreflection.EnglewoodCliffs,NJ:Prentice-Hall.SKINNER, B. F. (1989).Recent issues in the analysis of behavior. Columbus, OH:Merrill.

WHITEHEAD,A.N.(1926).Scienceandthemodernworld.NewYork:Macmillan.

1NatureandUseoftheCumulativeRecord(Segel,1938)—amonographpublishedthesameyearasTheBehaviorofOrganisms—wasconcernedwithrecordkeepinginschoolguidanceprograms!

21

PREFACESTOTHEEARLIEREDITIONS

FirstEdition(1959)

Thereaderwhoisnotfamiliarwiththeexpression“cumulativerecord”shouldtaketheroyal road to knowledge starting on page 178 [p. 212 in this volume]. FurtherillustrationsappearinPartII.Asthetitleofacollectionofpapers,theexpressionisnota metaphor, since the behavior of which the papers are a product can be plottedcumulatively.Certainfamiliarproblemsarise.Itishardtoidentifytheunitsofbehaviortobecounted,andthecurvemayneglectotherbehaviorofasimilarnatureoccurringatthesametime.Ratherarbitrarysolutionstotheseproblemsdonotwhollydestroythesignificanceoftheresult.Plottingtotalnumberofpagesagainsttheyearsinwhichthepapersreprintedherewerepublished,wegetFigure1. It revealsa relativelyconstantslope(asteadyoutput)forthetwenty-fiveyearperiodfrom1930to1955.Aperiodofheightenedactivitythenfollows.

FIG.1.Cumulativerecordoftheverbalbehaviorrecordedinthisbook.

Theprincipalcompetingbehaviorconsistedofwritingseveralbooks,themanuscriptsofwhichwerecompletedatthepointsindicated.Thesewillbereferredtothroughoutthepresentvolumebynameonly.Fullcitationsareasfollows:

TheBehaviorofOrganisms.NewYork:Appleton-Century,1938.WaldenTwo.NewYork:Macmillan,1948.ScienceandHumanBehavior.NewYork:Macmillan,1953.

22

SchedulesofReinforcement(withC.B.Ferster).NewYork:Appleton-Century-Crofts,1957.VerbalBehavior.NewYork:Appleton-Century-Crofts,1957.

Someofthesebooksincludedmaterialpreviouslypublishedinpapersnotreprintedhere(andhencenotincludedinthegraph).Thus,theBehaviorofOrganisms containsmaterial from19paperswhichwouldgreatlyhave increased theslopebetween1930and1937ifincludedinthecurve.SchedulesofReinforcementcontainsmaterialwhichwouldnormallyhaveappearedinpapersduringtheearly50’s.Asareportofmyverbalbehavior,therefore,theriseinslopeafter1954isundulydelayed.Thelowslopeduringthe40’sis“real,”however,andreflectsotheractivities—principallywarresearchand,later,administration.Thesharplyincreasedslopeduringthelastfouryearsofthegraphisduetoachange

in the variables controlling my behavior. Most of the papers published during thisperiod were “occasional” pieces—that is, they were written because I was asked towrite them. Evidently my verbal behavior is strongly controlled by the audiencevariable(seeVerbalBehavior,Chapter7).Mythanksareduetotheeditorsandpublisherswhohavekindlygivenpermissionto

reprint. Specific acknowledgement ismade in the introduction to each article. I alsohavetothankMrs.PatriciaPershanforhercarefulhelpinpreparingthemanuscript.Cambridge,Mass.

B.F.S.

EnlargedEdition(1961)

Iftheexpression“cumulativerecord”isnotfamiliartothereader,hemaywishtotakethe royal road to knowledge starting on page 178 [p. 212 in this volume]. FurtherillustrationsappearinPartII.Ontheprinciplethatturnaboutisfairplay,thebehaviorofwhich thepresentpapersareaproducthasbeenplottedcumulatively inFigure1.Certainfamiliarproblemsarise.It ishardtoidentifyunitsofbehaviortobecounted,and thecurveneglectsotherbehaviorofasimilarnatureoccurringat thesame time.Arbitrary solutions to these problems do not wholly destroy the significance of theresult.When total number of pages is plotted against year of publication, the curveshowsarelativelyconstantslope(indicatingasteadyoutput)forthetwenty-fiveyearperiodfrom1930to1955andalaterperiodofheightenedactivity.

23

FIG.1.Cumulativerecordoftheverbalbehaviorrecordedinthisbook.

The principal competing behavior consisted of the following books, referred tothroughoutthepresentvolumebynameonly:

TheBehaviorofOrganisms.NewYork:Appleton-Century,1938.WaldenTwo.NewYork:Macmillan,1948.ScienceandHumanBehavior.NewYork:Macmillan,1953.SchedulesofReinforcement(withC.B.Ferster).NewYork:Appleton-Century-Crofts,1957.VerbalBehavior.NewYork:Appleton-Century-Crofts,1957.

Three papers have been added to this second printing: as an addendum to Part I,“The Design of Cultures;” as an addendum to Part III, “Why We Need TeachingMachines;”andasanaddendumtoPartVIII,“PigeonsinaPelican.”

Cambridge,Mass.B.F.S.

ThirdEdition(1972)

Eighteenpapershavebeenaddedtothisedition,andtwowhichmaynowbefoundinTheTechnologyofTeachinghavebeenremoved.Notallthenewmaterialisrecent,andthereseemstobenopointinextendingmytitle.Preparationofpapers6,13,14,18,19,22,and29 [thesecorrespond to thearticles

thatbeginonpp.58,254,271,322,329,379,and467 in thisvolume]andeditorialwork on this edition as a whole have been supported by a Career Award from theNationalInstitutesofMentalHealth(GrantK6-MH-21,775-01).

24

B.F.Skinner

25

PARTI

TheImplicationsofaScienceofBehaviorforHumanAffairs,EspeciallyfortheConceptofFreedom

FreedomandtheControlofMenTheControlofHumanBehavior(Abstract)SomeIssuesConcerningtheControlofHumanBehaviorTheDesignofCultures“Man”TheDesignofExperimentalCommunities

26

FreedomandtheControlofMen

Originallywritten for a special issue of TheAmerican Scholar (Winter, 1955-56)devoted to “The Human Situation Today,” at the suggestion of the editor, HiramHaydn,thisarticlehasbeenreprintedinPerspectivesU.S.A.and,intranslation,intheFrenchandItalianeditionsofthatquarterly.

The second half of the twentieth century may be remembered for its solution of acuriousproblem.AlthoughWesterndemocracycreatedtheconditionsresponsibleforthe riseofmodern science, it isnowevident that itmaynever fullyprofit from thatachievement. The so-called “democratic philosophy” of human behavior towhich italsogaveriseisincreasinglyinconflictwiththeapplicationofthemethodsofscienceto human affairs. Unless this conflict is somehow resolved, the ultimate goals ofdemocracymaybelongdeferred.

I

Justasbiographersandcriticslookforexternalinfluencestoaccountforthetraitsandachievementsofthementheystudy,soscienceultimatelyexplainsbehaviorintermsof“causes”orconditionswhichliebeyondtheindividualhimself.Asmoreandmorecausal relations are demonstrated, a practical corollary becomes difficult to resist: itshould be possible to produce behavior according to plan simply by arranging theproper conditions. Now, among the specifications which might reasonably besubmittedtoabehavioraltechnologyarethese:Letmenbehappy,informed,skillful,wellbehaved,andproductive.Thisimmediatepracticalimplicationofascienceofbehaviorhasafamiliarring,for

it recalls the doctrine of human perfectibility of eighteenth- and nineteenth-centuryhumanism. A science of man shares the optimism of that philosophy and suppliesstrikingsupportfortheworkingfaiththatmencanbuildabetterworldand,throughit,bettermen.Thesupportcomes just in time,for therehasbeenlittleoptimismof lateamong thosewho speak from the traditional point of view.Democracy has become“realistic,” and it is only with some embarrassment that one admits today toperfectionisticorutopianthinking.Theearliertemperisworthconsidering,however.Historyrecordsmanyfoolishand

unworkable schemes for human betterment, but almost all the great changes in ourculture which we now regard as worthwhile can be traced to perfectionisticphilosophies. Governmental, religious, educational, economic, and social reformsfollowacommonpattern.Someonebelieves thatachangeinaculturalpractice—forexample, in the rules of evidence in a court of law, in the characterizationofman’srelationtoGod,inthewaychildrenaretaughttoreadandwrite,inpermittedratesofinterest, or in minimal housing standards—will improve the condition of men: by

27

promoting justice, permittingmen to seek salvationmore effectively, increasing theliteracy of a people, checking an inflationary trend, or improving public health andfamily relations, respectively. The underlying hypothesis is always the same: that adifferentphysicalorculturalenvironmentwillmakeadifferentandbetterman.The scientific study of behavior not only justifies the general pattern of such

proposals; itpromisesnewandbetterhypotheses.Theearliestculturalpracticesmusthaveoriginatedinsheeraccidents.Thosewhichstrengthenedthegroupsurvivedwiththegroupinasortofnaturalselection.Assoonasmenbegantoproposeandcarryoutchanges in practice for the sake of possible consequences, the evolutionary processmusthaveaccelerated.Thesimplepracticeofmakingchangesmusthavehadsurvivalvalue. A further acceleration is now to be expected. As laws of behavior are morepreciselystated,thechangesintheenvironmentrequiredtobringaboutagiveneffectmay bemore clearly specified.Conditionswhich have been neglected because theireffectswereslightorunlookedformaybeshowntoberelevant.Newconditionsmayactuallybecreated,asinthediscoveryandsynthesisofdrugswhichaffectbehavior.This is no time, then, to abandon notions of progress, improvement, or, indeed,

humanperfectibility.Thesimplefactisthatmanisable,andnowasneverbefore,tolifthimselfbyhisownbootstraps.Inachievingcontrolof theworldofwhichheisapart,hemaylearnatlasttocontrolhimself.

II

Timeworn objections to the planned improvement of cultural practices are alreadylosingmuchoftheirforce.MarcusAureliuswasprobablyrightinadvisinghisreaderstobecontentwithahaphazardameliorationofmankind.“NeverhopetorealizePlato’srepublic,” he sighed, “… for who can change the opinions of men? And without achangeofsentimentswhatcanyoumakebutreluctantslavesandhypocrites?”Hewasthinking,nodoubt,ofcontemporarypatternsofcontrolbaseduponpunishmentorthethreat of punishmentwhich, as he correctly observed, breed only reluctant slaves ofthosewhosubmitandhypocritesofthosewhodiscovermodesofevasion.Butweneednot sharehispessimism, for theopinionsofmencanbechanged.The techniquesofindoctrinationwhichwerebeingdevisedbytheearlyChristianChurchattheverytimeMarcus Aurelius was writing are relevant, as are some of the techniques ofpsychotherapy and of advertising and public relations. Other methods suggested byrecentscientificanalysesleavelittledoubtofthematter.Thestudyofhumanbehavioralsoanswersthecynicalcomplaintthatthereisaplain

“cussedness” inmanwhichwill always thwart efforts to improvehim.Weareoftentoldthatmendonotwanttobechanged,evenforthebetter.Trytohelpthem,andtheywilloutwityouandremainhappilywretched.Dostoevskyclaimedtoseesomeplaninit.“Outofsheeringratitude,”hecomplained,orpossiblyboasted,“manwillplayyouadirtytrick,justtoprovethatmenarestillmenandnotthekeysofapiano….Andevenifyoucouldprovethatamanisonlyapianokey,hewouldstilldosomethingoutofsheerperversity—hewouldcreatedestructionandchaos—justtogainhispoint….Andifall thiscould in turnbeanalyzedandpreventedbypredicting that itwouldoccur,thenmanwoulddeliberatelygomadtoprovehispoint.”Thisisaconceivableneuroticreaction to inept control. A few men may have shown it, and many have enjoyedDostoevsky’s statement because they tend to show it. But that such perversity is a

28

fundamental reaction of the human organism to controlling conditions is sheernonsense.So is the objection thatwe have noway of knowingwhat changes tomake even

thoughwehavethenecessarytechniques.Thatisoneofthegreathoaxesofthecentury—asortofboobytrapleftbehindintheretreatbeforetheadvancingfrontofscience.Scientists themselves have unsuspectingly agreed that there are two kinds of usefulprepositionsaboutnature—factsandvaluejudgments—andthatsciencemustconfineitselfto“whatis,”leaving“whatoughttobe”toothers.Butwithwhatspecialsortofwisdomisthenonscientistendowed?Scienceisonlyeffectiveknowing,nomatterwhoengagesinit.Verbalbehaviorprovesuponanalysistobecomposedofmanydifferenttypesofutterances, frompoetry and exhortation to logic and factualdescription, butthese are not all equally useful in talking about cultural practices.Wemay classifyuseful propositions according to the degrees of confidencewithwhich theymay beasserted.Sentencesaboutnaturerangefromhighlyprobable“facts” tosheerguesses.In general, future events are less likely to be correctly described than past.When ascientist talks about a projected experiment, for example, he must often resort tostatements having only a moderate likelihood of being correct; he calls themhypotheses.Designinganewculturalpatternisinmanywayslikedesigninganexperiment.In

drawing up a new constitution, outlining a new educational program, modifying areligious doctrine, or setting up a new fiscal policy,many statementsmust be quitetentative.Wecannotbesurethatthepracticeswespecifywillhavetheconsequenceswepredict, or that the consequenceswill rewardour efforts.This is in thenatureofsuchproposals.Theyarenotvaluejudgments—theyareguesses.Toconfuseanddelaytheimprovementofculturalpracticesbyquibblingaboutthewordimproveisitselfnotausefulpractice.Letusagree, tostartwith, thathealthisbetter thanillness,wisdombetterthanignorance,lovebetterthanhate,andproductiveenergybetterthanneuroticsloth.Another familiar objection is the “political problem.” Though we know what

changes to make and how to make them, we still need to control certain relevantconditions,butthesehavelongsincefallenintothehandsofselfishmenwhoarenotgoingtorelinquishthemforsuchpurposes.Possiblyweshallbepermittedtodevelopareaswhichatthemomentseemunimportant,butatthefirstsignsofsuccessthestrongmen will move in. This, it is said, has happened to Christianity, democracy, andcommunism.Therewillalwaysbemenwhoarefundamentallyselfishandevil,andinthe long run innocent goodness cannot have its way. The only evidence here ishistorical, and itmay bemisleading. Because of theway inwhich physical sciencedeveloped,historycoulduntilvery recentlyhave“proved” that theunleashingof theenergyoftheatomwasquiteunlikely,ifnotimpossible.Similarly,becauseoftheorderinwhichprocessesinhumanbehaviorhavebecomeavailableforpurposesofcontrol,history may seem to prove that power will probably be appropriated for selfishpurposes. The first techniques to be discovered fell almost always to strong, selfishmen.HistoryledLordActontobelievethatpowercorrupts,buthehadprobablyneverencounteredabsolutepower,certainlynotinallitsforms,andhadnowayofpredictingitseffect.An optimistic historian could defend a different conclusion. The principle that if

there are not enoughmen of goodwill in theworld the first step is to createmoreseems to be gaining recognition. TheMarshall Plan (as originally conceived), Point

29

Four,theofferofatomicmaterialstopower-starvedcountries—thesemayormaynotbewhollynewinthehistoryofinternationalrelations,buttheysuggestanincreasingawarenessofthepowerofgovernmentalgoodwill.Theyareproposalstomakecertainchanges in the environments ofmen for the sake of consequences which should berewardingforallconcerned.Theydonotexemplifyadisinterestedgenerosity,butaninterestwhich is the interestofeveryone.Wehavenotyet seenPlato’sphilosopher-king,andmaynotwantto,butthegapbetweenrealandutopiangovernmentisclosing.

III

Butwearenotyetintheclear,foranewandunexpectedobstaclehasarisen.Withaworld of their ownmaking almostwithin reach,men of goodwill have been seizedwithdistastefortheirachievement.Theyhaveuneasilyrejectedopportunitiestoapplythe techniques and findings of science in the service of men, and as the import ofeffective cultural design has come to be understood, many of them have voiced anoutrightrefusaltohaveanypartinit.Sciencehasbeenchallengedbeforewhenithasencroached upon institutions already engaged in the control of human behavior; butwhat are we to make of benevolent men, with no special interests of their own todefend, who nevertheless turn against the very means of reaching long-dreamed-ofgoals?Whatisbeingrejected,ofcourse,isthescientificconceptionofmanandhisplacein

nature. So long as the findings andmethods of science are applied to human affairsonly in a sort of remedial patchwork,wemay continue to hold any view of humannature we like. But as the use of science increases, we are forced to accept thetheoreticalstructurewithwhichscience represents its facts.Thedifficulty is that thisstructure is clearly at oddswith the traditionaldemocratic conceptionofman.Everydiscoveryofaneventwhichhasapartinshapingaman’sbehaviorseemstoleavesomuchthelesstobecreditedtothemanhimself;andassuchexplanationsbecomemoreandmore comprehensive, the contribution which may be claimed by the individualhimself appears to approach zero. Man’s vaunted creative powers, his originalaccomplishments in art, science, andmorals, his capacity to choose andour right toholdhimresponsiblefortheconsequencesofhischoice—noneoftheseisconspicuousin this new self-portrait.Man,we once believed,was free to express himself in art,music,andliterature,toinquireintonature,toseeksalvationinhisownway.Hecouldinitiateactionandmakespontaneousandcapriciouschangesofcourse.Underthemostextreme duress some sort of choice remained to him. He could resist any effort tocontrolhim,thoughitmightcosthimhislife.Butscienceinsiststhatactionisinitiatedby forces impinging upon the individual, and that caprice is only another name forbehaviorforwhichwehavenotyetfoundacause.In attempting to reconcile these views it is important to note that the traditional

democraticconceptionwasnotdesignedasadescriptioninthescientificsensebutasaphilosophytobeusedinsettingupandmaintainingagovernmentalprocess. Itaroseunderhistoricalcircumstancesandservedpoliticalpurposesapartfromwhichitcannotbe properly understood. In rallying men against tyranny it was necessary that theindividual be strengthened, that he be taught that he had rights and could governhimself.Togivethecommonmananewconceptionofhisworth,hisdignity,andhispower to save himself, both here and hereafter, was often the only resource of the

30

revolutionist.When democratic principleswere put into practice, the same doctrineswere used as a working formula. This is exemplified by the notion of personalresponsibility in Anglo-American law. All governments make certain forms ofpunishment contingent upon certain kinds of acts. In democratic countries thesecontingencies are expressedby thenotionof responsible choice.But thenotionmayhave nomeaning under governmental practices formulated in otherways andwouldcertainlyhavenoplaceinsystemswhichdidnotusepunishment.The democratic philosophy of human nature is determined by certain political

exigencies and techniques, not by the goals of democracy. But exigencies andtechniques change; and a conception which is not supported for its accuracy as alikeness—is not, indeed, rooted in fact at all—may be expected to change too. Nomatterhoweffectivewejudgecurrentdemocraticpracticestobe,howhighlywevaluethemorhow longweexpect them to survive, theyare almost certainlynot the finalform of government. The philosophy of human nature which has been useful inimplementingthemisalsoalmostcertainlynotthelastword.Theultimateachievementofdemocracymaybelongdeferredunlessweemphasizetherealaimsratherthantheverbaldevicesofdemocraticthinking.Aphilosophywhichhasbeenappropriatetooneset of political exigencies will defeat its purpose if, under other circumstances, itpreventsusfromapplyingtohumanaffairsthescienceofmanwhichprobablynothingbutdemocracyitselfcouldhaveproduced.

IV

Perhaps themostcrucialpartofourdemocraticphilosophy tobe reconsidered isourattitudetowardfreedom—oritsreciprocal,thecontrolofhumanbehavior.Wedonotopposeallformsofcontrolbecauseitis“humannature”todoso.Thereactionisnotcharacteristicofallmenunderall conditionsof life. It is anattitudewhichhasbeencarefullyengineered,inlargepartbywhatwecallthe“literature”ofdemocracy.Withrespect to some methods of control (for example, the threat of force), very littleengineering is needed, for the techniques or their immediate consequences areobjectionable. Society has suppressed these methods by branding them “wrong,”“illegal,”or “sinful.”But to encourage these attitudes towardobjectionable formsofcontrol, it has been necessary to disguise the real nature of certain indispensabletechniques, the commonest examples of which are education, moral discourse, andpersuasion.Theactualproceduresappearharmlessenough.Theyconsistofsupplyinginformation, presenting opportunities for action, pointing out logical relationships,appealing to reason or “enlightened understanding,” and so on. Through amasterfulpieceofmisrepresentation,theillusionisfosteredthattheseproceduresdonotinvolvethecontrolofbehavior;atmost,theyaresimplywaysof“gettingsomeonetochangehis mind.” But analysis not only reveals the presence of well-defined behavioralprocesses, itdemonstratesakindofcontrolno less inexorable, though insomewaysmoreacceptable,thanthebully’sthreatofforce.Letus suppose that someone inwhomweare interested isactingunwisely—he is

carelessinthewayhedealswithhisfriends,hedrivestoofast,orheholdshisgolfclubthewrongway.Wecouldprobablyhelphimbyissuingaseriesofcommands:don’tnag, don’t drive over sixty, don’t hold your club thatway.Much less objectionablewouldbe“anappeal to reason.”Wecouldshowhimhowpeopleareaffectedbyhis

31

treatmentof them,howaccident rates rise sharplyathigher speeds,howaparticulargripon theclubalters theway theball is struckandcorrectsa slice. Indoingsoweresorttoverbalmediatingdeviceswhichemphasizeandsupportcertain“contingenciesofreinforcement”—thatis,certainrelationsbetweenbehavioranditsconsequences—which strengthen the behavior we wish to set up. The same consequences wouldpossibly set up the behavior without our help, and they eventually take control nomatterwhichformofhelpwegive.Theappealtoreasonhascertainadvantagesoverthe authoritative command.A threat of punishment, nomatter how subtle, generatesemotionalreactionsandtendenciestoescapeorrevolt.Perhapsthecontrolleemerely“feelsresentment”atbeingmadetoactinagivenway,buteventhatistobeavoided.Whenwe“appealtoreason,”he“feelsfreertodoashepleases.”Thefactisthatwehaveexertedlesscontrolthaninusingathreat;sinceotherconditionsmaycontributetotheresult,theeffectmaybedelayedor,possiblyinagiveninstance,lacking.Butifwehaveworkedachangeinhisbehavioratall,itisbecausewehavealteredrelevantenvironmentalconditions,andtheprocesseswehavesetinmotionarejustasrealandjustasinexorable,ifnotascomprehensive,asinthemostauthoritativecoercion.“Arranginganopportunityforaction”isanotherexampleofdisguisedcontrol.The

power of the negative form has already been exposed in the analysis of censorship.Restrictionofopportunity isrecognizedasfarfromharmless.AsRalphBartonPerrysaid in an article which appeared in the Spring, 1953,PacificSpectator, “Whoeverdetermineswhatalternativesshallbemadeknowntomancontrolswhatthatmanshallchoose from.He is deprived of freedom in proportion as he is denied access toanyideas,orisconfinedtoanyrangeofideasshortofthetotalityofrelevantpossibilities.”But there is a positive side aswell.Whenwe present a relevant state of affairs,weincreasethelikelihoodthatagivenformofbehaviorwillbeemitted.Totheextentthatthe probability of action has changed, we have made a definite contribution. Theteacherofhistorycontrolsastudent’sbehavior(or,ifthereaderprefers,“depriveshimoffreedom”)justasmuchinpresentinghistoricalfactsasinsuppressingthem.Otherconditionswillnodoubtaffectthestudent,butthecontributionmadetohisbehaviorbythepresentationofmaterialisfixedand,withinitsrange,irresistible.The methods of education, moral discourse, and persuasion are acceptable not

becausetheyrecognizethefreedomoftheindividualorhisrighttodissent,butbecausetheymakeonlypartialcontributionstothecontrolofhisbehavior.Thefreedomtheyrecognize is freedom from amore coercive form of control. The dissentwhich theytolerate is the possible effect of other determiners of action. Since these sanctionedmethodsarefrequentlyineffective,wehavebeenabletoconvinceourselvesthattheydonotrepresentcontrolatall.Whentheyshowtoomuchstrengthtopermitdisguise,wegivethemothernamesandsuppressthemasenergeticallyaswesuppresstheuseofforce. Education grown too powerful is rejected as propaganda or “brain-washing,”while really effective persuasion is described as “undue influence,” “demagoguery,”“seduction,”andsoon.If we are not to rely solely upon accident for the innovations which give rise to

culturalevolution,wemustacceptthefactthatsomekindofcontrolofhumanbehavioris inevitable.Wecannotusegoodsense inhumanaffairsunlesssomeoneengages inthedesignandconstructionofenvironmentalconditionswhichaffect thebehaviorofmen.Environmentalchangeshavealwaysbeen theconditionfor the improvementofculturalpatterns,andwecanhardlyusethemoreeffectivemethodsofsciencewithoutmakingchangesonagrander scale.Weare all controlledby theworld inwhichwe

32

live,andpartofthatworldhasbeenandwillbeconstructedbymen.Thequestionisthis: Are we to be controlled by accident, by tyrants, or by ourselves in effectiveculturaldesign?Thedangerofthemisuseofpowerispossiblygreaterthanever.Itisnotallayedby

disguising the facts.We cannotmake wise decisions if we continue to pretend thathumanbehavior isnotcontrolled,or ifwerefuse toengage incontrolwhenvaluableresults might be forthcoming. Such measures weaken only ourselves, leaving thestrengthofscience toothers.The first step inadefenseagainst tyranny is the fullestpossible exposure of controlling techniques. A second step has already been takensuccessfullyinrestrictingtheuseofphysicalforce.Slowly,andasyetimperfectly,wehaveworkedout an ethical andgovernmentaldesign inwhich the strongman isnotallowed touse thepowerderiving fromhisstrength tocontrolhis fellowmen.He isrestrained by a superior force created for that purpose—the ethical pressure of thegroup, or more explicit religious and governmental measures. We tend to distrustsuperiorforces,aswecurrentlyhesitatetorelinquishsovereigntyinordertosetupaninternational police force. But it is only through such counter-control that we haveachievedwhatwecallpeace—acondition inwhichmenarenotpermitted tocontroleachotherthroughforce.Inotherwords,controlitselfmustbecontrolled.Sciencehasturnedupdangerousprocessesandmaterialsbefore.Tousethefactsand

techniquesof a scienceofman to the fullest extentwithoutmaking somemonstrousmistake will be difficult and obviously perilous. It is no time for self-deception,emotionalindulgence,ortheassumptionofattitudeswhicharenolongeruseful.Manisfacingadifficulttest.Hemustkeephisheadnow,orhemuststartagain—alongwayback.

V

Those who reject the scientific conception of man must, to be logical, oppose themethodsof scienceaswell.Theposition isoften supportedbypredictinga seriesofdire consequences which are to follow if science is not checked. A recent book byJosephWoodKrutch,TheMeasure ofMan, is in this vein. Mr. Krutch sees in thegrowing science of man the threat of an unexampled tyranny over men’s minds. Ifscienceispermittedtohaveitsway,heinsists,“wemayneverbeablereallytothinkagain.”Acontrolledculturewill, forexample, lack somevirtue inherent indisorder.Wehaveemergedfromchaosthroughaseriesofhappyaccidents,butinanengineeredcultureitwillbe“impossiblefortheunplannedtoeruptagain.”Butthereisnovirtueintheaccidentalcharacterofanaccident,andthediversitywhicharisesfromdisordercannot only be duplicated by design but vastly extended. The experimental method issuperior to simple observation just because it multiplies “accidents” in a systematiccoverageofthepossibilities.Technologyoffersmanyfamiliarexamples.Wenolongerwaitforimmunitytodiseasetodevelopfromaseriesofaccidentalexposures,nordowe wait for natural mutations in sheep and cotton to produce better fibers; but wecontinue to make use of such accidents when they occur, and we certainly do notprevent them.Manyof the thingswevaluehaveemergedfromtheclashof ignorantarmies on darkling plains, but it is not therefore wise to encourage ignorance anddarkness.Itisnotalwaysdisorderitselfwhichwearetoldweshallmissbutcertainadmirable

33

qualitiesinmenwhichflourishonlyinthepresenceofdisorder.Amanrisesaboveanunpropitiouschildhoodtoapositionofeminence,andsincewecannotgiveaplausibleaccountof theactionofsocomplexanenvironment,weattribute theachievement tosomeadmirablefacultyinthemanhimself.Butsuch“faculties”aresuspiciouslylikethe explanatory fictions against which the history of science warns us. We admireLincolnforrisingaboveadeficientschoolsystem,butitwasnotnecessarilysomethinginhimwhichpermittedhimtobecomeaneducatedmaninspiteofit.Hiseducationalenvironment was certainly unplanned, but it could nevertheless have made a fullcontributiontohismaturebehavior.Hewasarareman,but thecircumstancesofhischildhoodwereraretoo.WedonotgiveFranklinDelanoRooseveltthesamecreditforbecomingan educatedmanwith thehelpofGrotonandHarvard, although the samebehavioral processesmayhave been involved.The founding ofGroton andHarvardsomewhatreducedthepossibilitythatfortuitouscombinationsofcircumstanceswoulderupt to produce other Lincolns. Yet the founders can hardly be condemned forattackinganadmirablehumanquality.Anotherpredictedconsequenceofascienceofmanisanexcessiveuniformity.We

are told that effective control—whether governmental, religious, educational,economic, or social—will produce a race of men who differ from each other onlythroughrelatively refractorygeneticdifferences.Thatwouldprobablybebaddesign,but wemust admit that we are not now pursuing another course from choice. In amodern school, for example, there is usually a syllabus which specifies what everystudent is to learn by the end of each year. Thiswould be flagrant regimentation ifanyoneexpectedeverystudenttocomply.Butsomewillbepoorinparticularsubjects,others will not study, others will not remember what they have been taught, anddiversity is assured. Suppose, however, that we someday possess such effectiveeducational techniques that every studentwill in fact be put in possession of all thebehavior specified in a syllabus. At the end of the year, all students will correctlyanswerallquestionson the finalexaminationand“mustallhaveprizes.”Shouldwerejectsuchasystemon thegrounds that inmakingallstudentsexcellent ithasmadethem all alike? Advocates of the theory of a special faculty might contend that animportantadvantageofthepresentsystemisthatthegoodstudentlearnsinspiteof asystemwhichissodefectivethatitiscurrentlyproducingbadstudentsaswell.Butifreallyeffectivetechniquesareavailable,wecannotavoidtheproblemofdesignsimplyby preferring the status quo. At what point should education be made deliberatelyinefficient?Suchpredictionsofthehavoctobewreakedbytheapplicationofsciencetohuman

affairsareusuallymadewithsurprisingconfidence.Theynotonlyshowafaithintheorderlinessofhumanbehavior;theypresupposeanestablishedbodyofknowledgewiththehelpofwhichitcanbepositivelyassertedthatthechangeswhichscientistsproposetomakewill have quite specific results—albeit not the results they foresee. But thepredictions made by the critics of science must be held to be equally fallible andsubjectalsotoempiricaltest.Wemaybesurethatmanystepsinthescientificdesignofculturalpatternswillproduceunforeseenconsequences.Butthereisonlyonewaytofindout.Andthetestmustbemade,forifwecannotadvanceinthedesignofculturalpatterns with absolute certainty, neither can we rest completely confident of thesuperiorityofthestatusquo.

34

VI

Apartfromtheirpossiblyobjectionableconsequences,scientificmethodsseemtomakenoprovisionforcertainadmirablequalitiesandfacultieswhichseemtohaveflourishedin less explicitly planned cultures; hence they are called “degrading” or “lacking indignity.” (Mr.Krutchhascalled theauthor’sWaldenTwo an“ignobleUtopia.”)Theconditionedreflex is thecurrentwhippingboy.Becauseconditionedreflexesmaybedemonstrated in animals, they are spoken of as though they were exclusivelysubhuman.Itisimplied,aswehaveseen,thatnobehavioralprocessesareinvolvedineducation andmoral discourseor, at least, that theprocesses are exclusivelyhuman.Butmendoshowconditionedreflexes (forexample,when theyare frightenedbyallinstancesofthecontrolofhumanbehaviorbecausesomeinstancesengenderfear),andanimalsdoshowprocessessimilartothehumanbehaviorinvolvedininstructionandmoraldiscourse.WhenMr.Krutchassertsthat“‘Conditioning’isachievedbymethodswhich bypass or, as it were, short-circuit those very reasoning faculties whicheducationproposestocultivateandexercise,”heismakingatechnicalstatementwhichneedsadefinitionoftermsandagreatdealofsupportingevidence.Ifsuchmethodsarecalled“ignoble”simplybecausetheyleavenoroomforcertain

admirableattributes,thenperhapsthepracticeofadmirationneedstobeexamined.Wemightsaythatthechildwhoseeducationhasbeenskillfullyplannedhasbeendeprivedoftherighttointellectualheroism.Nothinghasbeenlefttobeadmiredinthewayheacquires an education. Similarly, we can conceive of moral training which is soadequatetothedemandsoftheculturethatmenwillbegoodpracticallyautomatically,buttothatextenttheywillbedeprivedoftherighttomoralheroism,sinceweseldomadmire automaticgoodness.Yet ifweconsider the endofmorals rather thancertainvirtuousmeans, isnot“automaticgoodness”adesirablestateofaffairs?Is itnot, forexample,theavowedgoalofreligiouseducation?T.H.Huxleyansweredthequestionunambiguously:“If somegreatpowerwouldagree tomakemealways thinkwhat istrueanddowhatisright,onconditionofbeingturnedintoasortofclockandwoundupeverymorningbeforeIgotoutofbed,Ishouldinstantlyclosewiththeoffer.”YetMr.Krutchquotesthisasthescarcelycrediblepointofviewofa“proto-modern”andseemshimselftoshareT.S.Eliot’scontemptfor“…systemssoperfect/Thatnoonewillneedtobegood.”“Having to be good” is an excellent example of an expendable honorific. It is

inseparable from a particular form of ethical and moral control. We distinguishbetween the thingswehave to do to avoidpunishment and thosewewant to do forrewarding consequences. In a culturewhich did not resort to punishmentwe shouldnever“have”todoanythingexceptwithrespecttothepunishingcontingencieswhicharisedirectlyinthephysicalenvironment.Andwearemovingtowardsuchaculture,becausetheneurotic,nottosaypsychotic,by-productsofcontrolthroughpunishmenthavelongsinceledcompassionatementoseekalternativetechniques.Recentresearchhas explained some of the objectionable results of punishment and has revealedresourcesofat leastequalpower in“positivereinforcement.”It is reasonable to lookforwardtoatimewhenmanwillseldom“have”todoanything,althoughhemayshowinterest, energy, imagination, and productivity far beyond the level seen under thepresentsystem(exceptforrareeruptionsoftheunplanned).Whatwehavetodowedowitheffort.Wecallit“work.”Thereisnootherwayto

35

distinguishbetweenexhaustinglaborandthepossiblyequallyenergeticbutrewardingactivityofplay. It ispresumablygoodculturaldesign to replace the formerwith thelatter. But an adjustment in attitudes is needed. We are much more practiced inadmiring the heroic labor of aHercules than the activity of onewhoworkswithouthavingto.Inatrulyeffectiveeducationalsystemthestudentmightnot“havetowork”atall,butthatpossibilityislikelytobereceivedbythecontemporaryteacherwithanemotionlittleshortofrage.Wecannotreconciletraditionalandscientificviewsbyagreeinguponwhatistobe

admiredorcondemned.Thequestioniswhetheranythingistobesotreated.Praiseandblame are cultural practices which have been adjuncts of the prevailing system ofcontrol in Western democracy. All peoples do not engage in them for the samepurposesortothesameextent,nor,ofcourse,arethesamebehaviorsalwaysclassifiedin the same way as subject to praise or blame. In admiring intellectual and moralheroism and unrewarding labor, and in rejecting a world in which these would beuncommon,wearesimplydemonstratingourownculturalconditioning.Bypromotingcertaintendenciestoadmireandcensure,thegroupofwhichweareaparthasarrangedfor the social reinforcement and punishment needed to assure a high level ofintellectualandmoralindustry.Underotherandpossiblybettercontrollingsystems,thebehavior which we now admire would occur, but not under those conditions whichmake it admirable, and we should have no reason to admire it because the culturewouldhavearrangedforitsmaintenanceinotherways.Tothosewhoarestimulatedbytheglamorousheroismofthebattlefield,apeaceful

worldmaynotbeabetterworld.Othersmayrejectaworldwithoutsorrow,longing,orasenseofguiltbecausetherelevanceofdeeplymovingworksofartwouldbelost.Tomanywhohavedevotedtheirlivestothestruggletobewiseandgood,aworldwithoutconfusion and evilmight be an empty thing.A nostalgic concern for the decline ofmoralheroismhasbeenadominatingthemeintheworkofAldousHuxley.InBraveNewWorldhecouldseeintheapplicationofsciencetohumanaffairsonlyatravestyonthenotionoftheGood(justasGeorgeOrwell,in1984,couldforeseenothingbuthorror).Writing inEsquire (August,1955)Huxleyhasexpressed thepoint thisway:“We have had religious revolutions,we have had political, industrial, economic andnationalisticrevolutions.Allofthem,asourdescendantswilldiscover,werebutripplesinanoceanofconservatism—trivialbycomparisonwiththepsychologicalrevolutiontowardwhichweare so rapidlymoving.Thatwill reallybea revolution.When it isover, the human race will give no further trouble.” (Footnote for the reader of thefuture:Thiswasnotmeantasahappyending.Upto1956menhadbeenadmired,ifatall,eitherforcausingtroubleoralleviatingit.Therefore—)Itwillbealongtimebeforetheworldcandispensewithheroesandhencewiththe

culturalpracticeofadmiringheroism,butwemoveinthatdirectionwheneverweacttopreventwar, famine, pestilence, and disaster. Itwill be a long time beforemanwillneverneedtosubmittopunishingenvironmentsorengageinexhaustinglabor,butwemove in that direction whenever we make food, shelter, clothing, and labor-savingdevices more readily available. We may mourn the passing of heroes but not theconditionswhichmake forheroism.Wecan spare the self-made saintor sage aswespare the laundress on the river’s bank struggling against fearful odds to achievecleanliness.

36

VII

The two great dangers in modern democratic thinking are illustrated in a paper byformerSecretaryofStateDeanAcheson.“Foralongtimenow,”writesMr.Acheson,“wehavegonealongwithsomewell-testedprinciplesofconduct:Thatitwasbettertotell the truth than falsehoods;… that duties were older than and as fundamental asrights; that, as JusticeHolmesput it, themodebywhich the inevitablecame topasswaseffort;thattoperpetuateaharmwaswrongnomatterhowmanyjoinedinit…andsoon….Ourinstitutionsarefoundedontheassumptionthatmostpeoplefollowtheseprinciplesmostofthetimebecausetheywantto,andtheinstitutionsworkprettywellwhenthisassumptionistrue.Morerecently,however,brightpeoplehavebeenfoolingwith themachinery in thehumanheadandtheyhavediscoveredquitea lot….Hitlerintroducednewrefinements[astheresultofwhich]awholepeoplehavebeenutterlyconfused and corrupted.Unhappily neither the possession of this knowledge nor thedesiretouseitwasconfinedtoHitler….Othersdipfromthissamedevil’scauldron.”1The first dangerous notion in this passage is that most people follow democratic

principlesofconduct“becausetheywantto.”Thisdoesnotaccountfordemocracyoranyotherformofgovernmentifwehavenotexplainedwhypeoplewanttobehaveingivenways.Although it is tempting to assume that it is human nature to believe indemocraticprinciples,wemustnotoverlookthe“culturalengineering”whichproducedand continues to maintain democratic practices. If we neglect the conditions whichproduce democratic behavior, it is useless to try to maintain a democratic form ofgovernment. And we cannot expect to export a democratic form of governmentsuccessfully ifwedonotalsoprovide for theculturalpracticeswhichwill sustain it.Our forebearsdidnotdiscover theessentialnatureofman; theyevolvedapatternofbehavior which worked remarkably well under the circumstances. The “set ofprinciples”expressedinthatpatternisnottheonlytruesetornecessarilythebest.Mr.Achesonhaspresumablylistedthemostunassailableitems;someofthemareprobablybeyond question, but others—concerning duty and effort—may need revision as theworldchanges.Thesecond—andgreater—threattothedemocracywhichMr.Achesonisdefending

ishisassumption thatknowledge isnecessarilyonthesideofevil.All theadmirablethingshementionsareattributed to the innategoodnessofman,all thedetestable to“foolingwith themachinery in thehumanhead.”This is reminiscentof theposition,taken by other institutions engaged in the control of men, that certain forms ofknowledgeare in themselvesevil.Buthowoutofplace inademocraticphilosophy!Havewecomethisfaronlytoconcludethatwell-intentionedpeoplecannotstudythebehavior ofmenwithout becoming tyrants or that informedmen cannot show goodwill?Letusforoncehavestrengthandgoodwillonthesameside.

VIII

FarfrombeingathreattothetraditionofWesterndemocracy,thegrowthofascienceof man is a consistent and probably inevitable part of it. In turning to the externalconditionswhichshapeandmaintainthebehaviorofmen,whilequestioningtherealityofinnerqualitiesandfacultiestowhichhumanachievementswereonceattributed,we

37

turnfromtheill-definedandremotetotheobservableandmanipulable.Thoughitisapainfulstep, ithasfar-reachingconsequences,for itnotonlysetshigherstandardsofhumanwelfarebutshowsushowtomeetthem.Achangeinatheoryofhumannaturecannotchangethefacts.Theachievementsofmaninscience,art,literature,music,andmorals will survive any interpretation we place upon them. The uniqueness of theindividual is unchallenged in the scientific view. Man, in short, will remain man.(There will be much to admire for those who are so inclined. Possibly the noblestachievementtowhichmancanaspire,evenaccordingtopresentstandards,istoaccepthimselfforwhatheis,asthatisrevealedtohimbythemethodswhichhedevisedandtestedonapartoftheworldinwhichhehadonlyasmallpersonalstake.)IfWesterndemocracydoesnotlosesightoftheaimsofhumanitarianaction,itwill

welcome the almost fabulous support of its own science ofman andwill strengthenitselfandplayanimportantroleinbuildingabetterworldforeveryone.Butifitcannotputits“democraticphilosophy”intoproperhistoricalperspective—if,underthecontrolofattitudesandemotionswhichitgeneratedforotherpurposes,itnowrejectsthehelpofscience—thenitmustbepreparedfordefeat.Forifwecontinuetoinsistthatsciencehasnothingtoofferbutanewandmorehorribleformoftyranny,wemayproducejustsucharesultbyallowingthestrengthofsciencetofallintothehandsofdespots.Andif,withluck,itweretofallinsteadtomenofgoodwillinotherpoliticalcommunities,it would be perhaps a more ignominious defeat; for we should then, through amiscarriageofdemocraticprinciples,beforcedtoleavetoothersthenextstepinman’slongstruggletocontrolnatureandhimself.

1ThePatternofResponsibility.Boston,1952.pages14–15.

38

TheControlofHumanBehavior(Abstract)

AshortenedversionofalecturegivenattheNewYorkAcademyofSciencesonApril18,1955,thispaperappearedintheTransactionsoftheAcademy(Series II,Vol.17,No.7,pp.547-551)inMayofthatyear.

Weareseldomwillingtoadmitthatweareengagedincontrollingthebehaviorofotherpeople.Thecommonest techniquesofcontroluseforceor the threatofforceandareobjectionable to the controllee and have come to be censured by society. But thecondoned techniquesofeducation,persuasion,andmoraldiscoursedifferonly in thebehavioralprocessesthroughwhichtheyoperateandintheminimizingofcertainsideeffects.Theyarestilldevicesthroughwhichonemancontrolsthebehaviorofanotherin some measure. Cajolery, seduction, incitement, and the various forms of whatbiographerscall“influence”suggestothertechniques.Familiar rules of thumb in controllingmen are embedded in folk wisdom and in

manygreatworksofliterature.Thisprescientifictechnologyisrapidlybeingextendedbythescientificstudyofhumanbehavior(therearethosewhorefusetoadmiteventhepossibilityofsuchascience,butIamspeakingheretothosewhoarenotonlyawareofthesciencebutshareadeepconcernfor itsconsequences). Incivilizedcountries, themore powerful controlling techniques have eventually been contained by a sort ofethicalcounter-control,whichpreventsexploitationbythoseinapositiontousethem.There is a real danger, however, that the rapid development of new techniqueswilloutstripappropriatemeasuresofcounter-control,withdevastatingresults.Wecanseehowcounter-controloriginatesinthecaseofforceorthethreatofforce.

In primitive literature, the hero is often themanwho canwhip everyone else in thegroup in open combat. He controls with the techniques of the bully. The relevantprocesseshavebeenanalyzedinthescientificstudyofbehaviorundertheheadingsofavoidanceandescape.Weseethesetechniquesexemplifiedtodayinthegovernmentofconqueredpeoples, indespoticgovernmentsofall sorts,by religiousagencieswhichlean heavily on the threat of punishment, by many parents in the control of theirchildren, and by most teachers. The technique is psychologically and biologicallyharmfultothecontrolleeand,forthisreason,hasgeneratedcounter-control.Theweakare, at least,morenumerous, andwenowgenerallyhold it tobe“wrong” tocontrolthroughtheuseofforceorthethreatofforce(althoughanimpartialobservermightnotcometothisconclusion).Formalizedgovernmentalandreligiouspreceptssupportthiscontainmentofthetechniquesofthebully.Theresult iscalledpeace—aconditioninwhichmenarenotpermittedtouseforceincontrollingeachother.A later type of popular hero is the cheat, who outwits the strong man by

misrepresentationanddeceit(ina technicalanalysis, therelevantprocesseswouldbeclassified under the extinction of conditioned reflexes). But the cheat, eventually, isalmostasobjectionableasthebully,andethicalcontrolaccordinglyarises.Itisheldtobe“wrong”tolie,cheat,orcry“Wolf”forone’samusement.

39

Therearetechniqueswhichmaybeaseffectiveasthesebutmaynotleadsodirectlytocounter-control.Thesetechniquesarebecomingmorepowerfulastheirprocessesarebetterunderstood.Afewexamplesfollow.1. Emotional conditioning. Aldous Huxley, in Brave New World, describes a

perfectly plausible process through which certain inferior types of citizens arepermanentlydissuadedfromwastingtimeonbooksandthebeautiesofnature.Babiesareallowedtocrawltowardbooksandflowersbutreceiveelectricshocksjustastheytouchthem.Theexampleappearstobeborrowed,notfromthescienceofconditionedreflexes,but fromcertain formsofmoraleducation inwhich, forexample,achild isspankedfortakinganinterestinpartsofhisownbody.Thesameprincipleisusedtogenerate strong reactions of rage and aggression toward the enemy in preparingservicemenforcombat.Itisthebasisofadvertisingwhichshowsaproductbeingusedbyorotherwiseassociatedwithprettygirlsoradmiredpublicfigures.Thecontrolleeisnot likely to revolt against such control, and he may carry the resulting prejudicescontentedlytohisgrave.2.Motivationalcontrol.Crudeinstances,suchas thestarvingofawholepeopleso

thatfoodmaybeusedtoreinforcethosewhobegintosupport thegovernment,bringtheirowneventualcontainment,buttheexploitationofprevailingdeprivationsmaybemoresubtleandpossiblyequallyeffective.Thedeliberatedesignofartandliterature(asinthemoviesand“comics”)toappealtopeoplewithsadistictendenciesiseasilydetected,butthesubtledesignofanautomobilesothatridinginitisinsomemeasureasexual experience is not so easily spotted. Neither practicemaymeet any objectionfromthepeoplesocontrolled.3. Positive reinforcement. Wages, bribes, and tips suggest a classical pattern in

whichwegeneratebehaviorinothersthroughreinforcementorreward.Betterwaysofusing reinforcement in shaping up new behavior and in maintaining the conditioncalledinterest,orenthusiasm,havebeenrecentlydiscovered.Thereinforcingeffectofpersonal attention and affection is coming to be better understood, especially byclinicalpsychologists.LordChesterfieldandDaleCarnegiehaverecommendedtheuseoffeignedattentionininfluencingpeople.4.Drugs.Weareenteringtheageofthechemicalcontrolofhumanbehavior.Drugs

havebeenusedforthispurposeeversincethefirstmanwasdeliberatelymadedrunk.Butbetterdrugsarenowavailable,notonlyforallayinganxietybutforotherpurposesofcontrol.Ourgovernmentwouldprobablynothesitatetouseadrugwhich,takenbyservicemen before combat, would eliminate all signs of fear, thus depriving theindividualoftheprotectivereflexeswhichmanhasacquiredthroughalongprocessofevolution. In thenot-too-distant future, themotivational and emotional conditionsofnormaldaily lifewillprobablybemaintained inanydesiredstate through theuseofdrugs.5.Knowledgeof the individual.Techniquesof control canbe effectiveonlywhen

certain facts about the controllee are known. Gathering information througheavesdropping, employing spies and informers, opening mail, and wiretapping has,fromtimetotime,comeunderethicalcounter-control,thoughthepresentstateofthisin our culture is uncertain. Meanwhile, new techniques have been developed.Somethingliketheprojectivetestsofclinicalpsychology,combinedwiththetechniqueof thepolitical trialballoon,mightmakeitpossible todiscover informationaboutanindividualorawholepeople,notonlywithouttheknowledgeofthecontrolleebutwithrespecttomattersofwhichthecontrolleehimselfhasnoclearunderstanding.

40

The doctrine that there is an absolute moral law applicable to all conditions ofhuman life discourages the analysis of controlling practices and obscures ourunderstanding of the need for counter-control. Themethods bywhichmen alter thebehaviorofothermenchange,andchangingethicalmeasuresarerequired.Atechniqueneednotbe immediatelyobjectionable to the controllee to engender counter-control.The gambler, for instance, is possibly the last person to ask for legal or moralrestrictions on gambling enterprises. The alcoholic does not usually advocate thecontrol of alcoholic beverages. Fewworkers object to being paid, even for kinds ofworkoraccording topayscheduleswhichsocietyproscribes. It is the raremanwhoobjects to the tyranny of the beautifulwoman. In all these cases, society appeals tolong-term consequences to justify measures of counter-control. Unfortunately, suchconsequencesdonotsupplyanyhard-and-fastrule.Wemustcontinuetoexperimentinculturaldesign,asnaturehasalreadyexperimented,testingtheconsequencesaswego.Wemaydealwithculturalpracticesasawhole,asin“utopian”thinking,orpiecemealby changing one counter-controlling technique at a time. Eventually, the practiceswhichmake for the greatest biological and psychological strength of the groupwillpresumably survive, aswill thegroupwhichadopts them.Survival is not a criterionwhichwearefreetoacceptorreject,butitis,nevertheless,theoneaccordingtowhichourcurrentdecisionswilleventuallybe tested. It is lessclear-cut thansomeabsolutecriterionofrightandwrong,butitismorereassuringinitsrecognitionofthechangingneedsofsociety.Suchanexperimentalattitudeissometimescriticizedbythosewhowanttodefend

someprincipleappropriatetoanearlierstageofourculturalhistory.AnexampleistherecentbookbyJosephWoodKrutch,TheMeasureofMan,which is inconsiderablepart an attack onmy utopian novel,WaldenTwo. While arguing that the notion ofbehavioral engineering is ultimately faulty, because man is in some sense free andhencemayescapecontrol,Krutchadmitsthathumanfreedomisunderattackandthat,ifscienceisnotchecked,freedommayvanishaltogether.Krutcharguesthatunlessweput a stop to the machinations of scientists “we may never really be able to thinkagain.”Byfreedom,Krutchseemstomeanmerelyalackoforder.Thevirtuesoftheprescientificerawerethevirtuesofaccident.ThegreatcrimeofthefounderofWaldenTwo,accordingtoKrutch,wasthedestructionofthepossibilityofthehappychance—evensuchasthatwhichgaverisetothefounderhimself,before“men’sthoughtswerecontrolledwithprecision.”Onthesamegrounds,wemightobjecttothesyntheticfibreindustry for circumventing the accidental evolutionary processes which producedcottonandwool. Ifwecanarrangebetterconditionsofhuman lifeandgrowth,whyshouldwewait for thehappyaccident,evenifpastaccidentshavebroughtus to thisverypointofpower?Krutch’s answer is essentially a mystical one: some vague power or faculty has

permitted man to transcend his chaotic environment, and this cannot continue tofunction in less chaotic circumstances.But the existence of suchpowers or facultiesgrowsmoredoubtfulasman’sactualachievementscometobeanalyzed.Nothingwillbelostifscienceisappliedtoeducationormoraldiscourse.Abetterwayofteachingachild to spell words meets the objection that he is not taught something called“spelling,”justasbettermoralandethicaltrainingmeetstheobjectionthatthechildnolonger “has” to be good. In the past, it was natural that some special honor shouldaccruetotheindividualwhorisesabovehisfaultyintellectualandethicaltrainingandiswise and good in spite of it.Men have been at times almost entirely occupied in

41

decidingwhat is right, intellectually andmorally.Aworld inwhich education is sosuccessful that one is naturally right in both these senses is criticized because itprovidesfornoheroismintranscendinganinadequateenvironment.Onemightaswellcriticizefireproofbuildingsbecausetheworldisthusdeprivedofbravefiremen.It is easy to object to the control of human behavior by applying the slogans of

democracy. But the democratic revolution in government and religion was directedagainstacertaintypeofcontrolonly.Menwerefreedfromautocraticrulersemployingtechniquesbasedupon forceor the threatof force. It doesnot follow thatmenwerethusfreedofallcontrol,andit ispreciselytheotherformsofcontrolwhichwemustnow learn to contain and to which the pattern of the democratic revolution isinappropriate. The democratic concept of “freedom” is no longer effective ininternationalpoliticsbecauseithaslostitspoint.Allmajorgovernmentsprofesstobegoverning for the people, and no government will bear close scrutiny of its actualpractices.Anewconceptionof the functionandpracticeofgovernment isneeded indealingwiththecounter-controloftechniquesagainstwhichthereisnorevolt.Mr.Krutchisjustifiablyconcernedlestanewtypeofdespotismarisewhichutilizes

themoreeffective techniquesofcontrolprovidedby the scienceofhumanbehavior.Buthissuggestionthatwedenythepossibilityofsuchascience,orthatweabandonit,woulddepriveusofimportanthelpinbuildingadequatesafeguardsagainstitsmisuse.Scienceposesproblems,butitalsosuggestssolutions.IncontendingthatthefounderofWaldenTwocouldaseasilyhavebeenamonster, insteadof thefairlybenevolentfigureheseemstobe,Krutchmissesthepointthat,inthelongrun,thestrengthofanygovernment depends upon the strength of the governed.Under present conditions ofcompetition,itisunlikelythatagovernmentcansurvivewhichdoesnotgoverninthebestinterestsofeveryone.Unless there is some unseen virtue in ignorance, our growing understanding of

humanbehaviorwillmake it all themore feasible todesignaworld adequate to theneedsofmen.Butwe cannotgain this advantage ifwe are towaste time defendingoutwornconceptionsofhumannature,conceptionswhichhavelongsinceservedtheiroriginal purpose of justifying special philosophies of government. A rejection ofscience at this time, in a desperate attempt to preserve a loved but inaccurateconceptionofman,wouldrepresentanunworthyretreatinman’scontinuingefforttobuildabetterworld.

42

SomeIssuesConcerningtheControlofHumanBehavior

ThisisonesideofadebatewithCarlR.Rogers,heldatameetingoftheAmericanPsychologicalAssociation on September 4, 1956.Part Iwas submitted inwriting toDr.Rogers,andhiscontributionandrebuttal(summarizedonpages34–35)weresenttomebeforethemeeting.PartIIIismyrebuttal.

43

PartI

Scienceissteadilyincreasingourpowertoinfluence,change,mold—inaword,control—humanbehavior.Ithasextendedour“understanding”(whateverthatmaybe)sothatwedealmoresuccessfullywithpeopleinnonscientificways,butithasalsoidentifiedconditionsorvariableswhichcanbeusedtopredictandcontrolbehaviorinanew,andincreasinglyrigorous,technology.Thebroaddisciplinesofgovernmentandeconomicsoffer examples of this, but there is special cogency in those contributions ofanthropology, sociology, and psychology which deal with individual behavior. CarlRogers has listed someof the achievements to date in a recent paper.1Thoseofhisexampleswhichshoworimplythecontrolofthesingleorganismareprimarilydue,asweshouldexpect,topsychology.Itistheexperimentalstudyofbehaviorwhichcarriesus beyond awkward or inaccessible “principles,” “factors,” and so on, to variableswhichcanbedirectlymanipulated.Itisalso,andformoreorlessthesamereasons,theconceptionofhumanbehavior

emerging from an experimental analysis which most directly challenges traditionalviews.Psychologists themselvesoftendonotseemtobeawareofhowfar theyhavemovedinthisdirection.Butthechangeisnotpassingunnoticedbyothers.Untilonlyrecently it was customary to deny the possibility of a rigorous science of humanbehaviorby arguing, either that a lawful sciencewas impossiblebecausemanwas afreeagent,orthatmerelystatisticalpredictionswouldalwaysleaveroomforpersonalfreedom. But those who used to take this line have become most vociferous inexpressingtheiralarmatthewaytheseobstaclesarebeingsurmounted.Now, thecontrolofhumanbehaviorhasalwaysbeenunpopular.Anyundisguised

effort to control usually arouses emotional reactions.We hesitate to admit, even toourselves,thatweareengagedincontrol,andwemayrefusetocontrol,evenwhenthiswouldbehelpful,forfearofcriticism.Thosewhohaveexplicitlyavowedaninterestincontrol have been roughly treated by history.Machiavelli is the great prototype.AsMacaulaysaidofhim,“Outofhissurnametheycoinedanepithetforaknaveandoutof his Christian name a synonym for the devil.” There were obvious reasons. ThecontrolwhichMachiavellianalyzedandrecommended,likemostpoliticalcontrol,usedtechniques aversive to the controllee. The threats and punishments of the bully, likethoseofthegovernmentoperatingonthesameplan,arenotdesigned—whatevertheirsuccess—toendearthemselvestothosewhoarecontrolled.Evenwhenthetechniquesthemselvesarenotaversive,controlisusuallyexercisedfortheselfishpurposesofthecontrollerand,hence,hasindirectlypunishingeffectsuponothers.Man’snaturalinclinationtorevoltagainstselfishcontrolhasbeenexploitedtogood

purposeinwhatwecallthephilosophyandliteratureofdemocracy.Thedoctrineoftherights ofman has been effective in arousing individuals to concerted action againstgovernmentalandreligioustyranny.Theliteraturewhichhashadthiseffecthasgreatlyextendedthenumberoftermsinourlanguagewhichexpressreactionstothecontrolofmen. But the ubiquity and ease of expression of this attitude spells trouble for anysciencewhichmaygivebirthtoapowerfultechnologyofbehavior.Intelligentmenandwomen,dominatedbythehumanisticphilosophyofthepasttwocenturies,cannotviewwith equanimitywhatAndrewHacker has called “the specter of predictableman.”2

44

Even the statistical or actuarial prediction of human events, such as the number offatalities to be expected on a holidayweekend, strikesmanypeople as uncanny andevil,while theprediction andcontrolof individualbehavior is regardedas little lessthan the work of the devil. I am not so much concerned here with the political oreconomicconsequencesforpsychology,althoughresearchfollowingcertainchannelsmaywellsufferharmfuleffects.Weourselves,asintelligentmenandwomen,andasexponentsofWesternthought,sharetheseattitudes.Theyhavealreadyinterferedwiththefreeexerciseofascientificanalysis,andtheirinfluencethreatenstoassumemoreseriousproportions.Three broad areas of human behavior supply good examples. The first of these

—personal control—may be taken to include person-to-person relationships in thefamily, among friends, in social and work groups, and in counseling andpsychotherapy.Otherfieldsareeducationandgovernment.Afewexamplesfromeachwill show how nonscientific preconceptions are affecting our current thinking abouthumanbehavior.

PERSONALCONTROL

Peoplelivingtogetheringroupscometocontroloneanotherwithatechniquewhichis not inappropriately called “ethical.” When an individual behaves in a fashionacceptable to the group, he receives admiration, approval, affection, andmany otherreinforcementswhich increase the likelihood that hewill continue to behave in thatfashion.When his behavior is not acceptable, he is criticized, censured, blamed, orotherwisepunished.Inthefirstcasethegroupcallshim“good”;inthesecond,“bad.”Thispracticeissothoroughlyingrainedinourculturethatweoftenfailtoseethatitisatechniqueofcontrol.Yetwearealmostalwaysengagedinsuchcontrol,eventhoughthereinforcementsandpunishmentsareoftensubtle.Thepracticeofadmirationisanimportantpartofaculture,becausebehaviorwhich

is otherwise inclined to be weak can be set up and maintained with its help. Theindividual is especially likely to be praised, admired, or lovedwhen he acts for thegroupinthefaceofgreatdanger,forexample,orsacrificeshimselforhispossessions,or submits to prolonged hardship, or suffers martyrdom. These actions are notadmirable inanyabsolutesense,but they requireadmiration if theyare tobestrong.Similarly,weadmirepeoplewhobehaveinoriginalorexceptionalways,notbecausesuch behavior is itself admirable, but because we do not know how to encourageoriginal or exceptional behavior in any otherway. The group acclaims independent,unaidedbehaviorinpartbecauseitiseasiertoreinforcethantohelp.Aslongasthistechniqueofcontrolismisunderstood,wecannotjudgecorrectlyan

environmentinwhichthereislessneedforheroism,hardship,orindependentaction.Wearelikelytoarguethatsuchanenvironmentisitselflessadmirableorproduceslessadmirable people. In the old days, for example, young scholars often lived inundesirablequarters,ateunappetizingorinadequatefood,performedunprofitabletasksforalivingortopayfornecessarybooksandmaterialsorpublication.Olderscholarsand othermembers of the group offered compensating reinforcement in the form ofapproval and admiration for these sacrifices. When the modern graduate studentreceives a generous scholarship, enjoys good living conditions, and has his researchandpublicationsubsidized,thegroundsforevaluationseemtobepulledfromunderus.Suchastudentnolongerneedsadmirationtocarryhimoveraseriesofobstacles(no

45

matterhowmuchhemayneed it forother reasons), and, inmissing certain familiarobjectsofadmiration,wearelikelytoconcludethatsuchconditionsarelessadmirable.Obstaclestoscholarlyworkmayserveasausefulmeasureofmotivation—andwemaygowrong unless some substitute is found—butwe can scarcely defend a deliberateharassmentof the student for thispurpose.Theproductivityof any set of conditionscanbeevaluatedonlywhenwehavefreedourselvesoftheattitudeswhichhavebeengeneratedinusasmembersofanethicalgroup.A similar difficulty arises from our use of punishment in the form of censure or

blame. The concept of responsibility and the related concepts of foreknowledge andchoice are used to justify techniques of control using punishment. Was So-and-Soawareoftheprobableconsequencesofhisaction,andwastheactiondeliberate?Ifso,wearejustifiedinpunishinghim.Butwhatdoesthismean?Itappearstobeaquestionconcerning the efficacy of the contingent relations between behavior and punishingconsequences.Wepunishbehaviorbecauseitisobjectionabletousorthegroup,butinaminorrefinementofratherrecentoriginwehavecometowithholdpunishmentwhenit cannotbeexpected tohaveanyeffect. If theobjectionableconsequencesofanactwereaccidentalandnot likely tooccuragain, there isnopoint inpunishing.Wesaythat the individual was not “aware of the consequences of his action” or that theconsequenceswerenot“intentional.”Iftheactioncouldnothavebeenavoided—iftheindividual “hadnochoice”—punishment is alsowithheld, as it is if the individual isincapable of being changed by punishment because he is of “unsoundmind.” In allthese cases—different as they are—the individual is held “not responsible” andgoesunpunished.Justaswesaythatitis“notfair”topunishamanforsomethinghecouldnothelp

doing,sowecallit“unfair”whenoneisrewardedbeyondhisdueorforsomethinghecouldnothelpdoing.Inotherwords,wealsoobjecttowastingreinforcerswheretheyarenotneededorwilldonogood.Wemake thesamepointwith thewords just andright. Thuswe have no right to punish the irresponsible, and aman has no right toreinforcershedoesnotearnordeserve.Butconceptsofchoice,responsibility,justice,and so on, provide amost inadequate analysis of efficient reinforcing and punishingcontingencies because they carry a heavy semantic cargo of a quite diffierent sort,whichobscuresanyattempttoclarifycontrollingpracticesortoimprovetechniques.Inparticular, they fail to prepare us for techniques based on other than aversivetechniques of control. Most people would object to forcing prisoners to serve assubjectsofdangerousmedicalexperiments,but fewobjectwhen theyare induced toserve by the offer of return privileges—even when the reinforcing effect of theseprivilegeshasbeencreatedbyforcibledeprivation.Inthetraditionalschemetherighttorefuseguarantees the individualagainstcoercionoranunfairbargain.But towhatextentcanaprisonerrefuseundersuchcircumstances?Weneednotgo so far afield tomake thepoint.Wecanobserveourownattitude

towardpersonalfreedominthewayweresentanyinterferencewithwhatwewanttodo. Suppose we want to buy a car of a particular sort. Then we may object, forexample,ifourwifeurgesustobuyalessexpensivemodelandtoputthedifferenceintoanewrefrigerator.Orwemayresentitifourneighborquestionsourneedforsuchacarorourabilitytopayforit.Wewouldcertainlyresentitifitwereillegaltobuysuchacar(rememberProhibition);andifwefindwecannotactuallyaffordit,wemayresentgovernmentalcontrolof theprice through tariffsand taxes.Weresent it ifwediscoverthatwecannotgetthecarbecausethemanufacturerisholdingthemodelin

46

deliberatelyshortsupplyinordertopushamodelwedonotwant.Inallthisweassertourdemocraticrighttobuythecarofourchoice.Wearewellpreparedtodosoandtoresentanyrestrictiononourfreedom.Butwhydowenotaskwhy it is thecarofourchoiceandresent theforceswhich

madeitso?Perhapsourfavoritetoyasachildwasacar,ofaverydifferentmodel,butnevertheless bearing the name of the car we now want. Perhaps our favorite TVprogramissponsoredbythemanufacturerofthatcar.Perhapswehaveseenpicturesofmany beautiful or prestigeful persons driving it—in pleasant or glamorous places.Perhapsthecarhasbeendesignedwithrespecttoourmotivationalpatterns:thedeviceonthehoodisaphallicsymbol,orthehorsepowerhasbeensteppeduptopleaseourcompetitive spirit in enablingus topassother cars swiftly (or, as the advertisementssay, “safely”). The concept of freedom which has emerged as part of the culturalpracticeofourgroupmakeslittleornoprovisionforrecognizingordealingwiththesekinds of control. Concepts like “responsibility” and “rights” are scarcely applicable.Weareprepared todealwithcoercivemeasures,butwehaveno traditional recoursewithrespect toothermeasureswhichinthelongrun(andespeciallywiththehelpofscience)maybemuchmorepowerfulanddangerous.

EDUCATION

The techniques of educationwere once frankly aversive. The teacherwas usuallyolderandstrongerthanhispupilsandwasableto“makethemlearn.”Thismeantthattheywerenotactuallytaughtbutweresurroundedbyathreateningworldfromwhichtheycould escapeonlyby learning.Usually theywere left to theirown resources indiscoveringhowtodoso.ClaudeColemanhaspublishedagrimlyamusingreminderoftheseolderpractices.3Hetellsofaschoolteacherwhopublishedacarefulaccountofhis servicesduring51yearsof teaching,duringwhichheadministered:“…911,527blowswith a cane; 124,010with a rod; 20,989with a ruler; 136,715with thehand;10,295overthemouth;7,905boxesontheear;[and]1,115,800slapsonthehead….”Progressiveeducationwasahumanitarianefforttosubstitutepositivereinforcement

forsuchaversivemeasures,butinthesearchforusefulhumanvaluesintheclassroomithasneverfullyreplacedthevariablesitabandoned.Viewedasabranchofbehavioraltechnology,educationremainsrelativelyinefficient.Wesupplementit,andrationalizeit, by admiring the pupilwho learns forhimself; andwe often attribute the learningprocess,orknowledgeitself, tosomething inside the individual.Weadmirebehaviorwhichseemstohaveinnersources.Thusweadmireonewhorecitesapoemmorethanonewhosimplyreads it.Weadmireonewhoknows theanswermore thanonewhoknowswheretolookitup.Weadmirethewriterratherthanthereader.Weadmirethearithmetician who can do a problem in his head rather than with a slide rule orcalculatingmachine,orin“original”waysratherthanbyastrictapplicationofrules.Ingeneral we feel that any aid or “crutch”—except those aids to which we are nowthoroughly accustomed—reduces the credit due. In Plato’s Phaedrus, Thamus, theking, attacks the invention of the alphabet on similar grounds! He is afraid “it willproduceforgetfulnessinthemindsofthosewholearntouseit,becausetheywillnotpractice theirmemories….” Inotherwords,heholds itmoreadmirable to rememberthantouseamemorandum.Healsoobjectsthatpupils“willreadmanythingswithoutinstruction… [and] will therefore seem to knowmany things when they are for themostpartignorant.”Inthesameveinwearetodaysometimescontemptuousofbook

47

learning,butaseducatorswecanscarcelyaffordtoadoptthisviewwithoutreservation.Byadmiring thestudent forknowledgeandblaminghimfor ignorance,weescape

someof the responsibilityof teachinghim.Weresistanyanalysisof theeducationalprocesswhichthreatensthenotionofinnerwisdomorquestionsthecontentionthatthefaultof ignorancelieswith thestudent.Morepowerful techniqueswhichbringaboutthe same changes in behavior by manipulating external variables are decried asbrainwashing or thought control. We are quite unprepared to judge effectiveeducationalmeasures.Aslongasonlyafewpupilslearnmuchofwhatistaught,wedonotworryaboutuniformityorregimentation.Wedonotfearthefeebletechnique;butweshouldviewwithdismaya systemunderwhichevery student learnedeverythinglistedinasyllabus—althoughsuchaconditionisfarfromunthinkable.Similarly,wedonotfearasystemwhichissodefectivethatthestudentmustworkforaneducation;butweareloathtogivecreditforanythinglearnedwithouteffort—althoughthiscouldwell be taken as an ideal result—and we flatly refuse to give credit if the studentalreadyknowswhataschoolteaches.Aworldinwhichpeoplearewiseandgoodwithouttrying,without“havingtobe,”

without“choosingtobe,”couldconceivablybeafarbetterworldforeveryone.Insuchaworldwe shouldnothave to “giveanyonecredit”—weshouldnotneed to admireanyone—forbeingwiseandgood.Fromourpresentpointofviewwecannotbelievethat such aworldwould be admirable.We do not even permit ourselves to imaginewhatitwouldbelike.

GOVERNMENT

Government has always been the special field of aversive control. The state isfrequently defined in terms of the power to punish, and jurisprudence leans heavilyupontheassociatednotionofpersonalresponsibility.Yet it isbecomingincreasinglydifficult to reconcile current practice and theory with these earlier views. Incriminology,forexample,thereisastrongtendencytodropthenotionofresponsibilityin favor of some such alternative as capacity or controllability. But no matter howstronglythefacts,orevenpracticalexpedience,supportsuchachange,itisdifficulttomake the change in a legal systemdesignedon a different plan.Whengovernmentsresort to other techniques (for example, positive reinforcement), the concept ofresponsibility is no longer relevant and the theory of government is no longerapplicable.TheconflictisillustratedbytwodecisionsoftheSupremeCourtinthe1930’swhich

dealtwith, anddisagreedon, thedefinitionofcontrolorcoercion.4 The AgriculturalAdjustment Act proposed that the Secretary of Agriculture make “rental or benefitpayments”tothosefarmerswhoagreedtoreduceproduction.Thegovernmentarguedthat the Act would be unconstitutional if the farmer had been compelled to reduceproductionbutwasnotsincehewasmerelyinvitedtodoso.JusticeRobertsexpressedthecontrarymajorityviewofthecourtthat“thepowertoconferorwithholdunlimitedbenefitsisthepowertocoerceordestroy.”Thisrecognitionofpositivereinforcementwaswithdrawnafewyears later inanothercase inwhichJusticeCardozowrote“Toholdthatmotiveortemptationisequivalenttocoercionistoplungethelawinendlessdifficulties.”Wemayagreewithhim,withoutimplyingthatthepropositionisthereforewrong.Soonerorlaterthelawmustbepreparedtodealwithallpossibletechniquesofgovernmentalcontrol.

48

The uneasiness with which we view government (in the broadest possible sense)when it does not use punishment is shown by the reception of my utopian novel,WaldenTwo.Thiswasessentiallyaproposal toapplyabehavioral technology to theconstruction of a workable, effective, and productive pattern of government. It wasgreetedwithwrathfulviolence.Lifemagazinecalled it “a travestyon thegood life,”and“amenace…atriumphofmortmainorthedeadhandnotenvisagedsincethedaysof Sparta… a slur upon a name, a corruption of an impulse.” JosephWoodKrutchdevoteda substantialpartofhisbook,TheMeasureofMan,5 toattackingmyviewsandthoseoftheprotagonist,Frazier,inthesamevein,andMorrisViteleshasrecentlycriticized the book in a similar manner in Science.6 Perhaps the reaction is bestexpressedinaquotationfromTheQuestforUtopiabyNegleyandPatrick.7

Halfwaythroughthiscontemporaryutopia,thereadermayfeelsure,aswedid,thatthisisabeautifullyironicsatireonwhathasbeencalled“behavioralengineering.”Thelongeronestays in this better world of the psychologist, however, the plainer it becomes that theinspiration isnotsatiric,butmessianic.This is indeed thebehaviorallyengineeredsociety,and while it was to be expected that sooner or later the principle of psychologicalconditioningwouldbemadethebasisofaseriousconstructionofutopia—BrownanticipateditinLimanora—yetnot even theeffective satireofHuxley is adequatepreparation for theshockinghorroroftheideawhenpositivelypresented.Ofall thedictatorshipsespousedbyutopists, this is themost profound, and incipient dictatorsmightwell find in this utopia aguidebookofpoliticalpractice.

Onewouldscarcelyguessthattheauthorsaretalkingaboutaworldinwhichthereisfood,clothing,andshelterforall,whereeveryonechooseshisownworkandworksontheaverageonlyfourhoursaday,wheremusicandthearts flourish,wherepersonalrelationships develop under the most favorable circumstances, where educationprepareseverychildforthesocialandintellectuallifewhichliesbeforehim,where—in short—people are truly happy, secure, productive, creative, and forward-looking.Whatiswrongwithit?Onlyonething:someone“planneditthatway.”Ifthesecriticshad comeupon a society in some remote corner of theworldwhichboasted similaradvantages,theywouldundoubtedlyhavehaileditasprovidingapatternweallmightwell follow—provided that it was clearly the result of a natural process of culturalevolution.Anyevidence that intelligencehadbeenused inarrivingat thisversionofthegoodlifewould,intheireyes,beaseriousflaw.NomatteriftheplannerofWaldenTwodivertsnoneoftheproceedsofthecommunitytohisownuse,nomatterifhehasno current control or is, indeed, unknown to most of the other members of thecommunity(heplannedthat,too),somewherebackofitallheoccupiesthepositionofprimemover.Andthis,tothechildofthedemocratictradition,spoilsitall.Thedangersinherentinthecontrolofhumanbehaviorareveryreal.Thepossibility

of the misuse of scientific knowledge must always be faced.We cannot escape bydenyingthepowerofascienceofbehaviororarrestingitsdevelopment.Itisnohelptocling to familiar philosophies of human behavior simply because they are morereassuring.As Ihavepointedoutelsewhere [page19], thenew techniquesemergingfroma scienceofbehaviormustbe subject to the explicit counter-controlwhichhasalready been applied to earlier and cruder forms. Brute force and deception, forexample, are now fairly generally suppressed by ethical practices and by explicit

49

governmentalandreligiousagencies.Asimilarcounter-controlofscientificknowledgeintheinterestsofthegroupisafeasibleandpromisingpossibility.Althoughwecannotsayhowdevious thecourseof itsevolutionmaybe,aculturalpatternofcontrolandcounter-controlwillpresumablyemergewhichwillbemostwidelysupportedbecauseitismostwidelyreinforcing.If we cannot foresee all the details (as we obviously cannot), it is important to

remember that this is trueof thecriticsofscienceaswell.Thedireconsequencesofnewtechniquesofcontrol,thehiddenmenaceinoriginalculturaldesigns—theseneedsomeproof.Thattheneedforproofissooftenoverlookedisonlyanotherexampleofmypresentpoint.Manhasgothimselfintosomeprettyfixes,anditiseasytobelievethathewilldosoagain.Butthereisamoreoptimisticpossibility.Theslowgrowthofthemethods of science, now for the first time being applied to human affairs,maymean a new and exciting phase of human life towhich historical analogieswill notapplyandinwhichearlierpoliticalsloganswillnotbeappropriate.Ifwearetousetheknowledge which a science of behavior is now making available with any hope ofsuccess,wemustlookathumannatureasitisbroughtintofocusthroughthemethodsofscienceratherthanasithasbeenpresentedtousinaseriesofhistoricalaccidents.Iftheadventofapowerfulscienceofbehaviorcausestrouble,itwillnotbebecause

science itself is inimical to human welfare but because older conceptions have notyieldedeasilyorgracefully.Weexpect resistance tonew techniquesof control fromthose who have heavy investments in the old, but we have no reason to help thempreserveaseriesofprincipleswhicharenotends in themselvesbut ratheroutmodedmeans to an end.What is needed is a new conception of human behavior which iscompatible with the implications of a scientific analysis. All men control and arecontrolled. The question of government in the broadest possible sense is not howfreedomistobepreservedbutwhatkindsofcontrolaretobeusedandtowhatends.Controlmustbeanalyzedandconsideredinitsproperproportions.Noscientist,Iamsure,wishestodevelopnewmaster-slaverelationshipsorbendthewillofthepeopletodespotic rulers in new ways. These are patterns of control appropriate to a worldwithout science.Theymaywellbe the first togowhen the experimental analysisofbehaviorcomesintoitsowninthedesignofculturalpractices.

50

PartII

Dr. Rogers presented his own point of view, together with comments on my paper,which had been submitted to him in manuscript. He argued that “in any scientificendeavor—whether‘pure’orappliedscience—thereisapriorsubjectivechoiceofthepurpose or valuewhich that scientificwork is perceived as serving,” and that “thissubjective value choice… must always lie outside of the scientific endeavor.” Heattributed certain value choices to me, including the decision to experiment withdifferent choices, and offered some alternative values which might guide scientificresearchin the fieldofhumanbehavior.“Wemight thenvalue:manasaprocessofbecoming,asaprocessofachievingworthanddignitythroughthedevelopmentofhispotentialities; theindividualhumanbeingasaself-actualizingprocess,movingontomore challenging and enriching experiences; the process by which the individualcreativelyadaptstoanever-newandchangingworld….”He illustrated this with client-centered therapy, where therapists establish “by

externalcontrolconditionswhichtheypredictwillbefollowedbyinternalcontrolbythe individual, in pursuit of internally chosen goals. We can choose to use thebehavioralsciencesinwayswhichwillfree,notcontrol.”

51

PartIII

Icannotquiteagreethatthepracticeofsciencerequiresapriordecisionaboutgoalsora prior choice of values. The metallurgist can study the properties of steel and theengineercandesignabridgewithoutraisingthequestionofwhetherabridgeistobebuilt. But such questions are certainly frequently raised and tentatively answered.Rogers wants to call the answers “subjective choices of values.” To me, such anexpressionsuggeststhatwehavehadtoabandonmorerigorousscientificpracticesinordertotalkaboutourownbehavior.IntheexperimentalanalysisofotherorganismsIwould use other terms, and I shall try to do so here. Any list of values is a list ofreinforcers—conditioned or otherwise. We are so constituted that under certaincircumstances food,water, sexual contact, and so onwillmake any behaviorwhichproducesthemmorelikelytooccuragain.Otherthingsmayacquirethispower.Wedonotneedtosaythatanorganismchoosestoeatratherthantostarve.Ifyoureplythatitisaverydifferentthingwhenamanchoosestostarve,Iamonlytoohappytoagree.Ifitwerenotso,weshouldhaveclearedupthequestionofchoicelongago.Anorganismcanbereinforcedby—canbemadeto“choose”—almostanygivenstateofaffairs.Rogers is concerned with choices which involve multiple and usually conflicting

consequences.IhavedealtwithsomeoftheseinScienceandHumanBehavior inananalysis of self-control. Shall I eat these delicious strawberries today if I will thensufferanannoyingrashtomorrow?ThedecisionIamtomakeusedtobeassignedtotheprovinceofethics.Butwearenowstudyingsimilarcombinationsofpositiveandnegativeconsequences,aswellascollateralconditionswhichaffect the result, in thelaboratory.Evenapigeoncanbetaughtsomemeasureofself-control!Andthisworkhelpsustounderstandtheoperationofcertainformulas—amongthemvaluejudgments—which folk-wisdom, religion, and psychotherapy have advanced in the interests ofself-discipline.Theobservableeffectofanystatementofvalue is toalter therelativeeffectiveness of reinforcers. We may no longer enjoy the strawberries for thinkingabouttherash.Ifrashesarebrandedsufficientlyshameful,illegal,sinful,maladjusted,or unwise, we may glow with satisfaction as we push the strawberries aside in agrandioseavoidanceresponsewhichwouldbringasmiletothelipsofMurraySidman.People behave in ways which, as we say, conform to ethical, governmental, or

religiouspatternsbecausetheyarereinforcedfordoingso.Theresultingbehaviormayhave far-reaching consequences for the survival of the pattern towhich it conforms.Andwhetherwelikeitornot,survivalistheultimatecriterion.Thisiswhere,itseemsto me, science can help—not in choosing a goal, but in enabling us to predict thesurvivalvalueofculturalpractices.Manhastoolongtriedtogetthekindofworldhewants by glorifying some brand of immediate reinforcement. As science points upmore and more of the remoter consequences, he may begin to work to strengthenbehavior,notinaslavishdevotiontoachosenvalue,butwithrespecttotheultimatesurvivalofmankind.DonotaskmewhyIwantmankindtosurvive.IcantellyouwhyonlyinthesenseinwhichthephysiologistcantellyouwhyIwanttobreathe.Oncetherelationbetweenagivenstepandthesurvivalofmygrouphasbeenpointedout,Iwilltakethatstep.Anditisthebusinessofsciencetopointoutjustsuchrelations.ThevaluesIhaveoccasionallyrecommended(andRogershasnotledmetorecant)

are transitional.Other thingsbeingequal, I ambettingon thegroupwhosepractices

52

makeforhealthy,happy,secure,productive,andcreativepeople.AndIinsistthatthevaluesrecommendedbyRogersaretransitional,too,forIcanaskhimthesamekindofquestion.Man as a process of becoming—what?Self-actualization—forwhat? Innercontrolisnomoreagoalthanexternalcontrol.WhatRogersseemstometobeproposing,bothhereandelsewhere,isthis:Letus

useourincreasingpowerofcontroltocreateindividualswhowillnotneedandperhapswillnolongerrespondtocontrol.Letussolvetheproblemofourpowerbyrenouncingit. At first blush this seems as implausible as a benevolent despot. Yet power hasoccasionallybeenforesworn.AnationhasburneditsReichstag,richmenhavegivenaway their wealth, beautiful women have become ugly hermits in the desert, andpsychotherapists have become nondirective. When this happens, I look to otherpossible reinforcements for a plausible explanation. A people relinquish democraticpowerwhenatyrantpromisesthemtheearth.Richmengiveawaywealthtoescapetheaccusing finger of their fellow men. A woman destroys her beauty in the hope ofsalvation. And a psychotherapist relinquishes control because he can thus help hisclientmoreeffectively.ThesolutionwhichRogersissuggestingisthusunderstandable.Butishecorrectly

interpreting the result?What evidence is there that a client ever becomes truly self-directing?Whatevidence is there thatheevermakesa truly inner choiceof idealorgoal?Eventhoughthetherapistdoesnotdothechoosing,eventhoughheencourages“self-actualization”—hehasnotceasedtocontrolaslongasheholdshimselfreadytostep inwhen occasion demands—when, for example, the client chooses the goal ofbecomingamoreaccomplishedliarormurderinghisboss.Butsupposingthetherapistdoeswithdrawcompletelyor isno longernecessary—whataboutall theother forcesacting upon the client? Is the self-chosen goal independent of his early ethical andreligious training? of the folk-wisdomof his group?of the opinions and attitudes ofotherswhoareimportanttohim?Surelynot.Thetherapeuticsituationisonlyasmallpartoftheworldoftheclient.Fromthetherapist’spointofviewitmayappeartobepossible to relinquishcontrol.But thecontrolpasses,not toa“self,”but to forces inother parts of the client’s world. The solution of the therapist’s problem of powercannotbeoursolution,forwemustconsideralltheforcesactingupontheindividual.Thechildwhomustbeproddedandnaggedissomethinglessthanafullydeveloped

humanbeing.Wewanttoseehimhurryingtohisappointment,notbecauseeachstepistaken in response toverbal reminders fromhismother, butbecause certain temporalcontingencies, in which dawdling has been punished and hurrying reinforced, haveworked a change in his behavior. Call this a state of better organization, a greatersensitivity to reality, orwhatyouwill.Theplain fact is that the childpasses fromatemporary verbal control exercised by his parents to control by certain inexorablefeaturesoftheenvironment.Ishouldsupposethatsomethingofthesamesorthappensinsuccessfulpsychotherapy.Rogersseemstometobesayingthis:Letusputanend,asquicklyaspossible, toanypatternofmaster-and-slave, toanydirectobedience tocommand, to the submissive following of suggestions. Let the individual be free toadjusthimself tomore rewarding featuresof theworldabouthim. In theend, lethisteachersandcounselors“witheraway,”liketheMarxiststate.Inotonlyagreewiththisasauseful idea, Ihaveconstructedafancifulworld todemonstrate itsadvantages. Itsaddensme to hear Rogers say that “at a deep philosophic level”WaldenTwo andGeorgeOrwell’s1984“seemindistinguishable.”Theycouldscarcelybemoreunlike—at any level. The book 1984 is a picture of immediate aversive control for vicious

53

selfishpurposes.ThefounderofWaldenTwo,ontheotherhand,hasbuiltacommunityinwhichneitherhenoranyotherpersonexertsanycurrentcontrol.Hisachievementlayinhisoriginalplan,andwhenheboastsofthis(“Itisenoughtosatisfythethirstiesttyrant”)wedonotfearhimbutonlypityhimforhisweakness.Another critic of Walden Two, Andrew Hacker,8 has discussed this point in

considering the bearing ofmass conditioning upon the liberal notion of autonomousman. In drawing certain parallels between the Grand Inquisitor passage inDostoevsky’sBrothersKaramazov,Huxley’sBraveNewWorld, andWaldenTwo,heattempts to setup adistinction tobedrawn in any societybetweenconditioners andconditioned. He assumes that “the conditioner can be said to be autonomous in thetraditional liberal sense.” But then he notes: “Of course the conditioner has beenconditioned.Buthehasnotbeenconditionedbytheconsciousmanipulationofanotherperson.”Buthowdoesthisaffect theresultingbehavior?Canwenotsoonforget theoriginsofthe“artificial”diamondwhichisidenticalwiththerealthing?Whetheritisan“accidental”culturalpattern,suchasissaidtohaveproducedthefounderofWaldenTwo,ortheengineeredenvironmentwhichisabouttoproducehissuccessors,wearedealingwith sets of conditions generating human behaviorwhichwill ultimately bemeasuredbytheircontributiontothestrengthofthegroup.Welooktothefuture,notthepast,forthetestof“goodness”oracceptability.Ifweareworthyofourdemocraticheritageweshall,ofcourse,be ready to resist

any tyrannical use of science for immediate, selfish purposes. But if we value theachievements and goals of democracy, we must not refuse to apply science to thedesignandconstructionofculturalpatterns,eventhoughwemaythenfindourselvesinsome sense in the position of controllers. Fear of control, generalized beyond anywarrant, has led to a misinterpretation of valid practices and the blind rejection ofintelligentplanningforabetterwayoflife.IntermswhichItrustRogerswillapprove,inconquering this fearweshallbecomemorematureandbetterorganizedandshall,thus,morefullyactualizeourselvesashumanbeings.

FromScience1956,124,1057–1066.AdebatewithCarlR.Rogers.1Rogers,C.R.TeachersCollegeRecord,1956,57,316.2Hacker,A.AntiochReview,1954,14,195.3Coleman,C.Bull.Am.Assoc.Univ.Professors,1953,39,457.4Freund,P.A.etal.ConstitutionalLaw:CasesandOtherProblems,Vol.1.Boston:Little,Brown&Company,1954.

5Krutch,J.W.TheMeasureofMan.Indianapolis:Bobbs-Merrill,1953.6Viteles,M.Science,1955,122,1167.7Negley,G.,andPatrick,J.M.TheQuestforUtopia.NewYork:Schuman,1952.8Hacker,A.J.Politics,1955,17,590.

54

TheDesignofCultures

Aseriesof threeconferenceson“EvolutionaryTheoryandHumanProgress”washeld at the American Academy of Arts and Sciences in the fall of 1960. The firstconsidered biological problems, the second, anthropological, and the third,psychological. The present paper, part of the third program, was published inDaedalus,summerissueof1961.Itisreprintedherebypermission.

Anyonewhoundertakestoimproveculturalpracticesbyapplyingascientificanalysisof human behavior is likely to be told that improvement involves a value judgmentbeyond the pale of his science and that he is exemplifying objectionable values byproposing to meddle in human affairs and infringe on human freedoms. ScientiststhemselvesoftenacceptthisstandardcontentionofWesternphilosophy,eventhoughitimpliesthatthereisakindofwisdomwhichismysteriouslydeniedtothemandeventhough the behavioral scientists among them would be hard pressed to give anempiricalaccountofsuchwisdomortodiscoveritssources.Thepropositiongainsunwarrantedstrengthfromthefactthatitappearstochampion

the natural against the artificial.Man is a product of nature, the argument runs, butsocietiesarecontrivedbymen.Manisthemeasureofallthings,andourplansforhim—ourcustomsandinstitutions—willsucceedonlyiftheyallowforhisnature.Tothisit might be answered that man is more than an immutable product of biologicalprocesses;heisapsychologicalentity,andassuchalsolargelyman-made.Hiscausemay be as contrived as society’s and possibly as weak. He is, nevertheless, anindividual,andhisdefendersareindividuals,too,whomayborrowzealinhisdefensefromtheirownroleinthegreatconflictbetweentheoneandthemany.Tosidewiththe individual against the state, to take a specific example, is reassuringly to defendone’s own, even though itmight be answered thatmankindhaswon its battles onlybecauseindividualmenhavelosttheirs.These are merely answers in kind, which can no doubt be met with plausible

rejoinders.Thedisputingofvalues isnotonlypossible, it is interminable.Toescapefrom itwemust get outside the system.We can do this by developing an empiricalaccount of the behavior of both protagonists. All objections to cultural design, likedesign itself,areformsofhumanbehaviorandmaybestudiedassuch. It ispossiblethataplausibleaccountofthedesignofcultureswillallayourtraditionalanxietiesandprepare theway for theeffectiveuseofman’s intelligence in theconstructionofhisownfuture.Itisreasonabletohopethatascientificanalysiswillsomedaysatisfactorilyexplain

howculturalpracticesariseandaretransmittedandhowtheyaffectthosewhoengagein them, possibly to further the survival of the practices themselves or at least tocontribute to their successors. Such an analysis will embrace the fact that men talkabouttheirculturesandsometimeschangethem.Changingacultureisitselfacultural

55

practice, and we must know as much as possible about it if we are to question itintelligently. Under what circumstances do men redesign—or, to use a discreditedterm,reform—theirwayoflife?Whatisthenatureoftheirbehaviorindoingso?Isthedeliberatemanipulationofacultureathreattotheveryessenceofmanor,attheotherextreme,anunfathomedsourceofstrengthfortheculturewhichencouragesit?Weneednotgointothedetailsofascientificaccountofbehaviortoseehowitbears

on this issue. Its contribution must, however, be distinguished from any help to bedrawnfromhistoricalanalogyortheextrapolationofhistoricaltrendsorcycles,aswellasfrominterpretationsbasedonsociologicalprinciplesorstructures.Suchanaccountmust make contact with biology, on the one hand, but serve in an interpretation ofsocialphenomena,ontheother.Ifitistoyieldasatisfactoryanalysisofthedesignandimplementationofsocialpractices,itmustbefreeofaparticulardefect.Evolutionarytheory,especiallyinitsappealtothenotionofsurvival,sufferedforalongtimefromcircularity. Itwasnot satisfying to argue that formsof lifewhichhad survivedmustthereforehavehadsurvivalvalueandhadsurvivedbecauseofit.Asimilarweaknessisinherentinpsychologiesbasedonadjustmentoradaptation.Itisnotsatisfyingtoarguethat a man adapts to a new environment because of his intelligence and emotionalstability if these are then defined in terms of capacities to adapt. It is true thatorganismsusuallydevelopindirectionswhichmaximize,phylogenetically,thesurvivalof the species and, ontogenetically, the adjustment of the individual; but themechanisms responsible for both kinds of change need to be explained withoutrecoursetotheselectiveeffectoftheirconsequences.In biology this is now being done. Genetics clarifies and supports evolutionary

theorywithnewkindsoffacts,andindoingsoeliminatesthecircularityintheconceptof survival. A comparable step in the study of human behavior is to analyze themechanisms of human action apart from their contribution to personal and culturaladjustment.Itisnotenoughtopointoutthatagivenformofbehaviorisadvantageoustotheindividualorthataculturalpracticestrengthensthegroup.Wemustexplaintheoriginandtheperpetuationofbothbehaviorandpractice.A scientific analysis which satisfies these conditions confines itself to individual

organismsratherthanstatisticalconstructsorinteractinggroupsoforganisms,eveninthestudyofsocialbehavior.Itsbasicdatumistheprobabilityoftheoccurrenceoftheobservableeventswecallbehavior(orofinferredeventshavingthesamedimensions).Theprobabilityofbehaviorisaccountedforbyappealtothegeneticendowmentoftheorganismand itspast andpresentenvironments,describedwholly in the languageofphysics and biology. The laboratory techniques of such an analysis, and theirtechnological applications, emphasize the prediction and control of behavior via themanipulationofvariables.Validationisfoundprimarilyinthesuccesswithwhichthesubjectmattercanbecontrolled.Anexampleofhowsuchananalysisdiffersfromitspredecessorsisconvenientlyat

hand. An important group of variables which modify behavior have to do with theconsequences of action.Rewards and punishments are variables of this sort, thoughratherinadequatelyidentifiedbythoseterms.Weareinterestedinthefact(apartfromany theory which explains it) that by arranging certain consequences—that is, bymakingcertainkindsofeventscontingentuponbehavior—weachieveahighdegreeofexperimental control. Our present understanding of the so-called “contingencies ofreinforcement” isundoubtedly incomplete,but itneverthelesspermitsus toconstructnew forms of behavior, to bring behavior under the control of new aspects of the

56

environment,andtomaintainitundersuchcontrolforlongperiodsoftime—andallofthisoftenwithsurprisingease.Extrapolation to less rigorouslycontrolledsamplesofbehavior outside the laboratory has already led to promising technologicaldevelopments.Buttheimportanceoftheprincipleisembarrassing.Almostanyinstanceofhuman

behaviorinvolvescontingenciesofreinforcement,andthosewhohavebeenalertedtotheirsignificancebylaboratorystudiesoftenseemfanaticalinpointingthemout.Yetbehaviorisimportantmainlybecauseofitsconsequences.Wemaymorereadilyacceptthisfactifwerecalltheubiquityoftheconceptofpurpose.Theexperimentalstudyofreinforcing contingencies is nothing more than a nonteleological analysis of thedirectedeffects of behavior, of relations which have traditionally been described aspurpose.Bymanipulating contingencies of reinforcement inwayswhich conform tostandardpracticesinthephysicalsciences,westudyandusethemwithoutappealingtofinalcauses.We can put this reinterpretation of purpose to immediate use, for it bears on a

confusion between the phylogenetic and the ontogenetic development of behaviorwhichhascloudedourthinkingabouttheoriginandgrowthofcultures.Contingenciesofreinforcementaresimilartowhatwemightcallcontingenciesofsurvival.Inheritedpatternsofbehaviormusthavebeenselectedbytheircontributionstosurvivalinwayswhicharenotunlikethoseinwhichthebehavioroftheindividualisselectedorshapedby its reinforcing consequences.Both processes exemplify adaptation or adjustment,butverydifferentmechanismsmustbeinvolved.Theevolutionofinheritedformsofbehaviorisasplausibleastheevolutionofany

functionoftheorganismwhentheenvironmentcanberegardedasreasonablystable.The internal environment satisfies this requirement, and a genetic endowment ofbehavior related to the internal economy—say, peristalsis or sneezing—is usuallyaccepted without question. The external environment is much less stable fromgenerationtogeneration,butsomekindsofresponsestoitarealsoplausiblyexplainedby evolutionary selection. The genetic mechanisms are presumably similar to thosewhichaccount forother functions.But environments change, and anyprocesswhichpermits an organism tomodify its behavior is then important. The structures whichpermitmodificationmusthaveevolvedwhenorganismswerebeingselectedby theirsurvivalinnovelenvironments.Although the mechanisms which permit modification of behavior are inherited,

learned behavior does not emerge from, and is not an extension of, the unlearnedbehavior of the individual.The organismdoes not simply refine or extend a geneticbehavioralendowmenttomakeitmoreeffectiveormoreinclusive.Instead,itdevelopscollateral behavior, which must be distinguished from an inherited response systemeven when both serve similar functions. It is important to remember this whenconsideringsocialbehavior.Inspiteofcertainintriguinganalogies,itisnotlikelythatthesocialinstitutionsofmanarefoundedonorthattheyemergedfromtheinstinctivepatternsofanimalsocieties.Theyaretheachievementsofindividuals,modifyingtheirbehavioras inheritedmechanismspermit.Theco-ordinatedactivitiesof theanthillorbeehiveoperateonverydifferentprinciplesfromthoseofafamily,alargecompany,oragreatcity.Thetwokindsofsocialbehaviormusthavedevelopedthroughdifferentprocesses,andtheyaremaintainedinforcefordifferentreasons.Totakeaspecificexample,verbalbehaviorisnotarefinementuponinstinctivecries

ofalarm,distress,andsoon,eventhoughthereinforcingcontingenciesintheonecase

57

are analogous to the conditions of survival in the other. Bothmay be said to servesimilaradaptivefunctions,but themechanisms involved inacquiringverbalbehaviorclearly set it apart from instinctive responses. The innate vocal endowment of anorganism is indeed particularly refractory to modification, most if not all verbalresponsesbeingmodificationsofanonspecificbehavioralendowment.In general, the evolution of man has emphasized modifiability rather than the

transmissionof specific formsofbehavior. Inheritedverbalorother social responsesare fragmentary and trivial. By far the greater part of behavior develops in theindividual through processes of conditioning, given a normal biological endowment.Man becomes a social creature only because other men are important parts of hisenvironment. The behavior of a child born into a flourishing society is shaped andmaintained by variables, most of which are arranged by other people. These socialvariablescomposethe“culture”inwhichthechildlives,andtheyshapehisbehaviorinconformitywiththatculture,usuallyinsuchawaythatheinturntendstoperpetuateit.The behavioral processes present no special problems. Nevertheless, a satisfactoryaccountcallsforsomeexplanationofhowasocialenvironmentcanhavearisenfromnonsocial precursors. This may seem to raise the hoary question of the origin ofsociety, but we have no need to reconstruct an actual historical event or even aspeculative beginning, such as a social compact from which conclusions about thenatureofsocietycanbedrawn.Wehaveonlytoshowthatasocialenvironmentcouldhaveemergedfromnonsocialconditions.Asinexplainingtheoriginoflife,wecannotdiscover an actual historical event but must be satisfied with a demonstration thatcertain structures with their associated functions could have arisen under plausibleconditions.The emergence of a given form of social behavior from nonsocial antecedents is

exemplified by imitation. Inherited imitative behavior is hard to demonstrate. Theparrotmaypossiblyoweitsdistinctiononlytoaninheritedcapacitytobereinforcedbythe production of imitative sounds. In any case, an inherited repertoire of imitativebehavior in man is insignificant, compared with the product of certain powerfulcontingencies of reinforcement which establish and maintain behaving-as-others-behave.Forexample,iforganismAseesorganismBrunninginobviousalarm,Awillprobablyavoidaversiveconsequencesbyrunninginthesamedirection.Or,ifAseesBpickingandeatingripeberries,Awillprobablybereinforcedforapproachingthesameberry patch. Thousands of instances of this sort compose a general contingencyproviding for the reinforcement of doing-as-others-do. In this sense, behaviorexemplifying imitation is acquired, yet it is practically inevitable whenever two ormoreorganisms live in contactwithone another.The essential conditions arenot inthemselvessocial.Mostsocialbehavior,however,arisesfromsocialantecedents.Transmissionismore

important than social invention. Unlike the origin of cultural practices, theirtransmissionneednotbeamatterforspeculation,since theprocesscanbeobserved.Deliberatetransmission(thatis,transmissionachievedbecauseofpracticeswhichhavebeenreinforcedbytheirconsequences)isnotneeded.Forexample,somepracticesareperpetuated as the members of a group are severally replaced. If A has alreadydeveloped specific controlling behavior with respect to B, depending partly uponincidental characteristicsofB’sbehavior,hemay impose the samecontrolonanewindividual,C,whomightnothimselfhavegenerated just the samepractices inA. Amotherwhohas shaped the vocal responses of her first baby into a primitive verbal

58

repertoiremay bring already established contingencies to bear on a second child.Aleader who has acquired aversive controlling practices in his interactions with asubmissive followermay take by storm a second follower even though,without thispreparation,theleader-followerrelationmighthavebeenreversedinthesecondcase.Overlappinggroupmembershipis,ofcourse,onlyonefactorcontributingtomanners,customs,folkways,andotherabidingfeaturesofasocialenvironment.These simple examples are offered not as solutions to important problems but to

illustrateanapproachtotheanalysisofsocialbehaviorandtothedesignofaculture.AspecialkindofsocialbehavioremergeswhenArespondsinadefinitewaybecauseoftheeffectonthebehaviorofB.WemustconsidertheimportanceofBtoAaswellasofA toB.Forexample,whenA seesB looking into a storewindow, he is likely to bereinforcedifhelookstoo,asintheexampleoftheberrypatch.Butif thislookingisimportant toB,or toa thirdpersonwhocontrolsB, a changemay takeplace inB’sbehavior.BmaylookintothewindowinordertoinduceAtodothesame.Thecarnivalshill plays on the behavior of prospective customers in thisway.B’s behavior is nolongercontrolledbywhatisseeninthewindowbut(directlyorindirectly)bytheeffectofthatbehavioronA.(TheoriginalcontingenciesforAbreakdown:thewindowmaynot now be “worth looking into.”) Action taken by B because of its effect on thebehaviorofA may be called “personal control.” An important subdivision is verbalbehavior,thepropertiesofwhichderivefromthefactthatreinforcementsaremediatedbyotherorganisms.Anothersubdivisionisculturaldesign.Inanalyzinganysocialepisodefromthispointofviewacompleteaccountmustbe

givenofthebehaviorsofbothpartiesastheycontributetotheoriginandmaintenanceof the behavior of each other. For example, in analyzing a verbal episode,wemustaccount for both speaker and listener. This is seldom done in the case of nonverbalpersonal control. In noticing how themaster controls the slave or the employer theworker,we commonly overlook reciprocal effects and, by considering action in onedirectiononly,areledtoregardcontrolasexploitation,oratleastthegainingofaone-sided advantage; but the control is actuallymutual.The slave controls themaster ascompletely as the master the slave, in the sense that the techniques of punishmentemployedby themaster have been selectedby the slave’s behavior in submitting tothem.Thisdoesnotmeanthatthenotionofexploitationismeaninglessorthatwemaynotappropriatelyask,Cuibono? Indoingso,however,wegobeyond theaccountofthesocialepisodeitselfandconsidercertainlong-termeffectswhichareclearlyrelatedtothequestionofvaluejudgments.Acomparableconsiderationarisesintheanalysisofanybehaviorwhichaltersaculturalpractice.Wemay not be satisfied with an explanation of the behavior of two parties in a

social interaction.Theslaves inaquarrycutting stone forapyramidwork toescapepunishmentordeath,andtherisingpyramidissufficientlyreinforcingtothereigningPharaoh to induce him to devote part of hiswealth tomaintaining the forceswhichpunishorkill.Anemployerpayssufficientwagestoinducementoworkforhim,andtheproductsoftheirlaborreimbursehimwith,letussay,agreatdealtospare.Theseare on-going social systems, but in thus analyzing them we may not have takeneverything into account.The systemmaybe alteredbyoutsiders inwhomsympathywith, or fear of, the lot of the slave or exploited worker may be generated. Moreimportant,perhaps,isthepossibilitythatthesystemmaynotactuallybeinequilibrium.Itmaybreedchangeswhich lead to itsdestruction.Control throughpunishmentmayleadtoincreasingviciousness,withaneventuallossofthesupportofthoseneededto

59

maintainit;andtheincreasingpovertyoftheworkerandtheresultingincreaseintheeconomicpoweroftheemployermayalsoleadtocounter-controllingaction.Aculturewhichraisesthequestionofcollateralordeferredeffectsismostlikelyto

discoverandadoptpracticeswhichwillsurviveor,asconditionschange,will leadtomodificationswhichinturnwillsurvive.Thisis an important step inculturaldesign,butitisnoteasilytaken.Long-termconsequencesareusuallynotobvious,andthereislittleinducementtopayanyattentiontothem.Wemayadmireamanwhosubmitstoaversive stimulation for the sake of later reinforcement or who eschews immediatereinforcement to avoid laterpunishment, but the contingencieswhich leadhim tobe“reasonable” in this sense (our admiration is part of them) are by no meansoverpowering.Ithastakencivilizedsocietiesalongtimetoinventtheverbaldevices—the precepts of morals and ethics—which successfully promote such an outcome.Ultimateadvantagesseemtobeparticularlyeasytooverlookinthecontrolofbehavior,where a quick though slight advantagemay have undue weight. Thus, although weboast that the birch rod has been abandoned, most school children are still underaversive control—not because punishment is more effective in the long run, butbecauseityieldsimmediateresults.Itiseasierfortheteachertocontrolthestudentbythreateningpunishmentthanbyusingpositivereinforcementwithitsdeferred,thoughmorepowerful,effects.Aculturewhichhasbecomesensitivetothelong-termconsequencesofitsmeasures

is usually supported by a literature or philosophywhich includes a set of statementsexpressingtherelationsbetweenmeasuresandconsequences.Totheculturaldesigner,these statements functionasprescriptions foreffectiveaction; to themembersof thegroup,theyareimportantvariablesfurtheringeffectiveself-management.(Toboth,andto theneutralobserver, theyare sometimes said to“justify”ameasure,but thismaymeannothingmorethanstrengtheningthemeasurebyclassifyingitwithcertainkindsofeventscharacteristicallycalled“good”or“right.”)Thus,agovernmentmayinduceitscitizenstosubmittothehardshipandtragedyofwarbypicturingafutureinwhichthe world is made safe for democracy or free of Communism, or to a program ofausteritybypointingtoeconomicchangeswhichwilleventuallyleadtoanabundanceof good things for all. In so doing, it strengthens certain behavior on the part of itscitizenswhichisessentialtoitspurposes,andtheresultinggaininpowerreinforcesthegovernment’sownconcernfordeferredeffectsanditseffortstoformulatethem.The scientific study of behavior underlines the collateral effects of controlling

practicesandrevealsunstablefeaturesofagiveninteractionwhichmayleadtolong-deferred consequences. It may dictate effective remedial or preventive measures. Itdoesnotdothis,however,bytakingthescientistoutofthecausalstream.Thescientistalso is theproductofagenericendowmentandanenvironmentalhistory.Healso iscontrolled by the culture or cultures to which he belongs. Doing-something-about-human-behavior is a kind of social action and its products and by-productsmust beunderstoodaccordingly.A reciprocal relationship between the knower and the known, common to all the

sciences, is important here. A laboratory for the study of behavior contains manydevicesforcontrollingtheenvironmentandforrecordingandanalyzingthebehaviorof organisms. With the help of these devices and their associated techniques, wechangethebehaviorofanorganisminvariousways,withconsiderableprecision.Butnote that the organism changes our behavior in quite as precise a fashion. Ourapparatuswasdesignedbytheorganismwestudy,foritwastheorganismwhichledus

60

to choose a particularmanipulandum, particular categories of stimulation, particularmodes of reinforcement, and so on, and to record particular aspects of its behavior.Measures which were successful were for that reason reinforcing and have beenretained,while others have been, aswe say, extinguished. The verbal behaviorwithwhichwe analyzeour data has been shaped in a similarway: order and consistencyemerged to reinforce certain practices which were adopted, while other practicessufferedextinctionandwereabandoned.(Allscientifictechniques,aswellasscientificknowledgeitself,aregeneratedinthisway.Acyclotronis“designed”bytheparticlesitistocontrol,andatheoryiswrittenbytheparticlesitistoexplain,asthebehavioroftheseparticlesshapesthenonverbalandverbalbehaviorofthescientist.)A similarly reciprocal effect is involved in social action, especially in cultural

design. Governmental, religious, economic, educational, and therapeutic institutionshavebeenanalyzedinmanyways—forexample,assystemswhichexaltsuchentitiesassovereignty,virtue,utility,wisdom,andhealth.Thereisaconsiderableadvantageinconsidering these institutions simply as behavioral technologies. Each one uses anidentifiablesetof techniques for thecontrolofhumanbehavior,distinguishedby thevariablesmanipulated.Thediscoveryandinventionofsuchtechniquesandtheirlaterabandonmentorcontinueduse—inshort, theirevolution—are,orshouldbe,apartofthe history of technology. The issues they raise, particularly with respect to thebehaviorofthediscovererorinventor,arecharacteristicoftechnologyingeneral.Bothphysicalandbehavioraltechnologieshaveshownprogressorimprovementin

thesensethatnewpracticeshavebeendiscoveredorinventedandtestedandthatsomeof themhavesurvivedbecause their effectswere reinforcing.Menhave foundbetterways,notonlytodyeaclothorbuildabridge,buttogovern,teach,andemploy.Theconditions underwhich all such practices originate range from sheer accident to theextremelycomplex behaviors called thinking. The conditions under which they aretested and selected are equally diverse. Certain immediate personal advantagesmaywellhavebeentheonlyimportantvariablesinthebehavioroftheprimitiveinventorsofbothphysicalandculturaldevices.Buttheelaborationofmoralandethicalpracticeshasreducedtheimportanceofpersonalaggrandizement.Thehonorificreinforcementswithwhichsocietyencouragesactionforthecommonweal,aswellasthesanctionsitapplies to selfish behavior, generate a relatively disinterested creativity. Even in thefield of personal control, improvements may be proposed, not for immediateexploitation,but—asbyreligiousleaders,benevolentrulers,politicalphilosophers,andeducators—for“thegoodofall.”Only an analysis of moral and ethical practices will clarify the behavior of the

cultural designer at this stage.He has faced a special difficulty in the fact that it iseasiertodemonstratetherightwaytobuildabridgethantherightwaytotreatone’sfellowmen(thedifferencereducingto the immediacyandclarityof theresults).Theculturalinventor,eventhoughrelativelydisinterested,hasfounditnecessarytoappealfor support to secular or divine authorities, supposedly inviolable philosophicalpremises,andeventomilitarypersuasion.Nothingofthesorthasbeenneededforthegreaterpartofphysicaltechnology.Thewheelwasnotpropagatedbytheswordorbypromisesofsalvation—itmadeitsownway.Culturalpracticeshavesurvivedorfallenonlyinpartbecauseoftheireffectonthestrengthofthegroup,andthosewhichhavesurvived are usually burdened with unnecessary impedimenta. By association, thecurrentdesignerishandicappedbythefactthatmenlookbehindanyculturalinventionforirrelevant,ingenuous,orthreateningforces.

61

Thereisanotherstepinphysicaltechnology,however,whichmusthaveaparallelinculturaldesign.Thepracticalapplicationofscientificknowledgeshowsanewkindofdisinterestedness. The scientist is usually concernedwith the control of nature apartfrom his personal aggrandizement. He is perhaps not wholly “pure,” but he seekscontrolmainly for itsownsakeor for the sakeof furtheringother scientificactivity.There are practical as well as ethical reasons for this: as technology becomesmorecomplex,forexample,thescientisthimselfislessandlessabletopursuethepracticalimplications of his work. There is very little personal reimbursement for the mostprofitable ideas of modern science. As a result, a new idea may yield immediatetechnologicalimprovementswithoutbringingthescientistundersuspicionofplottingapersonal coup. But social technology has not yet reached this stage.A disinterestedconsideration of cultural practices from which suggestions for improvement mayemergeisstilloftenregardedasimpossible.Thisis thepricewepayforthefact thatmen(1)havesooftenimprovedtheircontrolofothermenforpurposesofexploitation,(2)havehadtobolstertheirsocialpracticeswithspuriousjustifications,and(3)havesoseldomsharedtheattitudesofthebasicscientist.Most people would subscribe to the proposition that there is no value judgment

involved indecidinghow tobuild an atomicbomb,butwould reject thepropositionthat there is none involved in deciding to build one.Themost significant differenceheremaybethatthescientificpracticeswhichguidethedesignerofthebombareclear,whilethosewhichguidethedesigneroftheculturewhichbuildsabombarenot.Wecannotpredictthesuccessorfailureofaculturalinventionwiththesameaccuracyaswedothatofaphysicalinvention.Itisforthisreasonthatwearesaidtoresorttovaluejudgmentsinthesecondcase.Whatweresorttoisguessing.Itisonlyinthissensethatvalue judgments takeupwhere science leavesoff.Whenwecandesignsmall socialinteractions and, possibly, whole cultures with the confidence we bring to physicaltechnology,thequestionofvaluewillnotberaised.Sofar,menhavedesignedtheircultureslargelybyguesswork,includingsomevery

luckyhits;butwearenotfarfromastageofknowledgeinwhichthiscanbechanged.Thechangedoesnotrequirethatwebeabletodescribesomedistantstateofmankindtowardwhichwearemovingor“deciding”tomove.Earlyphysicaltechnologycouldnothaveforeseenthemodernworld,thoughitledtoit.Progressandimprovementarelocalchanges.Webetterourselvesandourworldaswego.Wechangeourculturalpracticesbecauseitisinournatureasmentobereinforced

incertainways.Thisisnotaninfallibleguide.Itcould,indeed,leadtofatalmistakes.For example, we have developed sanitation and medical science to escape fromaversive events associatedwith illness and death, yet a newvirus could conceivablyarisetowipeouteveryoneexceptthosetowhomchronicillnessandfilthhadgrantedimmunity.On thepresent evidence,ourdecision in favorof sanitation andmedicineseemstomakeforsurvival,butinthelightofunforeseeabledevelopmentswemayintimelookbackuponitashavinghadnosurvivalvalue.From time to time, men have sought to reassure themselves about the future by

characterizingprogressastheworkingoutofsomesuchprincipleasthegeneralwill,universalor collective reason,or thegreatest good.Suchaprinciple, if valid,wouldseemtoguaranteeaninevitable,ifdevious,improvementinthehumancondition.Nosuchprincipleisclearlysupportedbyascientificanalysisofhumanbehavior.Yetthenatureofmantellsussomething.Justasanultimategeneticeffectcannotbereachedifimmediate effects are not beneficial, so we must look only to the immediate

62

consequencesofbehaviorformodificationsinaculturalpattern.Nevertheless,culturalinventions have created current conditions which have at least a probabilisticconnectionwithfutureconsequences.Itiseasytosaythatmenworkforpleasureandtoavoidpain,as thehedonistswouldhaveit.Theseare, indeed,powerfulprinciples;but inaffecting theday-to-daybehaviorofmen, theyhave led to theconstructionofcultural devices which extend the range of both pleasure and pain almost beyondrecognition. It is the sameman, biologically speaking,who acts selfishly or for thegoodofthegroup,anditisthesamemanwho,asadisinterestedscientist,willmakehumanbehaviorvastlymoreeffectivethroughculturalinvention.

63

“Man”

Manhaslongsoughttoexplainhisbehaviorbysearchingforitscauses.Historiansandbiographershave tracedhumanachievements to conditionsof birth, climate, culture,and personal contacts, and some of them have joined philosophers and essayists inmore sweeping generalizations. Science has naturallyworked in the same direction.The social sciences specialize in statistical demonstrations, but psychology andphysiologyare closer tohistory andbiography in concentratingon the individual. Inany case, more and more of the behavior of organisms, including man, is beingplausiblyrelatedtoeventsintheirgeneticandenvironmentalhistories.Ifothersciencesareanyguide,humanbehaviormayultimatelybeaccountedforentirelyinsuchterms.Thetraditionalconception,ofcourse,isverydifferent.Itholdsthatamanbehavesas

hedoesbecauseofhiswishes,impulses,emotions,attitudes,andsoon.Hisbehaviorisimportantonlyastheexpressionofaninnerlife.Manypsychologistsstillsubscribetothis view. The good Freudian attributes observable behavior to a drama played innonphysical space by an immanent triumvirate scarcely to be distinguished from thespirits and demons of early animism. Other psychologists merely divide the innerpersonae intoparts,eachofwhichstillcarrieson its littleshareofmental life.Thus,whereascientificanalysisrelatesbehaviortothephysicalenvironment,thementalistmay insist that themind observes only a none-too-reliable copy of the environmentcalledsubjectiveexperience.Whereascientificanalysisshowsthatwereactinagivenwaybecausesimilaractionsinourpasthavehadparticularconsequences,thementalistmay insist thatweactbecausewehavestoredmemoriesofpast actionsandof theirconsequences,whichwenowscaninordertoreachcertainexpectationsleadingtoanactofwillwhichinitiatesbehavior.Whereascientificanalysistracescertaindisturbingpatterns of behavior to a history of punishment, the mentalist may argue that thedisturbance is in the personality and that it is the effect of anxiety, just possiblygenerated by punishment. The traditional conception of man is an example of anexplanatory strategy which was once common in other sciences. It has survived inpsychology,possiblybecauseoftheextraordinarycomplexityofthesubjectmatter.Asplausible connections with external variables are demonstrated in spite of thatcomplexity,however,theneedforinnerexplanationsisreduced.Aneffectivescientificanalysiswouldpresumablydispensewiththemaltogether.That such an analysis will be simpler, more expedient, and more useful will not

necessarily mean its adoption, because the older view served other than scientificfunctions.Abehavioristicreinterpretationofmentallifeisnotafundamentalissueformanypeople,buteveryonehasastakeinhumanbehavior,andthereareotherreasonswhy the scientific picturemay not seem to be a picture ofman at all.Certain long-admiredcharacteristicsofhumanbehaviorseemtobeneglected,andtheirabsenceismore threatening than any implication about the nature of consciousness or theexistenceoffreewill.C.S.Lewis,forexample,hasgonesofarastoarguethatscienceisembarkedupon

64

“theabolitionofman.”1Heisconcernedwiththeneglectofafamiliarfeatureof thetraditionalpicture—anindwellingsenseofjustice,afeltstandardofrightness,aninnersourceofvalues.Tothetraditionalistahumanactisnotsimplyaphysicalmovement,it is a judgment, or the expression of a judgment, reached only by applying certainstandardsofconduct.It isnottheactwhichisessentiallyhuman(morallyacceptablethough itmay be), but the application of the standard.Wemay condition aman tobehave in virtuousways aswe condition animals to behave according to any set ofspecifications,butsuchamanwillnotbevirtuous.According to thisviewhecanbevirtuous only if he has not been conditioned to behave well automatically but hasarrivedatgivenformsofvirtuousconductbyconsultinghissenseof rightness. (Theargument is reminiscent of the complaint that a rational religion destroys piety, thatproof of the existence ofGod deprivesmen of the opportunity to demonstrate theirfaith.)Ifthistraditionalconceptionofmanistocontinuetochallengethescientificview,

however,somethornyquestionsneedtobeanswered.Whatishappeningwhenamanrefers to a standard of rightness?Can this formof behavior be analyzed?Where dostandards come from? If the answer is that they come from the genetic orenvironmentalhistory,thenthescientificviewisnotindanger.Andthisappearstobethe case. Lewis, for example, acknowledges that the sentiments he so highly valuesmustbelearned.“Thelittlehumananimal,”hesays,“wouldnotatfirsthavetherightresponses”—indeed,inthatsensewouldnotyetbehuman.AndhequotesPlatowithapprovaltotheeffectthatsuchthingsastasteandcompassionmustbetaughtbeforeachild is“ofanage to reason.”Theseare thecontentionsofanenvironmentalist.Thevaluestowhichamanmustbeabletoappealinordertobehumanarenotoriginallyhis,andsomethingbeyondhimisthereforeultimatelyresponsibleforhisaction.(Thesame unhappy story can be told of all inner explanations of human conduct, for theexplanationsmustthemselvesbeexplained—possiblyintermsofotherinnereventsbuteventually,andnecessarily,intermsofforcesactinguponamanfromwithout.)Asmall issuesurvivesata technical level.Howarewe to teachachild tobehave

well?We can begin by conditioning him to make so many purely automatic, rightresponses,butweshallfindthatthenumberwhichmustthusbetaughtisdistressinglylarge.Itismoreefficient,ifnotactuallynecessary,toteachhimtoexamineeachnewoccasionasitarisesand,byapplyingcertainrules,toarriveatanappropriateresponse.Suchisourpracticeinteachingmultiplication.Uptotwelve-times-twelveweconditionspecificresponses,eachofwhichcanbequiteautomatic,implyingnounderstandingofmultiplication.Beyondthat,wefinditexpedienttoconditioncertainprocedureswhichpermitthechildtoarriveatavastnumberofspecificproductswhichitwouldnotbeefficienttoconditionseparately.It is sometimes argued that there is an element of freedom in the application of

standardswhichislackingintheautomaticexecutionofrightresponses.Butasenseoffreedomisanotherof thoseinnerattributeswhichlosetheirforceaswemoreclearlyunderstand man’s relation to his environment. Freedom—or, rather, behavior which“feelsfree”—isalsotheproductofahistoryofconditioning.Inthatremarkablebook,Émile, Jean JacquesRousseau tried to find replacements for thepunitivemethodsofthe schools of his time.He insisted that students shouldbehave as theywant, ratherthan as they are forced to behave through physical coercion. He showed anextraordinary ingenuity in substituting positive inducements for punishment. But he

65

wasnotturningeducationovertothepupilhimself.

Let [thechild]believe thathe isalways incontrol, though it isalwaysyou[the teacher]whoreallycontrols.Thereisnosubjugationsoperfectasthatwhichkeepstheappearanceoffreedom,forinthatwayonecapturesvolitionitself.Thepoorbaby,knowingnothing,abletodonothing,havinglearnednothing,ishenotatyourmercy?Canyounotarrangeeverythingin theworldwhichsurroundshim?Canyounot influencehimasyouwish?Hiswork,hisplay, his pleasures, his pains, are not all these in your hands andwithout his knowing it?Doubtlessheoughttodoonlywhathewants;butheoughttowanttodoonlywhatyouwanthimtodo;heoughtnottotakeastepwhichyouhavenotforeseen;heoughtnottoopenhis

mouthwithoutyourknowingwhathewillsay.2

Thusspokeagreatchampionofhumanfreedom!Likeasenseofrightnessorjustice,the dispositions whichmake a given act feel free come from the environment. Thesurvivingquestionisagaintechnical.Whatisthebestwaytobringaboutthosechangeswhicharetheobjectofeducation?Therearemanyadvantagesinarrangingmatterssothatthepupildoeswhathewantstodo,buthemustbecarefullypreparedtowanttodothosethingswhicharerequiredforeffectiveinstruction.Another human attribute which seems to be missing from the scientific picture

concernswhatonedoesnotwanttodo.Inthetraditionalviewamanhasdutiesaswellasrights:therearethingshemustdoorsuffertheconsequences.Heisresponsibleforhisconductinthesensethat,ifhedoesnotbehaveinagivenway,itisonlyfairthathebe punished. To escape punishment—either the natural punishments of the physicalenvironment or the social punishments of society—he engages in an activity calledself-control. When the same ultimate “good” behavior is achieved without usingpunishment,self-controlinthissenseisunnecessary.The omission of personal responsibility from the scientific conception ofman has

beenparticularlydeploredbyJosephWoodKrutch.3Whenweregardacriminalasinneedoftreatmentratherthanpunishment,forexample,wedeprivehimof“thehumanattribute of responsibility.” Treatment is only one way of generating good behaviorwithoutpunishment.Preventivestepsarelikelytobemorevaluable.Forexample,wemightcontrolstealingbycreatingaworldfreeofincitingcircumstances(forexample,aworldinwhichthereisnothingonedoesnotalreadyhaveorwherenothingiswithinreachtobestolen)orbyconditioningbehaviorwhichisincompatiblewithstealingordisplaces it (for example, we might strongly reinforce “respecting the property ofothers”or teacheasier, legalwaysofgetting things).Whenwesolve theproblem inany of theseways,we leave no room for personal responsibility or self-control.Weleave no room for moral struggle; and if to struggle is human, we have indeeddestroyedsomethingofman.Thesameargumentholdsfornongovernmentalpunishments.Smokingcigarettesis

“naturally” punished by lung cancer or the threat of lung cancer, as overeating ispunished by obesity, illness, and the threat of an early death. Aggressive action ispunished by retaliative measures. All these aversive consequences normally lead tosomemeasureofself-control.Butwecanreducetheinclinationtosmoke,eat,oractaggressivelyinotherways—andwithittheneedtocontroloneself.Appropriatedrugshave this effect. A tranquilizer reduces the need to control aggression, an appetite-suppressant reduces the need to control eating, and a drug whichwould reduce the

66

tendencytosmokecigaretteswouldreducetheneedtocontrolone’ssmokinghabits.Anotherformofcontrolwouldbetobuildaworldinwhichthepositivereinforcementsnowaccordedthesebehaviorsarecarefullymanaged.Insuchaworldamanwouldbeeither naturally wise and good or at least easily taught to be wise and good. Therewouldbenoplaceforintellectualandmoralstruggle.Anytechnology,physicalorsocial,whichreducespunishingconsequencesreduces

theneedforself-controlandpersonalresponsibility.Ifthesameacceptableconductisachieved, it is difficult to see why anyone should object. The trouble is that thecharacteristicswhich are thus dismissed have long been admired.We admire peoplewho apply ethical and moral standards, who accept responsibility, and who controlthemselves.Weadmire theminpartbecause the resultsare reinforcing tous, for theindividual is thus induced toconformto the interestsofothers.Wealsoadmiresuchbehavior just inorder to support it.Admiration isa socialpracticeused toekeoutadefectivecontrol.Therearecertainkindsofheroism,forexample,whichsocietycanengenderonlybyeffusivelyadmiringthem.Weinducementodiefortheircountrybyconvincing them that it is sweet and decorous to do so. Students work hard to beadmiredbytheirteachers.Menundergoexhaustinglaborandsufferpainwithpatiencebecause theyareadmiredfordoingso.Yet technologicalprogress isdirected towardmakingall thisunnecessary. Inaworldatpeace therewillbenomilitaryheroismtoadmire.Weshallnolongeradmirepatientsufferingwhenmenseldomneedtosuffer.Wedonotevennowgivemencreditforexhaustinglaborifthelaborcanbe“saved,”and we shall not admire students who work hard when there are techniques ofeducationinwhich theyneednot“work”atall.Weshallno longeradmirewrestlingwith the devil, if it turns out that the devil is simply a slight disturbance in thehypothalamuswhichcanbeallayedbyasuitabledrug.In turning to external andmanipulablevariables, a scientific analysismoves away

fromsupposed inneractivitieswhichwehave tried to reach throughadmiration.Theinner activity, needing to be admired, naturally seems admirable. Thuswe admire aman who can multiply by applying rules more than one who merely recites themultiplication table inanautomatic fashion,butweadmire the latter farbeyondonewhosimplyusesacalculatingmachine.Thecalculatingmachinehasbeendesignedtoreduce thebehavior required inmultiplication to external, sharplydefined, relativelyinfallible, and almost effortless responses. It improvesmultiplication, butmakes themultiplierlessadmirable.Platorecordsanobjectiontotheinventionofthealphabetonsimilargrounds: if textsweregenerallyavailable,amanwouldseemtoknowthingswhichhehadmerelyread.4Butthealphabetwasinventedpreciselytoenableonemantoprofitfromthedirectknowledgeofanother.Mustwedestroyallphysicalandsocialinventionsinordertorecaptureamanwecanwholeheartedlyadmire?Twoimportant featuresoftensaid tobemissingfromthescientificpictureofman

are actually emphasized in it. Ifmanhasno freedomof choice, if he can initiatenoactionwhich alters the causal streamof his behavior, then hemay seem to have nocontroloverhisowndestiny.ThescientificviewofmanaccordingtoKrutchisa“deadend.”5 The fact is, however, that men control both their genetic and environmentalhistories,andinthatsensetheydo,indeed,controlthemselves.Scienceandtechnologyare concerned with changing the world in which men live, and changes are madepreciselybecauseof theireffectsonhumanbehavior.Wehavereachedthestage, farfromadeadend,inwhichmancandeterminehisfuturewithanentirelyneworderof

67

effectiveness.C.S.Lewiswouldstillprotest;inTheAbolitionofManhewrote,“…thepowerofman tomakehimselfwhat he pleasesmeans… thepower of somemen tomake other men what they please.” But it has always been thus. Men controlthemselves by controlling theworld inwhich they live.Theydo this asmuchwhentheyexerciseself-control,aswhentheymakechangesintheirculturewhichaltertheconductofothers.Anotherfeatureofthetraditionalconceptwhichisemphasizedratherthanabolished

isindividuality.Somepracticesderivedfromascientificknowledgeofhumanbehaviorcould no doubt lead to regimentation, as practices consonant with traditionalconceptionshaveoftendone,butthereisnothinginthescientificpositionwhichmakesthis inevitable.On thecontrary,as theproductofasetofgeneticandenvironmentalvariablesman ismost reassuringly unique.The uniqueness of the human fingerprintonce came as a surprise and, because of its practical usefulness, is still a familiarsymbolofindividuality.Butthebodywhicheachmanderivesfromhisgenetichistoryisavastsystemofuniquestructuresofwhichthewhorlsontheballofthethumbarearidiculouslytrivialexample.Equallyidiosyncraticareallthosecharacteristicswhichaman derives from his environment. It is true that certain scientific practices aresimplifiedwhen these sourcesof individualityareminimized,but there isnothing inscientificpracticeor theorywhich threatens individualityor questions thepossibilitythatsomecollocationsofvariablesarisingfromthesesourceswillhavetheoutstandingresultsweattributetotalentorgenius.It is not easy to abandon notions like a sense of justice, a sense of freedom, and

personalresponsibilityortoacceptanewinterpretationofman’sindividualityandhispowertocontrolhisowndestiny.Yetitwouldberemarkableifanyconceptionofmandidnotoccasionallyneedrevision.Humanbehaviorisextraordinarilycomplex,anditis unlikely that a true definitive account has been reached so soon. The traditionalconceptionhascertainlynotmadeusconspicuouslysuccessfulindealingwithhumanaffairs. The alternative picture which a science of behavior asks us to accept is notreallyfrightening.Mansurvivesunchanged.Physicsdoesnotchangethenatureoftheworld it studies, andno scienceof behavior can change the essential nature ofman,even though both sciences yield technologieswith a vast power tomanipulate theirsubjectmatters.Scienceleadsustoseemaninadifferentlight,butheisneverthelessthesamemanweoncesawinanotherlight.Ifwemusthavesomethingtoadmire,letitbe man’s willingness to discard a flattering portrait of himself in favor of a moreaccurateandhencemoreusefulpicture.Evenhereadmirationissuperfluous.Thehardfact is that the culture whichmost readily acknowledges the validity of a scientificanalysis ismost likely to be successful in that competition between cultures which,whetherwelikeitornot,willdecideallsuchissueswithfinality.

FromProceedingsoftheAmericanPhilosophicalSociety,1964,108,482–485.1Lewis,C.S.TheAbolitionofMan.NewYork,1947.2Rousseau,JeanJacques.ÉmileoudeL’Éducation.AmsterdametFrancfort,1762.Page121intheClassiquesGarnierEdition.

3Krutch,JosephWood.TheMeasureofMan.Indianapolis:Bobbs-Merrill,1953.4Plato.Phaedrus.Jowetttranslation,III:274e–275b.

68

5Krutch,JosephWood.WhatIlearnedaboutexistentialism.SaturdayReview,45(April21,1962).

69

TheDesignofExperimentalCommunities

Acommunitymay be thought of as a small state, even aminiatureworld, inwhichsomeof theproblemsof implementingawayof lifearereducedtomanageablesize.Manykindsofcommunitieshaveservedthispurpose.Althoughseeminglysuccessfulunplannedcultureshaveoftenbeentakenasmodels(ArcadiabytheGreeks,theSouthSea islands by the eighteenth-century social philosophers), this article is concernedwithcommunitieswhichhavebeenormightbeexplicitlydesigned.Some of the rules of the Qumran Community were set forth in the Manual of

Discipline found among the Dead Sea Scrolls, which the community helped topreserve. The rules of Benedict and Augustine governed life in similar monasticcommunities. Semireligious and secular communities flourished in the nineteenthcentury in America (the Oneida Community is a particularly interesting example).1Explicitly designed, or intentional, communities of the twentieth century range fromthe intensely religious Bruderhof to the essentially secular kibbutzim in Israel. TheSovietcollectivesandmikroraionsandtheChinesecommunes,thoughpartsoflargergovernmental structures, are other examples. Fictional communities—for example,thosedescribed inThomasMore’sUtopia2 and FrancisBacon’sNewAtlantis3 havealsocapturedmen’simaginations.In its relation to government in the broadest sense, a community, speculative or

attempted,servessomethingofthefunctionofapilotexperimentinscienceorapilotmodel or plant in technology. It is constructed on a small scale. Certain problemsarising from sheer size—such as communication and transportation—can then beneglected,but themainadvantageis thatcloserattentioncanbegiventothelivesofindividualmembers.Suchacommunityisalsoalmostalwaysgeographicallyisolated.Utopias have often occupied islands, butwalls isolate almost aswell aswater. (Themembers of a sect, no matter how well organized, are not usually regarded as acommunity if they are widely dispersed geographically.) There is also a certainisolation from tradition. The eighteenth-century European could expect to abandonmuchofhisculturewhenhereachedTahiti;lifeinamonasterymaybeginwitharitualof rebirth. All this makes it easier to think about such a community as a viable orperishableentity—asanorganismwithalifeofitsown.Itssuccessorfailure,unliketheriseandfalloferasornations,islikelytobequickandconspicuous.Newwaysofdoingthingsaretestedfortheirbearingonitssuccess.Suchacommunity,inshort,isanexperiment.Menfound,join,ordreamofsuchcommunitiesformanyreasons.Somearemoved

by intellectual interests: they want to prove a theory (for example, that men arenaturally noble or that they are incomplete without “community” or “love”) or tohasten a prophesied stage in history.Others havemore immediate personal reasons:they seek simple pleasures, the satisfaction of basic needs, political order, economicstability,helpinself-discipline,andsoon.Suchgoalsareoftenformalizedas“values.”The goal of the community is tomaximize happiness, security, sanctity, or personal

70

fulfillment.Themoregeneralthegoal,however,themoredebatableitseemstobe.Inconceivingofacommunityasapilotexperiment,thedesignermayturndirectlytotwopracticalquestions:Whatbehavioronthepartofthemembersofacommunityismostlikely to contribute to its success? How may that behavior be generated andmaintained?Someanswerstothefirstquestionarequiteobvious.Itisimportanttoacommunity

thatitsmembersdefenditagainstitsenemies,producethefood,shelter,clothing,andotherthingsitneeds,andmaintaininternalorder.Itisalsoobviouslyimportantthatitsmembersteacheachother,and,particularly,newmembers,howtobehaveinnecessaryways. Other kinds of behavior—for example, in the uses of leisure—often figureprominentlyamongexpressedgoals,buttheirrelevancetothesuccessofacommunityis not always clear. These behaviors are thingsmembers “want to do,” and variousreasonsmaybegivenfordoingthem,butthedesignermayproceedmosteffectivelybyconfininghimselftobehaviorsthataredemonstrablyrelatedtosuccessorsurvival.Thesecondquestionhasusuallybeenansweredbyappealtohistoricalanalogy.Men

havelivedpeacefully,productively,stably,andhappilyundermanyobservedsystemsor structures of government, economics, society, family life, and so on. There is astrong presumption that a given system generates the behavior observed under it, aspolitical science, economics, sociology, and other social disciplines usually contend.Wemightconclude,therefore,thatthedesignerhasonlytochooseamongsystemsorstructures. Should the government of a community be authoritarian or democratic?Should the society be open or closed? Should the social structure be classless orstratified?Shouldtheeconomybeplannedorlaissez-faire?Shouldthefamilybestrongorweak?Questions at this level of analysis offer little practical help in designing acommunity. Terms like “authoritarian” and “laissez-faire” seldom refer to propertieswhichadesigner canbuild into a social environment, and terms like “peaceful” and“stable”donotsharplycharacterizebehaviorwhichcanbeshowntocontributetothesuccessofsuchanenvironment.There isamoreuseful levelofanalysis.Everydeveloped languagecontains terms

which describe in great detail the social environment and the behavior it generates.Rules of thumb useful in modifying behavior are expressed in such terms. Thus,everyoneknowshowtoattractaman’sattention,toarousehimemotionally,torewardandpunishhim,andsoon.Communitiesareusuallydesignedwithaneyetothislevelof human behavior. The designer is concerned not with a hypothetical type ofeconomicsystembutwithactualworkingconditions,notwithahypothetical typeofgovernment but with ethical practices and instructions in self-discipline, not with aformalconceptionofsocialorfamilystructurebutwithspecificinteractionsamongthemembersofagroup.The relations between behavior and environment at this level have only recently

been formulated in a systematic way. It is significant that statements expressing anunderstanding of human nature or a skill in handling people—for example, in theessays of such men as Bacon or Montaigne or in sporadic comments by politicalscientists, economists, and others—have remained aphoristic. They have never beenbrought together in a coherent, consistent account. Psychology is the scientificdiscipline relevant here, but it has only recently been able to supply an effectivealternative to folklore and personal experience. A special branch of psychology hasnow reached the point at which promising technological applications are becomingfeasible. The principles derived from an experimental analysis of behavior offer the

71

designerconsiderablehelp insettingupanenvironmentunderwhichbehaviorwhichwillcontributetothesuccessofthecommunitymaybegenerated.At any level of analysis, certain conditions either lie beyond the control of the

designer or, if used by him to advantage, limit the significance of his design as ageneralsolution.Hecannotactuallyinstituteanewcultureallatonce:theearliersocialenvironmentsof themembersofacommunitywillplayarole, ifonlyinprovidingacontrast to a new way of life. Members may show personal idiosyncrasies orbackground differences. They may have been explicitly selected—and will almostcertainly be self-selected—with respect to some such trait as cooperativeness orintelligence.Thesiteofthecommunity—itsclimate,soil,andexistingfloraandfauna—willbefavorableorunfavorable.Thecommunitywillbeginwithacertainamountofstartingcapital, itwillhavenatural resources, and itmaycontinue to receiveoutsidesupport in the form of charity or philanthropy. All these conditions limit thesignificanceofasuccessfulresult,butthereisstillscopeforextensivedesign.Afewexamplesmustsufficehere.

NegativeReinforcement

An important element in any culture is the use of force. The state is often definedprimarily in terms of the power to punish. We say that punishment requires forcebecauseitsimpositionisresisted.Inpoliticaltheorytherightandpowertopunisharediscussed under some such concept as “sovereignty.” The behavioral processes areobvious and easily related to the role of punishment. The term applies, strictlyspeaking,onlytothesuppressionofunwantedbehavior,butthepunishingeventsusedforthatpurposecanbeusedtogeneratebehavior—toinducepeopletobehaveingivenways by “punishing them for not behaving.” The technique is particularly useful inoffsettingotheraversiveconsequences,asinforcingmentofightortofillproductionquotas.Effectivelyused, punishment in thisbroad sense canmakemen law-abiding,obedient,anddutiful.But there are inevitable side effects.Onewho isbehavingwell inorder to escape

punishmentmaysimplyescapeinotherways,asexemplifiedbymilitarydesertionandreligious apostasy. Extensive use of punishment will cost a community some of itsmembers.Itmayalsoleadtocounterattack—asinrevolutionorreligiousreformation—or to stubborn resistance to all forms of control. These are familiar, predictablereactionsuponwhichanexperimentalanalysisofbehaviorthrowsconsiderablelight.Aslow,erratictrendtowardminimizingaversivecontrolinthedesignofacommunityisactually an example of such a by-product. This trend is exemplifiedwhen powerfulmilitaryorpoliceactionisreplacedbyethicalcontrolimposedbythosewithwhomthecitizen is in immediatecontactorwheneducationalprogramsaredesigned to reducethe frequency with which aversive behavior occurs or to prepare the individual toadjust more effectively to any remaining forcible control. An example of a moreextreme alternative is the cloister, an environment in which unwanted behavior isunlikelyorimpossibleandinwhichwantedbehaviorisparticularlylikelytooccur.

PositiveReinforcement

72

Averydifferentexampleoftherelevanceofananalysisofbehaviortothedesignofacommunityistheuseofso-calledrewards.Acommunitymayneedasmuchpowertorewardastopunish,butit isnotsaidtobeusingforcebecauseitsoperationsarenotresisted. Reward refers very loosely to the “positive reinforcers” which have beenextensivelyanalyzedinlaboratoryresearch.Itisabasicprinciplethatbehaviorwhichis followed by certain kinds of consequences is more likely to occur again, butreinforcementsmaybecontingentonbehaviorinmanysubtleandcomplexways,andextensivetechnologicalknowledgeisneededtouse theprincipleeffectively inall itsramifications.Althoughitisgenerallytruethatthegreaterthereinforcementthemoreit is productive of behavior, the amount of behavior generated is not related in anysimplewaytotheamountofreinforcement.Thenetgainorutilityofanactionhaslittlerelationtotheprobabilitythattheactionwilloccur.Indeed,undercertaincontingenciesof reinforcement—for example, ingambling—behaviormaybemaintained at a highlevelforlongperiodsoftimeeventhoughthenetmonetarygainisnegative.A community may resort to positive reinforcement to generate any behavior

importanttoitssuccess.Forexample,itmayarrangeforreinforcementthroughgroupapprovalofacceptedbehaviorasanalternative tocoercive legalorethicalcontrol. Itwill also be interested, of course, in the classical problemofmaintaining productivelabor. (If there is any established discipline which is most closely concerned withpositivereinforcement,itiseconomics.)Thedesignerofeffectiveworkingconditionsinasmallcommunityisinafavorablepositiontouseatechnologyofreinforcement.The immediate temporal contingencies are crucial. Many communities have givenspecial attention to rewarding productive labor. Some have returned to conditionswhich prevail in the life of the craftsman —that is, they have used the naturalreinforcingconsequencesof labor. It isnotaveryenlightenedsolution.Furthermore,the use of money as a reinforcement is admittedly not as simple as it may at firstappear.Thevalueofmoneymust,ofcourse,betaught—butsomustthevalueofearlystagesofcraftwork.Themaindifficulty is thatwagesareartificiallycontingentuponthebehaviorwhichproducesthem,andithasbeendifficulttoconstructcontingencieswhichmaintainproductivelaborwithoutundesirablesideeffects.Itwasoncethoughtthatthedeficiencymustbeoffsetbymakingwagesmorepowerfulasreinforcers—forexample,bymaintainingahungry labor force.Anothersolutionhasbeen to increasethe actual amount of reinforcement (by raising wages). The contingencies ofreinforcement have remained poorly analyzed, however. Current systems of rewardsare largelyaversive, the threatened lossofastandardof livingbeingmore importantthan the receipt ofwages. Effective reinforcement of productive labor is one of themoreinterestingareasinwhichthedesignerofanexperimentalcommunitymayapplyrecentscientificdiscoveries.Whengoodsandserviceswhichmaybeusedasreinforcersareallowedtobecome

availableforotherreasons—when,forexample,theyaresuppliedbyabountifulnatureor a bountiful government concerned with welfare or happiness—much of theirreinforcingeffectislost.Wemakeexplicituseofthisprinciplewhen,asanalternativetopunishment,wedeliberatelydestroycontingenciesbysupplyingreinforcersgratis—for example,whenwegivemen the things theywouldotherwisebehave illegally toget. If the community does not need productivework, reinforcing contingencies cansafely be neglected, but a long-standing conflict between welfare and incentivesuggeststhattheissuehasnotbeenwhollyresolved.

73

LEISURE

Positivereinforcementoccupiesanespeciallyimportantplaceinsolvingtheproblemofleisure.Withmoderntechnologyitisconceivablethatamanneednotspendmuchtimeinmakinghiscontributiontopeaceandprosperity.Whatishetodowiththerestof his time? Perhaps it does not matter. If the community has solved the essentialproblemsofdailylife,itmayleaveeachmemberfreetodoashepleases.Butheisfreeonlytocomeunderotherformsofcontrol.Iftherearenoeffectivereinforcers,hemayspendallhiswakinghoursdoingnothing.Orhemaycomeunderthesustainedcontrolof biological reinforcers, such as food, sex, aggressive damage to others, or drug-induced euphoric states. Weaker reinforcers will take control when they occur onpowerfulschedules:leisureisoftenspentinrepetitiveandcompulsiveactivities,suchassolitaireorothersimplegames.These are all forms of behavior which flourish when behaviors having a more

specificrelevancetothesuccessofacommunityarenotneeded.Acommunitymaybeabletoaffordacertainnumberofthem,butitstandstoprofitmorefromotherusesoffreetime.Sports,games,andotherformsofcomplexplay;artsandcrafts,music,andthe dance; literature and the theater; and the contemplation, observation, andexploration of naturewhich constitute “science” in the broadest sense are importantactivitiestothedesignerbecausetheybearonthesuccessofthecommunity.Someofthemmakethecommunitymoreattractive in thesensethat theyreinforcesupportingbehavioranddiscouragedefection.Forexample,theyreinforcethesimplebehaviorofremaininginthecommunity.Otheractivitiesdevelopextraordinaryskillswhichmakeit possible for members to meet emergencies with maximum effectiveness. Thosewhich advance science yield the physical and cultural technologies needed for themaintenanceandimprovementofthecommunityasawayoflife.Theserelationstothesuccessofacommunityareoverlookedinsayingthatleisureis

tobedevotedtothepursuitofhappiness,forthisemphasizesthereinforcersratherthanthebehaviorsreinforced.Theconceptof“happiness”(or,lessfrivolously,“fulfillment”or “enrichment”) is often felt to be a necessary, if admittedly troublesome, value inexplainingman’ssearchforawayoflife.Fromthepointofviewofanexperimentalanalysisofbehavior,itappearstobemerelyanawkwardwayofrepresentingtherolesofpositiveandnegativereinforcers.Itsmainfaultisitsneglectofthecontingenciesofreinforcement.Asked todescribeaworld inwhichhewould like to live, amanwilloftenreferdirectly to reinforcingconditions—freedomfromaversivestimulationandanabundanceofpositivereinforcers—buthethenfindshimselfunpreparedformanyparadoxes,suchastheoftenencounteredunhappinessofthose“whohaveeverything”or,inthatotherfieldofutopianspeculation,man’sfailuretoconceiveofaninterestingheaven.

Insummary,then,acommunityismuchmorecomplexthanalaboratoryexperimentinhumanbehaviorbutmuchsimplerthanthelarge-scaleenterprisesanalyzedinpoliticalscience,economics,andothersocialdisciplines.Forthisreasonitisespeciallyhelpfulinstudyingtheeffectsofasocialenvironmentonhumanbehaviorand,inreturn, therelevanceofthatbehaviortothemaintenanceanddevelopmentoftheenvironment.Itis a favorable ground for social invention. A surprising number of practices firstdescribed in utopian thinking have eventually been adopted on a broader scale. Inwriting the New Atlantis Francis Bacon could imagine that scientists might be

74

organized to solve the problemsof the community.Only after he hadmade such anorganizationplausiblewas theRoyalSociety founded—andquiteclearlyonBacon’smodel. More general principles are also encouraged. The success or failure of acommunity, for example, is easily seen to mean the success or failure of all itsmembers, whether or not its social structure is egalitarian; but it is hard to reach asimilarsenseofcommunityinthinkingaboutanationortheworldasawhole.Ithasbeensuggestedthatthewell-governedGreekcity-state,bypermittingmento

conceiveofanorderlyworldofnature,ledtothedevelopmentofGreekscience.Littlein theworld todaycouldhave thateffect, for theorder isnowclearlyon the sideofscience.Butiftheprincipleswhichareemergingfromthelaboratorystudyofhumanbehavior can be shown to be relevant, then sciencemay repay its debt by bringingorderbackintohumanaffairs.

FromTheInternationalEncyclopediaoftheSocialSciences.DavidL.Sills,editor.NewYork:CrowellCollierandMacmillan,1968.

1Noyes,Pierpont.MyFather’sHouse:AnOneidaBoyhood.NewYork:Farrar&Rinehart,1937.

2More,Thomas(1516).TheCompleteWorksofSt.ThomasMore,vol.4:Utopia.NewHaven:YaleUniversityPress,1966.

3Bacon,Francis(1627).NewAtlantis.

75

PARTII

AMethodfortheExperimentalAnalysisofBehavior–ItsTheoryandPractice,ItsHistory,andaGlimpseofItsFuture

AreTheoriesofLearningNecessary?TheAnalysisofBehaviorACaseHistoryinScientificMethodTheExperimentalAnalysisofBehaviorReinforcementToday

76

…saucehispalate

WiththymostoperantpoisonTIMONOFATHENS,IV,iii

77

AreTheoriesofLearningNecessary?Thispaper,presentedatameetingoftheMidwesternPsychologicalAssociationin

May, 1949, was mainly concerned with the nature and function of theory, with ananalysisofthe“measures”usedintracingthelearningprocess,withthequestionofauseful dependent variable, with rate of responding as a measure of probability ofresponse,andwiththeindependentvariableswhichgovernbehavior.Italsoprovidedanopportunitytoreportillustrativematerialfromaseriesofexperimentscarriedoutduringtheprecedingdecade.Insomeofthese,supportedbyGeneralMills,Inc.,andby the Office of Scientific Research and Development, Keller Breland, NormanGuttman,W.K. Estes, andMarionBreland collaborated.Otherswere conducted atIndianaUniversitywiththeassistanceofClaytonK.Bishop.ThepaperwaspublishedinPsychologicalReview,1950,57,193-216andisreprintedherebypermission.

Certain basic assumptions, essential to any scientific activity, are sometimes calledtheories.Thatnatureisorderlyratherthancapriciousisanexample.Certainstatementsarealsotheoriessimplytotheextentthattheyarenotyetfacts.Ascientistmayguessattheresultofanexperimentbeforetheexperimentiscarriedout.Thepredictionandthelaterstatementofresultmaybecomposedofthesametermsinthesamesyntacticarrangement,thedifferencebeinginthedegreeofconfidence.Noempiricalstatementiswhollynontheoretical in thissensebecauseevidence isnevercomplete,nor isanypredictionprobablyevermadewhollywithoutevidence.Thetermtheorywillnotreferheretostatementsofthesesortsbutrathertoanyexplanationofanobservedfactwhichappeals to events taking place somewhere else, at some other level of observation,describedindifferentterms,andmeasured,ifatall,indifferentdimensions.Three types of theory in the field of learning satisfy this definition. The most

characteristic is to be found in the field of physiological psychology. We are allfamiliarwiththechangeswhicharesupposedtotakeplaceinthenervoussystemwhenan organism learns. Synaptic connections are made or broken, electrical fields aredisruptedorreorganized,concentrationsofionsarebuiltuporallowedtodiffuseaway,andsoon.Inthescienceofneurophysiologystatementsofthissortarenotnecessarilytheoriesinthepresentsense.Butinascienceofbehavior,whereweareconcernedwithwhetherornotanorganismsecretessalivawhenabellrings,orjumpstowardagraytriangle, or says bik when a card reads tuz, or loves someone who resembles hismother,allstatementsaboutthenervoussystemaretheoriesinthesensethattheyarenotexpressedinthesametermsandcouldnotbeconfirmedwiththesamemethodsofobservationasthefactsforwhichtheyaresaidtoaccount.A second type of learning theory is in practice not far from the physiological,

although there is less agreement about themethodofdirect observation.Theoriesofthis type have always dominated the field of human behavior. They consist ofreferences to “mental” events, as in saying that an organism learns to behave in acertainwaybecauseit“findssomethingpleasant”orbecauseit“expectssomethingtohappen.” To the mentalistic psychologist these explanatory events are no moretheoretical than synaptic connections to the neurophysiologist, but in a science of

78

behaviortheyaretheoriesbecausethemethodsandtermsappropriatetotheeventstobeexplaineddifferfromthemethodsandtermsappropriatetotheexplainingevents.Ina third typeof learning theory theexplanatoryeventsarenotdirectlyobserved.

The writer’s suggestion [in The Behavior of Organisms] that the letters CNS beregardedasrepresenting,nottheCentralNervousSystem,buttheConceptualNervousSystemseemstohavebeentakenseriously.Manytheoristspointoutthattheyarenottalking about the nervous system as an actual structure undergoing physiological orbiochemicalchangesbutonlyasasystemwithacertaindynamicoutput.Theoriesofthissortaremultiplyingfast,andsoareparalleloperationalversionsofmentalevents.A purely behavioral definition of expectancy has the advantage that the problem ofmentalobservationisavoidedandwithittheproblemofhowamentaleventcancausea physical one. But such theories do not go so far as to assert that the explanatoryevents are identical with the behavioral facts which they purport to explain. Astatementaboutbehaviormaysupportsuchatheorybutwillneverresembleitintermsor syntax. Postulates are good examples. True postulates cannot become facts.Theoremsmaybededuced from themwhich,as tentative statementsaboutbehavior,may or may not be confirmed, but theorems are not theories in the present sense.Postulatesremaintheoriestotheend.Itisnotthepurposeofthispapertoshowthatanyofthesetheoriescannotbeputin

goodscientificorder,orthattheeventstowhichtheyrefermaynotactuallyoccurorbestudied by appropriate sciences. It would be foolhardy to deny the achievements oftheories of this sort in the history of science. The question of whether they arenecessary, however, has other implications and isworth asking. If the answer is no,then itmaybepossible toargueeffectivelyagainst theory in the fieldof learning.Ascience of behavior must eventually deal with behavior in its relation to certainmanipulablevariables. Theories—whether neural, mental, or conceptual—talk aboutinterveningstepsintheserelationships.Butinsteadofpromptingustosearchforandexplore relevant variables, they frequently have quite the opposite effect.When weattributebehaviortoaneuralormentalevent,realorconceptual,wearelikelytoforgetthatwestillhavethetaskofaccountingfortheneuralormentalevent.Whenweassertthatananimalactsinagivenwaybecauseitexpectstoreceivefood,thenwhatbeganas the task of accounting for learned behavior becomes the task of accounting forexpectancy.Theproblemisatleastequallycomplexandprobablymoredifficult.Wearelikelytocloseoureyestoitandtousethetheorytogiveusanswersinplaceoftheanswers we might find through further study. It might be argued that the principalfunctionoflearningtheorytodatehasbeen,nottosuggestappropriateresearch,buttocreateafalsesenseofsecurity,anunwarrantedsatisfactionwiththestatusquo.Researchdesignedwithrespecttotheoryisalsolikelytobewasteful.Thatatheory

generates research does not prove its value unless the research is valuable. Muchuselessexperimentationresultsfromtheories,andmuchenergyandskillareabsorbedbythem.Mosttheoriesareeventuallyoverthrown,andthegreaterpartoftheassociatedresearch is discarded.This could be justified if itwere true that productive researchrequiresatheory—asis,ofcourse,oftenclaimed.It isarguedthatresearchwouldbeaimless and disorganized without a theory to guide it. The view is supported bypsychological texts which take their cue from the logicians rather than empiricalscienceanddescribethinkingasnecessarilyinvolvingstagesofhypothesis,deduction,experimental test, and confirmation. But this is not thewaymost scientists actuallywork. It is possible to design significant experiments for other reasons, and the

79

possibilitytobeexaminedis thatsuchresearchwill leadmoredirectlytothekindofinformationwhichascienceusuallyaccumulates.The alternatives are at least worth considering. How much can be done without

theory? What other sorts of scientific activity are possible? And what light doalternativepracticesthrowuponourpresentpreoccupationwiththeory?Itwouldbeinconsistenttotrytoanswerthesequestionsatatheoreticallevel.Letus

thereforeturntosomeexperimentalmaterialinthreeareasinwhichthoriesoflearningnow flourish and raise the question of the function of theory in a more concretefashion.

TheBasicDatuminLearning

What actually happens when an organism learns is not an easy question to answer.Thosewhoareinterestedinascienceofbehaviorwillinsistthatlearningisachangeinbehavior, but they tend to avoid explicit references to responses or acts as such.“Learningisadjustmentoradaptationtoasituation.”Butofwhatstuffareadjustmentsand adaptations made? Are they data, or inferences from data? “Learning isimprovement.”Butimprovementinwhat?Andfromwhosepointofview?“Learningis restoration of equilibrium.” But what is in equilibrium and how is it put there?“Learningisproblemsolving.”Butwhatarethephysicaldimensionsofaproblem—orof a solution? Definitions of this sort show an unwillingness to take what appearsbeforetheeyesinalearningexperimentasabasicdatum.Particularobservationsseemtootrivial.Anerrorscorefalls;butwearenotreadytosaythatthisislearningratherthanmerely the result of learning. An organismmeets a criterion of ten successfultrials;butanarbitrarycriterionisatvariancewithourconceptionofthegeneralityofthelearningprocess.Thisiswheretheorystepsin.Ifitisnotthetimerequiredtogetoutofapuzzlebox

whichchanges in learning,but rather thestrengthofabond,or theconductivityofaneuralpathway,or theexcitatorypotentialofahabit, thenproblemsseem tovanish.Getting out of a box faster and faster is not learning; it ismerely performance.Thelearninggoesonsomewhereelse,inadifferentdimensionalsystem.Andalthoughthetime requireddependsuponarbitrary conditions, oftenvaries discontinuously, and issubject to reversal of magnitude, we feel sure that the learning process itself iscontinuous, orderly, and beyond the accidents ofmeasurement.Nothing could betterillustratetheuseoftheoryasarefugefromthedata.Butwemusteventuallygetbacktoanobservabledatum.Iflearningistheprocess

wesupposeittobe,thenitmustappearsointhesituationsinwhichwestudyit.Evenifthebasicprocessbelongstosomeotherdimensionalsystem,ourmeasuresmusthaverelevantandcomparableproperties.Butproductiveexperimentalsituationsarehardtofind,particularlyifweacceptcertainplausiblerestrictions.Toshowanorderlychangein the behavior of theaverage rat or ape or child is not enough, since learning is aprocessinthebehavioroftheindividual.Torecordthebeginningandendoflearningof a fewdiscrete stepswill not suffice, since a series of cross-sectionswill not givecompletecoverageofacontinuousprocess.Thedimensionsofthechangemustspringfromthebehavioritself;theymustnotbeimposedbyanexternaljudgmentofsuccessorfailureoranexternalcriterionofcompleteness.Butwhenwereviewthe literaturewiththeserequirementsinmind,wefindlittlejustificationforthetheoreticalprocess

80

inwhichwetakesomuchcomfort.The energy level or work-output of behavior, for example, does not change in

appropriate ways. In the sort of behavior adapted to the Pavlovian experiment(respondentbehavior)theremaybeaprogressiveincreaseinthemagnitudeofresponseduring learning. But we do not shout our responses louder and louder as we learnverbalmaterial,nordoesaratpressaleverharderandharderasconditioningproceeds.In operant behavior the energy ormagnitude of response changes significantly onlywhensomearbitraryvalueisdifferentiallyreinforced—whensuchachangeiswhatislearned.Theemergenceofarightresponse incompetitionwithwrongresponses isanother

datumfrequentlyusedin thestudyof learning.Themazeandthediscriminationboxyield resultswhichmay be reduced to these terms.But a behavior-ratio of right vs.wrongcannotyieldacontinuouslychangingmeasureinasingleexperimentonasingleorganism.Thepointatwhichoneresponsetakesprecedenceoveranothercannotgiveusthewholehistoryofthechangeineitherresponse.Averagingcurvesforgroupsoftrialsororganismswillnotsolvethisproblem.Increasingattentionhasrecentlybeengiventolatency,therelevanceofwhich,like

thatofenergy level, is suggestedby thepropertiesofconditionedandunconditionedreflexes.But inoperantbehavior therelation toastimulus isdifferent.Ameasureoflatency involves other considerations, as inspection of any case will show. Mostoperant responses may be emitted in the absence of what is regarded as a relevantstimulus. In such a case the response is likely to appear before the stimulus ispresented.Itisnosolutiontoescapethisembarrassmentbylockingaleversothatanorganism cannot press it until the stimulus is presented, since we can scarcely becontent with temporal relations which have been forced into compliance with ourexpectations.Runwaylatenciesaresubjecttothisobjection.Inatypicalexperimentthedoorofastartingboxisopenedandthetimewhichelapsesbeforearatleavestheboxismeasured.Opening the door is not only a stimulus, it is a change in the situationwhich makes the response possible for the first time. The time measured is by nomeansassimpleasa latencyandrequiresanother formulation.Agreatdealdependsupon what the rat is doing at the moment the stimulus is presented. Someexperimenterswaituntil therat is facing thedoor,but todoso is to tamperwith themeasurementbeingtaken.If,ontheotherhand,thedoorisopenedwithoutreferencetowhat the rat is doing, the first major effect is the conditioning of favorable waitingbehavior. The rat eventually stays near and facing the door. The resulting shorterstarting-time is not due to a reduction in the latency of a response, but to theconditioningoffavorablepreliminarybehavior.Latencies in a single organism do not follow a simple learning process. Relevant

data on this point were obtained as part of an extensive study of reaction time. Apigeon,enclosedinabox,isconditionedtopeckatarecesseddiscinonewall.Foodispresented as reinforcement by exposing a hopper through a hole below the disc. Ifresponses are reinforcedonly after a stimulushasbeenpresented, responses at othertimes disappear. Very short reaction times are obtained by differentially reinforcingresponseswhichoccur very soon after the stimulus.1But responses also come to bemadeveryquicklywithoutdifferentialreinforcement.Inspectionshowsthatthisisduetothedevelopmentofeffectivewaiting.Thebirdcomestostandbeforethediscwithits head in good striking position. Under optimal conditions, without differential

81

reinforcement,themeantimebetweenstimulusandresponsewillbeoftheorderof⅓second.Thisisnotatruereflexlatency,sincethestimulusisdiscriminativeratherthaneliciting,butitisafairexampleofthelatencyusedinthestudyoflearning.Thepointisthatthismeasuredoesnotvarycontinuouslyorinanorderlyfashion.Bygivingthebird more food, for example, we induce a condition in which it does not alwaysrespond. But the responses which occur show approximately the same temporalrelationtothestimulus(Figure1,middlecurve).Inextinction,ofspecialinteresthere,thereisascatteringoflatenciesbecauselackofreinforcementgeneratesanemotionalcondition. Some responses occur sooner and others are delayed, but the commonestvalue remainsunchanged (bottomcurve inFigure1).The longer latencies are easilyexplained by inspection. Emotional behavior, of which examples will bementionedlater, is likely to be in progress when the ready-signal is presented. It is often notdiscontinuedbeforethe“go”signalispresented,andtheresultisalongstarting-time.Casesalsobegin toappear inwhich thebirdsimplydoesnot respondatallduringaspecifiedtime.Ifweaveragealargenumberofreadings,eitherfromonebirdormany,wemaycreatewhatlookslikeaprogressivelengtheningoflatency.Butthedataforanindividualorganismdonotshowacontinuousprocess.Anotherdatumtobeexaminedistherateatwhicharesponseisemitted.Fortunately

the story here is different. We study this rate by designing a situation in which aresponse may be freely repeated, choosing a response (for example, touching orpressing a small lever or key) which may be easily observed and counted. Theresponsesmayberecordedonapolygraph,butamoreconvenientformisacumulativecurvefromwhichrateofrespondingisimmediatelyreadasslope.Therateatwhicharesponseisemittedinsuchasituationcomesclosetoourpreconceptionofthelearningprocess. As the organism learns, the rate rises. As it unlearns (for example, inextinction) the rate falls.Various sortsofdiscriminative stimulimaybebrought intocontrol of the response with corresponding modifications of the rate. Motivationalchanges alter the rate in a sensitiveway. So do those eventswhichwe speak of asgenerating emotion.The range throughwhich the ratevaries significantlymaybe asgreat as of the order of 1000:1. Changes in rate are satisfactorily smooth in theindividualcase,sothatitisnotnecessarytoaveragecases.Agivenvalueisoftenquitestable: in the pigeon a rate of four or five thousand responses per hour may bemaintainedwithoutinterruptionforaslongasfifteenhours.

82

FIG.1

Rateof respondingappears tobe theonlydatumwhichvaries significantlyand intheexpecteddirectionunderconditionswhichare relevant to the“learningprocess.”Wemay,therefore,betemptedtoacceptitasourlong-sought-formeasureofstrengthofbond,excitatorypotential,etc.Onceinpossessionofaneffectivedatum,however,wemayfeellittleneedforanytheoreticalconstructofthissort.Progressinascientificfieldusuallywaitsuponthediscoveryofasatisfactorydependentvariable.Untilsuchavariablehasbeendiscovered,weresort to theory.Theentitieswhichhavefiguredsoprominently in learning theory have served mainly as substitutes for a directlyobservableandproductivedatum.Theyhavelittlereasontosurvivewhensuchadatumhasbeenfound.It is no accident that rate of responding is successful as a datum because it is

particularlyappropriatetothefundamental taskofascienceofbehavior.Ifwearetopredictbehavior(andpossiblytocontrolit),wemustdealwithprobabilityofresponse.Thebusiness of a scienceof behavior is to evaluate this probability and explore the

83

conditionswhichdetermineit.Strengthofbond,expectancy,excitatorypotential,andso on, carry the notion of probability in an easily imagined form, but the additionalproperties suggested by these terms have hindered the search for suitablemeasures.Rate of responding is not a “measure” of probability, but it is the only appropriatedatuminaformulationintheseterms.As other scientific disciplines can attest, probabilities are not easy to handle.We

wishtomakestatementsaboutthelikelihoodofoccurrenceofasinglefutureresponse,butourdataareintheformoffrequenciesofresponseswhichhavealreadyoccurred.These responses were presumably similar to each other and to the response to bepredicted.But this raises the troublesomeproblemof response-instancevs. response-class. Precisely what responses are we to take into account in predicting a futureinstance?Certainlynottheresponsesmadebyapopulationofdifferentorganisms,forsuchastatisticaldatumraisesmoreproblemsthanitsolves.Toconsiderthefrequencyof repeated responses in an individual demands something like the experimentalsituationjustdescribed.Thissolutionof theproblemofabasicdatumisbasedupon theview thatoperant

behavior isessentiallyanemissivephenomenon.Latencyandmagnitudeof responsefail as measures because they do not take this into account. They are conceptsappropriatetothefieldofthereflex,wheretheallbutinvariablecontrolexercisedbytheelicitingstimulusmakesthenotionofprobabilityofresponsetrivial.Consider,forexample, the case of latency. Because of our acquaintance with simple reflexes weinferthataresponsewhichismorelikelytobeemittedwillbeemittedmorequickly.Butisthistrue?Whatcanthewordquicklymean?Probabilityofresponse,aswellaspredictionof response, is concernedwith themomentofemission.This is apoint intime,butitdoesnothavethetemporaldimensionofalatency.Theexecutionmaytaketime after the response has been initiated, but the moment of occurrence has noduration.2 In recognizing the emissive character of operant behavior and the centralpositionofprobabilityof responseasadatum, latency is seen tobe irrelevant toourpresenttask.Variousobjectionshavebeenmadetotheuseofrateofrespondingasabasicdatum.

Forexample,suchaprogrammayseemtobarusfromdealingwithmanyeventswhichare unique occurrences in the life of the individual. Aman does not decide upon acareer,getmarried,makeamilliondollars,orgetkilledinanaccidentoftenenoughtomakearateofresponsemeaningful.Buttheseactivitiesarenotresponses.Theyarenotsimple unitary events lending themselves to prediction as such. If we are to predictmarriage, success, accidents, and so on, in anythingmore than statistical terms, wemustdealwiththesmallerunitsofbehaviorwhichleadtoandcomposetheseunitaryepisodes.Iftheunitsappearinrepeatableform,thepresentanalysismaybeapplied.Inthe field of learning a similar objection takes the form of asking how the presentanalysis may be extended to experimental situations in which it is impossible toobserve frequencies. It does not follow that learning is not taking place in suchsituations. The notion of probability is usually extrapolated to cases in which afrequencyanalysiscannotbecarriedout.Inthefieldofbehaviorwearrangeasituationin which frequencies are available as data, but we use the notion of probability inanalyzing and formulating instances or even types of behavior which are notsusceptibletothisanalysis.Another commonobjection is that a rate of response is just a set of latencies and

84

hencenotanewdatumatall.Thisiseasilyshowntobewrong.Whenwemeasurethetimeelapsingbetweentworesponses,weareinnodoubtastowhattheorganismwasdoingwhenwestartedourclock.Weknowthatitwasjustexecutingaresponse.Thisisanaturalzero—quiteunlikethearbitrarypointfromwhichlatenciesaremeasured.The free repetition of a response yields a rhythmic or periodic datum very differentfromlatency.Manyperiodicphysicalprocessessuggestparallels.Wedonotchooserateofrespondingasabasicdatummerelyfromananalysisofthe

fundamentaltaskofascienceofbehavior.Theultimateappeal is to itssuccess inanexperimentalscience.Thematerialwhichfollowsisofferedasasampleofwhatcanbedone.Itisnotintendedasacompletedemonstration,butitshouldconfirmthefactthatwhenweareinpossessionofadatumwhichvariesinasignificantfashion,wearelesslikelytoresorttotheoreticalentitiescarryingthenotionofprobabilityofresponse.

WhyLearningOccurs

We may define learning as a change in probability of response, but we must alsospecifytheconditionsunderwhichitcomesabout.Todothiswemustsurveysomeoftheindependentvariablesofwhichprobabilityofresponseisafunction.Herewemeetanotherkindoflearningtheory.AneffectiveclassroomdemonstrationoftheLawofEffectmaybearrangedinthe

followingway.Apigeon,reducedto80percentofitsadlibweight,ishabituatedtoasmall,semicircularamphitheatreandisfedthereforseveraldaysfromafoodhopper,whichtheexperimenterpresentsbyclosingahandswitch.Thedemonstrationconsistsofestablishingaselectedresponsebysuitablereinforcementwithfood.Forexample,by sighting across the amphitheatre at a scale on the oppositewall, it is possible topresent the hopperwhenever the topof the pigeon’s head rises above a givenmark.Higherandhighermarksarechosenuntil,withinafewminutes,thepigeoniswalkingaboutthecagewithitsheadheldashighaspossible.Inanotherdemonstrationthebirdisconditionedtostrikeamarbleplacedontheflooroftheamphitheatre.Thismaybedoneinafewminutesbyreinforcingsuccessivesteps.Foodispresentedfirstwhenthebirdismerelymovingnearthemarble,laterwhenitlooksdowninthedirectionofthemarble,laterstillwhenitmovesitsheadtowardthemarble,andfinallywhenitpecksit. Anyone who has seen such a demonstration knows that the Law of Effect is notheory.Itsimplyspecifiesaprocedureforalteringtheprobabilityofachosenresponse.Butwhenwetrytosaywhyreinforcementhasthiseffect,theoriesarise.Learningis

said to takeplacebecause the reinforcement ispleasant, satisfying, tension reducing,andsoon.Theconverseprocessofextinctionisexplainedwithcomparabletheories.Iftherateofrespondingisfirstraisedtoahighpointbyreinforcementandreinforcementthenwithheld,theresponseisobservedtooccurlessandlessfrequentlythereafter.Onecommontheoryexplainsthisbyassertingthatastateisbuiltupwhichsuppressesthebehavior.This“experimentalinhibition”or“reactioninhibition”mustbeassignedtoadifferent dimensional system, since nothing at the level of behavior corresponds toopposedprocessesofexcitationandinhibition.Rateofrespondingissimplyincreasedbyone operation and decreased by another.Certain effects commonly interpreted asshowing release from a suppressing force may be interpreted in other ways.Disinhibition,forexample,isnotnecessarilytheuncoveringofsuppressedstrength:itmaybeasignofsupplementarystrengthfromanextraneousvariable.Theprocessof

85

spontaneous recovery, often cited to support the notion of suppression, has analternativeexplanation,tobenotedinamoment.Letusevaluate thequestionofwhy learning takesplaceby turningagain tosome

data. Since conditioning is usually too rapid to be easily followed, the process ofextinction will provide us with a more useful case. A number of different types ofcurveshavebeenconsistentlyobtainedfromratsandpigeonsusingvariousschedulesof prior reinforcement. By considering some of the relevant conditionswemay seewhatroomisleftfortheoreticalprocesses.Themerepassageof timebetweenconditioningandextinction isavariablewhich

hassurprisinglylittleeffect.Theratistooshort-livedtomakeanextendedexperimentfeasible,butthepigeon,whichmaylivetenorfifteenyears,isanidealsubject.Morethanfiveyearsago,twentypigeonswereconditionedtostrikealargetranslucentkeyupon which a complex visual pattern was projected. Reinforcement was contingentupon the maintenance of a high and steady rate of responding and upon striking aparticular feature of the visual pattern. These birds were set aside in order to studyretention. They were transferred to the usual living quarters, where they served asbreeders.Smallgroupswere tested forextinctionat theendofsixmonths,oneyear,twoyears,andfouryears.Beforethetesteachbirdwastransferredtoaseparatelivingcage.Acontrolledfeedingschedulewasusedtoreducetheweighttoapproximately80percentoftheadlibweight.Thebirdwasthenfedinthedimlylightedexperimentalapparatusintheabsenceofthekeyforseveraldays,duringwhichemotionalresponsestotheapparatusdisappeared.Onthedayofthetestthebirdwasplacedinthedarkenedbox.Thetranslucentkeywaspresentbutnotlighted.Noresponsesweremade.Whenthe pattern was projected upon the key, all four birds responded quickly andextensively.Figure2showsthelargestcurveobtained.Thisbirdstruckthekeywithintwosecondsafterpresentationofavisualpatternwhichithadnotseenforfouryears,and at the precise spot upon which differential reinforcement had previously beenbased.Itcontinuedtorespondforthenexthour,emittingabout700responses.Thisisof the order of one-half to one-quarter of the responses it would have emitted ifextinctionhadnotbeendelayedfouryears,butotherwisethecurveisfairlytypical.

FIG.2

Levelofmotivationisanothervariabletobetakenintoaccount.Anexampleoftheeffectofhungerhasbeen reportedelsewhere.3The responseof pressing a leverwas

86

establishedineightratswithascheduleofperiodicreinforcement.Theywerefedthemainpartoftheirrationonalternatedayssothattheratesofrespondingonsuccessivedayswerealternatelyhighandlow.Twosubgroupsoffourratseachwerematchedonthebasisof the ratemaintainedunderperiodic reinforcementunder theseconditions.Theresponsewasthenextinguished—inonegrouponalternatedayswhenthehungerwashigh, in theothergrouponalternatedayswhen thehungerwas low. (Thesameamountoffoodwaseatenonthenonexperimentaldaysasbefore.)TheresultisshowninFigure3.Theuppergraphgivestherawdata.ThelevelsofhungerareindicatedbythepointsatPontheabscissa,theratesprevailingunderperiodicreinforcement.Thesubsequent points show the decline in extinction. If we multiply the lower curvethroughbya factorchosen tosuperimpose thepointsatP, thecurvesare reasonablycloselysuperimposed,asshowninthelowergraph.Severalotherexperimentsonbothratsandpigeonshaveconfirmedthisgeneralprinciple. Ifagivenratioofrespondingprevails under periodic reinforcement, the slopes of later extinction curves show thesame ratio. Level of hunger determines the slope of the extinction curve but not itscurvature.Another variable, difficulty of response, is especially relevant because it hasbeen

used to test the theory of reaction inhibition,4 on the assumption that a responserequiring considerable energy will build up more reaction inhibition than an easyresponseandlead,therefore,tofasterextinction.Thetheoryrequiresthatthecurvatureof the extinction curve be altered, not merely its slope. Yet there is evidence thatdifficultyof responseacts like levelofhunger simply toalter the slope.Apigeon issuspendedina jacketwhichconfines itswingsand legsbut leaves itsheadandneckfree to respond to a key and a food magazine.5 Its behavior in this situation isquantitativelymuch like thatofabirdmovingfreely inanexperimentalbox,but theuseofthejackethastheadvantagethattheresponsetothekeymaybemadeeasyordifficult by changing the distance the bird must reach. In one experiment thesedistanceswereexpressedinsevenequalbutarbitraryunits.Atdistance7thebirdcouldbarely reach the key, at 3 it could strike without appreciably extending its neck.Periodicreinforcementgaveastraightbaselineuponwhichitwaspossibletoobservethe effect of difficulty by quickly changing position during the experimental period.EachofthefiverecordsinFigure4coversafifteen-minuteexperimentalperiodunderperiodic reinforcement.Distancesof thebird fromthekeyare indicatedbynumeralsabove the records. It will be observed that the rate of responding at distance 7 isgenerally quite lowwhile that at distance 3 is high. Intermediate distances produceintermediate slopes. It should also be noted that the change from one position toanotherisfeltimmediately.Ifrepeatedrespondinginadifficultpositionweretobuildaconsiderable amount of reaction inhibition, we should expect the rate to be low forsomelittletimeafterreturningtoaneasyresponse.Contrariwise,ifaneasyresponseweretobuildlittlereactioninhibition,weshouldexpectafairlyhighrateofrespondingforsometimeafteradifficultpositionisassumed.Nothinglikethisoccurs.The“morerapidextinction”ofadifficultresponseisanambiguousexpression.Theslopeconstantis affectedandwith it thenumberof responses in extinction to a criterion,but theremaybenoeffectuponcurvature.

87

88

FIG.3

Onewayofconsideringthequestionofwhyextinctioncurvesarecurvedistoregardextinctionasaprocessofexhaustioncomparabletothelossofheatfromsourcetosinkorthefallinthelevelofareservoirwhenanoutletisopened.Conditioningbuildsupapredisposition torespond—a“reserve”—whichextinctionexhausts.This isperhapsadefensibledescriptionatthelevelofbehavior.Thereserveisnotnecessarilyatheoryinthepresentsense,sinceitisnotassignedtoadifferentdimensionalsystem.Itcouldbeoperationally defined as a predicted extinction curve, even though, linguistically, itmakes a statement about the momentary condition of a response. But it is not aparticularlyusefulconcept,nordoestheviewthatextinctionisaprocessofexhaustionaddmuchtotheobservedfactthatextinctioncurvesarecurvedinacertainway.

89

FIG.4

Thereare,however,twovariableswhichaffecttherate,bothofwhichoperateduringextinctiontoalterthecurvature.Oneofthesefallswithinthefieldofemotion.Whenwe fail to reinforce a response which has previously been reinforced, we not onlyinitiateaprocessofextinction,wesetupanemotionalresponse—perhapswhatisoftenmeantbyfrustration.Thepigeoncoos inanidentifiablepattern,movesrapidlyaboutthe cage, defecates, or flaps itswings rapidly in a squatting positionwhich suggests

90

treading (mating)behavior.Thiscompeteswith the responseof strikingakeyand isperhapsenoughtoaccountforthedeclineinrateinearlyextinction.Itisalsopossiblethat theprobabilityof a responsebasedupon fooddeprivation isdirectly reducedaspartof suchan emotional reaction.Whatever itsnature, the effectof thisvariable iseliminated through adaptation. Repeated extinction curves become smoother, and insome of the schedules to be described shortly there is little or no evidence of anemotionalmodificationofrate.A second variable has a much more serious effect. Maximal responding during

extinction is obtained only when the conditions under which the response wasreinforcedarepreciselyreproduced.Aratconditionedinthepresenceofalightwillnotextinguishfullyintheabsenceofthelight.Itwillbegintorespondmorerapidlywhenthe light is again introduced.This is true forotherkindsof stimuli, as the followingclassroom experiment illustrates. Nine pigeons were conditioned to strike a yellowtriangle under intermittent reinforcement. In the session represented by Figure5 thebirdswerefirstreinforcedonthisschedulefor30minutes.Thecombinedcumulativecurveisessentiallyastraightline,showingmorethan1100responsesperbirdduringthisperiod.Aredtrianglewasthensubstitutedfor theyellowandno responseswerereinforced thereafter. The effect was a sharp drop in responding, with only a slightrecoveryduringthenextfifteenminutes.Whentheyellowtrianglewasreplaced,rapidresponding began immediately, and the usual extinction curve followed. Similarexperiments have shown that the pitch of an incidental tone, the shape of a patternbeingstruck,orthesizeofapattern,ifpresentduringconditioning,willtosomeextentcontrol the rate of responding during extinction. Some properties aremore effectivethanothers,andaquantitativeevaluationispossible.Bychangingtoseveralvaluesofa stimulus in randomorder repeatedlyduring the extinctionprocess, thegradient forstimulus generalization may be read directly in the rates of responding under eachvalue.

91

FIG.5

Somethingverymuchlikethismustgoonduringextinction.Letussupposethatall

92

responses to a keyhavebeen reinforced and that eachhas been followedby a shortperiod of eating. When we extinguish the behavior, we create a situation in whichresponses are not reinforced, inwhich no eating takes place, and inwhich there areprobably new emotional responses. The situation could easily be as novel as a redtriangleafterayellow.Ifso,itcouldexplainthedeclineinrateduringextinction.Wemight have obtained a smooth curve, shaped like an extinction curve, between theverticallinesinFigure5bygraduallychangingthecolorofthetrianglefromyellowtored. Thismight have happened even though no other sort of extinctionwere takingplace.Theveryconditionsofextinctionseemtopresupposeagrowingnoveltyintheexperimentalsituation.Isthiswhytheextinctioncurveiscurved?Some evidence comes from the data of “spontaneous recovery.” Even after

prolongedextinctionanorganismwilloftenrespondatahigherrateforatleastafewmoments at the beginning of another session. One theory contends that this showsspontaneousrecoveryfromsomesortofinhibition,butanotherexplanationispossible.No matter how carefully an animal is handled, the stimulation coincident with thebeginning of an experimentmust be extensive and unlike anything occurring in thelaterpartofanexperimentalperiod.Responseshavebeenreinforcedinthepresenceof,or shortly following, this stimulation. In extinction it is present for only a fewmoments. When the organism is again placed in the experimental situation thestimulation is restored; further responses are emitted as in the case of the yellowtriangle.Theonlywaytoachievefullextinctioninthepresenceofthestimulationofstartinganexperimentistostarttheexperimentrepeatedly.Other evidence of the effect of novelty comes from the study of periodic

reinforcement. The fact that intermittent reinforcement produces bigger extinctioncurvesthancontinuousreinforcementisatroublesomedifficultyforthosewhoexpecta simple relation between number of reinforcements and number of responses inextinction. But this relation is actually quite complex. One result of periodicreinforcement is that emotional changes adapt out. This may be responsible for thesmoothnessof subsequentextinctioncurvesbutprobablynot for theirgreaterextent.The lattermaybe attributed to the lackof novelty in the extinction situation.Underperiodicreinforcementmanyresponsesaremadewithoutreinforcementandwhennoeating has recently taken place. The situation in extinction is therefore not whollynovel.

93

FIG.6

Periodic reinforcement is not, however, a simple solution. If we reinforce on aregularschedule—say,everyminute—theorganismsoonformsadiscrimination.Littleor no responding occurs just after reinforcement, since stimulation from eating iscorrelatedwithabsenceof subsequent reinforcement.Howrapidly thediscriminationmaydevelopisshowninFigure6,whichreproducesthefirstfivecurvesobtainedfromapigeonunderperiodicreinforcementinexperimentalperiodsoffifteenminuteseach.Inthefifthperiod(orafteraboutonehourofperiodicreinforcement)thediscriminationyieldsapauseaftereachreinforcement, resulting inamarkedlystepwisecurve.Asaresult of this discrimination the bird is almost always responding rapidly whenreinforced.This is thebasis for anotherdiscrimination.Rapid respondingbecomes afavorable stimulating condition. A good example of the effect upon the subsequentextinctioncurve is shown in Figure7. This pigeon had been reinforced once everyminuteduringdailyexperimentalperiodsoffifteenminuteseachforseveralweeks.Intheextinctioncurveshown,thebirdbeginstorespondattherateprevailingundertheprecedingschedule.Aquickpositiveaccelerationatthestartislostinthereductionofthe record. The pigeon quickly reaches and sustains a ratewhich is higher than theover-all rate during periodic reinforcement. During this period the pigeon creates astimulatingconditionpreviouslyoptimallycorrelatedwith reinforcement.Eventually,as some sort of exhaustion intervenes, the rate falls off rapidly to amuch lowerbutfairlystablevalueandthentopracticallyzero.Aconditionthenprevailsunderwhicharesponseisnotnormallyreinforced.Thebirdisthereforenotlikelytobegintorespondagain. When it does respond, however, the situation is slightly improved and, if itcontinues to respond, the conditions rapidly become similar to those under whichreinforcement has been received. Under this “autocatalysis” a high rate is quicklyreached, andmore than 500 responses are emitted in a second burst. The rate thendeclines quickly and fairly smoothly, again to nearly zero.This curve is not by anymeansdisorderly.Mostofthecurvatureissmooth.Buttheburstofrespondingatforty-

94

fiveminutes showsaconsiderable residual strengthwhich, if extinctionweremerelyexhaustion, should have appeared earlier in the curve.The curvemaybe reasonablyaccountedforbyassumingthatthebirdislargelycontrolledbytheprecedingspuriouscorrelationbetweenreinforcementandrapidresponding.

FIG.7

This assumption may be checked by constructing a schedule of reinforcement inwhich a differential contingency between rate of responding and reinforcement isimpossible.Inonesuchscheduleofwhatmaybecalled“aperiodicreinforcement”oneinterval between successive reinforced responses is so short that no unreinforcedresponses intervene,while the longest interval isabout twominutes.6Other intervalsaredistributedarithmeticallybetweenthesevalues,theaverageremainingoneminute.Theintervalsareroughlyrandomizedtocomposeaprogramofreinforcement.Underthisprogramtheprobabilityofreinforcementdoesnotchangewithrespecttopreviousreinforcements, and the curves never acquire the stepwise character of curve E inFigure6. (Figure9 showscurves froma similarprogram.)Asa resultnocorrelationbetweendifferent ratesof respondinganddifferentprobabilitiesof reinforcementcandevelop.Anextinctioncurvefollowingabriefexposuretoaperiodicreinforcementisshown

in Figure 8. It begins characteristically at the rate prevailing under aperiodicreinforcementand,unlikethecurvefollowingregularperiodicreinforcement,doesnotaccelerate to a higher over-all rate. There is no evidence of the “autocatalytic”production of an optimal stimulating condition. Also characteristically, there are nosignificant discontinuities or sudden changes in rate in either direction. The curveextendsoveraperiodofeighthours,asagainstnotquite twohours inFigure7, andseems to representa singleorderlyprocess.The totalnumberof responses ishigher,perhapsbecauseofthegreatertimeallowedforemission.Allofthiscanbeexplainedby the single fact thatwe havemade it impossible for the pigeon to form a pair ofdiscriminations based, first, upon stimulation from eating and, second, uponstimulationfromrapidresponding.

95

FIG.8

Since the longest interval between reinforcementwas only twominutes, a certainnovelty must still have been introduced as time passed. Whether this explains thecurvature in Figure 8 may be tested to some extent with other programs ofreinforcement containing much longer intervals. A geometric progression wasconstructed by beginning with 10 seconds as the shortest interval and repeatedlymultiplyingby1.54.Thisyieldedasetofintervalsaveraging5minutes,thelongestofwhich was more than 21 minutes. Such a set was randomized in a program ofreinforcement repeated every hour. In changing to this program from the arithmeticseries, the rates first declinedduring the longer intervals, but the pigeonswere soonabletosustainaconstantrateofrespondingunderit.TworecordsintheforminwhichtheywererecordedareshowninFigure9.(Thepenresetstozeroaftereverythousandresponses.Inordertoobtainasinglecumulativecurveitwouldbenecessarytocuttherecordandpiecethesectionstogethertoyieldacontinuousline.Therawformmaybereproduced with less reduction.) Each reinforcement is represented by a horizontaldash.Thetimecoveredisaboutthreehours.Recordsareshownfortwopigeonswhichmaintaineddifferentover-allratesunderthisprogramofreinforcement.

96

FIG.9

Under such a schedule a constant rate of responding is sustained for at least 21minutes without reinforcement, after which a reinforcement is ceived. Less noveltyshould therefore develop during succeeding extinction. In Curve 1 of Figure10 thepigeonhadbeenexposedtoseveralsessionsofseveralhourseachwiththisgeometricsetofintervals.Thenumberofresponsesemittedinextinctionisabouttwicethatofthecurve inFigure8 after the arithmetic set of intervals averaging oneminute, but thecurvesareotherwisemuchalike.Furtherexposuretothegeometricschedulebuildsuplonger runs during which the rate does not change significantly. Curve 2 followedCurve1 after twoandone-halfhoursof further aperiodic reinforcement.On thedayshown inCurve2 a fewaperiodic reinforcementswere first given, asmarked at thebeginningofthecurve.Whenreinforcementwasdiscontinued,afairlyconstantrateofresponding prevailed for several thousand responses. After another experimentalsession of two and one-half hourswith the geometric series, Curve 3was recorded.Thissessionalsobeganwithashortseriesofaperiodicreinforcements,followedbyasustainedrunofmorethan6000unreinforcedresponseswithlittlechangeinrate(A).There seems to be no reason why other series averaging perhaps more than fiveminutesperintervalandcontainingmuchlongerexceptionalintervalswouldnotcarrysuchastraightlinemuchfurther.

97

FIG.10

Inthisattackupontheproblemofextinctionwecreateascheduleofreinforcementwhich is so much like the conditions which will prevail during extinction that nodeclineinratetakesplaceforalongtime.Inotherwordswegenerateextinctionwithno curvature. Eventually some kind of exhaustion sets in, but it is not approachedgradually. The last part of Curve 3 (unfortunatelymuch reduced in the figure)maypossiblysuggestexhaustionintheslightover-allcurvature,butitisasmallpartofthewhole process. The record is composedmainly of runs of a few hundred responseseach,mostof thematapproximately thesamerateas thatmaintainedunderperiodicreinforcement.The pigeon stops abruptly;when it starts to respond again, it quicklyreachestherateofrespondingunderwhichitwasreinforced.Thisrecallsthespuriouscorrelationbetween rapid respondingand reinforcementunder regular reinforcement.Wehavenot, of course, entirely eliminated this correlation.Even though there is nolonger a differential reinforcement of high against low rates, practically allreinforcementshaveoccurredunderaconstantrateofresponding.Further study of reinforcing schedules may or may not answer the question of

whetherthenoveltyappearingintheextinctionsituationisentirelyresponsibleforthecurvature. Itwould appear to be necessary tomake the conditions prevailing duringextinction identical with the conditions prevailing during conditioning. Thismay beimpossible,butinthatcasethequestionisacademic.Thehypothesis,meanwhile,isnot

98

atheoryinthepresentsense,sinceitmakesnostatementsaboutaparallelprocessinanyotheruniverseofdiscourse.It is true that it appeals to stimulation generated in part by the pigeon’s own

behavior.Thismaybedifficulttospecifyormanipulate,butitisnottheoreticalinthepresent sense. So long as we are willing to assume a one-to-one correspondencebetweenactionandstimulation,aphysicalspecificationispossible.The study of extinction after different schedules of aperiodic reinforcement is not

addressedwholly to this hypothesis. The object is an economical description of theconditionsprevailingduringreinforcementandextinctionandoftherelationsbetweenthem.Inusingrateofrespondingasabasicdatumwemayappealtoconditionswhichare observable andmanipulable and wemay express the relations between them inobjectiveterms.Totheextentthatourdatummakesthispossible,itreducestheneedfor theory. When we observe a pigeon emitting 7000 responses at a constant ratewithout reinforcement, we are not likely to explain an extinction curve containingperhapsafewhundredresponsesbyappealtothepilingupofreactioninhibitionoranyother fatigueproduct.Researchwhich isconductedwithoutcommitment to theory ismore likely to carry the study of extinction into new areas and new orders ofmagnitude.Byhasteningtheaccumulationofdata,wespeedthedepartureoftheories.If the theorieshaveplayednopart in thedesignofourexperiments,weneednotbesorrytoseethemgo.

ComplexLearning

A third type of learning theory is illustrated by terms like preferring, choosing,discriminating,andmatching.Aneffortmaybemadetodefinethesesolelyintermsofbehavior, but in traditional practice they refer to processes in another dimensionalsystem. A response to one of two available stimuli may be called choice, but it iscommonertosaythatitistheresultofchoice,meaningbythelatteratheoreticalpre-behavioralactivity.Thehighermentalprocessesare thebestexamplesof theoriesofthissort;neurologicalparallelshavenotbeenwellworkedout.Theappealtotheoryisencouragedbythefactthatchoosing(likediscriminating,matching,andsoon)isnotaparticularpieceofbehavior. It isnota responseoranactwithspecified topography.The term characterizes a larger segment of behavior in relation to other variables orevents.Canweformulateandstudythebehaviortowhichthesetermswouldusuallybeappliedwithoutrecoursetothetheorieswhichgenerallyaccompanythem?Discrimination is a relatively simple case.Supposewe find that theprobabilityof

emissionofagivenresponseisnotsignificantlyaffectedbychangingfromoneoftwostimulitotheother.Wethenmakereinforcementoftheresponsecontingentuponthepresenceofoneofthem.Thewell-establishedresultisthattheprobabilityofresponseremainshighunderthisstimulusandreachesaverylowpointundertheother.Wesaythat the organism now discriminates between the stimuli. But discrimination is notitself an action, or necessarily even a unique process. Problems in the field ofdiscrimination may be stated in other terms. How much induction obtains betweenstimuliofdifferentmagnitudesorclasses?Whatarethesmallestdifferencesinstimuliwhich yield a difference in control? And so on. Questions of this sort do notpresupposetheoreticalactivitiesinotherdimensionalsystems.A somewhat larger segment must be specified in dealing with the behavior of

99

choosing one of two concurrent stimuli. This has been studied in the pigeon byexaminingresponsestotwokeysdifferinginposition(rightorleft)orinsomepropertylikecolorrandomizedwithrespecttoposition.Byoccasionallyreinforcingaresponseon one key or the other without favoring either key, we obtain equal rates ofrespondingon the twokeys.The behavior approaches a simple alternation from onekeytotheother.Thisfollowstherulethattendenciestorespondeventuallycorrespondtotheprobabilitiesofreinforcement.Givenasysteminwhichonekeyortheotherisoccasionallyconnectedwith themagazinebyanexternalclock, then if the rightkeyhasjustbeenstruck,theprobabilityofreinforcementviatheleftkeyishigherthanthatvia the rightsinceagreater intervalof timehaselapsedduringwhich theclockmayhaveclosedthecircuittotheleftkey.Butthebird’sbehaviordoesnotcorrespondtothis probabilitymerely out of respect formathematics. The specific result of such acontingency of reinforcement is that changing-to-the-other-key-and-striking is moreoften reinforced than striking-the-same-key-a-second-time.We are no longer dealingwith just tworesponses. Inorder toanalyze“choice”wemustconsiderasinglefinalresponse,striking,withoutrespect to thepositionorcolorof thekey,andinadditiontheresponsesofchangingfromonekeyorcolortotheother.Quantitative results are compatiblewith this analysis. Ifwe periodically reinforce

responsestotherightkeyonly,therateofrespondingontherightwillrisewhilethaton the left will fall. The response of changing-from-right-to-left is never reinforcedwhiletheresponseofchanging-from-left-to-rightisoccasionallyso.Whenthebirdisstrikingontheright,thereisnogreattendencytochangekeys;whenitisstrikingontheleft,thereisastrongtendencytochange.Manymoreresponsescometobemadetotherightkey.Theneedforconsideringthebehaviorofchangingoverisclearlyshownifwenow reverse these conditions and reinforce responses to the left keyonly.Theultimateresultisahighrateofrespondingontheleftkeyandalowrateontheright.Byreversingtheconditionsagainthehighratecanbeshiftedbacktotherightkey.InFigure11agroupofeightcurveshavebeenaveragedtofollowthischangeduringsixexperimental periods of 45minutes each.Beginningon the secondday in the graphresponses to the right key (RR) decline in extinctionwhile responses to the left key(RL) increase through periodic reinforcement. The mean rate shows no significantvariation, sinceperiodic reinforcement iscontinuedon the sameschedule.Themeanrate shows the condition of strength of the response of striking a key regardless ofposition.Thedistributionofresponsesbetweenrightandleftdependsupontherelativestrengthoftheresponsesofchangingover.Ifthisweresimplyacaseoftheextinctionofone responseand theconcurrent reconditioningofanother, themeancurvewouldnot remain approximately horizontal since reconditioning occursmuchmore rapidlythan extinction. (Two topographically independent responses, capable of emission atthe same time and hence not requiring change-over, show separate processes ofreconditioningandextinction,andthecombinedrateofrespondingvaries.)

100

101

FIG.11

The ratewithwhich thebirdchanges fromonekey to theotherdependsupon thedistancebetweenthekeys.Thisdistanceisaroughmeasureofthestimulus-differencebetweenthe twokeys.Italsodetermines thescopeof theresponseofchanging-over,with an implied difference in sensory feedback. It also modifies the spread ofreinforcement to responses supposedly not reinforced, since if the keys are closetogether, a response reinforced on one side may occur sooner after a precedingresponseontheotherside.InFigure11thetwokeyswereaboutoneinchapart.Theywerethereforefairlysimilarwithrespecttopositionintheexperimentalbox.Changingfromonetotheotherinvolvedaminimumofsensoryfeedback,andreinforcementofaresponsetoonekeycouldfollowveryshortlyuponaresponsetotheother.Whenthekeys are separated by asmuch as four inches, the change in strength ismuchmorerapid. Figure 12 shows two curves recorded simultaneously from a single pigeonduringoneexperimentalperiodofabout40minutes.Ahighratetotherightkeyandalowrate to the lefthadpreviouslybeenestablished. In thefigurenoresponses to therightwerereinforced,butthosetotheleftwerereinforcedeveryminuteasindicatedbytheverticaldashesabovecurveL.TheslopeofRdeclinesinafairlysmoothfashionwhilethatofLincreases,alsofairlysmoothly,toavaluecomparabletotheinitialvalueofR.Thebirdhasconformedtothechangedcontingencywithinasingleexperimentalperiod.Themeanrateofrespondingisshownbyadottedline,whichagainshowsnosignificantcurvature.

102

FIG.12

Whatiscalled“preference”entersintothisformulation.Atanystageoftheprocessshown in Figure12 preference might be expressed in terms of the relative rates ofresponding to the twokeys.Thispreference,however, isnot in strikingakeybut inchanging from one key to the other. The probability that the bird will strike a keyregardless of its identifying properties behaves independently of the preferentialresponseofchangingfromonekeytotheother.Severalexperimentshaverevealedanadditionalfact.Apreferenceremainsfixedifreinforcementiswithheld.Figure13isanexample. It shows simultaneous extinction curves from two keys during seven dailyexperimentalperiodsofonehoureach.Prior toextinctiontherelative strengthof theresponsesofchanging-to-Randchanging-to-Lyieldeda“preference”ofabout3 to1forR.Theconstancyoftheratethroughouttheprocessofextinctionhasbeenshowninthe figurebymultiplyingL throughbyasuitableconstantandentering thepointsassmall circles onR. If extinction altered the preference, the two curves could not besuperimposedinthisway.

103

FIG.13

These formulations of discrimination and choosing enable us to dealwithwhat isgenerallyregardedasamuchmorecomplexprocess—matchingtosample.Supposewearrange three translucent keys, each ofwhichmay be illuminatedwith red or greenlight.Themiddlekey functionsas the sample, andwecolor it either redorgreen inrandomorder.Wecolorthetwosidekeysoneredandonegreen,alsoinrandomorder.The“problem”istostrikethesidekeywhichcorrespondsincolortothemiddlekey.Thereareonlyfour three-keypatterns insuchacase,and it ispossible thatapigeoncouldlearntomakeanappropriateresponsetoeachpattern.Thisdoesnothappen,atleastwithinthetemporalspanoftheexperimentstodate.Ifwesimplypresentaseriesofsettingsofthethreecolorsandreinforcesuccessfulresponses,thepigeonwillstrikethe side keyswithout respect to color or pattern andbe reinforced50percent of thetime.Thisis,ineffect,ascheduleof“fixedratio”reinforcementwhichisadequatetomaintainahighrateofresponding.Nevertheless it is possible to get a pigeon tomatch to sample by reinforcing the

discriminative responses of striking-red-after-being-stimulated-by-red and striking-green-after-being-stimulated-by-greenwhile extinguishing the other twopossibilities.Thedifficulty is in arranging theproper stimulation at the timeof the response.Thesamplemightbemadeconspicuous—forexample,byhaving thesamplecolor in thegeneralilluminationoftheexperimentalbox.Insuchacasethepigeonwouldlearntostrike red keys in a red light and green keys in a green light (assuming a neutralilluminationofthebackgroundofthekeys).Butaprocedurewhichholdsmorecloselytothenotionofmatchingistoinducethepigeonto“lookatthesample”bymeansofaseparatereinforcement.Wemaydothisbypresentingthecoloronthemiddlekeyfirst,leaving the side keys uncolored. A response to the middle key is then reinforced

104

(secondarily)byilluminatingthesidekeys.Thepigeonlearnstomaketworesponsesinquicksuccession—tothemiddlekeyandthentoonesidekey.Theresponsetothesidekey follows quickly upon the visual stimulation from the middle key, which is therequisiteconditionforadiscrimination.Successfulmatchingwasreadilyestablishedinalltenpigeonstestedwiththistechnique.Choosingtheoppositeisalsoeasilysetup.Thediscriminativeresponseofstriking-red-after-being-stimulated-by-redisapparentlyno easier to establish than striking-red-after-being-stimulated-by-green. When theresponseis toakeyofthesamecolor,however,generalizationmaymakeitpossibleforthebirdtomatchanewcolor.Thisisanextensionofthenotionofmatchingwhichhasnotyetbeenstudiedwiththismethod.Evenwhenmatchingbehaviorhasbeenwellestablished, thebirdwillnot respond

correctlyifallthreekeysarenowpresentedatthesametime.Thebirddoesnotpossessstrongbehaviorof lookingat thesample.Theexperimentermustmaintainaseparatereinforcementtokeepthisbehaviorinstrength.Inmonkeys,apes,andhumansubjectstheultimatesuccessinchoosingisapparentlysufficienttoreinforceandmaintainthebehavioroflookingatthesample.Itispossiblethatthisspeciesdifferenceissimplyadifferenceinthetemporalrelationsrequiredforreinforcement.Thebehaviorofmatchingsurvivesunchangedwhenall reinforcement iswithheld.

An intermediatecasehasbeenestablished inwhich thecorrectmatching response isonlyperiodicallyreinforced.Inoneexperimentonecolorappearedonthemiddlekeyforoneminute; itwasthenchangedornotchanged,atrandom,totheothercolor.Aresponsetothiskeyilluminatedthesidekeys,oneredandonegreen,inrandomorder.Aresponsetoasidekeycutofftheilluminationtobothsidekeys,untilthemiddlekeyhadagainbeensrtuck.Theapparatusrecordedallmatchingresponsesononegraphandall non-matching on another. Pigeonswhich have acquiredmatching behavior undercontinuous reinforcement havemaintained this behavior when reinforced no oftenerthanonceperminuteontheaverage.Theymaymakethousandsofmatchingresponsesperhourwhilebeingreinforcedfornomorethansixtyofthem.Thisschedulewillnotnecessarilydevelopmatchingbehaviorinanaivebird,fortheproblemcanbesolvedinthreeways.Thebirdwillreceivepracticallyasmanyreinforcementsif itrespondsto(1)onlyonekeyor(2)onlyonecolor,sincetheprogrammingoftheexperimentmakesanypersistentresponseeventuallythecorrectone.

105

FIG.14

AsampleofthedataobtainedinacomplexexperimentofthissortisgiveninFigure14.Althoughthispigeonhadlearnedtomatchcolorundercontinuousreinforcement,itchanged to the spurious solution of a color preference under periodic reinforcement.

106

Whenever the sample was red, it struck both the sample and the red side key andreceived all reinforcements.When the samplewas green, it did not respond and theside keys were not illuminated. The result shown at the beginning of the graph inFigure14 is a high rate of responding on the upper graph, which recordsmatchingresponses.(Therecordisactuallystep-wise,followingthepresenceorabsenceoftheredsample,butthisislostinthereductioninthefigure.)Acolorpreference,however,is not a solution to the problem of opposites. By changing to this problem, it waspossible to change thebird’sbehavior as shownbetwen the twovertical lines in thefigure.Theuppercurvebetweenthese linesshowsthedeclineinmatchingresponseswhichhadresultedfromthecolorpreference.Thelowercurvebetweenthesamelinesshows the development of responding to and matching the opposite color. At thesecondverticallinethereinforcementwasagainmadecontingentuponmatching.Theupper curve shows the re-establishment ofmatching behaviorwhile the lower curveshowsadeclineinstrikingtheoppositecolor.Theresultwasatruesolution:thepigeonstruckthesample,nomatterwhatitscolor,andthenthecorrespondingsidekey.Thelighter line connects themeans of a series of points on the two curves. It seems tofollowthesameruleasinthecaseofchoosing:changesinthedistributionofresponsesbetweentwokeysdonot involve theover-all rateofresponding toakey.Thismeanrate will not remain constant under the spurious solution achieved with a colorpreference,asatthebeginningofthisfigure.These experiments on a few higher processes have necessarily been very briefly

described.Theyarenotofferedasprovingthattheoriesoflearningarenotnecessary,buttheymaysuggestanalternativeprograminthisdifficultarea.Thedatainthefieldofthehighermentalprocessestranscendsingleresponsesorsinglestimulus-responserelationships. But they appear to be susceptible to formulation in terms of thedifferentiationofconcurrentresponses,thediscriminationofstimuli,theestablishmentofvarioussequencesofresponses,andsoon.Thereseemstobenoapriorireasonwhya complete account is not possible without appeal to theoretical processes in otherdimensionalsystems.

Conclusion

Perhapstodowithouttheoriesaltogetherisatourdeforcewhichistoomuchtoexpectasageneralpractice.Theoriesarefun.Butit ispossiblethat themostrapidprogresstowardanunderstandingoflearningmaybemadebyresearchwhichisnotdesignedtotesttheories.Anadequateimpetusissuppliedbytheinclinationtoobtaindatashowingorderlychangescharacteristicofthelearningprocess.Anacceptablescientificprogramistocollectdataof thissortandtorelate themtomanipulablevariables,selectedforstudythroughacommon-senseexplorationofthefield.This does not exclude the possibility of theory in another sense. Beyond the

collectionofuniformrelationshipsliestheneedforaformalrepresentationofthedatareduced to aminimal number of terms.A theoretical constructionmay yield greatergenerality than any assemblage of facts. But such a construction will not refer toanotherdimensionalsystemandwillnot,therefore,fallwithinourpresentdefinition.Itwillnotstandinthewayofoursearchforfunctionalrelationsbecauseitwillariseonlyafter relevant variables have been found and studied. Though it may be difficult tounderstand, it will not be easily misunderstood, and it will have none of the

107

objectionableeffectsofthetheorieshereconsidered.Wedonotseemtobereadyfortheoryinthissense.Atthemomentwemakelittle

effective use of empirical, let alone rational, equations.A few of the present curvescould have been fairly closely fitted. But the most elementary preliminary researchshows that there are many relevant variables, and until their importance has beenexperimentally determined, an equation which allows for them will have so manyarbitraryconstants thatagoodfitwillbeamatterofcourseandcause forvery littlesatisfaction.

1Anexperimenton“differentialreinforcementwithrespecttotime”wasreportedatameetingoftheAmericanPsychologicalAssociation,September,1946.AnabstractappearsinTheAmericanPsychologist,1946,1,274–275.

2Itcannot,infact,beshortenedorlengthened.Wherealatencyappearstobeforcedtowardaminimalvaluebydifferentialreinforcement,anotherinterpretationiscalledfor.Althoughwemaydifferentiallyreinforcemoreenergeticbehaviororthefasterexecutionofbehaviorafteritbegins,itismeaninglesstospeakofdifferentiallyreinforcingresponseswithshortorlonglatencies.Whatweactuallyreinforcedifferentiallyare(a)favorablewaitingbehaviorand(b)morevigorousresponses.Whenweaskasubjecttorespond“assoonaspossible”inthehumanreaction-timeexperiment,weessentiallyaskhim(a)tocarryoutasmuchoftheresponseaspossiblewithoutactuallyreachingthecriterionofemission,(b)todoaslittleelseaspossible,and(c)torespondenergeticallyafterthestimulushasbeengiven.Thismayyieldaminimalmeasurabletimebetweenstimulusandresponse,butthistimeisnotnecessarilyabasicdatumnorhaveourinstructionsaltereditassuch.Aparallelinterpretationofthedifferentialreinforcementoflong“latencies”isrequired.Thisiseasilyestablishedbyinspection.Intheexperimentswithpigeonspreviouslycited,preliminarybehaviorisconditionedwhichpostponestheresponsetothekeyuntilthepropertime.Behaviorwhich“markstime”isusuallyconspicuous.

3TheexperimentfromwhichthefollowingdataaretakenwasreportedatameetingoftheAmericanPsychologicalAssociation,September1940.AnabstractappearsinthePsychologicalBulletin,1940,37,243.

4Mowrer,O.H.,&Jones,H.M.Extinctionandbehaviorvariabilityasfunctionsofeffortfulnessoftask.J.exp.Psychol.,1943,33,369–386.

5ThisexperimentwasreportedatameetingoftheMidwesternPsychologicalAssociation,March,1946.AnabstractappearsintheAmericanPsychologist,1946,1,462.

6Whatiscalled“periodicreinforcement”inthispaperhassincecometobeknownas“fixed-intervalreinforcement”and“aperiodic”as“variable-interval.”(SeeSchedulesofReinforcement.)

108

TheAnalysisofBehavior

ThisdiscussionoftheoreticalissuesintheanalysisofbehaviorformedthebeginningandendofalecturegivenonJuly17,1951,intheRiksdaghuset,Stockholm,Sweden,at the Thirteenth International Congress of Psychology. The rest of the lecture wasdevotedtoillustrativeexperimentsonintermittentreinforcementwhichhavesincebeenreportedinSchedulesofReinforcement.Similarmaterialmaybefoundbeginningonpage 136. The whole lecture was reprinted with adaptations in the AmericanPsychologist(1953,8,69-79)underthetitle,“SomeContributionsofanExperimentalAnalysisofBehaviortoPsychologyasaWhole.”

THEBEHAVIOR of an organism is not an easy thing to describe. It is not an objectwhichmaybeheldstillforinspection.Itisaprocess,acontinuouschange.Evenwhenan accurate account has been given of it as such, a different aspect remains to betreated. A science must achieve more than a description of behavior as anaccomplishedfact.Itmustpredictfuturecoursesofaction;itmustbeabletosaythatanorganismwillengageinbehaviorofagivensortatagiventime.Butthisraisesaspecialproblem.Wewanttobelievethatapredictionisinsomesenseadescriptionofaconditionatthemoment—beforethepredictedeventhastakenplace.Thus,wespeakof tendencies or readinesses to behave as if they corresponded to something in theorganism at the moment. We have given this something many names—from thepreparatorysetofexperimentalpsychologytotheFreudianwish.Habitsandinstincts,dispositions and predispositions, attitudes, opinions—even personality itself—are allwaysofrepresenting,inthepresentorganism,somethingofitsfuturebehavior.Thisproblemcannotbe avoided in any scientific account, but it canbe expressed

much more rigorously. We are dealing here with a question of probability—specifically, the probability that an organismwill emit behavior of a given sort at agiven time. But probability is always a difficult concept, nomatter inwhat field ofscienceitarises.Whatisaprobability?Whereisit?Howmayweobserveit?Wehavetriedtoanswerthesedifficultquestionsbygivingprobabilitythestatusofathing—byembodying it, so to speak,within theorganism.We look forneurologicalorpsychicstatesoreventswithwhichhabits,wishes,attitudes,andsoon,maybe identified. Indoingsowe force extraneousproperties onbehaviorwhich arenot supportedby thedataandwhichmaybequitemisleading.Thepracticalproblemoftakingprobabilityasabasicdatummaynotbeasdifficult

aswesuppose.The“physicalreferent”mustbeamongourdata;otherwisetheproblemwouldnothavebeensopersistent.Themistakeweseemtohavemadeisinlookingforitasnecessarilyapropertyofasingleevent,occupyingonlyonepointintime.Asthemathematicianshavepointedout,perhapsnotunanimously, aprobability is simplyawayofrepresentingafrequencyofoccurrence.Wecandealwithprobabilityofactionby turning our attention to the repeated appearance of an act during an appreciableintervaloftime.

109

Some such practice is demanded in defining any of the concepts which haveforeshadowedanexplicit recognitionofprobabilityofresponseasabasicdatum.Anorganismpossessesa“habit”totheextentthatacertainformofbehaviorisobservedwith a special frequency—in this case attributable to events in the history of theindividual. It possesses an “instinct” to the extent that a certain form of behavior isobserved with a special frequency—in this case because of membership in a givenspecies. An “attitude” expresses the special frequency of a number of forms ofbehavior.Andsoon.Dozensoflesstechnicaltermsservethesamepurpose,andtheirexistencepointsto

our abiding practical and theoretical interest in frequency as a datum.We say thatsomeoneisatennis“fan”ifhefrequentlyplaystennisunderappropriatecircumstances.He is “enthusiastic” about skating, if he frequently goes skating. He is “greatlyinterested” in music if he plays, listens to, and talks about music frequently. The“inveterate”gamblergamblesfrequently.“Highlysexed”peoplefrequentlyengageinsexual behavior.And so on.The practical problems associatedwith these aspects ofhumannaturecanbeexpressedalsoasproblemsinchangingfrequenciesofresponse,but in each case we quicklymove from an observation of frequency to an inferredmomentarycondition.Thisisthelinguisticeffectoftermsofthissort,asitisofmoretechnicalterms,butalinguisticdeviceshouldnotbeallowedtoinfluencethedirectionofourresearch.Thebasicfactsaboutbehaviorcanbediscoveredonlybyexaminingbehaviorduringappreciableintervalsoftime.Itispossibletostudyprobability—inthelightofthisinterpretation—bydesigninga

laboratorysituationinwhichfrequencyofresponsemaybeeasilyexamined.Therearecertain considerations to be observed.Wemust choose a sample of behavior whichmay be so easily identified that repeated instances may be reliably counted. If ourexperiment is to be automatic—andwe shall see that there aremany advantages inmakingitso—ourresponsemustoperateanapparatus.Thebehaviorshouldnotrequiremuchtime,anditshouldleavetheorganismreadytorespondagain.Theseconditionsareratherarbitrary(andourresultsmustbequalifiedaccordingly),buttheyareeasilymet. Sometimes such a response is found ready-made—as in studying so-calledinstinctivebehavior.Otherwiseitmustbe,sotospeak,constructed.Inthecaseofarat,forexample,wehavefounditconvenienttousearesponselikedepressingahorizontalbar.Themovementof thebar isusuallyclear-cut,and itmaybemade tooperateanapparatusbyclosinganelectriccircuit. In thecaseofa fish, theresponseofpushinglightlyagainstaplatehasproveduseful.Thefishattackstheplate,backsaway,andistheninpositiontoattackitagain,themovementoftheplateclosingacircuit.Inbirds—for example, thepigeon—aconvenient response ispeckingadisc througha smallhole in the wall of the experimental space. The disc is delicatelymounted, and theslightestcontactclosesacircuit.Alltheseresponsesareeasilyspecified.Theycanbereadily repeated. In the case of the pigeon, for example, the discmay be pecked asrapidlyasfifteentimespersec.Torecordfrequencyofresponsewecould,ofcourse,usethestandardpolygraph,but

another sort of curvehas proved to bemuchmore convenient.Apen is arranged tomoveonestepacrossastripofpapereachtimetheorganismresponds(Seepages178-182). The result is a step-like diagonal line. Frequency is converted into slope. Co-ordinates are chosen which convert the commonest frequencies into the mostconvenient slopes. If theorganism is responding rapidly, the line is fairly steep. If itresponds slowly, the slope is low. If it does not respond at all, the pen draws a

110

horizontalline.Withalittlepracticeitiseasytoestimatefrequenciesfromtheslopesofsuchgraphsandtofollowchangesinfrequencywithfairaccuracy.InFigure1someactualrecordsshowtherangeoffrequenciesencounteredin thepigeon.Theseparatesteps of the pen cannot be seen on this scale.One record shows a sustained rate of18,000 responses per hour—five responses per second. Another record, by way ofcomparison,showsonly300respondsperhouroroneresponseeverytwelveseconds,yieldingaverylowslope.Iamconcernedherewithdemonstratingthatfrequencyofresponse,sorecorded,isa

useful and significant datum in the experimental analysis of behavior—that it is asensitive “dependent variable” which is a function of many subtle experimentalconditions.

Experiments were described at this point concerning: (1) fixed-intervalreinforcement, (2) extinction after fixed-interval reinforcement, (3) fixed-intervalreinforcementwith“addedclock,” (4)variable-interval reinforcement, (5)extinctionafter variable-interval reinforcement, (6) fixed-ratio reinforcement, (7) fixed-ratioreinforcementwith“addedcounter,”(8)two-valuedratioreinforcement,(9)variable-ratioreinforcement,and(10)extinctionaftervariable-ratioreinforcement.

111

FIG.1.Cumulativecurvesshowingstableperformancesatawidevarietyofratesofresponding.

The following points seem to be justified. First, frequency of response is anextremely orderly datum. The curveswhich represent its relations tomany types ofindependentvariablesareencouraginglysimple.Secondly,theyareeasilyreproduced.Itisseldomnecessarytoresorttogroupsofsubjectsatthisstage.Themethodpermitsadirectviewofprocesseswhichhavehithertobeenonlyinferred,andweoftenhaveaslittleuseforstatisticalcontrolasinthesimpleobservationofobjectsintheworldaboutus. If the essential features of a given curve are not readily duplicated in a later

112

experiment—in either the same or another organism—we take this, not as a cue toresort to averages, but as a warning that some relevant condition has still to bediscoveredandcontrolled.Theusualuniformityoftheresultsencouragesusto turn—nottosamplingprocedures—buttomorerigorousexperimentalcontrol.Asaresult,thirdly,theconceptsandlawswhichemergefromthissortofstudyhave

animmediatereferencetothebehavioroftheindividualwhichislackinginconceptsorlawswhicharetheproductsofstatisticalmethods.Whenweextendanexperimentalanalysistohumanaffairsingeneral,itisagreatadvantagetohaveaconceptualsystemwhich refers to the single individual,preferablywithoutcomparisonwithagroup.Amoredirectapplicationtothepredictionandcontroloftheindividualisthusachieved.Thestudyoffrequencyofresponseappearstoleaddirectlytosuchasystem.Fourthly, frequency of response provides a continuous account of many basic

processes.Wecanfollowacurveofextinction,forexample,formanyhours,andtheconditionoftheresponseateverymomentisapparentinourrecords.Thisisinmarkedcontrasttomethodsandtechniqueswhichmerelysamplealearningprocessfromtimeto time,where thewholeprocessmustbe inferred.The samples areoften sowidelyspacedthatthekindsofdetailswehaveseenherearecompletelyoverlooked.Fifthly,wemustnotforgettheconsiderableadvantageofadatumwhichlendsitself

toautomaticexperimentation.Manyprocessesinbehaviorcoverlongperiodsoftime.Therecordsweobtainfromanindividualorganismmaycoverhundredsofhoursandreportmillionsof responses.Wecharacteristicallyuseexperimentalperiodsofeight,ten,evenfifteenhours.Personalobservationofsuchmaterialisunthinkable.Finally,andperhapsmostimportantofall,frequencyofresponseisavaluabledatum

justbecauseitprovidesasubstantialbasisfortheconceptofprobabilityofresponse—aconcept toward which a science of behavior seems to have been groping for manydecades.Hereisaperfectlygoodphysicalreferentfor thatconcept.It is truethat themomentaryconditionoftheorganismasthetangentofacurveisstillanabstraction—theveryabstractionwhichbecameimportantinthephysicalscienceswithNewtonandLeibniz.Butwearenowabletodealwiththisinarigorousfashion.Thesuperfluoustrappingsoftraditionaldefinitionsoftermslikehabit,attitude,wish,andsoon,maybeavoided.Thepoints illustratedhere ina smallbranchof the fieldof learningapplyequally

well to other fields of behavior. Frequencyof response has alreadyproveduseful instudying the topographyofbehavior—theshapingofnewresponses. Itpermitsus toanswersuchaquestionas:DoestheemissionofResponseAaltertheprobabilityofaResponseBwhichresemblesAincertainways?Ithasprovedtobeausefuldatuminstudying the effect of discriminative stimuli. If we establish a given probability ofresponseunderStimulusA,whatistheprobabilitythattheresponsewillbemadeunderStimulusB, which resemblesA in certain ways? Is red as different from orange asgreenisfromblue?Wemayaskthepigeonaquestionofthissortquitemeaningfullyintermsofprobabilityofresponse.Patterndiscriminationandtheformationofconceptshavebeenstudiedwiththesamemethod.Frequency of response is also a useful datum when two responses are being

consideredatthesametime.Wecaninvestigate“choice”andfollowthedevelopmentofapreferenceforoneoftwostimuli.Thedatumhasprovedtobeespeciallyusefulinstudyingcomplexbehaviorinwhichtwoormoreresponsesarerelatedtotwoormorestimuli—forexample,inmatchingcolorfromsampleorinselectingtheoppositeofasample.Outside the fieldof learningconsiderableworkhasbeendone inmotivation

113

(wherefrequencyofresponsevarieswithdegreeofdeprivation),inemotion(where,forexample,rateofrespondingservesasausefulbaselineinobservingwhatwemaycall“anxiety”), in theeffectsofdrugs(evaluated,forexample,against thestablebaselineobtainedundervariable-intervalreinforcement),andsoon.Oneofthemostpromisingachievementshasbeenananalysisoftheeffectsofpunishment,whichconfirmsmuchof theFreudianmaterial on repressionand revealsmanydisadvantages in theuseofpunishmentasatechniqueofcontrol.The extension of such results to the world at large frequently meets certain

objections.Inthelaboratorywechooseanarbitraryresponse,holdtheenvironmentasconstant as possible, and so on. Can our results apply to behavior of much greatervariety emitted under conditions which are constantly changing? If a certainexperimentaldesignisnecessarytoobserveafrequency,canweapplytheresultstoasituationwherefrequencycannotbedetermined?Theanswerhereistheanswerwhichmustbegivenbyanyexperimentalscience.Laboratoryexperimentationisdesignedtomakeaprocessasobviousaspossible,toseparateprocessesonefromtheother,andtoobtain quantitative measures. These are indeed the very heart of an experimentalscience.Thehistoryofscienceshowsthattheresultscanbeeffectivelyextendedtotheworldatlarge.Forexample,wedeterminetheshapeofthecoolingcurvewiththeaidofthephysicallaboratory.Wehavelittledoubtthatthesameprocessisgoingonasourbreakfastcoffeegrowscold,butwehavenoevidenceforthisandprobablycouldnotprove it under genuine breakfast-table conditions. What we transfer from ourexperimentstoacasualworldinwhichsatisfactoryquantificationisimpossibleistheknowledge that certain basic processes exist, that they are lawful, and that theyprobablyaccountfortheunpleasantlychaoticfactswithwhichwearefaced.Thegaininpracticaleffectivenesswhichisderivedfromsuchtransferredknowledgemaybe,asthephysicalsciencesshow,enormous.Another common objection is that if we identify probability of response with

frequencyofoccurrence,wecannotlegitimatelyapplythenotiontoaneventwhichisneverrepeated.Amanmaymarryonlyonce.Hemayengageinabusinessdealonlyonce.Hemaycommitsuicideonlyonce.Isbehaviorofthissortbeyondthescopeofsuchananalysis?Theanswerhereconcernsthedefinitionoftheunittobepredicted.Complexactivities arenot always “responses” in the senseof repeatedor repeatableevents.Theyarecomposedofresponses,however,whicharerepeatableandcapableofbeingstudiedintermsoffrequency.Theproblemisagainnotpeculiartothefieldofbehavior.Was it possible to assign a given probability to the explosion of the firstatomicbomb?Theprobabilitiesofmanyofthecomponenteventsweresoundlybasedupondataintheformoffrequencies.Buttheexplosionofthebombasawholewasauniqueeventinthehistoryoftheworld.Thoughtheprobabilityofitsoccurrencecouldnotbestatedintermsofthefrequencyofaunitatthatlevel,itcouldstillbeevaluated.Theproblemofpredictingthatamanwillcommitsuicideisofthesamenature.In summary, then, the basic datum in the analysis of behavior has the status of a

probability. The actual observed dependent variable is frequency of response. Anexperimental situation is required in which frequencymay be studied.When this isarranged, important processes in behavior are revealed in a continuous, orderly, andreproducible fashion. Concepts and laws derived from such data are immediatelyapplicable to the behavior of the individual, and they permit us to move on to theinterpretationofbehaviorintheworldatlargewiththegreatestpossiblespeed.

114

ACaseHistoryinScientificMethod

In exploring the status and development of psychology in the United States theAmericanPsychologicalAssociation,withhelpfromtheNationalScienceFoundation,setupProjectAunderthedirectionofSigmundKoch.AmongotherthingstheProjectsponsored “the preparation of analyses of given systematizations by the actualoriginatorsof theseformulations.”Inotherwords, theproprietorsofseveralcurrentsystems of psychology were asked to describe their wares. In the instruction tocontributors“systematicformulation”wasdefinedas“anysetofsentencesformulatedasatoolfororderingempiricalknowledgewithrespecttosomespecifiabledomainofevents.”Topics tobecovered includedBackgroundFactorsandOrientingAttitudes,InitialEvidentialGroundsforAssumptionsofSystem,DegreeofProgrammaticity,andIntermediateandLongRangeStrategyfortheDevelopmentoftheSystem.My reactionwas the present paper. It waswritten primarily for ameeting of the

EasternPsychologicalAssociationinApril,1955,andwaspublishedintheAmericanPsychologist(1956,11,221-233).Itwassubmittedwithconsiderablediffidencetothedirector of Project A, who generously included it in the report of the project,Psychology: A Study of a Science, Vol. II (New York, McGraw-Hill, 1958). It isreprintedherebypermission.Intheprojectreportthepaperbeginswiththefollowingparagraph:

Ascientistisanextremelycomplexorganism,andhisbehaviorislikelytoresisttotheverylastanyefforttowardanempiricalanalysis.Nevertheless,ifanythingusefulisto be said about him, either in trying to understand his behavior or in inculcatingsimilarbehaviorinothers,itwillbeinthenatureofanempiricalratherthanaformalanalysis. As an anti-formalist it would be inconsistent of me to describe my ownscientificactivityintheformalframeworkofProjectA.Ihavethereforereactedtotheproposalofthedirectorbyillustratingmyownphilosophyofsciencewithapersonalhistory.

It has been said that college teaching is the only profession for which there is noprofessional training, and it is commonly argued that this is because our graduateschoolstrainscholarsandscientistsratherthanteachers.Wearemoreconcernedwiththediscoveryofknowledge thanwith itsdissemination.Butcanwe justifyourselvesquite so easily? It is a bold thing to say that we know how to train a man to be ascientist.Scientific thinking is themostcomplexandprobably themost subtleofallhuman activities. Do we actually know how to shape up such behavior, or do wesimply mean that some of the people who attend our graduate schools eventuallybecomescientists?Exceptforalaboratorycoursewhichacquaintsthestudentwithstandardapparatus

and standard procedures, the only explicit training in scientific method generallyreceivedbyayoungpsychologistisacourseinstatistics—nottheintroductorycourse,

115

whichisoftenrequiredofsomanykindsofstudentsthatitisscarcelyscientificatall,butanadvancedcoursewhich includes“modelbuilding,”“theoryconstruction,”and“experimental design.” But it is a mistake to identify scientific practice with theformalized constructions of statistics and scientific method. These disciplines havetheirplace,butitdoesnotcoincidewiththeplaceofscientificresearch.Theyofferamethodof sciencebut not, as is sooften implied, themethod.As formal disciplinestheyaroseverylateinthehistoryofscience,andmostofthefactsofsciencehavebeendiscoveredwithouttheiraid.IttakesagreatdealofskilltofitFaradaywithhiswiresandmagnetsintothepicturewhichstatisticsgivesusofscientificthinking.Andmostcurrent scientific practice would be equally refractory, especially in the importantinitialstages.Itisnowonderthatthelaboratoryscientistispuzzledandoftendismayedwhenhediscovershowhisbehaviorhasbeenreconstructedintheformalanalysesofscientificmethod.He is likely toprotest that this isnotatalla fair representationofwhathedoes.Buthisprotest isnot likely tobeheard.For theprestigeofstatisticsandscientific

methodologyisenormous.Muchofitisborrowedfromthehighreputeofmathematicsand logic, but much of it derives from the flourishing state of the art itself. Somestatisticiansareprofessionalpeopleemployedbyscientificandcommercialenterprises.Some are teachers and pure researcherswho give their colleagues the same kind ofservice fornothing—oratmostanoteofacknowledgment.Manyarezealouspeoplewho,withthebestofintentions,areanxioustoshowthenonstatisticalscientisthowhecandohisjobmoreefficientlyandassesshisresultsmoreaccurately.Therearestrongprofessional societies devoted to the advancement of statistics, and hundreds oftechnicalbooksandjournalsarepublishedannually.Against this, the practicing scientist has very little to offer. He cannot refer the

youngpsychologisttoabookwhichwilltellhimhowtofindoutallthereistoknowabouta subjectmatter,how tohave thegoodhunchwhichwill leadhim todeviseasuitablepieceofapparatus,howtodevelopanefficientexperimental routine,howtoabandonanunprofitablelineofattack,howtomoveonmostrapidlytolaterstagesofhisresearch.Theworkhabitswhichhavebecomesecondnaturetohimhavenotbeenformalizedbyanyone,andhemayfeel that theypossiblyneverwillbe.AsRichter1haspointedout,“Someofthemostimportantdiscoverieshavebeenmadewithoutanyplanofresearch,”and“thereareresearcherswhodonotworkonaverbalplane,whocannotputintowordswhattheyaredoing.”Ifweareinterestedinperpetuatingthepracticesresponsibleforthepresentcorpusof

scientific knowledge, we must keep in mind that some very important parts of thescientific process do not now lend themselves tomathematical, logical, or anyotherformal treatment.We do not know enough about human behavior to know how thescientistdoeswhathedoes.Althoughstatisticiansandmethodologistsmayseemtotellus,orat least imply,howthemindworks—howproblemsarise,howhypothesesareformed,deductionsmade,andcrucialexperimentsdesigned—weaspsychologistsareinapositiontoremindthemthattheydonothavemethodsappropriatetotheempiricalobservation or the functional analysis of such data. These are aspects of humanbehavior,andnooneknowsbetterthanwehowlittlecanatthemomentbesaidaboutthem.Some day we shall be better able to express the distinction between empirical

analysis and formal reconstruction, for we shall have an alternative account of the

116

behaviorofManThinking.Suchanaccountwillnotonlyplausiblyreconstructwhataparticularscientistdidinanygivencase,itwillpermitustoevaluatepracticesand,Ibelieve, to teachscientific thinking.But thatday issome littledistance in the future.Meanwhilewecanonlyfallbackonexamples.WhenthedirectorofProjectAoftheAmericanPsychologicalAssociationaskedme

to describe and analyze my activities as a research psychologist, I went through atrunkful of old notes and records and, for my pains, reread some of my earlierpublications. This has made me all the more aware of the contrast between thereconstructionsofformalizedscientificmethodandatleastonecaseofactualpractice.InsteadofamplifyingthepointsIhavejustmadebyresortingtoageneralizedaccount(principallybecauseitisnotavailable),Ishouldliketodiscussacasehistory.Itisnotoneofthecasehistoriesweshouldmostliketohave,butwhatitlacksinimportanceisperhapssomewhatoffsetbyaccessibility.Ithereforeaskyoutoimaginethatyouareallclinicalpsychologists—a taskwhichbecomeseasier and easier as theyearsgoby—whileIsitacrossthedeskfromyouorstretchoutuponthiscomfortableleathercouch.The first thing I can remember happenedwhen I was only twenty-two years old.

Shortly after I was graduated from college Bertrand Russell published a series ofarticles in the old Dial magazine on the epistemology of John B. Watson’sBehaviorism. I had had no psychology as an undergraduate but I had had a lot ofbiology,andtwoofthebookswhichmybiologyprofessorhadputintomyhandswereLoeb’sPhysiologyof theBrain and the newly publishedOxford edition of Pavlov’sConditionedReflexes. And now here was Russell extrapolating the principles of anobjectiveformulationofbehaviortotheproblemofknowledge!ManyyearslaterwhenI toldLordRussell that his articleswere responsible formy interest in behavior, hecould only exclaim, “Good Heavens! I had always supposed that those articles haddemolishedBehaviorism!”ButatanyratehehadtakenWatsonseriously,andsodidI.When I arrived at Harvard for graduate study, the air was not exactly full of

behavior,butWalterHunterwascominginonceaweekfromClarkUniversitytogiveaseminar,andFredKeller,alsoagraduatestudent,wasanexpertinboththetechnicaldetailsandthesophistryofBehaviorism.ManyatimehesavedmeasIsankintothequicksandsofanamateurishdiscussionof“What isan image?”or“Where is red?”IsooncameintocontactwithW.J.Crozier,whohadstudiedunderLoeb. Ithadbeensaid of Loeb, and might have been said of Crozier, that he “resented the nervoussystem.”Whether thiswas trueornot, the factwas thatboth thesemen talkedaboutanimalbehaviorwithoutmentioning thenervous systemandwith surprisingsuccess.So far as Iwas concerned, theycancelledout thephysiological theorizingofPavlovand Sherrington and thus clarified what remained of the work of these men as thebeginningsofan independent scienceofbehavior.Mydoctoral thesiswas inpart anoperational analysis of Sherrington’s synapse, in which behavioral laws weresubstitutedforsupposedstatesofthecentralnervoussystem.Butthepartofmythesisatissueherewasexperimental.SofarasIcansee,Ibegan

simplybylookingforlawfulprocessesinthebehavioroftheintactorganism.Pavlovhadshown theway;but Icouldnot then,as Icannotnow,movewithouta jolt fromsalivaryreflexestotheimportantbusinessoftheorganismineverydaylife.Sherringtonand Magnus had found order in surgical segments of the organism. Could notsomething of the same sort be found, to use Loeb’s phrase, in “the organism as awhole”?IhadthecluefromPavlov:controlyourconditionsandyouwillseeorder.It is not surprising that my first gadget was a silent release box, operated by

117

compressedairanddesignedtoeliminatedisturbanceswhenintroducingaratintoanapparatus.Iusedthisfirstinstudyingthewayaratadaptedtoanovelstimulus.Ibuilta soundproofed box containing a specially structured space. A rat was released,pneumatically,atthefarendofadarkenedtunnelfromwhichitemergedinexploratoryfashionintoawell-lightedarea.Toaccentuateitsprogressandtofacilitaterecording,the tunnel was placed at the top of a flight of steps, something like a functionalParthenon (Figure 1). The rat would peek out from the tunnel, perhaps glancingsuspiciouslyattheone-waywindowthroughwhichIwaswatchingit,thenstretchitselfcautiouslydownthesteps.Asoftclick(carefullycalibrated,ofcourse)wouldcauseittopullbackintothetunnelandremainthereforsometime.Butrepeatedclickshadlessandlessofaneffect.Irecordedtherat’sadvancesandretreatsbymovingapenbackandforthacrossamovingpapertape.Themajorresultofthisexperimentwasthatsomeofmyratshadbabies.Ibeganto

watch young rats. I saw them right themselves and crawl about verymuch like thedecerebrateorthalamiccatsandrabbitsofMagnus.SoIsetaboutstudyingtheposturalreflexesofyoungrats.Herewasafirstprinciplenotformallyrecognizedbyscientificmethodologists:When you run onto something interesting, drop everything else andstudyit.ItoreuptheParthenonandstartedover.

118

FIG.1

Ifyouholdayoungratononehandandpullitgentlybythetail,itwillresistyoubypullingforwardandthen,withasuddensharpspringwhichusuallydisengagesitstail,itwillleapoutintospace.Idecidedtostudythisbehaviorquantitatively.Ibuiltalightplatformcoveredwithclothandmounteditontightlystretchedpianowires(Figure2).Here was a version of Sherrington’s torsion-wire myograph, originally designed torecordtheisometriccontractionofthetibialisanticusofacat,buthereadaptedtotheresponseofawholeorganism.Whenthetailoftheyoungratwasgentlypulled,theratclungtotheclothfloorandtuggedforward.Byamplifyingthefinemovementsoftheplatform,itwaspossibletogetagoodkymographrecordofthetremorinthismotionandthen,asthepullagainstthetailwasincreased,ofthedesperatespringintotheair(Figure3).

119

FIG.2

120

FIG.3

Now,babyratshaveverylittlefuture,exceptasadultrats.Theirbehaviorisliterallyinfantile and cannot be usefully extrapolated to everyday life. But if this techniquewouldworkwithababy,whynottryitonamaturerat?Toavoidattachinganythingtotherat,itshouldbepossibletorecord,notapullagainstthesubstrate,buttheballisticthrustexertedas the rat runs forwardor suddenly stops in response tomycalibratedclick. So, invoking the first principle of scientific practice again, I threw away thepiano-wireplatformandbuiltarunway,eightfeetlong.Thiswasconstructedoflightwood,intheformofaUgirder,mountedrigidlyonverticalglassplates,theelasticityof which permitted a very slight longitudinal movement (Figure 4). The runwaybecame the floor of a long tunnel, not shown, at one end of which I placed mysoundless release box and at the other endmyself, prepared to reinforce the rat forcomingdowntherunwaybygivingitabitofwetmash,tosoundaclickfromtimetotimewhenithadreachedthemiddleoftherunway,andtoharvestkymographrecordsofthevibrationsofthesubstrate.Nowforasecondunformalizedprincipleofscientificpractice:Somewaysofdoing

121

researchareeasierthanothers.Igottiredofcarryingtheratbacktotheotherendoftherunway.Abackalleywas thereforeadded (Figure5).Now the rat couldeatabitofmashatpointC,godownthebackalleyA,aroundtheendasshown,andbackhomebyrunwayB.TheexperimenteratEcouldcollectrecordsfromthekymographatDincomfort.Inthiswayagreatmanyrecordsweremadeoftheforcesexertedagainstthesubstratumasratsrandownthealleyandoccasionallystoppeddeadintheirtracksasaclicksounded(Figure6).

FIG.4

122

Therewasoneannoyingdetail,however.Theratwouldoftenwaitan inordinatelylongtimeatCbeforestartingdownthebackalleyonthenextrun.Thereseemedtobeno explanation for this.When I timed these delayswith a stopwatch, however, andplottedthem,theyseemedtoshoworderlychanges(Figure7).Thiswas,ofcourse,thekindofthingIwaslookingfor.Iforgotallaboutthemovementsofthesubstratumandbegantorunratsforthesakeofthedelaymeasurementsalone.Buttherewasnownoreasonwhytherunwayhadtobeeightfeetlongand,asthesecondprinciplecameintoplayagain,Isawnoreasonwhytheratcouldnotdeliveritsownreinforcement.

123

FIG.5

124

FIG.6

FIG.7

Anewapparatuswasbuilt.InFigure8weseetherateatingapieceoffoodjustaftercompletingarun.Itproducedthefoodbyitsownaction.AsitrandownthebackalleyAtothefarendoftherectangularrunway,itsweightcausedthewholerunwaytotiltslightlyontheaxisCandthismovementturnedthewoodendiscD,permittingapieceoffood inoneof theholesarounditsperimeter todrop throughafunnel intoa fooddish.The foodwaspearl tapioca, theonlykind I could find in thegrocery stores inreasonablyuniformpieces.TherathadonlytocompleteitsjourneybycomingdownthehomestretchBtoenjoyitsreward.Theexperimenterwasabletoenjoyhis rewardatthesametime,forhehadonlytoloadthemagazine,putinarat,andrelax.Eachtiltwasrecordedonaslowlymovingkymograph.A third unformalized principle of scientific practice: Some people are lucky. The

disc of wood from which I had fashioned the food magazine was taken from astoreroom of discarded apparatus. It happened to have a central spindle, whichfortunatelyIhadnotbotheredtocutoff.OnedayitoccurredtomethatifIwoundastring around the spindle and allowed it to unwind as the magazine was emptied(Figure9), Iwouldgetadifferentkindofrecord. Insteadofamerereportof theup-and-downmovementoftherunway,asaseriesofpipsasinapolygraph,Iwouldgetacurve.AndIknewthat sciencemadegreatuseofcurves,although,so faras Icoulddiscover, very little of pips on a polygram. The difference between the old type ofrecordatA(Figure10)andthenewatBmaynotseemgreat,butasitturnedoutthecurve revealed things in the rate of responding, and in changes in that rate, whichwouldcertainlyotherwisehavebeenmissed.Byallowingthestringtounwindratherthantowind,IhadgotmycurveinanawkwardCartesianquadrant,butthatwaseasilyremedied.Psychologistshaveadoptedcumulativecurvesonlyveryslowly,butIthinkit is fair to say that they have become an indispensable tool for certain purposes ofanalysis.

125

FIG.8

FIG.9

126

FIG.10

Eventually,ofcourse,therunwaywasseentobeunnecessary.Theratcouldsimplyreach into a covered tray for pieces of food, and eachmovement of the cover couldoperate a solenoid to move a pen one step in a cumulative curve. The first majorchangeinrateobservedin thiswaywasdue to indigestion.Curvesshowinghowtherateofeatingdeclinedwith the timeofeatingcomprised theotherpartofmythesis.Butarefinementwasneeded.ThebehavioroftheratinpushingopenthedoorwasnotanormalpartoftheingestivebehaviorofRattusrattus.Theactwasobviouslylearnedbutitsstatusaspartofthefinalperformancewasnotclear.Itseemedwisetoaddaninitialconditionedresponseconnectedwithingestioninaquitearbiraryway.Ichosethefirstdevicewhichcametohand—ahorizontalbarorleverplacedwhereitcouldbeconveniently depressed by the rat to close a switch which operated a magneticmagazine.Ingestioncurvesobtainedwiththisinitialresponseinthechainwerefoundtohavethesamepropertiesasthosewithoutit.Now, as soon as you begin to complicate an apparatus, you necessarily invoke a

fourthprincipleofscientificpractice:Apparatusessometimesbreakdown.Ihadonlytowaitforthefoodmagazinetojamtogetanextinctioncurve.AtfirstItreatedthisasadefectandhastenedtoremedythedifficulty.Buteventually,ofcourse,Ideliberatelydisconnected the magazine. I can easily recall the excitement of that first completeextinctioncurve(Figure11).IhadmadecontactwithPavlovatlast!Herewasacurveuncorruptedbythephysiologicalprocessofingestion.Itwasanorderlychangeduetonothingmorethanaspecialcontingencyofreinforcement.Itwaspurebehavior!IamnotsayingthatIwouldnothavegotaroundtoextinctioncurveswithoutabreakdownintheapparatus;Pavlovhadgiventoostrongaleadinthatdirection.Butitisstillnoexaggerationtosaythatsomeofthemostinterestingandsurprisingresultshaveturnedupfirstbecauseofsimilaraccidents.Foolproofapparatusisnodoubthighlydesirable,butCharlesFerster and I in recently reviewing thedata froma five-yearprogramofresearchfoundmanyoccasionstocongratulateourselvesonthefallibilityofrelaysandvacuumtubes.

127

FIG.11

Ithenbuiltfoursoundproofedventilatedboxes,eachcontainingaleverandafoodmagazineandsuppliedwithacumulativerecorder,andwasonmywaytoanintensivestudyofconditionedreflexesinskeletalbehavior.Iwouldreinforceeveryresponseforseveral days and then extinguish for a day or two, varying the number ofreinforcements,theamountofpreviousmagazinetraining,andsoon.AtthispointImademyfirstuseofthedeductivemethod.Ihadlongsincegivenup

pearl tapioca as too unbalanced a diet for steady use. A neighborhood druggist hadshownmehispillmachine,andIhadhadonemadealongthesamelines(Figure12). Itconsistedofaflutedbrassbedacrosswhichonelaidalongcylinderofstiffpaste(inmycaseaMacCollumformulaforanadequateratdiet).Asimilarlyflutedcutterwasthenloweredontothecylinderandrolledslowlybackandforth,convertingthepasteintoaboutadozensphericalpellets.Theseweredriedforadayorsobeforeuse.Theprocedurewaspainstakingandlaborious.Eightratseatingahundredpelletseachperdaycouldeasilykeepupwithproduction.OnepleasantSaturdayafternoonIsurveyedmy supply of dry pellets and, appealing to certain elemental theorems in arithmetic,deducedthatunlessIspenttherestofthatafternoonandeveningatthepillmachine,thesupplywouldbeexhaustedbyten-thirtyMondaymorning.

128

FIG.12

Since I do notwish to deprecate the hypothetico-deductivemethod, I am glad totestifyhere to itsusefulness. It ledmetoapplyoursecondprincipleofunformalizedscientificmethodandtoaskmyselfwhyeverypressoftheleverhadtobereinforced.Iwas not then aware of what had happened at the Brown laboratories, as HaroldSchlosberglatertoldthestory.Agraduatestudenthadbeengiventhetaskofrunningacat through a difficult discrimination experiment.One Sunday the student found thesupplyofcatfoodexhausted.Thestoreswereclosed,andso,withabeautifulfaithinthefrequency-theoryoflearning,heranthecatasusualandtookitbacktoitslivingcage unrewarded. Schlosberg reports that the cat howled its protest continuously fornearlyforty-eighthours.Unawareof thisIdecidedtoreinforcearesponseonlyonceevery minute and to allow all other responses to go unreinforced. There were tworesults:(a)mysupplyofpelletslastedalmostindefinitely;and(b)eachratstabilizedatafairlyconstantrateofresponding.Now,asteadystatewassomethingIwasfamiliarwithfromphysicalchemistry,and

I thereforeembarkedupon thestudyofperiodic reinforcement. I soonfound that theconstantrateatwhichtheratstabilizeddependeduponhowhungryitwas.Hungryrat,high rate; less hungry rat, lower rate. At that time I was bothered by the practicalproblemofcontrollingfooddeprivation.IwasworkinghalftimeattheMedicalSchool(onchronaxieof subordination!)andcouldnotmaintainagoodschedule inworkingwiththerats.Therateofrespondingunderperiodicreinforcementsuggestedaschemefor keeping a rat at a constant level of deprivation. The argument went like this:Suppose you reinforce the rat, not at the end of a given period, but when it hascompletedthenumberofresponsesordinarilyemittedinthatperiod.Andsupposeyouuse substantial pellets of food andgive the rat continuous access to the lever.Then,except for periodswhen the rat sleeps, it should operate the lever at a constant ratearoundtheclock.For,wheneveritgrowsslightlyhungrier,itwillworkfaster,getfoodfaster, andbecome lesshungry,whilewhenever itgrowsslightly lesshungry, itwillrespondatalowerrate,getlessfood,andgrowhungrier.Bysettingthereinforcementatagivennumberofresponsesitshouldevenbepossibletoholdtheratatanygiven

129

levelofdeprivation. Ivisualizedamachinewithadialwhichonecould set tomakeavailable,atanytimeofdayornight,aratinagivenstateofdeprivation.Ofcourse,nothing of the sort happens. This is “fixed-ratio” rather than “fixed-interval”reinforcement and, as I soon found out, it produces a very different type ofperformance.Thisisanexampleofafifthunformalizedprincipleofscientificpractice,but one which has at least been named. Walter Cannon described it with a wordinventedbyHoraceWalpole:serendipity—theartof findingone thingwhile lookingforsomethingelse.This account of my scientific behavior up to the point at which I published my

resultsinabookcalledTheBehaviorofOrganismsisasexactinletterandspiritasIcannowmakeit.Thenotes,data,andpublicationswhichIhaveexamineddonotshowthatIeverbehavedinthemannerofManThinkingasdescribedbyJohnStuartMillorJohn Dewey or in reconstructions of scientific behavior by other philosophers ofscience.IneverfacedaProblemwhichwasmorethantheeternalproblemoffindingorder. I never attacked a problem by constructing a Hypothesis. I never deducedTheorems or submitted them to Experimental Check. So far as I can see, I had nopreconceivedModelofbehavior—certainlynotaphysiologicalormentalisticoneand,I believe, not a conceptual one. The “reflex reserve” was an abortive, thoughoperational,conceptwhichwasretractedayearorsoafterpublicationinapaperatthePhiladelphiameetingoftheAPA.It liveduptomyopinionoftheoriesingeneralbyprovingutterlyworthlessinsuggestingfurtherexperiments.Ofcourse,IwasworkingonabasicAssumption—thattherewasorderinbehaviorifIcouldonlydiscoverit—butsuchanassumptionisnottobeconfusedwiththehypothesesofdeductivetheory.ItisalsotruethatIexercisedacertainSelectionofFactsbutnotbecauseofrelevanceto theory but because one fact was more orderly than another. If I engaged inExperimentalDesignatall,itwassimplytocompleteorextendsomeevidenceoforderalreadyobserved.Mostof theexperimentsdescribed inTheBehaviorofOrganisms were donewith

groupsoffourrats.Afairlycommonreactiontothebookwasthatsuchgroupsweretoo small.Howdid I know that other groups of four ratswould do the same thing?Keller,indefendingthebook,counteredwiththechargethatgroupsoffourweretoobig.Unfortunately, however, I allowedmyself to be persuaded of the contrary.Thiswasdue inpart tomyassociationat theUniversityofMinnesotawithW.T.Heron.Through him I came into close contact for the first time with traditional animalpsychology.Heronwas interested in inheritedmaze behavior, inherited activity, andcertain drugs—the effects of which could then be detected only through the use offairly largegroups.Wedidanexperiment togetheron theeffectof starvationon therateofpressingaleverandstartedthenewerawithagroupofsixteenrats.Butwehadonlyfourboxes,andthiswassoinconvenientthatHeronappliedforagrantandbuiltabatteryoftwenty-fourlever-boxesandcumulativerecorders.Isuppliedanattachmentwhichwouldrecord,notonlythemeanperformanceofalltwenty-fourratsinasingleaveraged curve, but mean curves for four subgroups of twelve rats each and foursubgroupsofsixratseach.2Wethusprovidedforthedesignofexperimentsaccordingto theprinciplesofR.A.Fisher,whichwere thencoming intovogue.Wehad,so tospeak,mechanizedtheLatinsquare.WiththisapparatusHeronandIpublishedastudyofextinctioninmaze-brightand

maze-dullratsusingninety-fivesubjects.LaterIpublishedmeanextinctioncurvesfor

130

groupsoftwenty-four,andW.K.EstesandIdidourworkonanxietywithgroupsofthe same size. But although Heron and I could properly voice the hope that “thepossibility of using large groups of animals greatly improves upon the method aspreviously reported, since tests of significance are provided for and properties ofbehaviornotapparentinsinglecasesmaybemoreeasilydetected,”inactualpracticethat is notwhat happened.The experiments I have justmentioned are almost allwehave to show for this elaborate battery of boxes. Undoubtedlymore work could bedone with it and would have its place, but something had happened to the naturalgrowth of the method. You cannot easily make a change in the conditions of anexperimentwhentwenty-fourapparatuseshavetobealtered.Anygaininrigorismorethanmatchedbyalossinflexibility.Wewereforcedtoconfineourselvestoprocesseswhichcouldbestudiedwiththebaselinesalreadydevelopedinearlierwork.Wecouldnotmoveontothediscoveryofotherprocessesoreventoamorerefinedanalysisofthose we were working with. No matter how significant might be the relations weactually demonstrated, our statistical Leviathan had swum aground. The art of themethodhadstuckataparticularstageofitsdevelopment.Anotheraccidentrescuedmefrommechanizedstatisticsandbroughtmebacktoan

even more intensive concentration on the single case. In essence, I suddenly foundmyselffacetofacewiththeengineeringproblemoftheanimaltrainer.Whenyouhavethe responsibility ofmaking absolutely sure that a given organismwill engage in agiven sort of behavior at a given time, you quickly grow impatientwith theories oflearning. Principles, hypotheses, theorems, satisfactory proof at the .05 level ofsignificancethatbehavioratachoicepointshowstheeffectofsecondaryreinforcement—nothingcouldbemoreirrelevant.Noonegoestothecircustoseetheaveragedogjump throughahoopsignificantlyoftener thanuntraineddogs raisedunder the samecircumstances,ortoseeanelephantdemonstrateaprincipleofbehavior.Perhaps I can illustrate this without giving aid and comfort to the enemy by

describingaRussiandevicewhichtheGermansfoundquiteformidable.TheRussiansuseddogs toblowup tanks.Adogwas trained tohidebehind a treeorwall in lowbrushorothercover.Asatankapproachedandpassed,thedogranswiftlyalongsideit,andasmallmagneticmineattachedtothedog’sbackwassufficienttocripplethetankorsetitafire.Thedog,ofcourse,hadtobereplaced.NowIaskyou toconsidersomeof the technicalproblemswhich thepsychologist

facesinpreparingadogforsuchanactofunintentionalheroism.Thedogmustwaitbehind the tree for an indefinite length of time. Very well, it must therefore beintermittently reinforced for waiting. But what schedule will achieve the highestprobabilityofwaiting?Ifthereinforcementistobefood,whatistheabsolutelyoptimalscheduleofdeprivationconsistentwiththehealthofthedog?Thedogmustruntothetank—that can be arranged by reinforcing it with a practice tank—but it must startinstantlyifitistoovertakeaswifttank,andhowdoyoudifferentiallyreinforceshortreaction times, especially in counteracting the reinforcement for sitting andwaiting?Thedogmustreactonlytotanks,nottoarefugeedrivinghisoxcartalongtheroad,butwhatarethedefiningpropertiesofatanksofarasadogisconcerned?Ithinkitcanbesaidthatafunctionalanalysisprovedadequateinitstechnological

application.Manipulationof environmental conditions alonemadepossible awhollyunexpectedpracticalcontrol.Behaviorcouldbeshapedupaccordingtospecificationsandmaintained indefinitely almost atwill.One behavioral technologistwhoworkedwithmeatthetime(KellerBreland)isnowspecializingintheproductionofbehavior

131

as a salable commodity and has described this new profession in the AmericanPsychologist.3Therearemanyusefulapplicationswithinpsychologyitself.RatliffandBloughhave

recentlyconditionedpigeonstoserveaspsychophysicalobservers.Intheirexperimentapigeonmayadjustoneoftwospotsoflightuntilthetwoareequallybrightoritmayhold a spot of light at the absolute threshold during dark adaptation.The techniqueswhichtheyhavedevelopedto inducepigeons todo thisareonly indirectlyrelated tothe point of their experiments and hence exemplify the application of a behavioralscience.4Thefieldinwhichabettertechnologyofbehaviorisperhapsmosturgentlyneeded is education. I cannotdescribehere the applicationswhicharenowpossible,but perhaps I can indicate my enthusiasm by hazarding the guess that educationaltechniques at all age levels are on the threshold of revolutionary changes (see page217).Theeffectofabehavioraltechnologyonscientificpracticeistheissuehere.Faced

with practical problems in behavior, you necessarily emphasize the refinement ofexperimentalvariables.Asaresult,someofthestandardproceduresofstatisticsappeartobecircumvented.Letmeillustrate.Supposethatmeasurementshavebeenmadeontwogroupsofsubjectsdifferinginsomedetailofexperimentaltreatment.Meansandstandarddeviations for the twogroupsaredetermined,andanydifferencedue to thetreatment is evaluated. If the difference is in the expected direction but is notstatisticallysignificant,thealmostuniversalrecommendationwouldbetostudylargergroups. But our experience with practical control suggests that we may reduce thetroublesomevariabilitybychangingtheconditionsoftheexperiment.Bydiscovering,elaborating,andfullyexploitingeveryrelevantvariable,wemayeliminateinadvanceofmeasurementtheindividualdifferenceswhichobscurethedifferenceunderanalysis.Thiswill achieve the same result as increasing the sizeofgroups, and itwill almostcertainlyyieldabonusinthediscoveryofnewvariableswhichwouldnothavebeenidentifiedinthestatisticaltreatment.The same may be said of smooth curves. In our study of anxiety, Estes and I

published several curves, the reasonable smoothness of which was obtained byaveraging the performances of twelve rats for each curve. The individual curvespublished at that time show that the mean curves do not faithfully represent thebehaviorofanyonerat.Theyshowacertaintendencytowardachangeinslopewhichsupportedthepointweweremaking,andtheymayhaveappearedtowarrantaveragingforthatreason.Butanalternativemethodwouldhavebeen toexplore the individualcaseuntil an

equallysmoothcurvecouldbeobtained.Thiswouldhavemeantnotonlyrejectingthetemptation to produce smoothness by averaging cases, butmanipulating all relevantconditionsaswelaterlearnedtomanipulatethemforpracticalpurposes.Theindividualcurveswhichwepublishedatthattimepointtotheneednotforlargergroupsbutforimprovementinexperimentaltechnique.Here,forexample,isacurvethesmoothnessofwhichischaracteristicofcurrentpractice.SuchcurveswereshowninthemakinginademonstrationwhichFersterandIarrangedattheClevelandmeetingoftheAmericanPsychological Association (Figure 13). Here, in a single organism, three differentschedules of reinforcement are yielding corresponding performances with greatuniformity under appropriate stimuli alternating at random. One does not reach thiskindoforderthroughtheapplicationofstatisticalmethods.

132

FIG.13

InThe Behavior ofOrganisms I was content to deal with the over-all slopes andcurvature of cumulative curves and could make only a rough classification of thepropertiesofbehaviorshownbythefinergrain.Thegrainhasnowbeenimproved.Theresolving power of the microscope has been greatly increased, and we can seefundamental processes of behavior in sharper and sharper detail. In choosing rate ofrespondingasabasicdatumandinrecordingthisconvenientlyinacumulativecurve,wemakeimportanttemporalaspectsofbehaviorvisible.Oncethishashappened,ourscientificpracticeisreducedtosimplelooking.Anewworldisopenedtoinspection.

133

Weusesuchcurvesasweuseamicroscope,X-raycamera,ortelescope.Thisiswellexemplifiedbyrecentextensionsof themethod.Theseareno longerpartofmycasehistory,butperhapsyouwillpermitme toconsultyouaboutwhat somecriticshavedescribedasafolieàdeuxorgroupneurosis.An early application of themethod to the behavior of avoidance and escape was

madebyKellerinstudyingthelightaversionoftherat.ThiswasbrilliantlyextendedbyMurray Sidman in his shock-avoidance experiments. It is no longer necessary todescribe avoidance and escape by appeal to “principles,” for we may watch thebehaviordevelopwhenwehavearrangedthepropercontingenciesofreinforcement,aswelaterwatchitchangeasthesecontingenciesarechanged.Hunt and Brady have extended the use of a stable rate in the study of anxiety-

producing stimuli and have shown that the depression in rate is eliminated byelectroconvulsiveshockandbyothermeasureswhichareeffectiveinreducinganxietyin humanpatients.O.R.Lindsleyhas found the same thing for dogs, using insulin-shocktherapyandsedatives.Bradyhasrefinedthemethodbyexploringtherelevanceof various schedules of reinforcement in tracing the return of the conditioneddepression after treatment. In these experiments you see the effect of a treatment asdirectlyasyouseetheconstrictionofacapillaryunderthemicroscope.EarlyworkwithratsoncaffeineandBenzedrinehasbeenextendedbyLindsleywith

dogs. A special technique for evaluating several effects of a drug in a single shortexperimentalperiodyieldsarecordofbehaviorwhichcanbereadasaspecialistreadsan electrocardiogram. Dr. Peter Dews of the Department of Pharmacology at theHarvardMedicalSchoolisinvestigatingdose-responsecurvesandthetypesandeffectsof various drugs, using pigeons as subjects. In the Psychological Laboratories atHarvard additional work on drugs is being carried out by Morse, Herrnstein, andMarshall, and the technique is being adopted by drug manufacturers. There couldscarcely be a better demonstration of the experimental treatment of variability. In asingle experimental sessionwithasingle organism one observes the onset, duration,anddeclineoftheeffectsofadrug.The direct observation of defective behavior is particularly important. Clinical or

experimentaldamagetoanorganismischaracteristicallyunique.Hencethevalueofamethodwhichpermitsthedirectobservationofthebehavioroftheindividual.Lindsleyhasstudiedtheeffectsofnear-lethalirradiation,andtheeffectsofprolongedanesthesiaand anoxia are currently being examined by Thomas Lohr in co-operation with Dr.HenryBeecheroftheMassachusettsGeneralHospital.ThetechniqueisbeingappliedtoneurologicalvariablesinthemonkeybyDr.KarlPribramattheHartfordInstitute.The pattern of such research is simple: establish the behavior in which you areinterested, submit the organism to a particular treatment, and then look again at thebehavior. An excellent example of the use of experimental control in the study ofmotivation is somework on obesity by J. E.Anliker in collaborationwithDr. JeanMayer of the Harvard School of Public Health, where abnormalities of ingestivebehaviorinseveraltypesofobesemicecanbecomparedbydirectinspection.Thereisperhapsnofieldinwhichbehavioriscustomarilydescribedmoreindirectly

thanpsychiatry.InanexperimentattheMassachusettsStateHospital,O.R.Lindsleyiscarryingoutanextensiveprogramwhichmightbecharacterizedasaquantitativestudyofthetemporalpropertiesofpsychoticbehavior.5Hereagainitisaquestionofmakingcertaincharacteristicsofthebehaviorvisible.

134

Theextent towhichwecaneliminatesourcesofvariabilitybeforemeasurement isshownbya resultwhichhasanunexpected significance forcomparativepsychologyandthestudyofindividualdifferences.Figure14showstracingsofthreecurveswhichreport behavior in response to a multiple fixed-interval fixed-ratio schedule. Thehatches mark reinforcements. Separating them in some cases are short, steep linesshowingahighconstantrateonafixed-ratioscheduleand,inothers,somewhatlonger“scallops”showingasmoothaccelerationastheorganismshiftsfromaverylowratejustafterreinforcementtoahigherrateattheendofthefixedinterval.Thevaluesofthe intervals and ratios, the states of deprivation, and the exposures to the scheduleswere different in the three cases, but except for these details the curves are quitesimilar.Now,oneofthemwasmadebyapigeoninsomeexperimentsbyFersterandme,onewasmadebyaratinanexperimentonanoxiabyLohr,andthethirdwasmadebyamonkeyinKarlPribram’slaboratoryattheHartfordInstitute.Pigeon,rat,monkey,which is which? It doesn’t matter. Of course, these three species have behavioralrepertoireswhich are as different as their anatomies.But once youhave allowed fordifferences in theways inwhich theymakecontactwith theenvironment,and in theways inwhich theyactupon theenvironment,what remainsof theirbehaviorshowsastonishingly similar properties. Mice, cats, dogs, and human children could haveaddedothercurvestothisfigure.Andwhenorganismswhichdifferaswidelyasthisneverthelessshowsimilarpropertiesofbehavior,differencesbetweenmembersofthesame speciesmay be viewedmore hopefully. Difficult problems of idiosyncrasy orindividualitywillalwaysariseasproductsofbiologicalandculturalprocesses,butitistheverybusinessoftheexperimentalanalysisofbehaviortodevisetechniqueswhichreducetheireffectsexceptwhentheyareexplicitlyunderinvestigation.

135

FIG.14

We arewithin reach of a science of the individual. Thiswill be achieved, not byresorting to some special theory of knowledge in which intuition or understandingtakestheplaceofobservationandanalysis,butthroughanincreasinggraspofrelevantconditionstoproduceorderintheindividualcase.Asecondconsequenceofanimprovedtechnologyistheeffectuponbehaviortheory.

As I have pointed out elsewhere, it is the function of learning theory to create animaginaryworldoflawandorderandthustoconsoleusforthedisorderweobserveinbehavior itself. Scores on a T maze or jumping stand hop about from trial to trialalmostcapriciously.Thereforeweargue that if learning is,aswehope,acontinuousandorderlyprocess,itmustbeoccurringinsomeothersystemofdimensions—perhapsinthenervoussystem,orinthemind,orinaconceptualmodelofbehavior.Boththestatistical treatmentofgroupmeansandtheaveragingofcurvesencouragethebeliefthatwearesomehowgoingbehindtheindividualcasetoanotherwiseinaccessible,butmorefundamental,process.Thewholetenorofourpaperonanxiety,forexample,wastoimplythatthechangeweobservedwasnotnecessarilyapropertyofbehavior,butofsome theoretical state of the organism (“anxiety”) which was merely reflected in aslightmodificationofperformance.Whenwehaveachievedapracticalcontrolovertheorganism,theoriesofbehavior

losetheirpoint.Inrepresentingandmanagingrelevantvariables,aconceptualmodelisuseless; we come to grips with behavior itself. When behavior shows order and

136

consistency,wearemuchlesslikelytobeconcernedwithphysiologicalormentalisticcauses. A datum emerges which takes the place of theoretical fantasy. In theexperimentalanalysisofbehaviorweaddressourselvestoasubjectmatterwhichisnotonlymanifestly thebehaviorofan individualandhenceaccessiblewithout theusualstatistical aids but also “objective” and “actual” without recourse to deductivetheorizing.Statistical techniques serve a useful function, but they have acquired a purely

honorific statuswhichmaybe troublesome.Theirpresenceorabsencehasbecomeashibbolethtobeusedindistinguishingbetweengoodandbadwork.Becausemeasuresof behavior have been highly variable, we have come to trust only results obtainedfrom large numbers of subjects. Because some workers have intentionally orunconsciouslyreportedonlyselectedfavorableinstances,wehavecometoputahighvalue on researchwhich is planned in advance and reported in its entirety. Becausemeasureshavebehavedcapriciously,wehavecometovalueskillfuldeductivetheorieswhich restore order. But although large groups, planned experiments, and validtheorizing are associated with significant scientific results, it does not follow thatnothingcanbe achieved in their absence.Here are twobrief examplesof the choicebeforeus.Howcanwedeterminethecourseofdarkadaptationinapigeon?Wemoveapigeon

fromabrightlighttoadarkroom.Whathappens?Presumablythebirdisabletoseefainterandfainterpatchesoflightastheprocessofadaptationtakesplace,buthowcanwe follow this process? One way would be to set up a discrimination apparatus inwhich choices would be made at specific intervals after the beginning of darkadaptation. The test patches of light could be varied over a wide range, and thepercentagesofcorrectchoicesateachvaluewouldenableuseventually to locate thethresholdfairlyaccurately.Buthundredsofobservationswouldbeneededtoestablishonly a few points on the curve and to prove that these show an actual change insensitivity.IntheexperimentbyBloughalreadymentioned,thepigeonholdsaspotoflightclosetothethresholdthroughouttheexperimentalperiod.Asinglecurve,suchasthe one sketched in Figure15, yields asmuch information as hundreds of readings,togetherwiththemeansandstandarddeviationsderivedfromthem.Theinformationismoreaccuratebecauseitappliestoasingleorganisminasingleexperimentalsession.Yetmanypsychologistswhowouldacceptthefirstasafinishedexperimentbecauseofthe tables ofmeans and standard deviationswould boggle at the second or call it apreliminary study. The direct evidence of one’s senses in observing a process ofbehaviorisnottrusted.

137

FIG.15

Asanotherexample,considerthebehaviorofseveraltypesofobesemice.Dotheyall suffer froma singleabnormality in their eatingbehaviororare theredifferences?Onemightattempttoanswerthiswithsomesuchmeasureofhungerasanobstructionapparatus.Thenumbers of crossings of a grid to get to food, counted after differentperiodsoffreeaccesstofood,wouldbethedata.Largenumbersofreadingswouldbeneeded,andtheresultingmeanvalueswouldpossiblynotdescribethebehaviorofanyonemouseinanyexperimentalperiod.Amuchbetterpicturemaybeobtainedwithonemouse of each kind in single experimental sessions, as Anliker has shown.6 In anexperiment reported roughly in Figure16, each mouse was reinforced with a smallpiece of food after completing a short “ratio” of responses.The hypothalamic-obesemouse shows an exaggerated but otherwise normal ingestion curve. The hereditary-obesemouseeatsslowlybutforanindefinitelengthoftimeandwithlittlechangeinrate.Thegold-poisonedobesemouseshowsasharposcillationbetweenperiodsofveryrapidrespondingandnorespondingatall.Thesethreeindividualcurvescontainmoreinformationthancouldprobablyeverbegeneratedwithmeasuresrequiringstatisticaltreatment,yettheywillbeviewedwithsuspicionbymanypsychologistsbecausetheyaresinglecases.

138

FIG.16

Itisperhapsnaturalthatpsychologistsshouldawakenonlyslowlytothepossibilitythatbehavioralprocessesmaybedirectlyobserved,orthattheyshouldonlygraduallyputtheolderstatisticalandtheoreticaltechniquesintheirproperprospective.Butitistime to insist that science does not progress by carefully designed steps called“experiments” each of which has a well-defined beginning and end. Science is acontinuousandoftenadisorderlyandaccidentalprocess.Weshallnotdo theyoungpsychologist any favor if we agree to reconstruct our practices to fit the patterndemandedbycurrentscientificmethodology.Whatthestatisticianmeansbythedesignof experiments is design which yields the kind of data to which his techniques areapplicable.Hedoesnotmean thebehaviorof the scientist inhis laboratorydevisingresearchforhisownimmediateandpossiblyinscrutablepurposes.The organismwhose behavior is most extensivelymodified andmost completely

controlled in research of the sort I have described is the experimenter himself. Thepointwaswellmadebyacartoonist in theColumbiaJester(Figure17).Thecaptionread:“Boy,haveIgotthisguyconditioned!EverytimeIpressthebardownhedropsinapieceoffood.”Thesubjectswestudyreinforceusmuchmoreeffectivelythanwereinforcethem.IhavebeentellingyousimplyhowIhavebeenconditionedtobehave.Andofcourse it isamistake toargue toomuchfromonecasehistory.Mybehaviorwouldnothavebeenshapedasitwaswereitnotforpersonalcharacteristicswhichallpsychologists fortunately do not share. Freud has had something to say about themotivation of scientists and has given us some insight into the type of person whoachievesthefullestsatisfactionfrompreciseexperimentaldesignandtheintricaciesofdeductive systems. Such a person tends to bemore concernedwith his success as ascientistthanwithhissubjectmatter,asisshownbythefactthatheoftenassumestheroleofarovingambassador.Ifthisseemsunfair,letmehastentocharacterizemyownmotivationinequallyunflatteringterms.SeveralyearsagoIspentapleasantsummer

139

writinganovelcalledWaldenTwo.Oneof thecharacters,Frazier, saidmany thingswhichIwasnotyetreadytosaymyself.Amongthemwasthis:

FIG.17

Ihaveonlyoneimportantcharacteristic,Burris:I’mstubborn.I’vehadonlyoneideainmylife—atrueidéefixe…toputitasbluntlyaspossible,theideaofhavingmyownway. “Control” expresses it, I think. The control of human behavior, Burris. Inmyearlyexperimentaldays itwasa frenzied,selfishdesire todominate. I remember therageIusedtofeelwhenapredictionwentawry.Icouldhaveshoutedatthesubjectsofmyexperiments,“Behave,damnyou,behaveasyouought!”EventuallyIrealizedthatthesubjectswerealwaysright.Theyalwaysbehavedastheyought.ItwasIwhowaswrong.Ihadmadeabadprediction.

(InfairnesstoFrazierandtherestofmyself,Iwanttoaddhisnextremark:“Andwhatastrangediscoveryforawould-betyrant,thattheonlyeffectivetechniqueofcontrolisunselfish.”Fraziermeans,ofcourse,positivereinforcement.)We have no more reason to say that all psychologists should behave as I have

behaved than that they should all behave like R. A. Fisher. The scientist, like anyorganism, is the product of a unique history. The practices which he finds mostappropriatewilldependinpartuponthishistory.Fortunately,personal idiosyncrasiesusuallyleaveanegligiblemarkonscienceaspublicproperty.Theyareimportantonlywhenwe are concernedwith the encouragement of scientists and the prosecution ofresearch.WhenwehaveatlastanadequateempiricalaccountofthebehaviorofManThinking,we shall understand all this.Until then, itmaybebest not to try to fit allscientistsintoanysinglemold.

140

1Richter,C.P.Freeresearchversusdesign.Science,1953,118,91–93.2Heron,W.T.,&Skinner,B.F.Anapparatusforthestudyofbehavior.Psychol.Rec.,1939,3,166–176.

3Breland,K.,&Breland,Marion.Afieldofappliedanimalpsychology.Amer.Psychologist,1951,6,202–204.

4Ratliff,F.,&Blough,D.S.Behavioralstudiesofvisualprocessesinthepigeon.ReportofContractN50ri-07663,PsychologicalLaboratories,HarvardUniversity,September,1954.

5Lindsley,O.R.Operantconditioningmethodsappliedtoresearchinchronicschizophrenia.Psychiat.Res.Rep.,1956,5,118–139.

6Anliker,J.,andMayer,J.Operantconditioningtechniqueforstudyingfeedingpatternsinnormalandobesemice.J.Appl.Physiol.,1956,8,667–670.

141

TheExperimentalAnalysisofBehavior

A Sigma Xi National Lecture, given in November and December of 1956 beforethirty-onechaptersoftheSocietyoftheSigmaXiortheResearchEngineersSocietyofAmerica,thefollowingpaperwaswrittenprimarilyforthenon-psychologicalscientist.ItwaspublishedintheAmericanScientist(1957,45,343-371)and is reprintedwithpermission.

Notsolongagotheexpression“ascienceofbehavior”wouldhavebeenregardedasacontradictioninterms.Livingorganismsweredistinguishedbythefactthattheywerespontaneous andunpredictable. If you saw somethingmovewithoutbeingobviouslypushedorpulled,youcouldbeprettysureitwasalive.Thiswassomuchthecasethatmechanical imitations of living things—singing birds which flapped their wings,figures on a clock tolling a bell—had an awful fascination which, in the age ofelectronic brains and automation, we cannot recapture or fully understand. Onehundredand fiftyyearsof scienceand inventionhave robbed livingcreaturesof thishighdistinction.Sciencehasnotdonethisbycreatingtrulyspontaneousorcapricioussystems.Ithas

simplydiscoveredandusedsubtleforceswhich,actinguponamechanism,giveitthedirection and apparent spontaneity which make it seem alive. Similar forces weremeanwhilebeingdiscoveredinthecaseofthelivingorganismitself.Bythemiddleoftheseventeenthcenturyitwasknownthatmuscle,excisedfromalivingorganismandout of reach of any “will,” would contract if pinched or pricked or otherwisestimulated, and during the nineteenth century larger segments of the organismweresubmittedtoasimilaranalysis.Thediscoveryofthereflex,apartfromitsneurologicalimplications, was essentially the discovery of stimuli—of forces acting upon anorganismwhichaccountedforpartofitsbehavior.For a long time the analysis of behavior took the form of the discovery and

collectionof reflexmechanisms.Early in thepresentcentury, theDutchphysiologistRudolph Magnus,1 after an exhaustive study of the reflexes involved in themaintenance of posture, put the matter this way: when a cat hears a mouse, turnstowardthesourceofthesound,seesthemouse,runstowardit,andpounces,itspostureat every stage, even to the selection of the foot which is to take the first step, isdetermined by reflexes which can be demonstrated one by one under experimentalconditions. All the cat has to do is to decide whether or not to pursue the mouse;everythingelseispreparedforitbyitsposturalandlocomotorreflexes.To pursue or not to pursue is a question, however, which has never been fully

answered on the model of the reflex, even with the help of Pavlov’s principle ofconditioning.Reflexes—conditioned or otherwise—are primarily concernedwith theinternaleconomyof theorganismandwithmaintainingvarious sortsofequilibrium.Thebehaviorthroughwhichtheindividualdealswiththesurroundingenvironmentandgetsfromitthethingsitneedsforitsexistenceandforthepropagationofthespecies

142

cannotbeforcedintothesimpleall-or-nothingformulaofstimulusandresponse.Somewell-defined patterns of behavior, especially in birds, fish, and invertebrates arecontrolledby“releasers”whichsuggestreflexstimuli,2butevenheretheprobabilityofoccurrence of such behavior varies over amuchwider range, and the conditions ofwhichthatprobabilityisafunctionaremuchmorecomplexandsubtle.Andwhenwecome to that vast repertoire of “operant” behavior which is shaped up by theenvironmentinthelifetimeoftheindividual,thereflexpatternwillnotsufficeatall.Instudyingsuchbehaviorwemustmakecertainpreliminarydecisions.Webeginby

choosing an organism—onewhichwe hopewill be representative butwhich is firstmerely convenient.Wemust also choose a bit of behavior—not for any intrinsic ordramaticinterestitmayhave,butbecauseitiseasilyobserved,affectstheenvironmentinsuchawaythatitcanbeeasilyrecorded,andforreasonstobenotedsubsequentlymayberepeatedmanytimeswithoutfatigue.Thirdly,wemustselectorconstructanexperimentalspacewhichcanbewellcontrolled.These requirements are satisfied by the situation shown in Figure 1. A partially

sound-shielded aluminum box is divided into two compartments. In the nearcompartment a pigeon, standing on a screen floor, is seen in the act of pecking atranslucentplasticplatebehindacircularopeninginthepartition.Theplateispartofadelicate electric key;when it is pecked, a circuit is closed to operate recording andcontrolling equipment. Colored lights can be projected on the back of the plate asstimuli.Theboxisventilated,andilluminatedbyadimceilinglight.Weareinterestedintheprobabilitythatinsuchacontrolledspacetheorganismwe

selectwillengageinthebehaviorwethusrecord.Atfirstblush,suchaninterestmayseemtrivial.Weshallsee,however,thattheconditionswhichaltertheprobability,andtheprocesseswhichunfoldasthatprobabilitychanges,arequitecomplex.Moreover,theyhavean immediate, importantbearingon thebehaviorofotherorganismsunderother circumstances, including the organism called man in the everyday world ofhumanaffairs.

143

FIG.1.Anexperimentalspaceshowingapigeonintheactofpeckingaplastickey.

Probabilityofrespondingisadifficultdatum.Wemayavoidcontroversialissuesbyturningatoncetoapracticalmeasure,thefrequencywithwhicharesponseisemitted.TheexperimentalsituationshowninFigure1wasdesignedtopermitthisfrequencytovary over a wide range. In the experiments to be described here, stable rates arerecorded which differ by a factor of about 600. In other experiments, rates havedifferedbyasmuchas2000:1.Rateofrespondingismostconvenientlyrecordedinacumulativecurve.Apenmovesacrossapapertape,steppingashortuniformdistancewitheachresponse.Appropriatepaperspeedsandunitstepsarechosensothattheratestobestudiedgiveconvenientslopes.

OperantConditioning

144

Amongtheconditionswhichalterrateofrespondingaresomeoftheconsequencesof behavior.Operant behavior usually affects the environment and generates stimuliwhich“feedback”totheorganism.Somefeedbackmayhavetheeffectsidentifiedbythe layman as reward and punishment. Any consequence of behavior which isrewarding or, more technically, reinforcing increases the probability of furtherresponding. Unfortunately, a consequence which is punishing has a much morecomplex result.3 Pecking the key in our experimental space has certain naturalconsequences.Itstimulatesthebirdtactuallyandaurally,andsuchstimulationmaybeslightly reinforcing.Westudy theeffectmoreexpediently,however,byarranginganarbitrary consequence which is clearly reinforcing. Food is reinforcing to a hungrypigeon(forourpresentpurposesweneednotinquirewhythisisso),andwethereforearrangetopresentfoodwithaspecialmagazine.Whenasolenoidisenergized,atraycontainingamixtureofgrainsisbroughtintopositioninthesquareopeningbelowthekeyinFigure1,wherethepigeonhasaccesstothegrainfor,say,fourseconds.We can demonstrate the effect of operant reinforcement simply by connecting the

key which the pigeon pecks to the solenoid which operates the food tray. A singlepresentationof food, following immediatelyupona response, increases the ratewithwhich responses to the key are subsequently emitted so long as the pigeon remainshungry. By reinforcing several responses, we may create a high probability ofresponding.Ifthemagazineisnowdisconnected,theratedeclinesto,andmayevengobelow,itsoriginallevel.Thesechangesaretheprocessesofoperantconditioningandextinction,respectively.Moreinterestingphenomenaaregeneratedwhenresponsesaremerely intermittently reinforced. It is characteristic of everyday life that few of thethingswe do always “pay off.” The dynamic characteristics of our behavior dependupontheactualschedulesofreinforcement.The effects of intermittent reinforcement have been extensively studied in the

laboratory.4Acommon sort of intermittency is basedon time.Reinforced responsescanbespaced,say,tenminutesapart.Whenonereinforcementisreceived,atimerisstartedwhichopensthereinforcingcircuitfortenminutes;thefirstresponseafterthecircuit is closed is reinforced. When an organism is exposed to this schedule ofreinforcementformanyhours,itdevelopsacharacteristicperformancewhichisrelatedin a rather complex way to the schedule. A short sample of such a performance isshown in Figure 2, obtained with a cumulative recorder. The scales and a fewrepresentative speeds are shown in the lower right-hand corner. The experimentalsessionbeginsata.Thefirstreinforcementwillnotoccuruntiltenminuteslater,andthebirdbeginsataverylowrateofresponding.Asthe10-minuteintervalpasses,therateincreases,acceleratingfairlysmoothlytoaterminalrateatreinforcementatb.Theratethendropstozero.Exceptforaslightabortivestartatc, itagainacceleratestoahighterminalvaluebytheendofthesecond10-minuteinterval.Athirdfairlysmoothaccelerationisshownatd.(Atethepeninstantlyresetstothestartingpositiononthepaper.)Theover-all patternofperformanceona “fixed-interval” schedule is a fairlysmoothly accelerating scallop in each interval, the acceleration beingmore rapid thelonger the initial pause. Local effects due to separate reinforcements are evident,however,whichcannotbediscussedhereforlackofspace.

145

FIG.2.Characteristicperformancebypigeonunderfixed-intervalreinforcement.

If the intervals between reinforcements are not fixed, the performance shown inFigure 2 cannot develop. If the length of interval is varied essentially at random,responding occurs at a single rate represented by a constant slope in the cumulativerecord.TwoexamplesareshowninFigure3. In theuppercurve,ahungrypigeon isreinforcedwithgrainonavariable-intervalschedule,wherethemeanintervalbetweenreinforcementsis3minutes.Reinforcementsoccurwheremarkedbypips.Inthelowercurve a hungry chimpanzee, operating a toggle switch, is reinforced on the sameschedulewithlaboratoryfood.Theover-allrateundervariable-intervalreinforcementis a function of the mean interval, the level of food-deprivation, and many othervariables. It tends to increase slowly under prolonged exposure to any one set ofconditions.Theconstant rate itself eventuallybecomesan importantconditionof theexperimentandresistsanychangetoothervalues.ForthisreasonthestraightlinesofFigure3arenotassuitableforbaselinesasmightbesupposed.

146

FIG.3.Performanceundervariable-intervalreinforcementforapigeon(uppercurve)andachimpanzee(lowercurve).

Reinforcementsmaybe scheduledwithacounter insteadofa timer.Forexample,wemaymaintainafixedratiobetweenresponsesandreinforcements. In industry thisscheduleisreferredtoaspieceworkorpiece-ratepay.Anyonewhohasseenworkerspaidonsuchascheduleisfamiliarwithsomefeaturesoftheperformancegenerated:ahighrateissustainedforlongperiodsoftime.Forthisreason,thescheduleisattractiveto employers, but it is generally recognized that the level of activity generated ispotentiallydangerousandjustifiedonlyinseasonalorotherperiodicemployment.Performancesofapigeonunderfixed-ratioreinforcementareshowninFigure4. In

the left-hand record reinforcements occur every 210 responses (at a, b, c, andelsewhere). The over-all rate is high. Most of the pauses occur immediately afterreinforcement.Attherightistheperformancegeneratedwhenthepigeonpecksthekey900 times for each reinforcement. This unusually high ratio was reached in someexperiments in the Harvard Psychological Laboratories by W. H. Morse and R. J.Herrnstein.Ashortpauseafterreinforcementistherule.

147

FIG.4.Typicalperformancebyapigeonunderfixed-ratioreinforcement.Attheleftevery210thresponseisreinforced;attherightevery900thresponse.

Avariable-ratio schedule programmed by a counter corresponds to the variable-interval schedule programmed by a timer. Reinforcement is contingent on a givenaveragenumberofresponsesbutthenumbersareallowedtovaryroughlyatrandom.Weareall familiarwith this schedulebecause it is theheartofallgamblingdevicesandsystems.Theconfirmedorpathologicalgamblerexemplifiestheresult:averyhighrateofactivityisgeneratedbyarelativelyslightnetreinforcement.Wherethe“cost”ofaresponsecanbeestimated(in terms,say,of thefoodrequired tosupply theenergyneeded,orofthemoneyrequiredtoplaythegamblingdevice),itmaybedemonstratedthatorganismswilloperateatanetloss.When the food magazine is disconnected after intermittent reinforcement, many

responses continue to occur in greater number and for a longer time than aftercontinuousreinforcement.Aftercertainschedules,theratemaydeclineinasmoothlyaccelerated extinction curve. After other schedules, when the rate itself entersprominently into the experimental conditions, it may oscillate widely. The potentialresponding built up by reinforcement may last a long time. We have obtainedextinctioncurvessixyearsafterprolongedreinforcementonavariable-ratioschedule.5Ratio schedules characteristically produce large numbers of responses in extinction.After prolonged exposure to a ratio of 900:1 (Figure 4) the bird was put in theapparatuswiththemagazinedisconnected.Duringthefirst4½hoursitemitted73,000responses.Interval and ratio schedules have different effects for several reasons. When a

reinforcement is scheduled by a timer, the probability of reinforcement increasesduring any pause, and the first responses after pauses are especially likely to bereinforced.Onratioschedules responseswhicharepartofshort runsare likely tobereinforced.Moreover,whenagivenscheduleofreinforcementhashadafirsteffect,theperformance which develops becomes itself an important part of the experimentalsituation. This performance, in combination with the schedule, arranges certainprobableconditionsatthemomentofreinforcement.Sometimesascheduleproducesaperformancewhichmaintainsjustthoseconditionswhichperpetuatetheperformance.

148

Some schedules generate a progressive change. Under still other schedules thecombination of schedule and performance yields conditions at reinforcement whichgenerateadifferentperformance,which in turnproducesconditionsat reinforcementwhichrestoretheearlierperformance.InSchedulesofReinforcementCharlesB.FersterandIcheckedthisexplanationof

the effect of schedules by controlling conditions more precisely at the moment ofreinforcement. For example, we guaranteed that all reinforced responses would beprecededbypausesinsteadofmakingthisconditionmerelyprobableunderanintervalschedule.Inavariable-intervalperformance,suchasthatshowninFigure3, it isnotdifficulttofindresponseswhichareprecededby,say,3-secondpauses.Wecanarrangethat only such responses will be reinforced without greatly disturbing the schedule.Whenthis isdone, theslopeoftherecordimmediatelydrops.Ontheotherhand,wemaychoosetoreinforceresponseswhichoccurduringshortrapidburstsofresponding,andwethennoteanimmediateincreaseinrate.If we insist upon a very long pause, wemay be able to reinforce every response

satisfying these conditions and still maintain a very low rate. The differentialreinforcement of low rates was first studied by Douglas Anger in the HarvardLaboratories. Wilson and Keller at Columbia have reported an independentinvestigation.6RecentlyW.H.MorseandIhavestudiedtheeffectofrelatively longenforced pauses. Figure5 shows the performance obtained in one such experiment.Any responsewhich followed a pause at least 3minutes in durationwas reinforced.Wheneveraresponsewasmadebefore3minuteshadelapsed,thetimerwasresetandanother 3-minute pause required. Under these conditions a very low stable rate ofrespondingobtains.Thefigureshowsacontinuousperformance(cutintosegmentsforeasierreproduction)inasingleexperimentalsessionof143hours,duringwhichtimethe pigeon received approximately 250 reinforcements. At no time did it pause formorethan15minutes,anditseldompausedformorethan5minutes.

FIG.5.Veryslowsustainedrespondingwhenonlyresponsesprecededbya3-minutepausearereinforced.

Thesituationunderthisscheduleisinherentlyunstable.Rateofrespondingincreaseswiththeseverityoffooddeprivationanddecreasesasthebirdbecomessatiated.Letusassumethatatsometimeduringtheexperiment—say,atainFigure5—reinforcementsareoccurringtooinfrequentlytomaintainthebird’sweight.Thebirdisoperating,sotospeak,ata loss.The increasingdeprivation then increases the rateof respondingandmakes it even less likely that the pigeon will wait 3 minutes in order to respond

149

successfullyforreinforcement.Nothingbutstarvationliesaheadinthatdirection.If,onthe other hand, the bird is receiving slightly more reinforcements than necessary tomaintainbodyweight, thelevelofdeprivationwillbedecreased.Thiswillproducealower rate of responding, which in turn means that the 3-minute pause is morefrequently satisfied and reinforcements stillmore frequently received. In such a casetheresultisafullysatiatedbird,andtheexperimentmustbebroughttoaclose.ThisactuallyhappenedatbinFigure5,wherereinforcementshadbecomesofrequentthatthe bird was rapidly gaining weight. This inherent instability can be corrected bychangingtherequiredpauseintermsoftheorganism’sperformance.Iftheover-allrateof reinforcement begins to drift in either direction, the required pause may beappropriatelychanged.ThustheexperimentinFigure5couldhavebeencontinuedifatpointc,say, therequiredintervalhadbeenincreasedto4minutes.Byanappropriateadjustmentoftheinterval,wehavebeenabletokeepapigeonrespondingcontinuouslyfor1500hours—that is,24hoursaday,7daysaweek,forapproximately2months.WendellLevi7hasadvanced the thesis thatpigeonsneversleep(roosting ismerelyaprecautionarydeviceagainstblindflying),andthestatementseemstobeconfirmedbyexperimentsofthepresentsort.Bydifferentially reinforcing high rates of responding, pigeons have beenmade to

respondasrapidlyas10to15responsespersecond.Heretechnicalproblemsbecomecrucial.Itisnotdifficulttoconstructakeywhichwillfollowrapidresponding,butthetopographyof the behavior itself changes.The excursions of head and beak becomeverysmall,anditisdoubtfulwhetheranysingle“response”canbeproperlycomparedwitharesponseatalowerrate.Inourstudyofdifferentkindsofschedulesofreinforcement,FersterandIfoundthat

itwaspossibletosetupseveralperformancesinasinglepigeonbybringingeachoneofthemunderstimuluscontrol.Severaldifferentcoloredlightswereprojectedonthetranslucent key, and responses were reinforced on several corresponding schedules.Figure6showsatypicalperformanceundersuchamultiplescheduleofreinforcement.When the key was red, the pigeon was reinforced on a 6-minute fixed-intervalschedule.Theusualintervalscallopsareseen,asataandb.Whenthekeywasgreen,thepigeonwas reinforceduponcompleting60 responses (a fixed ratioof60:1).Theusualhighratiorateisshownasatcandd.Whenthekeywasyellow,reinforcementsfollowed a variable-interval schedulewhere a pause of 6 secondswas required. Theresultinglowsteadyperformanceisshownate,f,andelsewhere.Inoneexperimentwewere able to show nine different performances under the control of nine differentpatterns on the key. (The performance shown in Figure5 is actually one part of amultipleschedule.Onceperhourduringthe154-hoursessionthekeycolorchangedtooneof twoother colors. Inone case a single ratiowas runoff, in theother a singleinterval.All reinforcementsare indicated inFigure5. The number of reinforcementsobtainedunderthedifferentialreinforcementofalowratewasthereforemuchsmallerthanFigure5 indicates.The additional scheduleswere inserted in an effort todetectincreasing“mentalfatigue”duringsuchalongsustainedsession.)

150

FIG.6.Performanceunderamultiplescheduleofreinforcement.

Theexperimentmaybecomplicatedstillfurtherbyintroducingmorethanonekeyandbyreinforcingontwoormoreschedulesconcurrently.Anexampleoftheresultingperformances is shown in Figure 7, from some research by Ferster, at the YerkesLaboratories forPrimateBiologyatOrangePark,Florida. InFerster’s experiment, achimpanzee operates two toggle switches, one with each hand. Responses with therighthandarereinforcedonafixedratioofapproximately210:1,andtheperformancerecordedfromtherighttoggleswitchisshownintheupperpartofFigure7.Asusualinmanyratioperformances,pausesoccurafterreinforcements.Responseswiththelefthandareatthesametimebeingreinforcedonavariable-intervalschedulewithameaninterval of 5minutes, and the performance is shown in the lower part of the figure.Thereissomeinteractionbetweentheperformances,forreinforcementsinthevariable-intervalrecordusuallycorrespondtoslightpausesintheratioperformance.Ingeneral,however,theexperimentshowsaremarkableindependenceoftworesponsesystemsinasingleorganism.

151

FIG.7.Simultaneousrecordsoftheperformancesbyachimpanzeereinforcedforoperatingatoggleswitchwithitsrighthandonafixed-ratioscheduleandforoperatinganotherswitchwith

itslefthandonavariable-intervalschedule(lowercurve).

StimulusControl

Inspeakingaboutcolorsprojectedonthekeyorthefact thatakeyisontherightorleft,weare,ofcourse,talkingaboutstimuli.Moreover,theyarestimuliwhichactpriortotheappearanceofaresponseandthusoccurinthetemporalordercharacteristicofthereflex.Buttheyarenotelicitingstimuli;theymerelymodifytheprobabilitythataresponsewilloccur,andtheydothisoveraverywiderange.Thegeneralruleseemstobe that the stimuli present at the moment of reinforcement produce a maximalprobabilitythattheresponsewillberepeated.Anychangeinthestimulatingsituationreducestheprobability.ThisrelationshipisbeautifullyillustratedinsomeexperimentsbyNormanGuttman8andhiscolleaguesatDukeUniversityontheso-calledstimulusgeneralizationgradient.Guttmanmakesuseofthefactthat,afterabriefexposuretoavariable-intervalschedule,alargenumberofresponseswillbeemittedbytheorganismwithout further reinforcement (the usual extinction curve) and that, while thesearebeing emitted, it is possible tomanipulate the stimuli present and to determine theirrelativecontrolovertheresponsewithoutconfusingtheissuebyfurtherreinforcement.Inatypicalexperiment,forexample,amonochromaticlightwithawavelengthof550millimicronswasprojectedon thekeyduringvariable-interval reinforcement.Duringextinction, monochromatic lights from other parts of the visible spectrum wereprojectedonthekeyforshortperiodsoftime,eachwavelengthappearingmanytimes

152

and each being present for the same total time. Simply by counting the number ofresponsesmadeinthepresenceofeachwavelength,GuttmanandhiscolleagueshaveobtainedstimulusgeneralizationgradientssimilartothoseshowninFigure8.Thetwocurves represent separateexperiments.Each isanaverageofmeasurementsmadeonsix pigeons. It will be seen that during extinction respondingwasmost rapid at theoriginalwave length of 550millimicrons.A color differing byonly 10millimicronscontrols a considerably lower rate of responding. The curves are not symmetrical.Colors toward the red end of the spectrum control higher rates than those equallydistantonthevioletend.WiththistechniqueGuttmanandhiscolleagueshavestudiedgradients resulting from reinforcement at two points in the spectrum, gradientssurviving after a discrimination has been set up by reinforcing onewave length andextinguishinganother,andsoon.

FIG.8.Stimulusgeneralizationgradientsgivenbypigeonsreinforcedunderamonochromaticlightof550millimicronsandthenextinguishedundertheotherwavelengthsshown(Guttman).

153

Thecontrolofbehaviorachievedwithmethodsbaseduponrateof respondinghasgivenrisetoanewpsychophysicsoflowerorganisms.Itappearstobepossibletolearnas much about the sensory processes of the pigeon as from the older introspectivemethodswithhumansubjects.AnimportantnewtechniqueofthissortisduetoD.S.Blough.9HisingeniousprocedureutilizestheapparatusshowninFigure9.Apigeon,behavingmostof the time in totaldarkness, thrusts itshead throughanopening inapartitionata,whichprovidesusefultactualorientation.Throughthesmallopeningb,the pigeon can sometimes see a faint patch of light indicated by thewordStimulus.(Howthisappearstothepigeonisshownattheright.)Thepigeoncanreachandpecktwokeysjustbelowtheopeningb,anditissometimesreinforcedbyafoodmagazinewhich rises within reach at c. Through suitable reinforcing contingencies BloughconditionsthepigeontopeckKeyBwheneveritcanseethelightandKeyAwheneveritcannot.ThepigeonisoccasionallyreinforcedforpeckingKeyAbythepresentationoffood(indarkness).Bloughguaranteesthatthepigeoncannotseethespotoflightatthe time this response ismadebecauseno light at all is thenon thekey.Byawell-establishedprincipleof“chaining,”thepigeonisreinforcedforpeckingKeyBbythedisappearance of the spot of light. This suffices to keep responses to both keys instrength.AfurtherfactabouttheapparatusisthatKeyBautomaticallyreducestheintensity

ofthespotoflight,whileKeyAincreasesit.Suppose,now,thatapigeonisplacedinabrightly lighted space for agiven intervalof timeand thenput immediately into theapparatus.Thespotoflightisatanintensityintheneighborhoodofthebright-adaptedthreshold.Ifthepigeoncanseethespot,itpecksKeyBuntilitdisappears.Ifitcannotseethespot,itpecksKeyAuntilitappears.Ineachcaseitthenshiftstotheotherkey.During an experimental session of one hour ormore, it holds the spot of light veryclose to its thresholdvalue,occasionallybeingreinforcedwithfood.Theintensityofthe light is recorded automatically. The result is the “dark-adaptation curve” for thepigeon’seye.Typicalcurvesshowabreakas thedark-adaptationprocessshiftsfromtheconeelementsintheretinatotherods.

154

FIG.9.Blough’sapparatusforthestudyofdark-adaptationandspectralsensitivityinthepigeon.

Byrepeatingtheexperimentwithaseriesofmonochromaticlights,Bloughhasbeenabletoconstructspectralsensitivitycurvesforthepigeonwhichareaspreciseasthoseobtainedwiththebesthumanobservers.AnexampleisshowninFigure10,wheredatafor thepigeon are comparedwithdata for an aphakichuman10—onewhohashadacrystalline lens removed for medical reasons. Such a person sees violet light moresensitively than normal subjects because the light is not absorbed by the lens. Evenwiththisadvantagethehumanobserverisnomoresensitivetolightatthevioletendofthespectrumthanthepigeon.Thediscontinuitiesinthephotopiccurves(thelowersetofopencircles)ofthepigeonappeartobereal.Thesurprisingcorrespondenceinthescotopic curves (after dark adaptation, and presumably mediated by the rods) isremarkablewhenwerecall that theavianandmammalianeyepartedcompanyin theevolutionaryscaleofthingsmanymillionsofyearsago.

155

FIG.10.Spectralsensitivitycurvesforscotopic(uppercurves)andphotopic(lowercurves)visioninthepigeonandaphakichuman.

156

Avoidance

Sofarourdatahavebeentakenfromthepleasantersideoflife—frombehaviorwhichproduces positive consequences. There are important consequences of another sort.Muchofwhatwedoduringthedayisdonenotbecauseofthepositivereinforcementswereceivebutbecauseofaversiveconsequencesweavoid.Thewholefieldofescape,avoidance,andpunishmentisanextensiveone,butorderisslowlybeingbroughtintoit.AnimportantcontributionhasbeentheresearchofMurraySidman11onavoidancebehavior. In the Sidman technique, a rat is placed in a box the floor ofwhich is anelectricgridthroughwhichtheratcanbeshocked.Thepatternofpolarityofthebarsofthegridischangedseveraltimespersecondsothattheratcannotfindbarsofthesamesigntoavoidtheshock.Inatypicalexperimentashockoccursevery20secondsunlesstheratpressesthelever,butsucharesponsepostponestheshockforafull20seconds.These circumstances induce a rat to respond steadily to the lever, the onlyreinforcementbeing thepostponementof shock.The ratmust occasionally receive ashock—thatis,itmustallow20secondstopasswithoutaresponse—ifthebehavioristo remain in strength. By varying the intervals between shocks, the time ofpostponement,andvariouskindsofwarningstimuli,Sidmanhasrevealedsomeoftheimportantpropertiesofthisall-too-commonformofbehavior.AsampleofbehaviorwhichW.H.MorseandthewriterobtainedwiththeSidman

procedure is shown in Figure 11. Here both the interval between shocks and thepostponementtimewere8seconds.(Whitespacehasbeencutoutoftherecordandtheseparatesegmentsbroughttogethertofacilitatereproduction.)Therecordsreporta7-hour experimental session during which about 14,000 responses were emitted.Occasional shocks are indicated by the downwardmovements of the pen (not to beconfusedwiththefragmentsoftheresetline).Asignificantfeatureoftheperformanceis the warm-up at a.When first put into the apparatus the rat “takes” a number ofshocksbeforeenteringuponthetypicalavoidancepattern.Thisoccurswheneveranewsessionisbegun.Itmayindicatethatanemotionalconditionisrequiredforsuccessfulavoidancebehavior.Theconditiondisappearsbetweensessionsandmustbereinstated.Thefigureshowsconsiderablevariationinover-allrateandmanylocalirregularities.At times small groups of shocks are taken, suggesting a return to the warm-upcondition.

157

FIG.11.Seven-hourperformanceofaratwhichavoidsashockbypressingaleverundertheSidmanprocedure.

Motivation

Theconsequencesofbehavior,whetherpositiveornegative,andthecontrolacquiredbyvariousstimulirelatedtothemdonotexhaustthevariablesofwhichbehaviorisafunction.Otherslieinthefieldcommonlycalledmotivation.Foodisareinforcementonlytoahungryorganism.Inpracticethismeansanorganismwhoseweighthasbeenreducedsubstantiallybelowitsvalueunderfreefeeding.Reinforcingstimuliarefoundinothermotivationalareas.Respondingtoakeycanbereinforcedwithwaterwhentheorganism is deprived of water, with sexual contact when the organism has beensexually deprived, and so on. The level of deprivation is in each case an importantconditiontobeinvestigated.Howdoesfooddeprivationincreasetherateofeatingorof engaging inbehavior reinforcedwith food?Howdoes satiationhave theoppositeeffect?Thefirststeptowardansweringsuchquestionsisanempiricalstudyofrateofresponding as a function of deprivation. An analysis of the internal mechanismsresponsiblefortherelationsthusdiscoveredmayrequiretechniquesmoreappropriatelyemployedinotherscientificdisciplines.Anexampleofhowthepresentmethodmaybeappliedtoaprobleminmotivationis

an experiment by Anliker and Mayer12 on the familiar and important problem ofobesity.Obeseanimalseatmorethannormal,butjusthowistheiringestivebehaviordisrupted?Anliker andMayer have studied several types of normal and obesemice.Therearestrainsofmice inwhichtheabnormality ishereditary:somemembersofalitter simply grow fat. A normal mouse may be made obese by poisoning it withgoldthioglucose or by damaging the hypothalamus. The food-getting behavior of allthesetypesofobesemicecanbeobservedintheapparatusshowninFigure12.A fatmouseisshowndepressingahorizontalleverwhichprojectsfromthepartitioninthebox.On a fixed-ratio schedule every 25th response produces a small pellet of food,deliveredbythedispenserseenbehindthepartition.Asupplyofwaterisavailableinabottle.Each mouse was studied continuously for several days. The resulting cumulative

curves(Figure13)showstrikingdifferencesamongthepatternsofingestion.CurveCshowsnormalcyclicchangesinrate.Thenon-obesemouseeatsasubstantialpartofits

158

dailyrationinasingleperiod(asataandb)andfortherestofeachdayrespondsonlyatalowover-allrate.Theresultisawave-likecumulativecurvewith24-hourcycles.Amouseof thesamestrainmadeobesebygoldthioglucosepoisoningdoesnotshowthisdailyrhythmbutcontinuestorespondatafairlysteadyrate(CurveA).Theslopeis no higher than parts of Curve C, but the mechanism which turns off ingestivebehavior in a normal mouse appears to be inoperative. Curve B is a fairly similarrecordproducedbyamouseofthesamestrainmadeobesebyahypothalamiclesion.CurvesDandEareforlittermatesfromastraincontaininganhereditary-obesefactor.Eistheperformanceofthenormalmember.CurveD,showingtheperformanceoftheobesemember, differsmarkedly fromCurvesA andB.The hereditary obesemouseeatsataveryhighrateforbriefperiods,whichareseparatedbypausesoftheorderofone or two hours. A different kind of disturbance in the physiological mechanismseemstobeindicated.

FIG.12.Obesemousepressingaleverforfood.

159

FIG.13.Ingestivepatternsforseveraltypesofnormalandobesemice(AnlikerandMayer).

WilliamsandTeitelbaum13have recentlyproduceda fourthkindofobeseanimal,with an apparatus inwhich a ratmust eat a small amount of liquid food to avoid ashock.TheavoidancecontingenciesspecifiedbySidmanand illustrated inFigure11areusedtoinducetherattoingestunusuallylargeamountsofevenunpalatablefood.Aconditionwhichmaybecalled“behavioralobesity”quicklydevelops.

TheEffectsofDrugsonBehavior

Other powerful variables which affect operant behavior are found in the field ofpharmacology.Somedrugswhichaffectbehavior—alcohol,caffeine,nicotine,andsoon—were discovered by accident and have had a long history. Others have beenproducedexplicitlytoyieldsucheffects.Thefieldisanactiveone(partlybecauseofthe importance of pharmacotherapy inmental illness), and available compounds aremultiplying rapidly.Most of the behavioral drugs now available have effectswhichwould be classified in the fields of motivation and emotion. There is no reason,however, why the effects of various contingencies of reinforcement could not besimulated by direct chemical action—why “intelligence” could not be facilitated orconfusion ormental fatigue reduced. In any case, the behavior generated by variouscontingenciesofreinforcement(includingthecontrolofthatbehaviorviastimuli)arethe baselines against whichmotivational and emotional effects are felt. The presenttechniqueforthestudyofoperantbehavioroffersaquantitative,continuousrecordofthe behavior of an individual organism, which is already being widely used—inindustry as well as the research laboratory—in screening psychopharmacologicalcompoundsandinvestigatingthenatureofpharmacologicaleffects.An example is some research by Peter B. Dews,14 of the Department of

Pharmacologyof theHarvardMedicalSchool.Dewshasstudiedtheeffectofcertainsedatives on the pigeon’s performance under a multiple fixed-interval fixed-ratioschedule.Astandardbaselineobtainedinashortdailyexperimentalsessionisshown

160

in theupperhalfofFigure14.Thepigeon is reinforcedon a fixed-interval schedulewhen the key is red and on a fixed-ratio schedule when the key is green, the twoschedulesbeingpresentedintheorder:oneinterval,oneratio,twointervals,tenratios,two intervals, four ratios. In addition to the usual characteristics of the multipleperformance, this brief program shows local effectswhich add to its usefulness as abaseline. For example, the period of slow responding after reinforcement is greaterwhentheprecedingreinforcementhasbeenonaratioschedule—thatis,thescallopsata and b are shallower than those at c and d. The effect of moderate doses ofbarbiturates,bromides, andother sedativesunderamultiple fixed-interval fixed-ratioschedule is to destroy the interval performance while leaving the ratio performanceessentiallyuntouched.ThelowerhalfofFigure14wasrecordedonthedayfollowingtheupperhalf.Threemilligramsofchlorpromazinehadbeeninjected2.5hourspriortotheexperiment.Thetranquilizingeffectofchlorpromazinedevelopsonlywithrepeateddoses;whatisshownhereistheimmediateeffectofadoseofthismagnitude,whichissimilartothatofasedative.Itwillbeseenthattheratiossurvive(ate,f,andg)butthatthe interval performances are greatly disturbed. There is responding where none isexpected,asath,butnotenoughwherearapidrateusuallyobtains.Thisfactprovidesa useful screening test, but it also throws important light on the actual nature ofsedation.The difference between intervals and ratiosmay explain some instances inwhichsedativesappeartohaveinconsistenteffectsonhumansubjects.Theintervalperformanceisalsodamagedbychlorpromazineinadifferenttypeof

compound schedule. Ferster and I have studied the effect of concurrent schedules inwhich two or more controlling circuits set up reinforcements independently. In oneexperimentaratwasreinforcedwithfoodatfixedintervalsof10minutesandalsobytheavoidanceofshock,whereashockoccurredevery20secondsunlesspostponedfor20secondsbyaresponse toa lever.Thenormalresultof thisconcurrentschedule isshownintheupperpartofFigure15.Whentheratis“workingforfoodandtoavoidashock,”itsperformancesuggeststheusualintervalscalloptiltedupwardsothatinsteadof pausing after reinforcement, the rat responds at a rate sufficient to avoid mostshocks.A one-milligramdose of chlorpromazine immediately before the experimenthas the effect shown in the lower part of the figure. The interval performance iseliminated, leaving the slow steady responding characteristic of avoidanceconditioning.

161

FIG.14.Performanceofapigeononamultiplefixed-intervalfixed-ratioscheduleinacontrol(uppercurve)andafteraninjectionofchlorpromazine(lowercurve).

Drugswhichalteremotionalconditionsmaybestudiedbyexaminingtheeffectofthe emotional variable upon operant behavior. An example is the condition usuallycalled“anxiety.”ManyyearsagoEstesandI (seepage558) showed that thenormalperformance under fixed-interval reinforcementwas suppressed by a stimuluswhichcharacteristicallyprecededashock.Inourexperiment,aratwasreinforcedonafixed-intervalscheduleuntilastablebaselinedeveloped.Astimuluswasthenintroducedfor3minutes and followed by a shock to the feet of the rat. In later presentations thestimulusbegantodepresstherateofresponding—aneffectcomparabletothewayinwhich“anxiety” interfereswith thedailybehaviorofaman.HuntandBrady15haveshown that some of the “treatments” for human anxiety (for example, electro-convulsive shock) temporarily eliminate the conditioned suppression in suchexperiments.

162

FIG.15.Performanceofaratonaconcurrentfixed-intervalandavoidancescheduleinacontrolsession(uppercurve)andafterbeinginjectedwithchlorpromazine(lowercurve).

Brady16hasrecentlyappliedthistechniquetothestudyoftranquilizingdrugs.Inhisexperiment, a rat is reinforced on a variable-interval schedule until respondingstabilizesataconstantintermediaterate.Stimuliarethenpresentedevery10minutes.Eachstimuluslastsfor3minutesandisfollowedbyashock.Conditionedsuppressionsoonappears.InFigure16eachsimplearrowshowstheonsetofthestimulus.Inorderto isolate the performance in the presence of the stimulus, the record is displaceddownward.Inthesalinecontrol,shortlyaftertheonsetofthestimulus,theratefallstozero, as shownby the horizontal portions of the displaced segments.As soon as theshockisreceived(atthebrokenarrows,wherethepenreturnstoitsnormalposition),respondingbeginsalmostimmediatelyatthenormalrate.Thebaselinebetweenstimuliis not smooth because a certain amount of chronic anxiety develops under thesecircumstances.A suitabledoseof a stimulant suchas amphetaminehas the effectofincreasing theover-all rate,asseen in themiddlepartofFigure16.Thesuppressingstimulusis, ifanything,moreeffective.Acourseoftreatmentwithreserpine,anothertranquilizer, has the effect of slightly depressing the over-all rate but restoringrespondingduringtheformerlysuppressingstimulus.Thus,inthelowerpartofFigure16,theslopesofthedisplacedsegmentsoftherecordareofthesameorderastheover-all record itself. The reserpine has eliminated an effect which, from a similarity ofincitingcauses,wemayperhapscallanxiety.

163

164

FIG.16.Effectofastimulantandatranquilizeronconditionedsuppressionintherat(Brady).

Another field in which important variables affecting behavior are studied isneurology. Performances under various schedules of reinforcement supply baselineswhichareasusefulhereasinthefieldofpsychopharmacology.Theclassicalpatternofresearchistoestablishaperformancecontainingfeaturesofinterest,thentoremoveordamagepartofthenervoussystem,andlatertohaveanotherlookatthebehavior.Thedamaged performance shows the effect of the lesion and helps in inferring thecontributionoftheareatonormalbehavior.

FIG.17.Performanceofaratpressingaleveltoshockitselfintheanteriorhypothalamus(Olds).

The procedure is, of course, negative. Another possibility is that neurologicalconditionsmaybearrangedwhichwillhaveapositiveeffect.AstepinthisdirectionhasbeentakenbyJamesOlds17withhisdiscoverythatweakelectricalstimulationofcertain parts of the brain, through permanently implanted electrodes, has an effectsimilar to thatofpositivereinforcement. InoneofOlds’experiments,aratpressesalevertogiveitselfmildelectricalstimulationintheanteriorhypothalamus.Whenevery

165

responseisso“reinforced,”behaviorissustainedinstrengthforlongperiodsoftime.OneofOlds’resultsisshowninFigure17.Theelectrical“reinforcement”wasbegunshortly after noon. The rat responded at approximately 2000 responses per hourthroughout the day and night until the following noon. There are only three or fourbriefpausesduringthisperiod.Whentheexperimentwascontinuedthefollowingday,however,theratfellasleepandsleptfor20hours.Thenitawokeandbeganagainatapproximately the same rate. Although there remain some puzzling differencesbetweenbehaviorsoreinforcedandbehaviorreinforcedwithfood,Olds’discoveryinanimportantsteptowardourunderstandingofthephysiologicalmechanismsinvolvedin the operation of the environmental variable. A similar reinforcing effect of brainstimulation has been found in cats by Sidman, Brady, Boren, and Conrad18 and inmonkeysbyLilly,oftheNationalInstitutesofHealth,andBrady,inthelaboratoriesoftheWalterReedArmyInstituteofResearch.

HumanBehavior

Whataboutman?Israteofrespondingstillanorderlyandmeaningfuldatumhere,orishuman behavior the exception in which spontaneity and and caprice still reign? Inwatching experiments of the sort described above, most people feel that they could“figure out” a schedule of reinforcement and adjust to it more efficiently than theexperimental organism. In saying this, they are probably overlooking the clocks andcalendars, thecounters,and thebehaviorofcountingwithwhichmanhassolved theproblem of intermittency in his environment. But if a pigeon is given a clock or acounter, it works more efficiently, and without these aids man shows little if anysuperiority.Parallelshavealreadybeensuggestedbetweenhumanandinfra-humanbehaviorin

notingthesimilarityoffixed-ratioschedulestopiece-ratepayandofvariableratiostotheschedules ingamblingdevices.Thesearemore thanmereanalogies.Comparableeffectsofschedulesofreinforcementinmanandtheotheranimalsaregraduallybeingestablishedbydirectexperimentation.AnexampleissomeworkbyJamesHolland19attheNavalResearchLaboratoriesonthebehaviorofobserving.Weoftenforgetthatlooking at a visual pattern or listening to a sound is itself behavior, becausewe arelikely to be impressed by the more important behavior which the pattern or soundcontrols. But any act which brings an organism into contact with a discriminativestimulus,or clarifiesor intensifies its effect, is reinforcedby this result andmustbeexplained in such terms. Unfortunately mere “attending” (as in reading a book orlisteningtoaconcert)hasdimensionswhicharedifficult tostudy.Butbehaviorwithcomparable effects is sometimes accessible, such as turning the eyes toward a page,tiltingapage tobring it intobetter light,or turningup thevolumeofaphonograph.Moreover,underexperimentalconditions,aspecificresponsecanbereinforcedbytheproductionorclarificationofastimuluswhichcontrolsotherbehavior.Thematterisofconsiderable practical importance. How, for example, can a radar operator or other“lookout”bekeptalert?Theansweris:byreinforcinghislookingbehavior.Hollandhasstudiedsuchreinforcementinthefollowingway.Hishumansubjectis

seatedinasmallroombeforeadial.Thepointeronthedialoccasionallydeviatesfromzero,andthesubject’staskistorestoreitbypressingabutton.Theroomisdark,and

166

thesubjectcanseethedialonlybypressinganotherbuttonwhichflashesalightforafractionofasecond.Pressingthesecondbuttonis,then,anactwhichpresentstothesubjectastimuluswhichisimportantbecauseitcontrolsthebehaviorofrestoringthepointertozero.Hollandhasonlytoschedulethedeviationsofthepointertoproducechangesinthe

rate of flashing the light comparable to the performances of lower organisms undercomparable schedules. InFigure18, for example, the upper curve shows a pigeon’sperformance on a fairly short fixed-interval. Each interval shows a rather irregularcurvature as the rate passes from a low value after reinforcement to a high, fairlyconstant, terminal rate. In the lower part of the figure is one of Holland’s curvesobtainedwhenthepointerdeflectedfromzeroeverythreeminutes.Afterafewhoursofexposuretotheseconditions, thesubjectflashedthelight(“lookedatthepointer”)onlyinfrequentlyjustafteradeflection,butastheintervalpassed,hisrateaccelerated,sometimes smoothly, sometimes abruptly, to a fairly constant terminal rate. (Aninterestingfeatureofthiscurveisthetendencyto“runthrough”thereinforcementandtocontinueatahighrateforafewsecondsafterreinforcementbeforedroppingtothelowratefromwhichtheterminalratethenemerges.Examplesofthisareseenata,b,andc. Examples in the case of the pigeon are also seen atd ande. In our study ofschedules, Ferster and I had investigated this effect in detail long before the humancurveswereobtained.)Otherexperimentsonhumansubjectshavebeenconductedinthefieldofpsychotic

behavior.InaprojectattheBehaviorResearchLaboratoriesoftheMetropolitanStateHospital,inWaltham,Massachusetts,20apsychoticsubjectspendsoneormorehourseachdayinasmallroomcontainingachairandaninstrumentpanelasseeninFigure19. At the right of the instrument board is a small compartment (a) into whichreinforcers (candy, cigarettes, coins) are dropped by an appropriate magazine. Theboard contains a plunger (b), similar to that of a vending machine. The controllingequipmentbehindaseriesofsuchroomsisshowninFigure20.Alongthewallatleft,asata,areseenfourmagazines,whichcanbeloadedwithvariousobjects.Alsoseenare periscopes (as atb) throughwhich the rooms can be observed through one-waylenses. At the right are cumulative recorders and behind them panels bearing thecontrollingequipmentwhicharrangesschedules.

167

FIG.18.Fixed-intervalperformancebyahumansubjectcomparedwiththatofapigeon.

It has been found that even deteriorated psychotics of long standing can, throughproperreinforcement,be inducedtopullaplungerforavarietyofreinforcersduringsubstantial daily experimental sessions and for long periods of time. Schedules ofreinforcementhave the expected effects, but the fact that theseorganismsare sick isalso apparent. In Figure21, for example, the record atA shows a “normal” humanperformanceonavariable-intervalschedulewherethesubject(ahospitalattendant)is

168

reinforcedwithnickelsonanaverageofonceperminute.Astraightline,similartotherecordsofthepigeonandchimpanzeeinFigure3,isobtained.RecordsB,C,andD arethe performances of three psychotics on the same schedule working for the samereinforcers.Behaviorissustainedduringthesession(asitisduringmanysessionsforlongperiodsof time),but therearemarkeddeviations fromstraight lines.Periodsofexceptionallyrapidrespondingalternatewithpausesorperiodsataverylowrate.

169

FIG.19.Arrangementforthestudyofthebehaviorofapsychoticsubject(Lindsley).

ThatascheduleisneverthelesseffectiveinproducingacharacteristicperformanceisshownbyFigure22.Afixed-ratioperformancegivenbyapigeonunderconditionsinwhich there is substantial pausing after reinforcement is shown atA. In spite of thepauses,thegeneralruleholds:assoonasrespondingbegins,thewholeratioisquicklyrunoff.Fixed-ratiocurvesfortwopsychoticsubjects,bothseverelyill,areshownatB

170

andC.Onlysmallratioscanbesustained(40and20,respectively),andpausesfollowall reinforcements. Nevertheless, the performance is clearly the result of a ratioschedule:oncerespondingbeings,thecompleteratioisrunoff.

FIG.20.ControllingequipmentusedinresearchonpsychoticbehaviorattheMetropolitanStateHospital,Waltham,Massachusetts.

171

FIG.21.Onenormal(A)andthreepsychotic(B,C,D)performancesonavariable-intervalschedule.

172

FIG.22.A“breaking”fixed-ratioperformancebyapigeon(A)andtwofixed-ratioperformancesbypsychoticsubjects(B,C).

Conclusion

It is unfortunate that a presentation of this sortmust be confined tomere examples.Littlemorecanbedonethantosuggesttherangeofapplicationofthemethodandtheuniformityofresultsoverafairlywiderangeofspecies.Theextentofwhichwearemoving towardaunified formulationof thisdifficultmaterial cannotbeproperly setforth.Perhapsenoughhasbeensaid,however, tomakeonepoint—that in turning toprobability of response or,more immediately, to frequency of respondingwe find adatumwhichbehavesinanorderlyfashionunderagreatvarietyofconditions.Sucha

173

datum yields the kind of rigorous analysis which deserves a place in the naturalsciences. Several features should not be overlooked.Most of the records reproducedherereportthebehaviorofsingleindividuals;theyarenotthestatisticalproductofan“averageorganism.”Changesinbehaviorarefollowedcontinuouslyduringsubstantialexperimental sessions. They often reveal changes occurring within a few secondswhichwouldbemissedbyanyprocedurewhichmerelysamplesbehaviorfromtimetotime. The properties of the changes seen in the cumulative curves cannot be fullyappreciatedinthenon-instrumentalobservationofbehavior.Thereproducibilityfromspecies to species is a product of the method. In choosing stimuli, responses, andreinforcersappropriatetothespeciesbeingstudied,weeliminatethesourcesofmanyspeciesdifferences.Havewebeenguiltyofanunduesimplificationofconditionsinordertoobtainthis

level of rigor? Have we really “proved” that there is comparable order outside thelaboratory?Itisdifficulttobesureoftheanswerstosuchquestions.Supposeweareobserving the rate at which a man sips his breakfast coffee. We have a switchconcealed in our hand,which operates a cumulative recorder in another room.Eachtimeoursubjectsips,weclosetheswitch.Itisunlikelythatweshallrecordasmoothcurve.Atfirstthecoffeeistoohot,andsippingisfollowedbyaversiveconsequences.As it cools, positive reinforcers emerge, but satiation sets in. Other events at thebreakfast table intervene.Sippingeventuallyceasesnotbecausethecupisemptybutbecausethelastfewdropsarecold.Butalthoughourbehavioralcurvewillnotbepretty,neitherwillthecoolingcurve

forthecoffeeinthecup.Inextrapolatingourresultstotheworldatlarge,wecandonomore than the physical and biological sciences in general. Because of experimentsperformedunderlaboratoryconditions,noonedoubtsthatthecoolingofthecoffeeinthecupisanorderlyprocess,eventhoughtheactualcurvewouldbeverydifficulttoexplain. Similarly, when we have investigated behavior under the advantageousconditionsof the laboratory,wecanaccept itsbasicorderliness in theworldat largeeventhoughwecannottherewhollydemonstratelaw.Inturningfromananalysisofthissortmanyfamiliaraspectsofhumanaffairstake

onnewsignificance.Moreover, aswemightexpect, scientificanalysisgivesbirth totechnology. The insight into human behavior gained from research of this sort hasalready proved effective in many areas. The application to personnel problems inindustry,topsychotherapy,to“humanrelations”ingeneral,isclear.Themostexcitingtechnological extension at the moment appears to be in the field of education. Theprinciplesemergingfromthisanalysis,andfromastudyofverbalbehaviorbaseduponit,arealreadybeingappliedinthedesignofmechanicaldevicestofacilitateinstructionin reading, spelling, and arithmetic in young children, and in routine teaching at thecollegelevel.Inthelongrunonemayenvisageafundamentalchangeingovernmentitself,taking

thatterminthebroadestpossiblesense.Foralongtimemenofgoodwillhavetriedtoimprovetheculturalpatternsinwhichtheylive.Itispossiblethatascientificanalysisofbehaviorwillprovideusatlastwiththetechniquesweneedforthistask—withthewisdomweneedtobuildabetterworldand,throughit,bettermen.

1Magnus,R.Köperstellung.Berlin,1924.

174

2Tinbergen,N.TheStudyofInstinct.Oxford,1951.3SeeTheBehaviorofOrganisms.4MuchoftheresearchfromwhichthefollowingexamplesaredrawnhasbeensupportedbytheOfficeofNavalResearchandtheNationalScienceFoundation.ItisreportedinSchedulesofReinforcement.

5Acurveafterfouryearsappearsonpage80.6Wilson,M.P.andKeller,F.S.Ontheselectivereinforcementofspacedresponses.J.Comp.andPhysiol.Psychol.,1953,46,190–193.

7Levi,Wendell.ThePigeon.Sumter,S.C.,1941.8Guttman,N.,Thepigeonandthespectrumandotherperplexities.Psychol.Reports,1956,2,449–460.

9Blough,D.S.Darkadaptationinthepigeon.J.Comp.andPhysiol.Psychol.,1956,49,425–430.

10Wald,G.J.Gen.Physiol.,1954,38,623–681.11Sidman,M.Avoidanceconditioningwithbriefshockandnoexteroceptivewarningsignal.Science,1953,118,157–158.

12Anliker,J.,andMayer,J.Operantconditioningtechniqueforstudyingfeedingpatternsinnormalandobesemice.J.Appl.Physiol.,1956,8,667–670.

13Williams,D.R.,andTeitelbaum,P.Controlofdrinkingbehaviorbymeansofanoperant-conditioningtechnique.Science,1956,124,1294–1296.

14Dews,P.B.Modificationbydrugsofperformanceonsimpleschedulesofpositivereinforcement.Ann.N.Y.Acad.Sci.,1956,65,268–281.

15Hunt,H.F.,andBrady,J.V.Someeffectsofelectro-convulsiveshockonaconditionedemotionalresponse(“anxiety”).J.Comp.andPhysiol.Psychol.,1955,48,305.

16Brady,J.V.Assessmentofdrugeffectsonemotionalbehavior.Science,1956,123,1033–1034.

17Olds,J.Pleasurecentersinthebrain.Scient.Amer.,1956,195,105–116.18Sidman,M.,Brady,J.V.,Boren,J.J.,andConrad,D.G.Rewardschedulesandbehaviormaintainedbyintracranialself-stimulation.Science,1955,122,830–831.

19Holland,J.G.Techniqueforbehavioralanalysisofhumanobserving.Science,1957,125,348–350.

20Lindsley,O.R.Operantconditioningmethodsappliedtoresearchinchronicschizophrenia.Psychiat.Res.Rep.,1956,5,118–139.

175

ReinforcementToday

This paper was part of a symposium on reinforcement held at a meeting of theAmerican Psychological Association, in September, 1957. It was published in TheAmericanPsychologist(1958,13,94-99)andisreprintedherebypermission.During the past twenty-five years the role of reinforcement in human affairs hasreceived steadily increasing attention—not through any changing fashion in learningtheorybutastheresultofthediscoveryoffactsandpracticeswhichhaveincreasedourpowertopredictandcontrolbehaviorandindoingsohaveleftnodoubtoftheirrealityand importance.The scope of reinforcement is still not fully grasped, evenby thosewho have donemost to demonstrate it, and elsewhere among psychologists culturalinertia is evident. This is understandable because the change has been little short ofrevolutionary: scarcely anything in traditional learning theory is left in recognizableform.InthispaperIshalltrytocharacterizesomeofthechangesinourconceptionofreinforcementwhichhavebeenforceduponusandtosuggestwhyithasbeensohardtoacceptthemandtorecognizetheirimport.

TheAcquisitionofBehavior

In1943KellerBreland,NormanGuttman, and Iwereworkingonawartimeprojectsponsored byGeneralMills, Inc.Our laboratorywas the top floor of a flourmill inMinneapolis,wherewespentagooddealoftimewaitingfordecisionstobemadeinWashington.Alldaylong,aroundthemill,wheeledgreatflocksofpigeons.Theywereeasily snared on the window sills and proved to be an irresistible supply ofexperimentalsubjects.Webuiltamagneticfood-magazine,whichdispensedgrainonthe principle of an automatic peanut vendor, and conditioned pigeons to turn at thesound it made and eat the grain it discharged into a cup. We used the device toconditionseveralkindsofbehavior.Forexample,webuiltagaugetomeasuretheforcewithwhichapigeonpeckedahorizontalblock,andbydifferentiallyreinforcingharderpecks we built up such forceful blows that the base of the pigeon’s beak quicklybecameinflamed.Thiswasseriousresearch,butwehadourlightermoments.Onedaywedecidedtoteachapigeontobowl.Thepigeonwastosendawoodenballdownaminiature alley toward a set of toy pins by swiping the ball with a sharp sidewardmovementof thebeak.Tocondition the response,weput theballon the floorofanexperimentalboxandpreparedtooperatethefood-magazineassoonasthefirstswipeoccurred.But nothinghappened.Thoughwehad all the time in theworld,wegrewtired of waiting. We decided to reinforce any response which had the slightestresemblancetoaswipe—perhaps,atfirst,merelythebehavioroflookingattheball—and then to select responses which more closely approximated the final form. Theresultamazedus.Inafewminutes,theballwascaromingoffthewallsoftheboxasifthe pigeon had been a champion squash player. The spectacle so impressed KellerBreland that he gave up a promising career in psychology and went into the

176

commercialproductionofbehavior.Whyhadthepigeonlearnedwithsuchsurprisingspeed?Threepointsseemrelevant:1.Inmagazine-trainingthepigeon—thatis,ingettingittorespondtothesoundof

themagazinebyturningimmediatelyandapproachingthefoodtray—wehadcreatedanauditoryconditionedreinforcer.Thisisagreathelpinoperantconditioningbecauseitcanfollowaresponseinstantly.Whenaratrunsdownanalleyandfindsfoodattheend, or when a performing seal bounces a ball off its nose and is thrown a fish,behavior is reinforced under relatively loose temporal conditions. The rat may notimmediately find the food, and the trainer may take a moment to throw the fish.Organisms will, of course, learn and continue to behave when reinforcement issubstantially delayed, but only when certain temporal contingencies have beenstrengthened. Unless the gap between the behavior and the ultimate reinforcer isbridged with a sequence of conditioned reinforcers, other behavior will occur andreceivethefullforceofthereinforcement.Ifthesealhastimetoturntowardthetrainerbefore receiving the visual reinforcement of the approaching fish, its behavior inturningismostpowerfullyreinforcedandmayinterferewiththebehaviorthetraineristryingtosetup.Eventuallyadiscriminationisformedsothatthesealturnsonlyafterhavingexecutedtheproperbehavior,butthiscanbeaslowprocess.Adelayofevenafractionofasecondissometimesimportant,aswehavefoundindesigningequipmentforthestudyofoperantbehaviorinthepigeon.Whentheresponseispeckingaplasticdisc, the controlling circuitmust act so rapidly that the soundof themagazine, as aconditionedreinforcer,willcoincidewithstrikingthediscratherthanpullingtheheadaway from it. This is amatter of perhaps a twentieth of a second, but such a delayproducesdisturbingchangesinthetopographyoftheresponse.2.Inearlyexperimentsonleverpressing,aquickresponsetothefood-magazinewas

alwayssetupbefore the leverwas introduced.Thiswasdoneforanotherreason—topermitemotionalresponsestothenoiseofthemagazinetoadaptout—butitmusthavebeenimportantinprovidinginstantaneousreinforcement.Theexplicitconditioningofanauditoryreinforcerwas,therefore,notnew;theremusthavebeensomethingelseinthe bowling experiment. In most experiments on learning, an organism producesreinforcement by direct action: a rat pushes over the door of a jumping stand anddiscoversfood,oramonkeyliftsacupandexposesagrape.Electricalcircuitsgreatlyincreasethepossibilities,buteventhentheorganismisusuallylefttoclosethecircuitby mechanical contact. In The Behavior of Organisms I describe an experiment inwhicha ratwasconditioned topullastring togetamarblefromarack,pickup themarblewith its forepaws,carry itacross thecage toavertical tuberising twoinchesabove the floor, lift themarble, and drop it into the tube. The behavior was set upthrough successive approximations, but every stage was reached by constructingmechanicalandelectricalsystemsoperatedby the rat. In theexperimentonbowling,however, we held the reinforcing switch in our hand and could reinforce any givenformofbehaviorwithout constructing amechanicalor electrical system to report itsoccurrence. The mechanical connection between behavior and reinforcement wasgreatlyattenuated.3.But thiswas not new, either.Thorndike had reinforced a catwhen it licked its

paw,andanimaltrainersusehandreinforcement.Thesurprisingresultinourbowlingexperiment may have been due to the combination of the temporal precision ofreinforcement provided by a conditioned reinforcer and the free selection oftopography resulting from hand reinforcement. In any event this combination must

177

haveenhancedtheeffectofthethird,andmain,featureoftheexperiment:thegradualshaping of behavior by reinforcing crude approximations of the final topographyinsteadofwaitingforthecompleteresponse.The techniqueofshapingbehavior isnowafamiliarclassroomdemonstration,but

theprincipleitdemonstrateshasnotyetfoundasecureplaceintextbookdiscussionsoflearning. Curiously enough, the acquisition of behavior has never been directlyattackedinclassicalresearch.Thestudyofmemory,fromEbbinghauson,hasnotbeenprimarilyconcernedwithhowbehaviorisacquiredbutonlywithhowitisretainedorhowoneforminterfereswithanotherinretention.Whydoesthesubjectsitinfrontofthememorydrum,whydoeshevouchsafeanticipatoryguesses,andhow(notwhen)doesheeventuallyarriveatthatfirstcorrectresponse?Thesequestionshavenotbeentheprimaryconcernof researchonmemory.Animal researchhasalmostalways leftthe shaping of behavior to mechanical devices. In both fields the acquisition ofbehavior has been reported by learning curves or, worse, by something called thelearningcurve.Whenonehaswatchedtheactualshapingofbehavior,itisobviousthatsuchcurvesdonot reflect any importantpropertyof the change inbehaviorbroughtabout by operant reinforcement. They summarize the arbitrary and often accidentalconsequences which arise when complex and largely unanalyzed conditions ofreinforcementactuponlargesamplesofbehavior.Thereareprobablyasmanylearningcurvesasthereareapparatusesforthestudyoflearning,andmathematicianswillstriveinvaintopullausefulorderoutofthischaos.Yettheprestigeofthelearningcurveissogreatthatpsychologistsareunabletobelievetheireyeswhentheprocessoflearningismadevisible.

TheMaintenanceofBehavior

Anobviousfactaboutbehavioristhatitisalmostneverinvariablyreinforced.Notsoobviousisthefactthatthepatternofintermittentreinforcementcontrolsthecharacterand level of a performance.Why this is so cannot be explained in a fewwords. InSchedulesofReinforcementCharlesB.FersterandIhavearguedasfollows.A schedule of reinforcement is arranged by a programming systemwhich can be

specified in physical terms. A clock is introduced into the circuit between key andmagazinesothatthefirstresponsemadetothekeyafteragivenintervaloftimewillbereinforced.Acounterintroducedintothecircuitestablishesacontingencyintermsofnumberofresponsesemittedperreinforcement.Varioussettingsofclockandcounterandcombinationsofthesegeneratealmostunlimitedpossibilities.A selected schedule usually generates a characteristic performance, expressed in

termsofrateofrespondingandchangesinrate.Oncethishashappened,theorganismis characteristically reinforced at the end of a particular pattern of responding. Itsbehavior at the moment of reinforcement and during the period precedingreinforcementispartofthestimulatingenvironment,aspectsofwhichacquirecontrolover subsequent behavior. To take a very simple example: if an organism ischaracteristicallyrespondingatahighrateatthemomentofreinforcement,behavioratthatratebecomesanoptimalstimulatingcondition,comparabletothepresenceofthereinforced stimulus in a discrimination, and the probability of further responding istherefore maximal. When the organism is not responding at all, the probability isminimal. Other rates and patterns of changes in rate come to serve similar

178

discriminative functions. Ferster and I have checked this explanation of theperformances characteristic of schedules in several ways. For example, instead oflettingaschedulegenerateaconditionmostofthetime,wehaveaddedspecialdevicestoassure agivenconditionofbehaviorat every reinforcement.Thus,wherea fixed-interval performance usually arranges a moderately high rate at the moment ofreinforcement, a special devicewill guarantee that reinforcements occuronly at thatrate.Wehavealsoaddedstimulitothephysicalenvironmentwhicharecorrelatedwith,and hence amplify, the aspects of the organism’s behavior appealed to in such anexplanation.This, then, iswhathappensunder intermittent reinforcement:Aschedulingsystem

sets up a performance, and the performance generates stimuli which enter into thecontroloftherateofresponding,eithermaintainingtheperformanceorchangingitinvariousways. Some schedules produce performanceswhich guarantee reinforcementunder conditions which continue to maintain that performance. Others produceprogressive changes. Still others yield oscillations: the first performance generatesconditionswhicheventuallyproduceadifferentperformance,whichinturngeneratesconditionsrestoringtheearlierperformance,andsoon.Boththecircuitandthebehavior,then,contributetothereinforcingcontingencies.It

followsthattheeffectofanycircuitdependsuponthebehaviortheorganismbringstoit. Some complex schedules can be studied only by taking the organism through aseriesofsimplerschedulesintothefinalperformance.Theperformance,aswellasthetopographyofaresponse,mayneedtobe“shaped.”Thisdoesnotmeanthatschedule-performancesvarygreatlybecauseofindividualhistories,foronlyafewoftheeffectsofschedulesarenotreadilyreversible.Onceaperformanceisreached,itusuallyshowsahighorderofuniformity,evenbetweenspecies.Thefactthatitisthecombinationofschedule and performance which generates reinforcing contingencies can easily beoverlooked.Aphysiologistonceasked toborrowoneofourapparatuses toshowhisclass the behavioral effects of certain drugs. We sent him an apparatus whichreinforcedapigeononamultiplefixed-ratiofixed-intervalschedule,togetherwithtwopigeons showing beautifully stable performances.When one pigeon died through anoverdoseofadrug, thephysiologistsimplyboughtanotherpigeonandputit intotheapparatus.Tohissurprise,nothinghappened.The same mistake is made in much traditional work on learning and problem

solving. In the usual study of problem solving, for example, the experimenterconstructsacomplexsetofcontingenciesandsimplywaitsforittotakehold.Thisisnotestofwhethertheorganismcanadjusttothesecontingencieswithaperformancewhich would be called a solution. All we can properly conclude is that theexperimenterhasnotconstructedanadequatesuccessionofperformances.Theabilityoftheexperimenterratherthanthatoftheorganismisbeingtested.Itisdangeroustoassertthatanorganismofagivenspeciesoragecannotsolveagivenproblem.Astheresult of careful scheduling,pigeons, rats, andmonkeyshavedone thingsduring thepast fiveyearswhichmembersof theirspecieshaveneverdonebefore. It isnot thattheir forebears were incapable of such behavior; nature had simply never arrangedeffectivesequencesofschedules.What we have learned about the shaping of response-topography and about the

techniqueswhichbringanorganismunderthecontrolofcomplexscheduleshasmadeitpossibletostudythebehaviorgeneratedbyarrangementsofresponses,stimuli,andreinforcementsonceclassifiedasthe“highermentalprocesses.”Anexperimentcanbe

179

designed in which two or more responses are emitted concurrently or in rapidalternation,underthecontrolofmultiplestimuli,oftenundertwoormoreschedulesofreinforcementor twoormore typesof reinforcementunderappropriateconditionsofmotivation. It has been found that a schedule, or rather the stimuli present when ascheduleisinforce,hasreinforcingoraversiveproperties.Anorganismwillrespondononescheduletoreachoravoidanother.Wecandeterminewhichoftwoschedulesapigeon“prefers”bycomparinghowfastitwillrespondonavariable-intervalscheduleto get into Schedule A with how fast it will respond on the same variable-intervalscheduletogetintoScheduleB.Theanalysisofavoidanceandescapebehaviorinthehands of Sidman, Brady, and others has made it possible to study combinations ofpositive and negative reinforcers in many interrelated patterns. The analysis ofpunishment in such terms has permitted a reformulation of the so-called Freudiandynamisms.1The technology resulting from the study of reinforcement has been extended into

otherfieldsofpsychologicalinquiry.IthaspermittedBlough,Guttman,andotherstoconvert pigeons into sensitive psychophysical observers. It has allowedpharmacologists and psychologists in pharmacological laboratories to constructbehavioralbaselinesagainstwhichtheeffectsofdrugsontheso-calledhighermentalprocessescanbeevaluated.IthasenabledLindsleyandhisco-workerstotestthelimitsof the environmental control of psychotic subjects. And so on, in a long list. Thetechnology is difficult. It cannot conveniently be learned from books; somethingresemblinganapprenticeshipisalmostnecessary.Possiblywemayexplainthefactthatpsychologistsingeneralhaveonlyslowlyacceptedthesenewmethodsbynotingthatundersuchconditionsknowledgeisdiffusedslowly.Many psychologists may never wish to acquire the competence necessary for

detailedresearchonreinforcement,butthereisanotherapplicationwhichisofbroadersignificance.Aclinicalpsychologistrecentlycomplained2thatlearningtheorytoldhimnothingaboutimportantaspectsofhumanbehavior.Itwouldnotexplain,forexample,whyamanwouldseek“littlebitsofpunishmentinordertoacceptabigpunishment.”Hemayberightinsayingthatlearningtheorydoesnottellhimmuch,buttheexamplehechoseisjustthekindofcomplexarrangementofcontingencieswhichisnowunderintensive investigation.Andhe isasking for just thekindof interpretationofhumanaffairs which is emerging from this work. The world in which man lives may beregarded as an extraordinarily complex set of positive and negative reinforcingcontingencies. In addition to the physical environment to which he is sensitivelyattunedandwithwhichhecarriesonanimportantinterchange,wehave(ashehas)tocontend with social stimuli, social reinforcers, and a network of personal andinstitutionalcontrolandcountercontrol—allofamazingintricacy.Thecontingenciesofreinforcementwhichmanhasmadeformanarewonderfultobehold.But theyarebynomeans inscrutable.Theparallelbetween thecontingenciesnow

being studied in the laboratory and those of daily life cry for attention—and forremedialaction.Inanysocialsituationwemustdiscoverwhoisreinforcingwhomwithwhat and towhateffect.As avery simple example, take the aggressive child.Whentwoyoungchildrenareleftaloneinaroomwithafewtoys,conditionsarealmostidealfor shaping selfish and aggressive behavior. Under these circumstances one child’sreinforcementistheotherchild’spunishment,andviceversa.WhenIoncediscussedthisexamplewithagroupofteachers,oneofthemexclaimed:“Yes,andthat’swhyin

180

thenurseryschoolsoftheSovietUnionthetoysaresobigittakestwochildrentoplaywith them!” Possibly that is one solution. Certainly there are many others. Whencontingenciesofreinforcementareproperlyunderstood,wecannotthoughtlesslyallowdamagingcontingenciestoariseorgounremedied.Bytakingalittlethoughtitisnowpossibletodesignsocialsituationswhichhavehappierconsequencesforeveryone.Iamnotsayingthatanyonesetofcontingenciesexplainsaggressioninchildrenor

thatittakesalongapprenticeshipinreinforcementresearchtounderstandthatcase.Itistheveryexistenceofreinforcingcontingencieswhichmustfirstberecognized—andthatisnotalwayseasy.Hereisaslightlylessobviousexample.Thecurrentnationwideproblemof school discipline is frequently, thoughpossibly erroneously, attributed toprogressiveeducation.Whateveritsexplanation,it isaseriousproblem.Howcanwerecapturetheorderlyconductonceattributedto“discipline”withoutreinstatingalltheundesirableby-productsofaninhumaneaversivecontrol?Theansweris:usepositivereinforcement instead of punishment. But how? A first step is to analyze thereinforcingcontingenciesintheclassroom.Inparticular,whatreinforcersareavailabletotheteacher?Theanswertothatquestionissometimesdiscouraging,butevenintheworstpossiblecaseshecanatleastreinforceaclassbydismissingit.Thepointisthatshemustunderstandthatdismissalisreinforcingifsheisnottothrowawaythesmallmeasureofpoweritoffersher.The“natural”thingisforateachertodismisstheclasswhenitsconductismostaversivetoher.Butthisisexactlythewrongthingtodo,forshe then differentially reinforces the very behavior shewants to suppress.A teacherwho understands reinforcement will survey the class during the final minutes of aperiodandchoosefordismissalthemomentatwhichthingsaregoingaswellascanbeexpected.Theeffectwillnotbeevidentthefirstday,itmaynotbeevidentthesecondorthird,anditmayneverbeenoughtosolveallherproblems;butacarefulhusbandingofsmallreinforcersandthenurturingofpropercontingenciesisaprogramwellworthexploring.As a final andmore technical example of the use of reinforcement in interpreting

humanaffairs,takethealwaysinterestingformofbehaviorcalledgambling.Gamblersappeartoviolatethelawofeffectbecausetheycontinuetoplayeventhoughtheirnetreward is negative. Hence it is often argued that they must be gambling for otherreasons.To thepsychoanalyst thegamblermay simplybepunishinghimself.Othersmayinsistthattheattractionisnotmoneybutexcitement,orthatpeoplegambletogetaway from a humdrum life. Now, all gambling devices arrange a variable-ratioscheduleof reinforcement, andourexplanationof theperformancegeneratedby thatschedule embraces the behavior of the gambler. It happens to be relatively excitedbehavior,butthis,aswellasthefactthatthereisnonetgain,isirrelevantinaccountingfor the performance. A pigeon, too, can become a pathological gambler, and it isunlikelythatitdoessotopunishitself,orfortheexcitement,ortogetawayfromitall.Suchexpressionsmaynotbemeaningless.Thecomplexcontingencies involved in

“self-punishment” may well be involved, although quantitative evidence would beneeded to show this. “Getting away from it all” reminds us that some schedules areaversive.HerrnsteinandMorsehaveshownthatapigeoncanbeconditionedtopeckonekeyifthisisoccasionallyfollowedbytheopportunitytotaketimeofffromanotherkey.Inturningtoavariable-ratiosystemofreinforcement,then,thegamblermaywellbe escaping from other schedules. Moreover, a variable-ratio schedule at suitablevalues isreinforcing.Thesefactsaccountforanybehaviorwhichbringsanorganismunder a variable-ratio schedule, but they do not explain the performance once this

181

scheduleisinforce.Theconditionswhichprevailundertheschedulearetherelevantfacts.Thesearenecessarilyfragmentaryexamplesof thecontributionofanexperimental

analysis of intermittent reinforcement to our understanding of human behavior, buttheymay serve tomake an important point.The relevance of reinforcement is oftenquiteunexpected.Theseexamplesarenotpartoftheclassicalfieldoflearning;theyaremattersofmotivation!Oneexpects to see themdiscussedbydynamicpsychologists,psychologistsofpersonality,orpsychoanalysts,notbypeoplewhostudywhiteratsandpigeons.True,learningtheoryhaslongbeenappliedtopsychotherapy,buttraditionalresearchinlearninghasnotmadeaveryhelpfulcontribution.Suddenly,reinforcementtakes on new dimensions.When Freudwas once askedwhether psychoanalysis andpsychologywerethesame,heinsistedthatpsychoanalysisembracedallofpsychologyexcept the physiology of the sense organs.3 This was an ambitious statement, andperhapsasimilarclaimforreinforcementwouldbeequallyunjustified.Yetthefactsofhuman behavior fall to the psychoanalyst and the student of reinforcement alike forexplanation.Butwheretheanalysthasstudiedbehaviorinagivenenvironmentasthemanifestation of hidden (even if eventually-to-be-revealed) forces, we can nowinterpret thesamebehaviorandenvironmentasasetof reinforcingcontingencies. Indoingsowegainatremendousadvantage,foralltermsnecessaryforsuchananalysisliewithin an observable and oftenmanipulable universe.Beyond the prediction andcontrol made possible by recent research in reinforcement lies the broader field ofinterpretation.And it is a kindof interpretation so closely alliedwithpredictionandcontrolthatpositiveandsuccessfulactionarefrequentlywithineasyreach.IfIhavesuggestedtopsychologistsingeneralthattheywillfindmuchofinterestin

themodernstudyofreinforcement, itwillbeappropriatetoendwithafewwordsofcaution.1. This kind of research is difficult and relatively expensive. In our book on

schedules of reinforcement, Ferster and I report on 70,000 hours of continuouslyrecordedbehaviorcomposedofaboutonequarterofabillionresponses.Thepersonalobservationofbehavioronsuchascale isunthinkable.Theresearchmustbeheavilyinstrumented. The programming of complex schedules demands not only a largebudget but considerable skill in relay engineering, neither of which is common inpsychologicallaboratories.2. It is usually single-organism research.Any other experimentalmethod is often

impossible.Whenanexperimentononepigeonrunstothousandsofhours,itcannotberepeatedonevenamodestgroupof,say,tensubjects—atleastifonewantstogetonwithothermatters.Fortunately,astatisticalprogramisunnecessary.Mostofwhatweknowabout the effectsof complex schedulesof reinforcementhasbeen learned in aseriesofdiscoveriesnooneofwhichcouldhavebeenproved to thesatisfactionofastudent inStatisticsA.Moreover,astatisticalapproach is justwrong.Thecurveswegetcannotbeaveragedorotherwisesmoothedwithoutdestroyingpropertieswhichweknow to be of first importance. These points are hard to make. The seasonedexperimenter can shrug off the protests of statisticians, but the young psychologistshouldbepreparedtofeelguilty,oratleaststrippedoftheprestigeconferreduponhimbystatisticalpractices,inembarkinguponresearchofthissort.3.Theresearch isnot theoretical in thesense thatexperimentsaredesignedto test

theories. As I have pointed out elsewhere [see page 69], when lawful changes in

182

behavior take place before our very eyes—or, at most, only one step removed in acumulative curve—we lose the taste, as we lose the need, for imagined changes insomefancifulworldofneurones, ideas,or interveningvariables.Hereagaintraditionthrowsuparoadblock.Certainpeople—amongthempsychologistswhoshouldknowbetter—haveclaimedtobeabletosayhowthescientificmindworks.Theyhavesetupnormativerulesofscientificconduct.Thefirststep foranyone interested instudyingreinforcement is to challenge that claim. Until a great deal more is known aboutthinking, scientific or otherwise, a sensible man will not abandon common sense.FersterandIwereimpressedbythewisdomofthiscourseofactionwhen,inwritingour book,we reconstructed our own scientific behavior.At one timewe intended—though,alas,wechangedourminds—toexpress thepoint in thisdedication:“Tothemathematicians,statisticians,andscientificmethodologistswithwhosehelpthisbookwouldneverhavebeenwritten.”Thedifficultieswhichhavestoodinthewayoftheadvancingstudyofreinforcement

will undoubtedly continue to cause trouble, but theywill bemore thanoffset by thepowerfulreinforcingconsequencesofworkinthisfield.Techniquesarenowavailableforanewandhighlyprofitableexplorationofthehumanbehavioratissueineducation,commerce and industry, psychotherapy, religion, and government. A program ofculturaldesigninthebroadestsenseisnowwithinreach.Sociologists,anthropologists,political scientists, economists, theologians, psychotherapists, andpsychologists havelong tried to reach an understanding of human behavior which would be useful insolvingpracticalproblems. In that technological raceadarkhorse iscomingup fast.The new principles andmethods of analysis which are emerging from the study ofreinforcementmayprove to be among themost productive social instruments of thetwentiethcentury.

1SeeScienceandHumanBehavior.2Sheehan,J.G.Themaritalstatusofpsychoanalysisandlearningtheory.Amer.Psychologist,1957,12,277–278.

3Wortis,J.FragmentsofananalysiswithFreud.NewYork:Simon&Schuster,1954.

183

PARTIII

TheTechnologyofEducation

TheScienceofLearningandtheArtofTeachingTeachingMachinesWhyWeNeedTeachingMachinesReflectionsonaDecadeofTeachingMachinesTeachingScienceinHighSchool—WhatIsWrong?ContingencyManagementintheClassroom

184

TheScienceofLearningandtheArtofTeaching

InMarch, 1954, the Annual Conference on Current Trends in Psychology at theUniversityofPittsburghwasdevotedto“PsychologyandtheBehavioralSciences.”Itprovided an opportunity to report some recent reflections on the technology ofeducation and to describe an early device designed to mechanize certain forms ofinstruction.ThepaperwaspublishedintheHarvardEducationalReview(1954,24,86-97) and in Psychology and the Behavioral Sciences (Pittsburgh, University ofPittsburghPress,1955),andisreprintedherewiththepermissionoftheUniversityofPittsburghPress.

SOME PROMISING ADVANCES have recently been made in the field of learning.Special techniques have been designed to arrangewhat are called “contingencies ofreinforcement”—therelationswhichprevailbetweenbehaviorontheonehandandtheconsequencesofthatbehaviorontheother—withtheresultthatamuchmoreeffectivecontrolofbehaviorhasbeenachieved.Ithaslongbeenarguedthatanorganismlearnsmainly by producing changes in its environment, but it is only recently that thesechanges have been carefully manipulated. In traditional devices for the study oflearning—in theserialmaze, forexample,or in theT-maze, theproblembox,or thefamiliardiscriminationapparatus—theeffectsproducedbytheorganism’sbehaviorareleft tomanyfluctuatingcircumstances.There ismanyaslipbetween the turn-to-the-rightandthefood-cupattheendofthealley.Itisnotsurprisingthattechniquesofthissorthaveyieldedonlyvery roughdata fromwhich theuniformitiesdemandedbyanexperimentalsciencecanbeextractedonlybyaveragingmanycases. Innoneof thiswork has the behavior of the individual organism been predicted in more than astatisticalsense.Thelearningprocesseswhicharethepresumedobjectofsuchresearcharereachedonlythroughaseriesofinferences.Currentpreoccupationwithdeductivesystemsreflectsthisstateofthescience.Recent improvements in the conditions which control behavior in the field of

learning are of two principal sorts. TheLaw of Effect has been taken seriously;wehavemadesurethateffectsdooccurand that theyoccurunderconditionswhichareoptimal for producing the changes called learning. Once we have arranged theparticular type of consequence called a reinforcement, our techniques permit us toshapeupthebehaviorofanorganismalmostatwill.Ithasbecomearoutineexercisetodemonstratethisinclassesinelementarypsychologybyconditioningsuchanorganismas a pigeon. Simply by presenting food to a hungry pigeon at the right time, it ispossible to shape up three or four well-defined responses in a single demonstrationperiod—suchresponsesasturningaround,pacingthefloorinthepatternofafigure-8,standing still in a corner of the demonstration apparatus, stretching the neck, orstamping the foot. Extremely complex performances may be reached throughsuccessive stages in the shaping process, the contingencies of reinforcement beingchangedprogressively in thedirectionof the requiredbehavior.Theresultsareoften

185

quitedramatic.Insuchademonstrationonecanseelearningtakeplace.Asignificantchangeinbehaviorisoftenobviousastheresultofasinglereinforcement.Asecondimportantadvancein techniquepermitsus tomaintainbehavior ingiven

statesofstrengthforlongperiodsoftime.Reinforcementscontinuetobeimportant,ofcourse, long after an organism has learned how to do something, long after it hasacquiredbehavior.Theyarenecessarytomaintainthebehaviorinstrength.Ofspecialinterestistheeffectofvariousschedulesofintermittentreinforcement.Mostimportanttypesofscheduleshavenowbeeninvestigated,andtheeffectsofschedulesingeneralhavebeenreducedtoafewprinciples.1On the theoreticalsidewenowhaveafairlygood idea of why a given schedule produces its appropriate performance. On thepractical sidewe have learned how tomaintain any given level of activity for dailyperiods limitedonlyby thephysicalexhaustionof theorganismandfromdaytodaywithout substantial change throughout its life. Many of these effects would betraditionally assigned to the field of motivation, although the principal operation issimplythearrangementofcontingenciesofreinforcement.Thesenewmethodsofshapingbehaviorandofmaintainingitinstrengthareagreat

improvementoverthetraditionalpracticesofprofessionalanimaltrainers,anditisnotsurprising that our laboratory results are already being applied to the production ofperforming animals for commercial purposes. In amore academic environment theyhave been used for demonstration purposes which extend far beyond an interest inlearning as such. For example, it is not too difficult to arrange the complexcontingencies which produce many types of social behavior. Competition isexemplifiedbytwopigeonsplayingamodifiedgameofping-pong.Thepigeonsdrivetheballbackandforthacrossasmalltablebypeckingatit.Whentheballgetsbyonepigeon, the other is reinforced. The task of constructing such a “social relation” isprobablycompletelyoutofreachofthetraditionalanimaltrainer.Itrequiresacarefullydesignedprogramofgraduallychangingcontingenciesandtheskillfuluseofschedulestomaintainthebehaviorinstrength.Eachpigeonisseparatelypreparedforitspartinthe total performance, and the “social relation” is then arbitrarily constructed. Thesequenceofeventsleadinguptothisstablestateareexcellentmaterialforthestudyofthefactorsimportantinnonsyntheticsocialbehavior.Itisinstructivetoconsiderhowasimilarseriesofcontingenciescouldariseinthecaseofthehumanorganismthroughtheevolutionofculturalpatterns.Co-operation can also be set up, perhapsmore easily than competition.We have

trained two pigeons to co-ordinate their behavior in a co-operative endeavor with aprecisionwhichequalsthatofthemostskillfulhumandancers.Inamoreseriousveinthese techniques have permitted us to explore the complexities of the individualorganism and to analyze some of the serial or co-ordinate behaviors involved inattention,problemsolving,varioustypesofself-control,andthesubsidiarysystemofresponses within a single organism called “personalities.” Some of these areexemplified inwhatwecallmultiple schedulesof reinforcement. Ingeneral, agivenschedulehasaneffectupontherateatwhicharesponseisemitted.Changesintheratefrommomenttomomentshowapatterntypicaloftheschedule.Thepatternmaybeassimple as a constant rate of responding at a given value, it may be a graduallyaccelerating rate between certain extremes, it may be an abrupt change from notresponding at all to a given stable high rate, and so on. It has been shown that theperformancecharacteristicofagivenschedulecanbebroughtunder thecontrolofa

186

particularstimulusandthatdifferentperformancescanbebroughtunderthecontrolofdifferent stimuli in the same organism. At a recent meeting of the AmericanPsychologicalAssociation,C.B.FersterandIdemonstratedapigeonwhosebehaviorshowed the pattern typical of “fixed-interval” reinforcement in the presence of onestimulusand,alternately,thepatterntypicaloftheverydifferentschedulecalled“fixedratio” in the presence of a second stimulus. In the laboratorywe have been able toobtain performances appropriate to nine different schedules in the presence ofappropriate stimuli in random alternation. When Stimulus 1 is present, the pigeonexecutes theperformanceappropriate toSchedule1.WhenStimulus2 ispresent, thepigeonexecutestheperformanceappropriatetoSchedule2.Andsoon.Thisresult isimportantbecauseitmakestheextrapolationofourlaboratoryresultstodailylifemuchmore plausible. We are all constantly shifting from schedule to schedule as ourimmediate environment changes, but the dynamics of the control exercised byreinforcementremainessentiallyunchanged.Itisalsopossibletoconstructverycomplexsequencesofschedules.Itisnoteasyto

describe these in a fewwords,but twoor three examplesmaybementioned. Inoneexperiment thepigeongeneratesaperformanceappropriate toScheduleAwhere thereinforcementissimplytheproductionofthestimuluscharacteristicofScheduleB,towhichthepigeonthenrespondsappropriately.Underathirdstimulus,thebirdyieldsaperformanceappropriatetoScheduleCwherethereinforcementinthiscaseissimplythe production of the stimulus characteristic of Schedule D, to which the bird thenrespondsappropriately. Inaspecialcase,first investigatedbyL.B.Wyckoff,Jr., theorganismrespondstoonestimuluswherethereinforcementconsistsoftheclarificationofthestimuluscontrollinganotherresponse.Thefirstresponsebecomes,sotospeak,an objective form of “paying attention” to the second stimulus. In one importantversionofthisexperiment,asyetunpublished,wecouldsaythatthepigeonistellinguswhetheritis“payingattention”totheshapeofaspotoflightortoitscolor.Oneofthemostdramaticapplicationsofthesetechniqueshasrecentlybeenmadein

theHarvardPsychologicalLaboratoriesbyFloydRatliffandDonaldS.Blough,whohave skillfullyusedmultiple and serial schedulesof reinforcement to study complexperceptual processes in the infrahuman organism. They have achieved a sort ofpsychophysics without verbal instruction. In a recent experiment by Blough, forexample, a pigeon draws a detailed dark-adaptation curve showing the characteristicbreaks of rod and cone vision. The curve is recorded continuously in a singleexperimentalperiodand isquitecomparablewith thecurvesofhumansubjects.Thepigeonbehavesinawaywhich,inthehumancase,wewouldnothesitatetodescribebysaying that itadjustsaveryfaintpatchof lightuntil itcan justbeseen[seepage145].Inallthiswork,thespeciesoftheorganismhasmadesurprisinglylittledifference.It

istruethattheorganismsstudiedhaveallbeenvertebrates,buttheystillcoverawiderange. Comparable results have been obtained with pigeons, rats, dogs, monkeys,human children, and most recently, by the author in collaboration with Ogden R.Lindsley,humanpsychoticsubjects.Inspiteofgreatphylogeneticdifferences,alltheseorganisms show amazingly similar properties of the learning process. It should beemphasizedthatthishasbeenachievedbyanalyzingtheeffectsofreinforcementandbydesigningtechniqueswhichmanipulatereinforcementwithconsiderableprecision.Only in thiswaycan thebehaviorof the individualorganismbebroughtunder suchprecisecontrol.Itisalsoimportanttonotethatthroughagradualadvancetocomplex

187

interrelationsamongresponses,thesamedegreeofrigorisbeingextendedtobehaviorwhichwouldusuallybeassignedtosuchfieldsasperception,thinking,andpersonalitydynamics.Fromthisexcitingprospectofanadvancingscienceoflearning,itisagreatshockto

turn to thatbranchof technologywhich ismostdirectlyconcernedwith the learningprocess—education. Let us consider, for example, the teaching of arithmetic in thelowergrades.Theschool is concernedwith imparting to thechilda largenumberofresponsesofaspecialsort.Theresponsesareallverbal.Theyconsistofspeakingandwritingcertainwords,figures,andsignswhich,toputitroughly,refertonumbersandtoarithmeticoperations.Thefirsttaskistoshapeuptheseresponses—togetthechildto pronounce and towrite responses correctly, but the principal task is to bring thisbehaviorundermany sortsof stimulus control.This iswhathappenswhen the childlearnstocount,torecitetables,tocountwhiletickingofftheitemsinanassemblageofobjects, to respond to spoken orwritten numbers by saying “odd,” “even,” “prime,”and so on. Over and above this elaborate repertoire of numerical behavior, most ofwhich is often dismissed as the product of rote learning, the teaching of arithmeticlooksforwardto thosecomplexserialarrangementsofresponses involvedinoriginalmathematical thinking. The child must acquire responses of transposing, clearingfractions,andsoon,whichmodifytheorderorpatternoftheoriginalmaterialsothattheresponsecalledasolutioniseventuallymadepossible.Now,howisthisextremelycomplicatedverbalrepertoiresetup?Inthefirstplace,

what reinforcementsareused?Fiftyyearsago theanswerwouldhavebeenclear.Atthattimeeducationalcontrolwasstillfranklyaversive.Thechildreadnumbers,copiednumbers, memorized tables, and performed operations upon numbers to escape thethreatofthebirchrodorcane.Somepositivereinforcementswereperhapseventuallyderivedfromtheincreasedefficiencyofthechildinthefieldofarithmetic,andinrarecasessomeautomaticreinforcementmayhaveresultedfromthesheermanipulationofthemedium—fromthesolutionofproblemsor thediscoveryof the intricaciesof thenumbersystem.Butfortheimmediatepurposesofeducationthechildactedtoavoidorescape punishment. It was part of the reform movement known as progressiveeducationtomakethepositiveconsequencesmoreimmediatelyeffective,butanyonewhovisitsthelowergradesoftheaverageschooltodaywillobservethatachangehasbeen made, not from aversive to positive control, but from one form of aversivestimulation to another. The child at his desk, filling in his workbook, is behavingprimarilytoescapefromthethreatofaseriesofminoraversiveevents—theteacher’sdispleasure, the criticism or ridicule of his classmates, an ignominious showing in acompetition,lowmarks,atriptotheoffice“tobetalkedto”bytheprincipal,orawordto the parent who may still resort to the birch rod. In this welter of aversiveconsequences,gettingtherightanswerisinitselfaninsignificantevent,anyeffectofwhich is lost amid the anxieties, the boredom, and the aggressions which are theinevitableby-productsofaversivecontrol.Secondly, we have to ask how the contingencies of reinforcement are arranged.

Whenisanumericaloperationreinforcedas“right”?Eventually,ofcourse, thepupilmay be able to check his own answers and achieve some sort of automaticreinforcement, but in the early stages the reinforcement of being right is usuallyaccordedby the teacher.Thecontingencies sheprovidesare far fromoptimal. It caneasily be demonstrated that, unless explicitmediating behavior has been set up, thelapseofonlyafewsecondsbetweenresponseandreinforcementdestroysmostofthe

188

effect. In a typical classroom, nevertheless, long periods of time customarily elapse.Theteachermaywalkupanddowntheaisle,forexample,whiletheclassisworkingonasheetofproblems,pausinghereandtheretosayrightorwrong.Manysecondsorminutes intervene between the child’s response and the teacher’s reinforcement. Inmanycases—forexample,whenpapersaretakenhometobecorrected—asmuchas24hoursmayintervene.Itissurprisingthatthissystemhasanyeffectwhatsoever.Athirdnotableshortcomingisthelackofaskillfulprogramwhichmovesforward

throughaseriesofprogressiveapproximationstothefinalcomplexbehaviordesired.Alongseriesofcontingencies isnecessary tobringtheorganisminto thepossessionofmathematicalbehaviormostefficiently.Buttheteacherisseldomabletoreinforceateachstepinsuchaseriesbecauseshecannotdealwiththepupil’sresponsesoneatatime. It is usually necessary to reinforce the behavior in blocks of responses—as incorrectingaworksheetorpagefromaworkbook.Theresponseswithinsuchablockmustnotbeinterrelated.Theanswertooneproblemmustnotdependupontheanswerto another. The number of stages through which onemay progressively approach acomplexpatternofbehavioristhereforesmall,andthetasksomuchthemoredifficult.Eventhemostmodernworkbookinbeginningarithmeticisfarfromexemplifyinganefficientprogramforshapingupmathematicalbehavior.Perhaps the most serious criticism of the current classroom is the relative

infrequencyofreinforcement.Sincethepupilisusuallydependentupontheteacherforbeing right, andsincemanypupilsareusuallydependentupon the same teacher, thetotalnumberofcontingencieswhichmaybearrangedduring,say,thefirstfouryearsisoftheorderofonlyafewthousand.Butaveryroughestimatesuggests thatefficientmathematical behavior at this level requires something of the order of 25,000contingencies.Wemaysupposethateveninthebrighterstudentagivencontingencymustbearrangedseveraltimestoplacethebehaviorwellinhand.Theresponsestobesetuparenotsimplythevariousitemsintablesofaddition,subtraction,multiplication,anddivision;wehavealsotoconsiderthealternativeformsinwhicheachitemmaybestated. To the learning of such material we should add hundreds of responsesconcerned with factoring, identifying primes, memorizing series, using short-cuttechniquesofcalculation,constructingandusinggeometricrepresentationsornumberforms,andsoon.Overandaboveallthis,thewholemathematicalrepertoiremustbebroughtunderthecontrolofconcreteproblemsofconsiderablevariety.Perhaps50,000contingencies is a more conservative estimate. In this frame of reference the dailyassignmentinarithmeticseemspitifullymeagre.The result of all this is, of course, well known. Even our best schools are under

criticismfortheirinefficiencyintheteachingofdrillsubjectssuchasarithmetic.Thecondition in the average school is amatter ofwidespread national concern.Modernchildren simply do not learn arithmetic quickly or well. Nor is the result simplyincompetence.Theverysubjectsinwhichmoderntechniquesareweakestarethoseinwhichfailureismostconspicuous,andinthewakeofanever-growingincompetencecome the anxieties, uncertainties, and aggressions which in their turn present otherproblems to the school.Mostpupils soonclaim the asylumofnotbeing“ready” forarithmetic at a given level or, eventually, of not having amathematicalmind. Suchexplanationsarereadilyseizeduponbydefensiveteachersandparents.Fewpupilseverreachthestageatwhichautomaticreinforcementsfollowasthenaturalconsequencesofmathematical behavior. On the contrary, the figures and symbols ofmathematicshavebecome standard emotional stimuli.Theglimpseof a columnof figures, not to

189

say an algebraic symbol or an integral sign, is likely to set off—not mathematicalbehavior—butareactionofanxiety,guilt,orfear.Theteacherisusuallynohappieraboutthisthanthepupil.Deniedtheopportunityto

controlviathebirchrod,quiteatseaastothemodeofoperationofthefewtechniquesat her disposal, she spends as little time as possible on drill subjects and eagerlysubscribes tophilosophiesofeducationwhichemphasizematerialofgreater inherentinterest.Aconfessionofweaknessisherextraordinaryconcernlestthechildbetaughtsomething unnecessary. The repertoire to be imparted is carefully reduced to anessential minimum. In the field of spelling, for example, a great deal of time andenergyhasgone intodiscovering just thosewordswhich theyoungchild isgoing touse,asifitwereacrimetowasteone’seducationalpowerinteachinganunnecessaryword.Eventually,weaknessoftechniqueemergesinthedisguiseofareformulationoftheaimsofeducation.Skillsareminimizedinfavorofvagueachievements—educatingfordemocracy,educatingthewholechild,educatingforlife,andsoon.Andtherethematterends;for,unfortunately,thesephilosophiesdonotinturnsuggestimprovementsintechniques.Theyofferlittleornohelpinthedesignofbetterclassroompractices.Therewouldbenopointinurgingtheseobjectionsifimprovementwereimpossible.

Buttheadvanceswhichhaverecentlybeenmadeinourcontrolofthelearningprocesssuggestathoroughrevisionofclassroompracticesand,fortunately,theytellushowtherevisioncanbebroughtabout.Thisisnot,ofcourse,thefirsttimethattheresultsofanexperimental science have been brought to bear upon the practical problems ofeducation. The modern classroom does not, however, offer much evidence thatresearchinthefieldoflearninghasbeenrespectedorused.Thisconditionisnodoubtpartlydue to the limitationsof earlier research.But it hasbeenencouragedbya toohastyconclusion that the laboratorystudyof learning is inherently limitedbecause itcannot take into account the realitiesof the classroom. In the lightofour increasingknowledgeof the learningprocessweshould, instead, insistupondealingwith thoserealities and forcing a substantial change in them. Education is perhaps the mostimportantbranchof scientific technology. Itdeeplyaffects the livesofallofus.Wecannolongerallowtheexigenciesofapracticalsituationtosuppressthetremendousimprovementswhicharewithinreach.Thepracticalsituationmustbechanged.Therearecertainquestionswhichhavetobeansweredinturningtothestudyofany

new organism.What behavior is to be set up?What reinforcers are at hand?Whatresponses are available in embarking upon a program of progressive approximationwhichwill lead to the final formof the behavior?How can reinforcements bemostefficiently scheduled to maintain the behavior in strength? These questions are allrelevantinconsideringtheproblemofthechildinthelowergrades.Inthefirstplace,whatreinforcementsareavailable?Whatdoestheschoolhavein

its possessionwhichwill reinforce a child?Wemay look first to thematerial to belearned, for it ispossible that thiswillprovideconsiderableautomatic reinforcement.Childrenplayforhourswithmechanicaltoys,paints,scissorsandpaper,noise-makers,puzzles—in short,with almost anythingwhich feeds back significant changes in theenvironmentandisreasonablyfreeofaversiveproperties.Thesheercontrolofnatureis itself reinforcing. This effect is not evident in the modern school because it ismasked by the emotional responses generated by aversive control. It is true thatautomaticreinforcementfromthemanipulationoftheenvironmentisprobablyonlyamildreinforcerandmayneedtobecarefullyhusbanded,butoneofthemoststrikingprinciplestoemergefromrecentresearchisthatthenetamountofreinforcementisof

190

little significance. A very slight reinforcement may be tremendously effective incontrollingbehaviorifitiswiselyused.If the natural reinforcement inherent in the subject matter is not enough, other

reinforcersmustbeemployed.Eveninschoolthechildisoccasionallypermittedtodo“what he wants to do,” and access to reinforcements of many sorts may be madecontingentupon themore immediateconsequencesof thebehavior tobeestablished.Those who advocate competition as a useful social motive may wish to use thereinforcementswhichfollowfromexcellingothers,althoughthereisthedifficultythatin thiscase thereinforcementofonechild isnecessarilyaversive toanother.Next inorderwemightplacethegoodwillandaffectionoftheteacher,andonlywhenthathasfailedneedweturntotheuseofaversivestimulation.In thesecondplace,howare these reinforcements tobemadecontingentupon the

desired behavior? There are two considerations here—the gradual elaboration ofextremely complex patterns of behavior and the maintenance of the behavior instrengthateachstage.Thewholeprocessofbecomingcompetentinanyfieldmustbedivided into a very large number of very small steps, and reinforcement must becontingent upon the accomplishment of each step. This solution to the problem ofcreatingacomplexrepertoireofbehavioralsosolves theproblemofmaintaining thebehavior in strength. We could, of course, resort to the techniques of schedulingalready developed in the study of other organisms but in the present state of ourknowledgeofeducationalpractices,schedulingappearstobemosteffectivelyarrangedthrough the designof thematerial to be learned.Bymaking each successive step assmallaspossible, thefrequencyofreinforcementcanberaisedtoamaximum,whilethepossiblyaversiveconsequencesofbeingwrongarereducedtoaminimum.Otherways of designing material would yield other programs of reinforcement. Anysupplementary reinforcement would probably have to be scheduled in the moretraditionalway.Theserequirementsarenotexcessive,but theyareprobably incompatiblewith the

current realities of the classroom. In the experimental study of learning it has beenfound that thecontingenciesof reinforcementwhicharemostefficient incontrollingtheorganismcannotbearrangedthroughthepersonalmediationof theexperimenter.An organism is affected by subtle details of contingencies which are beyond thecapacityofthehumanorganismtoarrange.Mechanicalandelectricaldevicesmustbeused.Mechanicalhelpisalsodemandedby thesheernumberofcontingencieswhichmay be used efficiently in a single experimental session. We have recorded manymillionsofresponsesfromasingleorganismduringthousandsofexperimentalhours.Personalarrangementof thecontingenciesandpersonalobservationof theresultsarequiteunthinkable.Now,thehumanorganismis,ifanything,moresensitivetoprecisecontingencies than the other organisms we have studied. We have every reason toexpect, therefore, that the most effective control of human learning will requireinstrumentalaid.Thesimplefactisthat,asamerereinforcingmechanism,theteacheris out of date. Thiswould be true even if a single teacher devoted all her time to asingle child, but her inadequacy is multiplied manyfold when she must serve as areinforcingdevicetomanychildrenatonce.Iftheteacheristotakeadvantageofrecentadvancesinthestudyoflearning,shemusthavethehelpofmechanicaldevices.Thetechnicalproblemofprovidingthenecessaryinstrumentalaidisnotparticularly

difficult.Therearemanywaysinwhichthenecessarycontingenciesmaybearranged,eithermechanically or electrically. An inexpensive devicewhich solvesmost of the

191

principalproblemshasalreadybeenconstructed.Itisstillintheexperimentalstage,buta description will suggest the kind of instrument which seems to be required. Thedeviceconsistsofaboxaboutthesizeofasmallrecordplayer.Onthetopsurfaceisaglazedwindowthroughwhichaquestionorproblemprintedonapaper tapemaybeseen.Thechildanswers thequestionbymovingoneormore slidersuponwhich thedigits0through9areprinted.Theanswerappearsinsquareholespunchedinthepaperuponwhich thequestion is printed.When the answerhas been set, the child turns aknob.Theoperationisassimpleasadjustingatelevisionset.Iftheanswerisright,theknob turns freely and can bemade to ring a bell or provide some other conditionedreinforcement.Iftheansweriswrong,theknobwillnotturn.Acountermaybeaddedtotallywronganswers.Theknobmustthenbereversedslightlyandasecondattemptat a right answer made. (Unlike the flash-card, the device reports a wrong answerwithoutgivingtherightanswer.)Whentheanswerisright,afurtherturnoftheknobengages a clutch which moves the next problem into place in the window. Thismovementcannotbecompleted,however,untiltheslidershavebeenreturnedtozero.Theimportantfeaturesofthedevicearethese:Reinforcementfortherightansweris

immediate.Themeremanipulationofthedevicewillprobablybereinforcingenoughto keep the average pupil atwork for a suitable period each day, provided traces ofearlieraversivecontrolcanbewipedout.A teachermaysuperviseanentireclassatworkonsuchdevicesat thesametime,yeteachchildmayprogressathisown rate,completingasmanyproblemsaspossiblewithintheclassperiod.Ifforcedtobeawayfromschool,hemayreturntopickupwhereheleftoff.Thegiftedchildwilladvancerapidly, but can be kept from getting too far ahead either by being excused fromarithmetic fora timeorbybeinggivenspecial setsofproblemswhich takehimintosomeoftheinterestingbypathsofmathematics.

192

FIG.1.Arecentmodelofateachingmachineforthelowergrades.Themachineoperatesontheprinciplesdescribedintheaccompanyingarticle.Materialispresentedinawindowwithafewlettersorfiguresmissing.Thepupilmovessliderswhichcauselettersorfigurestoappear.Whenananswerhasbeencomposed,thepupilturnsacrank.Iftheanswerwasright,anewframeofmaterialmovesintothewindowandtheslidersreturntotheirhomeposition.Ifthematerialwas

wrong,theslidersreturnbuttheframeremainsandmustbecompletedagain.

The devicemakes it possible to present carefully designedmaterial inwhich oneproblemcandependupontheanswertotheprecedingandwhere,therefore,themostefficientprogresstoaneventuallycomplexrepertoirecanbemade.Provisionhasbeenmade for recording the commonest mistakes so that the tapes can be modified asexperiencedictates.Additionalstepscanbeinsertedwherepupilstendtohavetrouble,andultimatelythematerialwillreachapointatwhichtheanswersoftheaveragechildwillalmostalwaysberight.Ifthematerialitselfprovesnottobesufficientlyreinforcing,otherreinforcersinthe

possessionoftheteacherorschoolmaybemadecontingentupontheoperationofthedevice or upon progress through a series of problems. Supplemental reinforcementwouldnotsacrificetheadvantagesgainedfromimmediatereinforcementandfromthepossibility of constructing an optimal series of steps which approach the complexrepertoireofmathematicalbehaviormostefficiently.A similar device in which the sliders carry the letters of the alphabet has been

designed to teach spelling. In addition to the advantageswhich can be gained from

193

precise reinforcement and careful programming, the devicewill teach reading at thesametime.Itcanalsobeusedtoestablishthelargeandimportantrepertoireofverbalrelationshipsencounteredinlogicandscience.Inshort,itcanteachverbalthinking.Astocontentinstruction,thedevicecanbeoperatedasamultiple-choiceself-rater.Someobjectionstotheuseofsuchdevicesintheclassroomcaneasilybeforeseen.

The cry will be raised that the child is being treated as a mere animal and that anessentially human intellectual achievement is being analyzed in unduly mechanisticterms. Mathematical behavior is usually regarded, not as a repertoire of responsesinvolvingnumbersandnumericaloperations,butasevidencesofmathematicalabilityortheexerciseofthepowerofreason.Itistruethatthetechniqueswhichareemergingfromtheexperimentalstudyoflearningarenotdesignedto“developthemind”ortofurthersomevague“understanding”ofmathematicalrelationships.Theyaredesigned,onthecontrary,toestablishtheverybehaviorswhicharetakentobetheevidencesofsuchmentalstatesorprocesses.Thisisonlyaspecialcaseofthegeneralchangewhichisunderwayintheinterpretationofhumanaffairs.Anadvancingsciencecontinuestooffermoreandmoreconvincingalternativestotraditionalformulations.Thebehaviorin termsofwhichhuman thinkingmusteventuallybedefined isworth treating in itsownrightasthesubstantialgoalofeducation.Ofcoursetheteacherhasamoreimportantfunctionthantosayrightorwrong.The

changesproposedwouldfreeherfortheeffectiveexerciseofthatfunction.Markingasetofpapersinarithmetic—“Yes,nineandsixarefifteen;no,nineandsevenarenoteighteen”—isbeneaththedignityofanyintelligentindividual.Thereismoreimportantworktobedone—inwhichtheteacher’srelationstothepupilcannotbeduplicatedbya mechanical device. Instrumental help would merely improve these relations. Onemight say that themain troublewith education in the lower grades today is that thechild is obviously not competent and knows it and that the teacher is unable to doanythingaboutitandknowsthattoo.Iftheadvanceswhichhaverecentlybeenmadeinourcontrolofbehaviorcangive thechildagenuinecompetence in reading,writing,spelling,andarithmetic,thentheteachermaybegintofunction,notinlieuofacheapmachine, but through intellectual, cultural, and emotional contacts of that distinctivesortwhichtestifytoherstatusasahumanbeing.Another possible objection is thatmechanized instructionwillmean technological

unemployment.We need notworry about this until there are enough teachers to goaroundanduntilthehoursandenergydemandedoftheteacherarecomparabletothosein other fields of employment.Mechanical deviceswill eliminate themore tiresomelaborsof the teacherbut theywillnotnecessarily shorten the timeduringwhichsheremainsincontactwiththepupil.Amorepracticalobjection:Canweaffordtomechanizeourschools?Theansweris

clearlyyes.ThedeviceIhavejustdescribedcouldbeproducedascheaplyasasmallradioorphonograph.Therewouldneed tobe far fewerdevices thanpupils, for theycouldbeusedinrotation.Butevenifwesupposethattheinstrumenteventuallyfoundtobemosteffectivewouldcostseveralhundreddollarsandthatlargenumbersofthemwould be required, our economy should be able to stand the strain. Once we haveaccepted the possibility and the necessity of mechanical help in the classroom, theeconomicproblemcaneasilybesurmounted.Thereisnoreasonwhytheschoolroomshould be any less mechanized than, for example, the kitchen. A country whichannually produces millions of refrigerators, dish-washers, automatic washing-machines, automatic clothes-driers, and automatic garbage disposers can certainly

194

affordtheequipmentnecessarytoeducateitscitizenstohighstandardsofcompetenceinthemosteffectiveway.There is a simple job to be done. The task can be stated in concrete terms. The

necessary techniques are known. The equipment needed can easily be provided.Nothing stands in the way but cultural inertia. But what is more characteristic ofAmerica thananunwillingness toaccept the traditionalas inevitable?Weareon thethresholdofanexcitingandrevolutionaryperiod,inwhichthescientificstudyofmanwillbeputtoworkinman’sbestinterests.Educationmustplayitspart.Itmustacceptthe fact that a sweeping revision of educational practices is possible and inevitable.Whenithasdonethis,wemaylookforwardwithconfidencetoaschoolsystemwhichisawareofthenatureofitstasks,secureinitsmethods,andgenerouslysupportedbytheinformedandeffectivecitizenswhomeducationitselfwillcreate.

1SeeSchedulesofReinforcement.

195

TeachingMachines

This paper, published in Science (1958, 128, 969-977) and reprinted here bypermission,waspartofareporttotheFundfortheAdvancementofEducationwhichhad sponsoreda two-year test ofmachine instructionat thehigh-schooland collegelevels.

THEREAREMOREpeopleintheworldthaneverbefore,andafargreaterpartofthemwantaneducation.Thedemandcannotbemet simplybybuildingmore schoolsandtrainingmore teachers.Educationmust becomemore efficient.To this end curriculamustberevisedandsimplified,andtextbooksandclassroomtechniquesimproved.Inany other field a demand for increased production would have led at once to theinventionoflabor-savingcapitalequipment.Educationhasreachedthisstageverylate,possibly through a misconception of its task. Thanks to the advent of television,however, the so-called audio-visual aids are being re-examined. Film projectors,television sets, phonographs, and tape recorders are finding theirway intoAmericanschoolsandcolleges.Audio-visualaidssupplementandmayevensupplantlectures,demonstrations,and

textbooks.Indoingsotheyserveonefunctionoftheteacher:theypresentmaterialtothe student and, when successful, make it so clear and interesting that the studentlearns.There is another function towhich they contribute little or nothing. It is bestseenintheproductiveinterchangebetweenteacherandstudentinthesmallclassroomortutorialsituation.MuchofthatinterchangehasalreadybeensacrificedinAmericaneducationinordertoteachlargenumbersofstudents.Thereisarealdangerthatitwillbewhollyobscuredifuseofequipmentdesignedsimplytopresentmaterialbecomeswidespread. The student is becoming more and more a mere passive receiver ofinstruction.

Pressey’sTeachingMachines

Thereisanotherkindofcapitalequipmentwhichwillencouragethestudenttotakeanactiveroleintheinstructionalprocess.Thepossibilitywasrecognizedinthe1920’s,when Sidney L. Pressey designed several machines for the automatic testing ofintelligenceandinformation.ArecentmodelofoneoftheseisshowninFigure1. Inusing the device the student refers to a numbered item in amultiple-choice test.Hepressesthebuttoncorrespondingtohisfirstchoiceofanswer.Ifheisright,thedevicemovesontothenextitem;ifheiswrong,theerroristallied,andhemustcontinuetomakechoicesuntilheisright.1Suchmachines,Presseypointedout,2 couldnotonlytestandscore,theycouldteach.Whenanexaminationiscorrectedandreturnedafteradelayofmanyhoursordays,thestudent’sbehaviorisnotappreciablymodified.Theimmediate report supplied by a self-scoring device, however, can have an important

196

instructional effect. Pressey also pointed out that such machines would increaseefficiencyinanotherway.Eveninasmallclassroomtheteacherusuallyknowsthatheismoving tooslowly for somestudentsand too fast forothers.Thosewhocouldgofaster are penalized, and those who should go slower are poorly taught andunnecessarilypunishedbycriticismandfailure.Machineinstructionwouldpermiteachstudenttoproceedathisownrate.

FIG.1.Pressey’sself-testingmachine.Thedevicedirectsthestudenttoaparticulariteminamultiple-choicetest.Thestudentpressesthekeycorrespondingtohischoiceofanswer.Ifcorrect,

thedeviceadvancestothenextitem.Errorsaretotaled.

The “industrial revolution in education” which Pressey envisioned stubbornlyrefused tocomeabout. In1932heexpressedhisdisappointment.3 “The problems ofinvention are relatively simple,” he wrote. “With a little money and engineeringresource,agreatdealcouldeasilybedone.Thewriterhasfoundfrombitterexperience

197

that one person alone can accomplish relatively little and he is regretfully droppingfurther work on these problems. But he hopes that enough may have been done tostimulateotherworkers,thatthisfascinatingfieldmaybedeveloped.”Pressey’smachinessuccumbedinparttoculturalinertia;theworldofeducationwas

notreadyfor them.But theyalsohadlimitationswhichprobablycontributedto theirfailure.Presseywasworkingagainstabackgroundofpsychologicaltheorywhichhadnot come to grips with the learning process. The study of human learning wasdominated by the “memory drum” and similar devices originally designed to studyforgetting.Rate of learningwas observed, but littlewas done to change it.Why thesubjectofsuchanexperimentbotheredtolearnatallwasoflittleinterest.“Frequency”and “recency” theories of learning, and principles of “massed and spaced practice,”concernedtheconditionsunderwhichresponseswereremembered.Pressey’smachinesweredesignedagainst this theoreticalbackground.Asversions

of the memory drum, they were primarily testing devices. They were used after acertainamountof learninghadalready takenplaceelsewhere.Byconfirmingcorrectresponses andbyweakening responseswhich shouldnothavebeen acquired, a self-testingmachinedoes, indeed, teach;but it isnotdesignedprimarilyfor thatpurpose.Nevertheless, Pressey seems to have been the first to emphasize the importance ofimmediatefeedbackineducationandtoproposeasysteminwhicheachstudentcouldmove at his own pace. He saw the need for capital equipment in realizing theseobjectives.Above all he conceived of amachinewhich (in contrastwith the audio-visual aidswhichwere beginning to be developed) permitted the student to play anactiverole.

AnotherKindofMachine

The learningprocess is nowmuchbetterunderstood.Muchofwhatweknowhascomefromstudyingthebehavioroflowerorganisms,buttheresultsholdsurprisinglywell for human subjects. The emphasis in this research has not been on proving ordisprovingtheoriesbutondiscoveringandcontrollingthevariablesofwhichlearningisafunction.Thispracticalorientationhaspaidoff,forasurprisingdegreeofcontrolhasbeenachieved.Byarrangingappropriate“contingenciesofreinforcement,”specificformsofbehavior canbe setupandbroughtunder the controlof specific classesofstimuli.Theresultingbehaviorcanbemaintainedinstrengthforlongperiodsoftime.A technology based on this work has already been put to use in neurology,pharmacology,nutrition,psychophysics,psychiatry,andelsewhere[seepage132].Theanalysisisalsorelevanttoeducation.Astudentis“taught”inthesensethatheis

induced to engage in new forms of behavior and in specific forms upon specificoccasions. It is not merely a matter of teaching him what to do; we are as muchconcerned with the probability that appropriate behavior will, indeed, appear at theproper time—an issue which would be classed traditionally under motivation. Ineducation thebehavior tobe shapedandmaintained isusuallyverbal, and it is tobebroughtunderthecontrolofbothverbalandnonverbalstimuli.Fortunately,thespecialproblemsraisedbyverbalbehaviorcanbesubmittedtoasimilaranalysis.4Ifourcurrentknowledgeoftheacquisitionandmaintenanceofverbalbehavioristo

be applied to education, some sort of teachingmachine is needed. Contingencies of

198

reinforcementwhichchangethebehavioroflowerorganismsoftencannotbearrangedbyhand;ratherelaborateapparatusisneeded.Thehumanorganismrequiresevenmoresubtle instrumentation.An appropriate teachingmachinewill have several importantfeatures. The student must compose his response rather than select it from a set ofalternatives,asinamultiple-choiceself-rater.Onereasonforthisisthatwewanthimto recall rather than recognize—to make a response as well as see that it is right.Anotherreasonisthateffectivemultiple-choicematerialmustcontainplausiblewrongresponses,whichareoutofplaceinthedelicateprocessof“shaping”behaviorbecausetheystrengthenunwantedforms.Althoughitismucheasiertobuildamachinetoscoremultiple-choiceanswersthantoevaluateacomposedresponse,thetechnicaladvantageisoutweighedbytheseandotherconsiderations.A second requirement of a minimal teaching machine also distinguishes it from

earlier versions. In acquiring complex behavior the student must pass through acarefullydesignedsequenceofsteps,oftenofconsiderablelength.Eachstepmustbesosmallthatitcanalwaysbetaken,yetintakingitthestudentmovessomewhatclosertofullycompetentbehavior.Themachinemustmakesurethatthesestepsaretakeninacarefullyprescribedorder.Severalmachineswiththerequiredcharacteristicshavebeenbuiltandtested.Setsof

separate presentations or “frames” of visual material are stored on disks, cards, ortapes.Oneframeispresentedatatime,adjacentframesbeingoutofsight.Inonetypeofmachinethestudentcomposesaresponsebymovingprintedfiguresorletters[seepage189].Hissetting iscomparedby themachinewithacodedresponse. If the twocorrespond, the machine automatically presents the next frame. If they do not, theresponse iscleared,andanothermustbecomposed.Thestudentcannotproceed toasecond step until the first has been taken.Amachine of this kind is being tested inteachingspelling,arithmetic,andothersubjectsinthelowergrades.For more advanced students—from junior high school, say, through college—a

machine which senses an arrangement of letters or figures is unnecessarily rigid inspecifyingformofresponse.Fortunately,suchstudentsmaybeaskedtocomparetheirresponses with printed material revealed by the machine. In the machine shown inFigure2,materialisprintedin30radialframesona12-inchdisk.Thestudentinsertsthediskandclosesthemachine.Hecannotproceeduntilthemachinehasbeenlocked,and,oncehehasbegun,themachinecannotbeunlocked.Allbutacornerofoneframeisvisiblethroughawindow.Thestudentwriteshisresponseonapaperstripexposedthrough a second opening.By lifting a lever on the front of themachine, hemoveswhathehaswrittenunderatransparentcoveranduncoversthecorrectresponseintheremaining corner of the frame. If the two responses correspond, hemoves the leverhorizontally. This movement punches a hole in the paper opposite his response,recordingthefactthathecalleditcorrect,andaltersthemachinesothattheframewillnotappearagainwhenthestudentworksaroundthediskasecondtime.Whethertheresponsewascorrectornot,asecondframeappearswhenthe lever is returnedto itsstartingposition.Thestudentproceedsinthiswayuntilhehasrespondedtoallframes.Hethenworksaroundthediskasecondtime,butonlythoseframesappeartowhichhehasnotcorrectlyresponded.Whenthediskrevolveswithoutstopping,theassignmentisfinished.(Thestudentisaskedtorepeateachframeuntilacorrectresponseismadetoallowforthefactthat,intellinghimthataresponseiswrong,suchamachinetellshimwhatisright.)The machine itself, of course, does not teach. It simply brings the student into

199

contact with the person who composed the material it presents. It is a labor-savingdevicebecauseitcanbringoneprogrammerintocontactwithanindefinitenumberofstudents. This may suggest mass production, but the effect upon each student issurprisingly like thatofaprivate tutor.Thecomparisonholds in several respects. (i)There is a constant interchange between program and student. Unlike lectures,textbooks,andtheusualaudio-visualaids,themachineinducessustainedactivity.Thestudentisalwaysalertandbusy.(ii)Likeagoodtutor,themachineinsiststhatagivenpointbethoroughlyunderstood,eitherframebyframeorsetbyset,beforethestudentmoveson.Lectures, textbooks,and theirmechanizedequivalents,on the other hand,proceedwithoutmakingsurethatthestudentunderstandsandeasilyleavehimbehind.(iii)Likeagoodtutor,themachinepresentsjustthatmaterialforwhichthestudentisready.Itaskshimtotakeonlythatstepwhichheisatthemomentbestequippedandmostlikelytotake.(iv)Likeaskillfultutor,themachinehelpsthestudenttocomeupwith the right answer. It does this in part through the orderly construction of theprogram and in part with techniques of hinting, prompting, suggesting, and so on,derivedfromananalysisofverbalbehavior.(v)Lastly,ofcourse,themachine,liketheprivate tutor, reinforces the student for every correct response, using this immediatefeedbacknotonlytoshapehisbehaviormostefficientlybuttomaintainitinstrengthinamannerwhichthelaymanwoulddescribeas“holdingthestudent’sinterest.”

200

FIG.2.Studentatworkintheself-instructionroom.Oneframeofmaterialispartlyvisibleintheleft-handwindow.Thestudentwriteshisresponseonastripofpaperexposedattheright.Hethen

liftsaleverwithhislefthand,advancinghiswrittenresponseunderatransparentcoveranduncoveringthecorrectresponseintheuppercorneroftheframe.Ifheisright,hemovesthelevertotheright,punchingaholealongsidetheresponsehehascalledrightandalteringthemachinesothatthatframewillnotappearagainwhenhegoesthroughtheseriesasecondtime.Anewframe

appearswhentheleverisreturnedtoitsstartingposition.

ProgrammingMaterial

The success of such a machine depends on the material used in it. The task of

201

programming a given subject is at first sight rather formidable. Many helpfultechniquescanbederivedfromageneralanalysisoftherelevantbehavioralprocesses,verbal and nonverbal. Specific forms of behavior are to be evoked and, throughdifferentialreinforcement,broughtunderthecontrolofspecificstimuli.Thisisnottheplaceforasystematicreviewofavailabletechniques,orofthekindof

researchwhichmaybeexpectedtodiscoverothers.However,themachinesthemselvescannotbeadequatelydescribedwithoutgivingafewexamplesofprograms.Wemaybeginwithasetofframes(seeTable1)designedtoteachathird-orfourth-gradepupiltospellthewordmanufacture.Thesixframesarepresentedintheordershownandthepupilmovessliderstoexposelettersintheopensquares.The word to be learned appears in bold face in frame 1, with an example and a

simpledefinition.Thepupil’sfirsttaskissimplytocopyit.Whenhedoessocorrectly,frame2appears.Hemustnowcopyselectively:hemustidentifyfact as thecommonpartofmanufactureandfactory.Thishelpshimtospellthewordandalsotoacquireaseparable“atomic”verbaloperant.Inframe3anotherrootmustbecopiedselectivelyfrommanual.Inframe4thepupilmustforthefirsttimeinsertletterswithoutcopying.Since he is asked to insert the same letter in two places, a wrong response will bedoubly conspicuous, and the chance of failure is thereby minimized. The sameprinciple governs frame 5. In frame 6 the pupil spells the word to complete thesentence used as an example in frame 1. Even a poor student is likely to do thiscorrectlybecausehehas just composedor completed theword five times,hasmadetwoimportantroot-responses,andhaslearnedthattwolettersoccurinthewordtwice.Hehasprobablylearnedtospellthewordwithouthavingmadeamistake.Teaching spelling is mainly a process of shaping complex forms of behavior. In

othersubjects—forexample,arithmetic—responsesmustbebroughtunderthecontrolof appropriate stimuli. Unfortunately, the material which has been prepared forteachingarithmetic,withthehelpofSusanR.Meyer,doesnotlenditselftoexcerpting.Thenumbers 0 through9 are generated in relation to objects, quantities, and scales.The operations of addition, subtraction, multiplication, and division are thoroughlydevelopedbefore thenumber10 is reached. In thecourseof this thepupilcomposesequationsandexpressionsinagreatvarietyofalternativeforms.Hecompletesnotonly5+4=□,but□+4=9,5□4=9,andsoon,aided inmostcasesby illustrativematerials.Noappealismadetorotememorizing,eveninthelateracquisitionofthetables.Thestudentisexpectedtoarriveat9×7=63,notbymemorizingitashewouldmemorizealineofpoetry,butbyputtingintopracticesuchprinciplesasthatninetimesanumberisthesameastentimesthenumberminusthenumber(bothofthesebeing“obvious”or already well learned), that the digits in a multiple of nine add to nine, that incomposingsuccessivemultiplesofnineonecountsbackwards(nine,eighteen,twenty-seven,thirty-six,andsoon),thatninetimesasingledigitisanumberbeginningwithone less than the digit (nine times six is fifty something), and possibly even that theproduct of two numbers separated by only one number is equal to the square of theseparatingnumberminusone(thesquareofeightalreadybeingfamiliarfromaspecialseriesofframesconcernedwithsquares).

TABLE1ASETOFFRAMESDESIGNEDTOTEACHATHIRD-ORFOURTH-GRADEPUPILTO

SPELLTHEWORDmanufacture

202

Programsofthissortruntogreatlength.Atfiveorsixframesperword,fourgradesofspellingmayrequire20,000or25,000frames,andthreeorfourgradesofarithmeticasmanyagain. If these figures seem large, it isonlybecauseweare thinkingof thenormal contact between teacher and pupil. Admittedly, a teacher cannot supervise10,000or15,000responsesmadebyeachpupilperyear.Butthepupil’stimeisnotsolimited. In any case, surprisingly little time is needed. Fifteenminutes per day on amachineshouldsufficeforeachoftheseprograms,themachinesbeingfreeforotherstudents for the rest of each day. (It is probably because traditionalmethods are soinefficientthatwehavebeenledtosupposethateducationrequiressuchaprodigiouspartofayoungperson’sday.)A simple technique used in programming material at the high-school or college

level,bymeansofthemachineshowninFigure2,isexemplifiedinteachingastudentto recite apoem.The first line ispresentedwith severalunimportant lettersomitted.The student must read the line “meaningfully” and supply the missing letters. Thesecond,third,andfourthframespresentsucceedinglinesinthesameway.Inthefifthframe the first line reappears with other letters also missing. Since the student hasrecentlyreadtheline,hecancompleteitcorrectly.Hedoesthesameforthesecond,third,andfourthlines.Subsequentframesareincreasinglyincomplete,andeventually—say, after 20 or 24 frames—the student reproduces all four lineswithout externalhelp, and quite possibly without having made a wrong response. The technique issimilartothatusedinteachingspelling:responsesarefirstcontrolledbyatext,butthisisslowlyreduced(colloquially,“vanished”)untiltheresponsescanbeemittedwithoutatext,eachmemberinaseriesofresponsesbeingnowunderthe“intraverbal”controlofothermembers.“Vanishing”canbeusedinteachingothertypesofverbalbehavior.Whenastudent

describes the geography of part of theworld or the anatomyof part of the body, ornamesplantsandanimalsfromspecimensorpictures,verbalresponsesarecontrolledby nonverbal stimuli. In setting up such behavior the student is first asked to reportfeaturesofafullylabeledmap,picture,orobject,andthelabelsarethenvanished.Inteachingamap,forexample,themachineasksthestudenttodescribespatialrelations

203

amongcities,countries,rivers,andsoon,asshownonafullylabeledmap.Heisthenasked to do the same with a map in which the names are incomplete or, possibly,lacking.Eventuallyheisaskedtoreport thesamerelationswithnomapatall. If thematerialhasbeenwellprogrammed,hecandosocorrectly. Instruction issometimesconcerned not somuchwith imparting a new repertoire of verbal responses aswithgetting the student to describe something accurately in any available terms. Themachinecan“makesurethestudentunderstands”agraph,diagram,chart,orpicturebyaskinghimtoidentifyandexplainitsfeatures—correctinghim,ofcourse,wheneverheiswrong.Inadditiontocharts,maps,graphs,models,andsoon,thestudentmayhaveaccess

toauditorymaterial.Inlearningtotakedictationinaforeignlanguage,forexample,heselectsashortpassageonanindexingphonographaccordingtoinstructionsgivenbythemachine.Helistenstothepassageasoftenasnecessaryandthentranscribesit.Themachine then reveals thecorrect text.Thestudentmay listen to thepassageagain todiscoverthesourcesofanyerror.Theindexingphonographmayalsobeusedwiththemachinetoteachotherlanguageskills,aswellastelegraphiccode,music,speech,partsofliteraryanddramaticappreciation,andothersubjects.A typicalprogramcombinesmanyof these functions.Thesetof frames shown in

Table2 isdesigned to induce the studentofhigh-schoolphysics to talk intelligently,and to some extent technically, about the emission of light from an incandescentsource. In using themachine the student will write aword or phrase to complete agiven item and then uncover the corresponding word or phrase shown here in thecolumn at the right. The readerwhowishes to get the “feel” of thematerial shouldcover the right-hand column with a card, uncovering each line only after he hascompletedthecorrespondingitem.Several programming techniques are exemplified by the set of frames in Table2.

Technical terms are introduced slowly. For example, the familiar term finewire inframe2isfollowedbyadefinitionofthetechnicaltermfilamentinframe4;filamentisthenaskedforinthepresenceofthenonscientificsynonyminframe5andwithoutthesynonyminframe9.Inthesamewayglow,giveofflight,andsendoutlight inearlyframes are followed by a definition of emit with a synonym in frame 7. Variousinflected forms of emit then follow, and emit itself is asked for with a synonym inframe16. It isaskedforwithoutasynonymbut inahelpfulphrase inframe30,andemitted and emission are asked for without help in frames 33 and 34. The relationbetween temperature and amount and color of light is developed in several framesbefore a formal statement using the word temperature is asked for in frame 12.Incandescentisdefinedandusedinframe13,isusedagaininframe14,andisaskedfor in frame 15, the student receiving a thematic prompt from the recurring phrase“incandescentsourceoflight.”Aformalpromptissuppliedbycandle.Inframe25thenewresponseenergyiseasilyevokedbythewordsformof…because theexpression“formof energy” is used earlier in the frame.Energy appears again in the next twoframesandisfinallyaskedfor,withoutaid,inframe28.Frames30through35discussthe limiting temperatures of incandescent objects, while reviewing several kinds ofsources.TheFigure800isusedin threeframes.Twointerveningframesthenpermitsometimetopassbeforetheresponse800isaskedfor.

TABLE2

204

PARTOFAPROGRAMINHIGHSCHOOLPHYSICS.THEMACHINEPRESENTSONEITEMATATIME.THESTUDENTCOMPLETESTHEITEMANDTHENUNCOVERS

THECORRESPONDINGWORDORPHRASESHOWNATTHERIGHT

Sentencetobecompleted Wordtobesupplied

1. Theimportantpartsofaflashlightarethebatteryandthebulb.Whenwe“turnon”aflashlight,wecloseaswitchwhichconnectsthebatterywiththe——.

bulb

2. Whenweturnonaflashlight,anelectriccurrentflowsthroughthefinewireinthe——andcausesittogrowhot.

bulb

3. Whenthehotwireglowsbrightly,wesaythatitgivesofforsendsoutheatand——.

light

4. Thefinewireinthebulbiscalledafilament.Thebulb“lightsup”whenthefilamentisheatedbythepassageofa(n)——current.

electric

5. Whenaweakbatteryproduceslittlecurrent,thefinewire,or——,doesnotgetveryhot.

filament

6. Afilamentwhichislesshotsendsoutorgivesoff——light. less7. “Emit”means“sendout.”Theamountoflightsentout,or

“emitted,”byafilamentdependsonhow——thefilamentis.hot

8. Thehigherthetemperatureofthefilamentthe——thelightemittedbyit.

brighter,stronger

9. Ifaflashlightbatteryisweak,the——inthebulbmaystillglow,butwithonlyadullredcolor.

filament

10. Thelightfromaveryhotfilamentiscoloredyelloworwhite.Thelightfromafilamentwhichisnotveryhotiscolored——.

red

11. Ablacksmithorothermetalworkersometimesmakessurethatabarofironisheatedtoa“cherryred”beforehammeringitintoshape.Heusesthe——ofthelightemittedbythebartotellhowhotitis.

color

12. Boththecolorandtheamountoflightdependonthe——oftheemittingfilamentorbar.

temperature

13. Anobjectwhichemitslightbecauseitishotiscalled“incandescent.”Aflashlightbulbisanincandescentsourceof——.

light

14. Aneontubeemitslightbutremainscool.Itis,therefore,notanincandescent——oflight.

source

15. Acandleflameishot.Itisa(n)——sourceoflight. incandescent16. Thehotwickofacandlegivesoffsmallpiecesorparticlesof

carbonwhichburnintheflame.Beforeorwhileburning,thehotemit

205

particlessendout,or——,light.17. Alongcandlewickproducesaflameinwhichoxygendoesnot

reachallthecarbonparticles.Withoutoxygentheparticlescannotburn.Particleswhichdonotburnriseabovetheflameas——.

smoke

18. Wecanshowthatthereareparticlesofcarboninacandleflame,evenwhenitisnotsmoking,byholdingapieceofmetalintheflame.Themetalcoolssomeoftheparticlesbeforetheyburn,andtheunburnedcarbon——collectonthemetalassoot.

particles

19. Theparticlesofcarboninsootorsmokenolongeremitlightbecausetheyare——thanwhentheywereintheflame.

cooler,colder

20. Thereddishpartofacandleflamehasthesamecolorasthefilamentinaflashlightwithaweakbattery.Wemightguessthattheyelloworwhitepartsofacandleflameare——thanthereddishpart.

hotter

21. “Puttingout”anincandescentelectriclightmeansturningoffthecurrentsothatthefilamentgrowstoo——toemitlight.

cold,cool

22. Settingfiretothewickofanoillampiscalled——thelamp. lighting23. Thesunisourprincipal——oflight,aswellasofheat. source24. Thesunisnotonlyverybrightbutveryhot.Itisapowerful——

sourceoflight.incandescent

25. Lightisaformofenergy.In“emittinglight”anobjectchanges,or“converts,”oneformof——intoanother.

energy

26. Theelectricalenergysuppliedbythebatteryinaflashlightisconvertedto——and——.

heat,light;light,heat

27. Ifweleaveaflashlighton,alltheenergystoredinthebatterywillfinallybechangedor——intoheatandlight.

converted

28. Thelightfromacandleflamecomesfromthe——releasedbychemicalchangesasthecandleburns.

energy

29. Anearly“dead”batterymaymakeaflashlightbulbwarmtothetouch,butthefilamentmaystillnotbehotenoughtoemitlight—inotherwords,thefilamentwillnotbe——atthattemperature.

incandescent

30. Objects,suchasafilament,carbonparticles,orironbars,becomeincandescentwhenheatedtoabout800degreesCelsius.Atthattemperaturetheybeginto————.

emitlight

31. Whenraisedtoanytemperatureabove800degreesCelsius,anobjectsuchasanironbarwillemitlight.Althoughthebarmaymeltorvaporize,itsparticleswillbe——nomatterhowhottheyget.

incandescent

32. About800degreesCelsiusisthelowerlimitofthetemperatureatwhichparticlesemitlight.Thereisnoupperlimitofthe——atwhichemissionoflightoccurs.

temperature

206

33. Sunlightis——byveryhotgasesnearthesurfaceofthesun. emitted34. Complexchangessimilartoanatomicexplosiongeneratethegreat

heatwhichexplainsthe——oflightbythesun.emission

35. Belowabout——degreesCelsiusanobjectisnotanincandescentsourceoflight.

800

Unwanted responses are eliminated with special techniques. If, for example, thesecond sentence in frame 24were simply “It is a(n)—— source of light,” the twovery’swould frequently lead the student to fill the blankwith strong or a synonymthereof. This is prevented by inserting the word powerful to make a synonymredundant.Similarly,inframe3thewordsheatandpre-empttheresponseheat,whichwouldotherwisecorrectlyfilltheblank.The net effect of such material is more than the acquisition of facts and terms.

Beginningwith a largelyunverbalized acquaintancewith flashlights, candles, and soon,thestudentisinducedtotalkaboutfamiliarevents,togetherwithafewnewfacts,with a fairly technicalvocabulary.Heapplies the same terms to factswhichhemayneverbeforehaveseentobesimilar.Theemissionoflightfromanincandescentsourcetakes shape as a topic or field of inquiry. An understanding of the subject emergeswhichisoftenquitesurprisinginviewofthefragmentationrequiredinitembuilding.Itisnoteasytoconstructsuchaprogram.Whereaconfusingorellipticalpassagein

a textbook is forgivable because it canbe clarified by the teacher,machinematerialmustbe self-contained andwholly adequate.There areother reasonswhy textbooks,lecture outlines, and film scripts are of little help in preparing a program. They areusuallynotlogicalordevelopmentalarrangementsofmaterialbutstrategemswhichtheauthorshavefoundsuccessfulunderexistingclassroomconditions.Theexamplestheygive are more often chosen to hold the student’s interest than to clarify terms andprinciples.Incomposingmaterialforthemachine,theprogrammermaygodirectlytothepoint.Afirststepistodefinethefield.Asecondistocollecttechnicalterms,facts,laws,

principles,andcases.Thesemustthenbearrangedinaplausibledevelopmentalorder—linearifpossible,branchingifnecessary.Amechanicalarrangement,suchasacardfiling system, helps. The material is distributed among the frames of a program toachieve an arbitrary density. In the final composition of an item, techniques forstrengthening asked-for responses and for transferring control from one variable toanother are chosen from a list according to a given schedule in order to prevent theestablishmentof irrelevantverbal tendenciesappropriate toasingle technique.Whenone set of frames has been composed, its terms and facts are seeded mechanicallyamongsucceedingsets,wheretheywillagainbereferredtoincomposinglateritemstomake sure that the earlier repertoire remains active. Thus, the technical terms, facts,and examples in Table 2 have been distributed for reuse in succeeding sets onreflection,absorption,andtransmission,wheretheyareincorporatedintoitemsdealingmainly with other matters. Sets of frames for explicit review can, of course, beconstructed.Furtherresearchwillpresumablydiscoverother,possiblymoreeffective,techniques. Meanwhile, it must be admitted that a considerable measure of art isneededincomposingasuccessfulprogram.Whethergoodprogrammingistoremainanartortobecomeascientifictechnology,

207

it is reassuring to know that there is a final authority—the student. An unexpectedadvantageofmachineinstructionhasprovedtobethefeedbacktotheprogrammer.Inthe elementary school machine, provision is made for discovering which framescommonly yield wrong responses, and in the high-school and college machine thepaperstripsbearingwrittenanswersareavailableforanalysis.Atrial runof thefirstversion of a programquickly reveals frameswhich need to be altered, or sequenceswhich need to be lengthened. One or two revisions in the light of a few dozenresponses work a great improvement. No comparable feedback is available to thelecturer,textbookwriter,ormakeroffilms.Althoughonetextorfilmmayseemtobebetterthananother,itisusuallyimpossibletosay,forexample,thatagivensentenceonagivenpageoraparticularsequenceinafilmiscausingtrouble.Difficultasprogrammingis,ithasitscompensations.Itisasalutarythingtotryto

guarantee a right response at every step in the presentation of a subjectmatter. Theprogrammerwillusuallyfindthathehasbeenaccustomedtoleavemuchtothestudent—thathehasfrequentlyomittedessentialstepsandneglectedtoinvokerelevantpoints.The responsesmade to hismaterialmay reveal surprising ambiguities. Unless he islucky,hemayfindthathestillhassomethingtolearnabouthissubject.Hewillalmostcertainlyfindthatheneedstolearnagreatdealmoreaboutthebehavioralchangesheis trying to induce in the student. This effect of the machine in confronting theprogrammer with the full scope of his task may in itself produce a considerableimprovementineducation.Composingasetofframescanbeanexcitingexerciseintheanalysisofknowledge.

Theenterprisehasobviousbearingsonscientificmethodology.Therearehopefulsignsthat the epistemological implications will induce experts to help in composingprograms. The expert may be interested for another reason.We can scarcely ask atopflightmathematiciantowriteaprimerinsecond-gradearithmeticifitistobeusedby the average teacher in the average classroom.But a carefully controlledmachinepresentationandtheresultingimmediacyofcontactbetweenprogrammerandstudentofferaverydifferentprospect,whichmaybeenoughtoinducethosewhoknowmostaboutthesubjecttogivesomethoughttothenatureofarithmeticalbehaviorandtothevariousformsinwhichsuchbehaviorshouldbesetupandtested.

CanMaterialBeTooEasy?

The traditional teacher may view these programs with concern. He may beparticularly alarmedby the effort tomaximize success andminimize failure.Hehasfoundthatstudentsdonotpayattentionunlesstheyareworriedabouttheconsequencesoftheirwork.Thecustomaryprocedurehasbeentomaintainthenecessaryanxietybyinducingerrors. Inrecitation, thestudentwhoobviouslyknowstheanswer isnot toooften asked; a test item which is correctly answered by everyone is discarded asnondiscriminating; problems at the end of a section in a textbook in mathematicsgenerally include one or two very difficult items; and so on. (The teacher-turned-programmermaybesurprisedtofindthisattitudeaffectingtheconstructionofitems.Forexample,hemayfinditdifficulttoallowanitemtostandwhich“givesthepointaway.”Yetifwecansolvethemotivationalproblemwithothermeans,whatismoreeffective than giving a point away?)Making sure that the student knows he doesn’tknow is a technique concerned with motivation, not with the learning process.

208

Machines solve the problem ofmotivation in otherways. There is no evidence thatwhat is easily learned ismore readily forgotten. If this should prove to be the case,retentionmaybeguaranteedbysubsequentmaterialconstructedforanequallypainlessreview.Thestandarddefenseof“hard”materialisthatwewanttoteachmorethansubject

matter. The student is to be challenged and taught to “think.” The argument issometimes little more than a rationalization for a confusing presentation, but it isdoubtless true that lectures and texts are often inadequate andmisleading by design.Buttowhatend?Whatsortof“thinking”doesthestudentlearninstrugglingthroughdifficultmaterial? It is true that thosewho learnunderdifficult conditions arebetterstudents, but are they better because they have surmounted difficulties or do theysurmount them because they are better? In the guise of teaching thinking we setdifficultandconfusingsituationsandclaimcreditforthestudentswhodealwiththemsuccessfully.Thetroublewithdeliberatelymakingeducationdifficultinordertoteachthinkingis

(i)thatwemustremaincontentwiththestudentsthusselected,eventhoughweknowthattheyareonlyasmallpartofthepotentialsupplyofthinkers,and(ii)thatwemustcontinuetosacrificetheteachingofsubjectmatterbyrenouncingeffectivebut“easier”methods.Amore sensible program is to analyze the behavior called “thinking” andproduce it according to specifications. A program specifically concerned with suchbehavior could be composed of material already available in logic, mathematics,scientificmethod,andpsychology.Muchwoulddoubtlessbeaddedincompletinganeffective program. Themachine has already yielded important relevant by-products.Immediatefeedbackencouragesamorecarefulreadingofprogrammedmaterialthanisthecase in studyinga text,where theconsequencesofattentionor inattentionare solong deferred that they have little effect on reading skills. The behavior involved inobservingorattendingtodetail—asininspectingchartsandmodelsorlisteningcloselyto recorded speech—is efficiently shaped by the contingencies arranged by themachine.Andwhenanimmediateresultisinthebalance,astudentwillbemorelikelyto learn how to marshal relevant material, to concentrate on specific features of apresentation,torejectirrelevantmaterials,torefusetheeasybutwrongsolution,andtotolerateindecision,allofwhichareinvolvedineffectivethinking.Partoftheobjectiontoeasymaterialisthatthestudentwillcometodependonthe

machine andwill be less able than ever to copewith the inefficient presentations oflectures,textbooks,films,and“reallife.”Thisisindeedaproblem.Allgoodteachersmust“wean”theirstudents,andthemachineisnoexception.Thebettertheteacher,themoreexplicitmusttheweaningprocessbe.Thefinalstagesofaprogrammustbesodesigned that the student no longer requires the helpful conditions arranged by themachine. This can be done in many ways—among others by using the machine todiscussmaterialwhichhasbeenstudiedinotherforms.Thesearequestionswhichcanbeadequatelyansweredonlybyfurtherresearch.Nolarge-scale“evaluation”ofmachineteachinghasyetbeenattempted.Wehaveso

farbeenconcernedmainlywithpracticalproblemsinthedesignanduseofmachines,andwith testingand revisingsampleprograms.Themachine shown inFigure2wasbuiltandtestedwithagrantfromtheFundfortheAdvancementofEducation.Materialhas been prepared and tested with the collaboration of Lloyd E. Homme, Susan R.Meyer,andJamesG.Holland.5Theself-instructionroomshowninFigure3wassetup

209

under this grant. It contains ten machines and was recently used to teach part of acourse in human behavior to Harvard and Radcliffe undergraduates. Nearly 200students completed 48 disks (about 1400 frames) preparedwith the collaboration ofHolland.Thefactualcoreofthecoursewascovered,correspondingtoabout200pagesofthetext.6Themediantimerequiredtofinish48diskswas14½hours.Thestudentswerenotexaminedonthematerialbutwereresponsibleforthetextwhichoverlappedit.Their reactions to thematerialand toself-instruction ingeneralhavebeenstudiedthrough interviews and questionnaires. Both themachines and thematerial are nowbeingmodifiedinthelightofthisexperience,andamoreexplicitevaluationwillthenbemade.

FIG.3.Self-instructionroominSeverHallatHarvard.Tenboothscontainteachingmachines,someequippedwithindexingphonographs.

Meanwhile,itcanbesaidthattheexpectedadvantagesofmachineinstructionweregenerously confirmed. Unsuspected possibilities were revealed which are nowundergoing further exploration. Although it is less convenient to report to a self-instructionroomthantopickupatextbookinone’sroomorelsewhere,moststudentsfelt that theyhadmuchtogain instudyingbymachine.Mostof themworkedforanhour or more with little effort, although they often felt tired afterwards, and theyreported that they learned much more in less time and with less effort than inconventionalways.Noattemptwasmadetopointouttherelevanceofthematerialtocrucialissues,personalorotherwise,butthestudentsremainedinterested.(Indeed,one

210

change in the reinforcing contingencies suggested by the experiment is intended toreduce themotivational level.)An importantadvantageproved tobe that the studentalwaysknewwherehestood,withoutwaitingforanhourtestorfinalexamination.

SomeQuestions

Several questions are commonly asked when teaching machines are discussed.Cannot the results of laboratory research on learning be used in education withoutmachines?Ofcoursetheycan.Theyshouldleadtoimprovementsintextbooks,films,and other teaching materials. Moreover, the teacher who really understands theconditionsunderwhichlearningtakesplacewillbemoreeffective,notonlyinteachingsubjectmatterbutinmanagingtheclass.Nevertheless,somesortofdeviceisnecessaryto arrange the subtle contingencies of reinforcement required for optimal learning ifeachstudentistohaveindividualattention.Innonverbalskillsthisisusuallyobvious;texts and instructor can guide the learner but they cannot arrange the finalcontingencieswhichsetupskilledbehavior.Itistruethattheverbalskillsatissuehereare especiallydependentupon social reinforcement, but itmustnot be forgotten thatthemachinesimplymediatesanessentiallyverbalrelation.Inshapingandmaintainingverbal knowledge we are not committed to the contingencies arranged throughimmediatepersonalcontact.Machinesmaystillseemunnecessarilycomplexcomparedwithothermediatorssuch

as workbooks or self-scoring test forms. Unfortunately, these alternatives are notacceptable. When material is adequately programmed, adjacent steps are often sosimilar thatoneframereveals theresponse toanother.Onlysomesortofmechanicalpresentationwillmakesuccessiveframesindependentofeachother.Moreover,inself-instructionanautomaticrecordofthestudent’sbehaviorisespeciallydesirable,andformany purposes it should be fool-proof. Simplified versions of the present machineshavebeenfounduseful—forexample,intheworkofFersterandSapon,ofPorter,andof Gilbert—but the mechanical and economic problems are so easily solved that amachinewithgreatercapabilitiesisfullywarranted.Willmachines replace teachers?On thecontrary, theyarecapital equipment tobe

usedbyteacherstosavetimeandlabor.Inassigningcertainmechanizablefunctionstomachines,theteacheremergesinhisproperroleasanindispensablehumanbeing.Hemayteachmorestudentsthanheretofore—thisisprobablyinevitableiftheworld-widedemand for education is to be satisfied—but hewill do so in fewer hours andwithfewerburdensomechores. In return forhisgreaterproductivityhecanasksociety toimprovehiseconomiccondition.The role of the teachermaywell be changed, formachine instruction will affect

several traditional practices. Students may continue to be grouped in “grades” or“classes,” but itwill be possible for each to proceed at his own level, advancing asrapidlyashecan.Theotherkindof“grade”willalsochangeitsmeaning.IntraditionalpracticeaCmeans thatastudenthasasmatteringofawholecourse.But ifmachineinstructionassuresmasteryateverystage,agradewillbeusefulonlyinshowinghowfar a student has gone. C might mean that he is halfway through a course. GivenenoughtimehewillbeabletogetanA;andsinceAisnolongeramotivatingdevice,this is fair enough. The quick student will meanwhile have picked up A’s in othersubjects.

211

Differences in ability raise other questions. A program designed for the sloweststudentintheschoolsystemwillprobablynotseriouslydelaythefaststudent,whowillbefreetoprogressathisownspeed.(Hemayprofitfromthefullcoveragebyfillinginunsuspectedgapsinhisrepertoire.)Ifthisdoesnotprovetobethecase,programscanbeconstructedattwoormorelevels,andstudentscanbeshiftedfromonetotheotherasperformancesdictate. If therearealsodifferences in“typesof thinking,” theextratime available for machine instruction may be used to present a subject in waysappropriate tomany types. Each student will presumably retain and use thosewayswhich he finds most useful. The kind of individual difference which arises simplybecauseastudenthasmissedpartofanessentialsequence(comparethechildwhohasno “mathematical ability” because hewas outwith themeasleswhen fractionswerefirsttakenup)willsimplybeeliminated.

OtherUses

Self-instruction by machine has many special advantages apart from educationalinstitutions. Home study is an obvious case. In industrial andmilitary training it isoften inconvenient to schedule students in groups, and individual instruction bymachineshouldbeafeasiblealternative.Programscanalsobeconstructedinsubjectsforwhichteachersarenotavailable—forexample,whennewkindsofequipmentmustbeexplainedtooperatorsandrepairmen,orwhereasweepingchangeinmethodfindsteachers unprepared.7 Education sometimes fails because students have handicapswhichmakeanormalrelationshipwithateacherdifficultorimpossible.(Manyblindchildrenaretreatedtodayasfeeble-mindedbecausenoonehashadthelimeorpatiencetomakecontactwiththem.Deaf-mutes,spastics,andotherssuffersimilarhandicaps.)A teaching machine can be adapted to special kinds of communication—as, forexample,Braille—and,aboveall,ithasinfinitepatience.

Conclusions

Ananalysisofeducationwithin the frameworkofascienceofbehaviorhasbroadimplications. Our schools, in particular our “progressive” schools, are often heldresponsibleformanycurrentproblems—includingjuveniledelinquencyandthethreatofamorepowerfulforeigntechnology.Oneremedyfrequentlysuggestedisareturntoolder techniques,especially toagreater“discipline”inschools.Presumablythis is tobe obtained with some form of punishment, to be administered either with certainclassicalinstrumentsofphysicalinjury—thedriedbullock’stailoftheGreekteacherorthe cane of theEnglish schoolmaster—or as disapproval or failure, the frequency ofwhichistobeincreasedby“raisingstandards.”Thisisprobablynotafeasiblesolution.Not only education but Western culture as a whole is moving away from aversivepractices.Wecannotprepareyoungpeopleforonekindoflifeininstitutionsorganizedonquitedifferentprinciples.Thedisciplineofthebirchrodmayfacilitatelearning,butwemustrememberthatitalsobreedsfollowersofdictatorsandrevolutionists.Inthelightofourpresentknowledgeaschoolsystemmustbecalledafailureif it

cannotinducestudentstolearnexceptbythreateningthemfornotlearning.Thatthishas always been the standard pattern simply emphasizes the importance of modern

212

techniques. JohnDeweywas speaking forhis culture andhis timewhenhe attackedaversiveeducationalpracticesandappealedtoteacherstoturntopositiveandhumanemethods.Whathe threwout shouldhavebeen thrownout.Unfortunatelyhehad toolittletoputinitsplace.Progressiveeducationhasbeenatemporizingmeasurewhichcannowbeeffectivelysupplemented.Aversivepracticescannotonlybereplaced,theycan be replaced with far more powerful techniques. The possibilities should bethoroughly explored if we are to build an educational system which will meet thepresentdemandwithoutsacrificingdemocraticprinciples.ThematerialinTable3istakenfromtheprogramfortheauthor’scourseinHuman

Behaviormentionedintheprecedingarticle.ItisaddedhereasafurtherillustrationandforpurposesmentionedinthePreface.

TABLE3

213

214

215

216

217

218

219

1TheNavy’s“Self-Rater,”isalargerversionofPressey’smachine.Theitemsareprintedoncode-punchedplasticcardsfedbythemachine.Thetimerequiredtoansweristakenintoaccountinscoring.

2Pressey,S.L.SchoolandSociety,23,586(1926).3Pressey,S.L.SchoolandSociety,36,934(1932).4SeeVerbalBehavior.5Dr.Hommepreparedsetsofframesforteachingpartofcollegephysics(kinematics),andMrs.Meyerhaspreparedandinformallytestedmaterialinremedialreadingandvocabularybuildingatthejuniorhighschoollevel.Otherswhohavecontributedtothedevelopmentofteachingmachinesshouldbementioned.NathanH.Azrincooperatedwithmeintestingaversionofamachinetoteacharithmetic.C.B.FersterandStanleyM.Saponusedasimple“machine”toteachGerman[see“AnapplicationofrecentdevelopmentsinpsychologytotheteachingofGerman,”HarvardEducationalRev.28,1(1958)].DouglasPorter,oftheGraduateSchoolofEducationatHarvard,hasmadeanindependentschoolroomtestofmachine

220

instructioninspelling[see“Teachingmachines,”HarvardGraduateSchoolofEduc.Assoc.Bull.3,1(1958)].DevraCooperhasexperimentedwiththeteachingofEnglishcompositionforfreshmenattheUniversityofKentucky.ThomasF.Gilbert,oftheUniversityofGeorgia,hascomparedstandardandmachineinstructioninanintroductorycourseinpsychology,andwiththecollaborationofJ.E.Jewetthaspreparedmaterialinalgebra.

6ScienceandHumanBehavior.7Menger,K.Newapproachtoteachingintermediatemathematics.Science,1273310(1958).

221

WhyWeNeedTeachingMachines

Currentsuggestionsfor improvingeducationarefamiliar toeveryone.Weneedmoreandbetterschoolsandcolleges.Wemustpaysalarieswhichwillattractandholdgoodteachers.Weshouldgroupstudentsaccordingtoability.Wemustbringtextbooksandothermaterials up-to-date, particularly in science andmathematics.And so on. It issignificantthatallthiscanbedonewithoutknowingmuchaboutteachingorlearning.Thosewhoaremostactivelyconcernedwithimprovingeducationseldomdiscusswhatishappeningwhenastudentreadsabook,writesapaper,listenstoalecture,orsolvesa problem, and their proposals are only indirectly designed to make these activitiesmoreproductive.Inshort,thereisageneralneglectofeducationalmethod.(Televisionisnoexception,foritisonlyawayofamplifyingandextendingoldmethods,togetherwiththeirshortcomings.)It is true that the psychology of learning has so far not been very helpful in

education. Its learning curves and its theories of learning have not yielded greatlyimprovedclassroompractices.Butitistooearlytoconcludethatnothingusefulistobelearnedaboutthebehaviorofteacherandstudent.Noenterprisecanimproveitselfveryeffectivelywithoutexaminingitsbasicprocesses.Fortunately,recentadvancesinthe experimental analysis of behavior suggest that a true technology of education isfeasible.Improvedtechniquesareavailable tocarryout the twobasicassignmentsofeducation: constructing extensive repertoires of verbal and nonverbal behavior andgeneratingthathighprobabilityofactionwhichissaidtoshowinterest,enthusiasm,orastrong“desiretolearn.”The processes clarified by an experimental analysis of behavior have, of course,

alwaysplayedapartineducation,buttheyhavebeenusedwithlittleunderstandingoftheir effects, wanted or unwanted.Whether by intention or necessity, teachers havebeenlessgiventoteachingthantoholdingstudentsresponsibleforlearning.Methodsare still basically aversive. The student looks, listens, and answers questions (and,incidentally,sometimeslearns)asagestureofavoidanceorescape.Agoodteachercanciteexceptions,butitisamistaketocallthemtypical.Thebirchrodandcanearegone,but their place has been taken by equally effective punishments (criticism, possiblyridicule,failure)usedinthesameway:thestudentmustlearn,orelse!By-products of aversive control in education range from truancy, early drop-outs,

andschool-vandalism to inattention,“mental fatigue,” forgetting,andapathy. Itdoesnottakeascientificanalysistotracethesetotheirsourcesineducationalpractice.Butmore acceptable techniques have been hard to find. Erasmus tells of an EnglishgentlemanwhotriedtoteachhissonGreekandLatinwithoutpunishment.Hetaughtthe boy to use a bow and arrow and set up targets in the shape ofGreek andLatinletters, rewarding each hit with a cherry. Erasmus suggested cutting letters (“fromdeliciousbiscuits”).As a result,wemayassume that theboy salivated slightlyuponseeingaGreekorLatintextandthathewasprobablyabetterarcher;butanyeffectonhisknowledgeofGreekandLatinisdoubtful.

222

Current efforts to use rewards in education show the same indirection. Textsgarnishedwithpicturesinfourcolors,excitingepisodesinascientificfilm,interestingclassroomactivities—thesewillmakeaschool interestingandevenattractive(justastheboyprobablylikedhisstudyofGreekandLatin),buttogeneratespecificformsofbehavior these thingsmustberelatedto thestudent’sbehavior inspecialways.Onlythenwilltheybetrulyrewardingor,technicallyspeaking,“reinforcing.”Wemakeareinforcingeventcontingentonbehaviorwhen,forexample,wedesigna

pieceofequipmentinwhichahungryratormonkeyorchimpanzeemaypressaleverandimmediatelyobtainabitoffood.Suchapieceofequipmentgivesusapowerfulcontroloverbehavior.Byschedulingreinforcements,wemaymaintainthebehaviorofpressing the lever in any given strength for periods of time. By reinforcing specialkindsofresponsestothelever—forexample,verylightorveryheavypressesorthosemade with one hand or the other—we “shape” different forms or topographies ofbehavior.Byreinforcingonlywhenparticularstimuliorclassesofstimuliarepresent,webringthebehaviorunderthecontroloftheenvironment.Alltheseprocesseshavebeen thoroughly investigated, and they have already yielded standard laboratorypracticesinmanipulatingcomplexformsofbehaviorforexperimentalpurposes.Theyareobviouslyappropriatetoeducationaldesign.In approaching the problemof the educatorwemaybegin by surveying available

reinforcers.Whatpositivereasonscanwegivethestudentforstudying?Wecanpointtotheultimateadvantagesofaneducation—tothewaysoflifewhichareopenonlytoeducatedmen—and the student himselfmay cite these to explain why he wants aneducation, but ultimate advantages are not contingent on behavior in ways whichgenerateaction.Manyastudentcantestifytotheresult.Nomatterhowmuchhemaywant to become a doctor or an engineer, say, he cannot force himself to read andrememberthepageoftextinfrontofhimatthemoment.Allnotionsofultimateutility(as,forexample,ineconomics)sufferfromthesameshortcoming:theydonotspecifyeffectivecontingenciesofreinforcement.The gap between behavior and a distant consequence is sometimes bridged by a

seriesof“conditionedreinforcers.”Inthelaboratoryexperimentjustdescribedadelayofevenafractionofasecondbetweentheresponsetotheleverandtheappearanceoffoodmayreducetheeffectivenessofthefoodbyameasurableamount.Itisstandardpracticetoletthemovementofaleverproducesomevisualstimulus,suchasachangein the illumination in the apparatus,which is then followedby food. In thisway thechange in illumination becomes a conditioned reinforcer which can be madeimmediatelycontingentontheresponse.Themarks,grades,anddiplomasofeducationare conditioned reinforcers designed to bring ultimate consequences closer to thebehavior reinforced.Likeprizes andmedals, they represent the approval of teachers,parents, and others; and they show competitive superiority, but they are mainlyeffective because they signalize progress through the system—toward some ultimateadvantageof,orat least freedomfrom,education.To thisextent theybridge thegapbetween behavior and its remote consequences; but they are still not contingent onbehaviorinaveryeffectiveway.Progressiveeducationtriedtoreplacethebirchrod,andatthesametimeavoidthe

artificialityofgradesandprizes,bybringing the reinforcersofeveryday life into theschools. Such natural contingencies have a kind of guaranteed effectiveness. But aschoolisonlyasmallpartofthestudent’sworld,andnomatterhowrealitmayseem,itcannotprovidenaturalreinforcingconsequencesforallthekindsofbehaviorwhich

223

educationis tosetup.Thegoalsofprogressiveeducationwereshifted toconformtothislimitation,andmanyworthwhileassignmentsweresimplyabandoned.Fortunately,wecansolvetheproblemofeducationwithoutdiscoveringorinventing

additional reinforcers.Wemerelyneed tomakebetteruseof thosewehave.Humanbehavior is distinguished by the fact that it is affected by small consequences.Describingsomethingwiththerightwordisoftenreinforcing.Soistheclarificationofa temporary puzzlement, or the solution of a complex problem, or simply theopportunity tomove forward after completingone stageof an activity.Weneednotstop to explain why these things are reinforcing. It is enough that, when properlycontingentuponbehavior,theyprovidethecontrolweneedforsuccessfuleducationaldesign. Proper contingencies of reinforcement, however, are not always easilyarranged.Amodernlaboratoryforthestudyofbehaviorcontainselaborateequipmentdesigned to control the environment of individual organisms during many hours ordaysofcontinuousstudy.Therequiredconditionsandchangesinconditionscannotbearranged by hand, not only because the experimenter does not have the time andenergy, but because many contingencies are too subtle and precise to be arrangedwithoutinstrumentalhelp.Thesameproblemarisesineducation.Consider, for example, the temporal patterning of behavior called “rhythm.”

Behavior is often effective only if properly timed. Individual differences in timing,ranging from the most awkward to the most skillful performances, affect choice ofcareer and of artistic interests and participation in sports and crafts. Presumably a“senseofrhythm”isworthteaching,yetpracticallynothingisnowdonetoarrangethenecessary contingencies of reinforcement. The skilled typist, tennis player, latheoperator, or musician is, of course, under the influence of reinforcing mechanismswhichgenerate subtle timing, butmanypeople never reach the point atwhich thesenaturalcontingenciescantakeover.TherelativelysimpledeviceshowninFigure1suppliesthenecessarycontingencies.

The student taps a rhythmicpattern inunisonwith thedevice. “Unison” is specifiedvery loosely at first (the student can be a little early or late at each tap) but thespecifications are slowly sharpened. The process is repeated for various speeds andpatterns.Inanotherarrangement,thestudentechoesrhythmicpatternssoundedbythemachine, though not in unison, and again the specifications for an accuratereproductionareprogressivelysharpened.Rhythmicpatternscanalsobebroughtunderthecontrolofaprintedscore.Another kind of teaching machine generates sensitivity to properties of the

environment.Wecallaneffectiveperson“discriminating.”Hecan tell thedifferencebetween the colors, shapes, and sizes of objects, he can identify three-dimensionalformsseenfromdifferentaspects,hecanfindpatternsconcealedinotherpatterns,hecan identify pitches, intervals, andmusical themes and distinguish between differenttempos and rhythms—and all of this in an almost infinite variety. Subtlediscriminationsofthissortareasimportantinscienceandindustryandineverydaylifeasinidentifyingtheschoolofapainterortheperiodofacomposer.

224

FIG.1.Amachinetoteach“agoodsenseofrhythm.”

The ability to make a given kind of discrimination can be taught. A pigeon, forexample,canbemadesensitivetothecolor,shape,andsizeofobjects,topitches,andrhythms,andsoon—simplybyreinforcingitwhenitrespondsinsomearbitrarywaytoone set of stimuli and extinguishing responses to all others. The same kinds ofcontingenciesofreinforcementareresponsibleforhumandiscriminativebehavior.Theremarkable fact is that they are quite rare in the environment of the average child.True,childrenareencouragedtoplaywithobjectsofdifferentsizes,shapes,andcolors,and are given a passing acquaintance with musical patterns; but they are seldomexposed to theprecise contingenciesneeded tobuild subtlediscriminations. It is notsurprisingthatmostofthemmoveintoadulthoodwithlargelyundeveloped“abilities.”The number of reinforcements required to build discriminative behavior in the

populationasawholeisfarbeyondthecapacityofteachers.Toomanyteacherswouldbe needed, andmany contingencies are too subtle to bemediated by even themostskillful.Yetrelativelysimplemachineswillsuffice.TheapparatusshowninFigure2 isadapted from research on lower organisms. It teaches an organism to discriminateselected properties of stimuli while “matching to sample.” Pictures or words areprojectedontranslucentwindowswhichrespondtoatouchbyclosingcircuits.Achildcanbemadeto“lookatthesample”byreinforcinghimforpressingthetopwindow.Anadequatereinforcementforthisresponseissimplytheappearanceofmaterialinthelowerwindows,fromwhichachoiceistobemade.Thechildidentifiesthematerialwhichcorrespondstothesampleinsomeprescribed

waybypressingoneofthelowerwindows,andheisthenreinforcedagain—possibly

225

simplybecauseanewsetofmaterialsnowappearsonthewindows.Ifhepressesthewrongwindow,allthreechoicesdisappearuntilthetopwindowhasbeenpressedagain—whichmeansuntilhehasagain lookedat thesample.Manyotherarrangementsofresponsesandreinforcementsare,ofcourse,possible.Inanauditoryversion,thechildlistenstoasamplepatternoftonesandthenexploresothersamplestofindamatch.

FIG.2.Amachinetoteachthematchingofcolors,shapes,sizes,aswellascorrespondencesbetweenpicturesandwords,wordsandotherwords,andsoon.

IfdevicessimilartotheseshowninFigures1and2weregenerallyavailableinournurseryschoolsandkindergartens,ourchildrenwouldbe farmoreskillful indealingwiththeirenvironments.Theywouldbemoreproductiveintheirwork,moresensitivetoartandmusic,betteratsports,andsoon.Theywouldleadmoreeffectivelives.We

226

cannotassertallthiswithcompleteconfidenceonthepresentevidence,butthereisnodoubtwhatsoeverthattheconditionsneededtoproducesuchastateofaffairsarenowlacking.Inthelightofwhatweknowaboutdifferentialcontingenciesofreinforcement,the world of the young child is shamefully impoverished. And only machines willremedythis,fortherequiredfrequencyandsubtletyofreinforcementcannototherwisebearranged.The teacher is, of course, at a disadvantage in teaching skilled and discriminative

behaviorbecausesuchinstructionislargelynonverbal.Itmaybethatthemethodsofthe classroom, in which the teacher is said to “communicate” with the student, to“impart information,” and to build “verbal abilities,” are better adapted to standardsubjectmatters,thelearningofwhichisusuallyregardedasmorethantheacquisitionof forms of behavior or of environmental control. Yet a second look may beworthwhile. Traditional characterizations of verbal behavior raise almost insuperableproblemsforateacher,andamorerigorousanalysissuggestsanotherpossibility.Wecandefinetermslike“information,”“knowledge,”and“verbalability”byreferencetothe behavior from which we infer their presence.We may then teach the behaviordirectly.Insteadof“transmittinginformationtothestudent”wemaysimplysetupthebehaviorwhichistakenasasignthathepossessesinformation.Insteadofteachinga“knowledge of French” we may teach the behavior from which we infer suchknowledge. Instead of teaching “an ability to read” we may set up the behavioralrepertoirewhichdistinguishes the childwhoknowshow to read fromonewhodoesnot.To take the last example, a child readsor “shows that heknowshow to read”by

exhibitingabehavioralrepertoireofgreatcomplexity.Hefindsaletterorwordinaliston demand; he reads aloud; he finds or identifies objects described in a text; herephrases sentences; he obeys written instructions; he behaves appropriately todescribedsituations;hereactsemotionallytodescribedevents;andsoon,inalonglist.Hedoesnoneof thisbefore learning to readandallof it afterwards.Tobringaboutsuchachangeisanextensiveassignment,anditistemptingtotrytocircumventitbyteachingsomethingcalled“anabilitytoread”fromwhichallthesespecificbehaviorswillflow.Butthishasneveractuallybeendone.“Teachingreading”isalwaysdirectedtowardsettingupspecificitemsinsucharepertoire.Itistruethatpartsoftherepertoirearenotindependent.Astudentmayacquiresome

kindsofresponsesmorereadilyforhavingacquiredothers,andhemayforatimeusesome in place of others (for example, he may follow written directions not byrespondingdirectlytoatextbutbyfollowinghisownspokeninstructionsashereadsthetextaloud).Inthelongrunallpartsoftherepertoiretendtobefilledin,notbecausethestudentisroundingoutanabilitytoread,butsimplybecauseallpartsareintheirseveralwaysuseful.Theyallcontinuetobereinforcedbytheworldatlargeaftertheexplicitteachingofreadinghasceased.Viewed in thisway, readingcan alsobemost effectively taughtwith instrumental

help.Apupilcanlearntodistinguishamonglettersandgroupsoflettersinanalphabetsimplyasvisualpatternsinusingthedeviceandproceduresjustdescribed.Hecanbetaughttoidentifyarbitrarycorrespondences(forexample,betweencapitalsandlower-case letters, or between handwritten and printed letters) in amore complex type ofstimulus control which is within reach of the same device. With a phonographicattachment,correspondencesbetweenprinted lettersandsounds,betweensoundsandletters,betweenwordsandsounds,betweensoundsandprintedwords,andsoon,can

227

besetup.(Thestudentcouldbetaughtallofthiswithoutpronouncingaword,anditispossiblethathewouldlearngoodpronunciationmorequicklyifhehadfirstdoneso.)The same device can teach correspondences betweenwords and the properties of

objects.Thepupilselectsaprintedorspokenwordwhichcorrespondsinthelanguageto, say, a pictured object or another printed or spoken word. These semanticcorrespondences differ in important respects from formal matches, but the sameprocessesofprogrammingandreinforcementcan—indeed,must—beused.Traditionalwaysofteachingreadingestablishall theserepertoires,buttheydosoindirectlyand,alas, inefficiently. In “building a child’s need to read,” in motivating “his mentalreadiness,”in“sharinginformation,”andsoon,theteacherarranges,sometimesalmostsurreptitiously, many of the contingencies just listed, and these are responsible forwhatever is learned. An explicit treatment clarifies the program, suggests effectiveprocedures,andguaranteesacoveragewhichisoftenlackingwithtraditionalmethods.Muchofwhatiscalledreadinghasnotbeencovered,ofcourse,butitmaynotneedtobe taught, for once these basic repertoires have been established, the child begins toreceiveautomaticreinforcementinrespondingtotextualmaterial.Thesameneedforabehavioraldefinitionarisesinteachingotherverbalskills(for

example, a second language) as well as the traditional subjects of education. Inadvancingtothatlevel,however,wemusttranscendalimitationofthedeviceinFigure2.Thestudentcanselectaresponsewithoutbeingabletospeakorwrite,butwewanthimtolearntoemittheresponse,sincethisisthekindofbehaviorwhichhewilllaterfindmostuseful.Theemissionofverbalbehavioristaughtbyanotherkindofmachine(seeTheTechnologyofTeaching,page25).Aframeoftextualmaterialappearinginasquare opening is incomplete: in place of certain letters or figures there are holes.Lettersorfigurescanbemadetoappearintheseholesbymovingsliders(akeyboardwouldbeanobviousimprovement).Whenthematerialhasbeencompleted,thestudentcheckshisresponsebyturningacrank.Themachinesensesthesettingsoftheslidersand,iftheyarecorrect,movesanewframeofmaterialintoplace,theslidersreturningtotheirhomeposition.Iftheresponseiswrong,theslidersreturnhome,andasecondsettingmustbemade.Themachinecantellthestudentheiswrongwithouttellinghimwhatisright.This

isanadvantage,butitisrelativelycostly.Moreover,correctbehaviorisratherrigidlyspecified.Suchamachineisprobablysuitableonlyforthelowergrades.Asimplerandcheaper procedure, with greater flexibility, is to allow the student to compare hiswrittenresponsewitharevealedtext.ThedeviceshowninFigure3usesthisprinciple.It issuitableforverbalinstructionbeyondthelowerprimarygrades—thatis, throughjunior high school, high school, and college, and in industrial and professionaleducation. Programmed material is stored on fan-folded paper tapes. One frame ofmaterial,thesizeofwhichmaybevariedwiththenatureofthematerial,isexposedata time.The studentwriteson a separatepaper strip.Hecannot lookat unauthorizedpartsofthematerialwithoutrecordingthefactthathehasdoneso,becausewhenthemachinehasbeen loadedandclosed, it canbeopenedonlybypunching the stripofpaper.The student sees printedmaterial in the large window at the left. Thismay be a

sentence tobe completed, aquestion tobe answered,or aproblem tobe solved.Hewriteshisresponseinanuncoveredportionofapaperstripattheright.Hethenmovesasliderwhichcoverstheresponsehehaswrittenwithatransparentmaskanduncoversadditionalmaterialinthelargeropening.Thismaytellhimthathisresponseiswrong

228

without tellinghimwhat is right. For example, it may list a few of the commonesterrors.Iftheresponsehewroteisamongthem,hecantryagainonanewlyuncoveredportionofthepaperstrip.Afurtheroperationofthemachinecovershissecondattemptanduncoversthecorrectresponse.Thestudentrecordsawrongresponsebypunchingahole alongside it, leaving a record for the instructor who may wish to review astudent’sperformance,andoperatingacounterwhichbecomesvisibleattheendoftheset.Thenthestudentrecordsthenumberofmistakeshehasmadeandmaycompareitwithaparscorefortheset.

229

FIG.3.Amachinetoteach“verbalknowledge.”

Exploratory research in schools and colleges indicates thatwhat is now taught by

230

teacher,textbook,lecture,orfilmcanbetaughtinhalfthetimewithhalftheeffortbyamachineofthisgeneraltype.Onehasonlytoseestudentsatworktounderstandwhythis is a conservative estimate. The student remains active. If he stops, the programstops(inmarkedcontrastwithclassroompracticeandeducationaltelevision);butthereisnocompulsionforheisnotinclinedtostop.Immediateandfrequentreinforcementsustains a lively interest. (The interest, incidentally, outlasts any effect of novelty.Noveltymayberelevanttointerest,butthematerialinthemachineisalwaysnovel.)Where current instructional procedures are highly efficient, the gain may not be sogreat.Inoneexperimentinvolvingindustrialeducationtherewasapproximatelya25%savinginthetimerequiredforinstructionandsomethingoftheorderofa10%increasein retention, and about 90% of the students preferred to study by the machine. Ingeneral,thestudentgenerallylikeswhatheisdoing;hemakesnoefforttoescape—forexample, by letting his attentionwander. He need not force himself to work and isusuallyfreeofthefeelingofeffortgeneratedbyaversivecontrol.Hehasnoreasontobe anxious about impending examinations, for none are required. Both he and hisinstructorknowwherehestandsatalltimes.Nolessimportantinexplainingthesuccessofteachingmachinesisthefactthateach

student is free toproceedathisown rate.Holdingstudents together for instructionalpurposesinaclassisprobablythegreatestsourceofinefficiencyineducation.Someefforts to mechanize instruction have missed this point. A language laboratorycontrolledfromacentralconsolepresupposesagroupofstudentsadvancingataboutthe same rate, even though some choice of material is permitted. Television ineducationhasmade thesamemistakeonacolossalscale.Aclassof twentyor thirtystudentsmovingatthesamepaceisinefficientenough,butwhatmustwesayofallthestudentsinhalfadozenstatesmarchinginasimilarlockstep?Intryingtoteachmorethanonestudentatonceweharmbothfastandslowlearners.

Theplightofthegoodstudenthasbeenrecognized,buttheslowlearnersuffersmoredisastrousconsequences.Theeffectofpressuretomovebeyondone’snaturalspeediscumulative.Thestudentwhohasnotfullymasteredafirstlessonislessabletomastera second. His ultimate failure may greatly exaggerate his shortcoming; a smalldifference in speedhasgrown to an immensedifference in comprehension.Someofthosemostactiveinimprovingeducationhavebeentemptedtodismissslowstudentsimpatientlyasawasteoftime,butitisquitepossiblethatmanyofthemarecapableofsubstantial,evenextraordinary,achievements ifpermittedtomoveat theirownpace.Manydistinguishedscientists,forexample,haveappearedtothinkslowly.One advantage of individual instruction is that the student is able to follow a

programwithoutbreaksoromissions.Amemberofaclassmovingatapproximatelythe same rate cannot always make up for absences, and limitations of contact timebetween student and teachermake it necessary to abbreviatematerial to the point atwhichsubstantialgapsare inevitable.Workingonamachine, thestudentcanalwaystakeupwhereheleftoffor,ifhewishes,reviewearlierworkafteralongerabsence.The coherence of the program helps to maximize the student’s success, for bythoroughlymasteringonestepheisoptimallypreparedforthenext.Manyyearsago,intheir Elementary Principles of Education, Thorndike and Gates considered thepossibilityofabook“soarrangedthatonlytohimwhohaddonewhatwasdirectedonpage onewould page two become visible, and so on.”With such a book, they felt,“muchthatnowrequirespersonalinstructioncouldbemanagedbyprint.”Theteachingmachineis,ofcourse,suchabook.

231

Insummary,then,machineteachingisunusuallyefficientbecause(1)thestudentisfrequentlyandimmediatelyreinforced,(2)heisfreetomoveathisnaturalrate,and(3)hefollowsacoherentsequence.Thesearethemoreobviousadvantages,andtheymaywell explain current successes. But there are more promising possibilities: theconditionsarrangedbyagoodteachingmachinemakeitpossibletoapplytoeducationwhatwehavelearnedfromlaboratoryresearchandtoextendourknowledgethroughrigorousexperimentsinschoolsandcolleges.The conceptions of the learning process which underlie classroom practices have

long been out of date. For example, teachers and textbooks are said to “impartinformation.” They expose the student to verbal and nonverbal material and callattention toparticular featuresof it,and insodoing theyaresaid to“tell thestudentsomething.”Inspiteofdiscouragingevidencetothecontrary,itisstillsupposedthatifyou tell a student something, he then knows it. In this scheme, teaching is thetransmission of information, a notion which, through a false analogy, has acquiredundue prestige from communication engineering. Something is undoubtedlytransmitted by teacher to student, for if communication is interrupted, instructionceases; but the teacher is notmerely a source fromwhich knowledge flows into thestudent. We cannot necessarily improve instruction by altering the conditions oftransmission—as,forexample,bychangingtoadifferentsensorymodality.This isamistake made by some so-called teaching machines which, accepting our failure toteach reading, have tried to restore communication by using recorded speech. Thestudentnolongerporesoverabook,asinthetraditionalportrait;hestaresintospacewith earphones on his head. For the same reasons, improvements in the coding ofinformationmaynotbeimmediatelyrelevant.The student is more than a receiver of information. He must take some kind of

action. The traditional view is that he must “associate.” The stream of informationflowingfromteachertostudentcontainspairsofitemswhich,beingclosetogetherorotherwiserelated,becomeconnectedinthestudent’smind.Thisistheolddoctrineofthe association of ideas, now strengthened by a scientific, if uncritical, appeal toconditioned reflexes: two things occurring together in experience somehow becomeconnectedsothatoneof themlaterremindsthestudentof theother.Theteacherhaslittlecontrolovertheprocessexcepttomakesurethatthingsoccurtogetheroftenandthatthestudentpaysattentiontothem—forexample,bymakingtheexperiencesvividor,aswesay,memorable.Somedevicescalledteachingmachinesaresimplywaysofpresenting things together in ways which attract attention. The student listens torecorded speech, for example, while looking at pictures. The theory is that he willassociatetheseauditoryandvisualpresentations.But the action demanded of the student is not some sort ofmental association of

contiguous experiences. It ismore objective and, fortunately,more controllable thanthat. To acquire behavior, the student must engage in behavior. This has long beenknown.Theprincipleisimpliedinanyphilosophyof“learningbydoing.”Butitisnotenoughsimplytoacknowledgeitsvalidity.Teachingmachinesprovidetheconditionsneededtoapplytheprincipleeffectively.Onlyintheearlystagesofeducationarewemainlyinterestedinestablishingforms

ofbehavior.Intheverbalfield,forexample,weteachachildtospeak,eventuallywithacceptableaccentandpronunciation,andlatertowriteandspell.Afterthat,topographyofbehaviorisassumed;thestudentcanspeakandwriteandmustnowlearntodosoappropriately—that is, he must speak or write in given ways under given

232

circumstances. How he comes to do so is widely misunderstood. Education usuallybeginsbyestablishingso-calledformalrepertoires.Theyoungchildistaughtto“echo”verbal behavior in the sense of repeating verbal stimuliwith reasonable accuracy.Alittle later he is taught to read—to emit verbal behavior under thecontrolof textualstimuli. These and other formal repertoires are used in later stages of instruction toevokenewresponseswithout“shaping”them.Inanimportantcaseofwhatwecallinstruction,controlissimplytransferredfrom

so-called formal to thematicstimuli.Whenastudent learns tomemorizeapoem, forexample,itisclearlyinadequatetosaythatbyreadingthepoemhepresentstohimselfitsvariouspartscontiguouslyand thenassociates them.Hedoesnot simply read thepoemagain andagainuntil heknows it. (It ispossible thathe couldnever learn thepoeminthatway.)Somethingelsemustbedone,asanyoneknowswhohasmemorizedapoemfromthetext.Thestudentmustmaketentativeresponseswhilelookingawayfromthetext.Hemustglanceatthetextfromtimetotimetoprovidefragmentaryhelpin emitting a partially learned response. If a recalled passage makes sense, it mayprovideitsownautomaticconfirmation,butifthepassageisfragmentaryorobscure,thestudentmustconfirmthecorrectnessofanemittedresponsebyreferringtothetextafterhehasemittedit.Ateachingmachinefacilitatesthisprocess.Itpresentsthepoemlinebylineandasks

the student to read it. The text is then “vanished”—that is, it becomes less and lessclearorlessandlesscompleteinsubsequentpresentations.Otherstimuli(arisingfromthestudent’sownbehaviorinthiscase)takeover.Inoneprocedureafewunimportantlettersareomittedinthefirstpresentation.Thestudentreadsthelinewithouttheirhelpandindicateshissuccessbywritingdowntheomittedletters,whichareconfirmedbythemachine.More of the line ismissingwhen it again appears, but because he hasrecently responded to a fuller text, the student can nevertheless read it correctly.Eventually,notextualstimulusremains,andhecan“recite”thepoem.(Ifthereaderwishestotrythismethodonafriendormemberofhisfamilywithouta

machine,hemaydosobywritingthepoemonachalkboardinaclearhand,omittingafewunimportantletters.Heshouldaskhissubjecttoreadthepoemaloudbuttomakenoefforttomemorizeit.Heshouldtheneraseanotherselectionofletters.Hewillhavetoguessathowfarhecangowithoutinterferingwithhissubject’ssuccessonthenextreading, but under controlled conditions this could be determined for the averagestudentquiteaccurately.Againthesubjectreadsthepoemaloud,makingnoefforttomemorize, thoughhemayhave tomake some effort to recall.Other letters are thenerased and the process repeated. For a dozen lines of averagematerial, four or fivereadingsshouldsufficetoeliminatethetextaltogether.Thepoemcanstillbe“read.”)Memorizedverbalbehavior is avaluable formofknowledgewhichhasplayed an

importantroleinclassicaleducation.Thereareother,andgenerallymoreuseful,formsinwhichthesameprocessesareinvolved.Consider,forexample,alabeledpicture.Tosaythatsuchaninstructionaldevice“tellsthestudentthenameofthepicturedobject”is highly elliptical—and dangerous if we are trying to understand the processesinvolved.Simplyshowingastudentalabeledpictureisnomoreeffectivethanlettinghim read a poem.Hemust take some sort of action.As a formal stimulus, the labelevokesaverbalresponse,notinthiscaseinthepresenceofotherverbalbehavioronthepartofthestudent,butinthepresenceofthepicture.Thecontroloftheresponseisto pass from the label to the picture; the student is to give the nameof the picturedobjectwithoutreadingit.

233

Thesteps taken in teachingwith labeledpicturescanalsobearrangedparticularlywellwithamachine.Supposeweareteachingmedical-schoolanatomyatthetextbooklevel.Certainlabeledchartsrepresentwhatistobelearnedinthesensethatthestudentwilleventually(1)givethenamesofindicatedpartsanddescriberelationsamongthemand(2)beabletopointto,draw,orconstructmodelsofparts,orrelationsamongthem,given their names. To teach the first of these, we induce the student to describerelationsamongthepartsshownonafullylabeledchart.Oneeffectofthisisthatheexecutes the verbal behavior at issue—he writes the names of the parts. Moreimportant,hedoes thiswhile,or justafter, lookingatcorrespondingpictureddetails.Hewillbeable towrite thenamesagainwhile lookingat achartwhichshowsonlyincompletenames,possiblyonly initial letters.Finally, hewill be able to supply thecompletenamesofparts identifiedonlybynumberon still another chart.Hisverbalresponseshavepassedfromthecontroloftextualstimulitothatofpicturedanatomicaldetails.Eventually,ashestudiesacadaver,thecontrolwillpasstotheactualanatomyofthehumanbody.Inthissensehethen“knowsthenamesofthepartsofthebodyandcandescriberelationsamongthem.”(ThedeviceshowninFigure3 isdesigned toskiponeor twosteps in“vanishing”

textualstimuli.Afullylabeledchartmaybefollowedbyamerelynumberedone.Thestudentwritesthenamecorrespondingtoanumberinthefirstspace.Ifhecannotdothis,heoperatesthemachinetouncover,notmerelysomeindicationthatheisrightorwrong,butadditionalhelp—say,afewlettersofthecorrectresponse.)Learningapoemorthenamesofpicturedobjectsisarelativelystraightforwardtask.

More complex forms of knowledge require other procedures. At an early point, themain problem becomes that of analyzing knowledge. Traditionally, for example,somethingcalleda“knowledgeofFrench”issaidtopermitthestudentwhopossessesittodomanythings.Onewhopossessesitcan(1)repeataFrenchphrasewithagoodaccent,(2)readaFrenchtextinallthesensesofreadinglistedabove,(3)takedictationin French, (4) find aword spoken in French on a printed list, (5) obey instructionsspoken in French, (6) comment in French upon objects or events, (7) give orders inFrench,andsoon.Ifhealso“knowsEnglish,”hecangivetheEnglishequivalentsofFrenchwordsorphrasesortheFrenchequivalentsofEnglishwordsorphrases.The concept of “a knowledge of French” offers very little help to the would-be

teacher. As in the case of reading, we must turn to the behavioral repertoiresthemselves, for these are all that have ever been taught when education has beeneffective. The definition of a subject matter in such terms may be extraordinarilydifficult. Students who are “competent in first-year college physics,” for example,obviouslydifferfromthosewhoarenot—butinwhatway?Evenatentativeanswertothatquestionshouldclarifytheproblemofteachingphysics.Itmaywelldomore.Inthe not-too-distant future much more general issues in epistemology may beapproached from the samedirection. It is possible thatwe shall fullyunderstand thenatureofknowledgeonlyafterhavingsolvedthepracticalproblemsofimpartingit.Untilwecandefinesubjectmattersmoreaccuratelyanduntilwehaveimprovedour

techniquesofbuildingverbalrepertoires,writingprogramsforteachingmachineswillremain something of an art. This is not wholly satisfactory, but there is someconsolationinthefactthatanimpeccableauthorityontheexcellenceofaprogramisavailable. The student himself can tell the programmer where he has failed. Byanalyzing the errorsmade by even a small number of students in a pilot study, it isusuallypossible toworkagreat improvement inanearlyversionofaprogram.(The

234

machine shown in Figure 3 is designed to supply the necessary feedback to theprogrammerinaconvenientform.Whenastudentpunchesanerror,hemarksthebackoftheprintedmaterial,whicheventuallycarriesanitem-by-itemrecordofthesuccessorfailureoftheprogrammer.Thisisobviouslyvaluableduringtheexperimentalstagesofprogramming,butitwillalsobedesirablewhenmachinesarewidelyusedinschoolsandcolleges,sincepublisherscanthenperiodicallycallinprogramstobestudiedandimprovedby their authors.The information suppliedmight be compared to a recordshowingthepercentageofstudentswhohavemisunderstoodeachsentenceinatext.)The teaching machine shown in Figure 3 falls far short of the “electronic

classrooms”oftenvisualizedfortheschoolsandcollegesofthefuture.Manyofthese,often incorporating small computers, arebasedonmisunderstandingsof the learningprocess.Theyaredesignedtoduplicatecurrentclassroomconditions.Wheninstructionis badly programmed, a student often goes astray, and a teacher must come to hisrescue.Hismistakesmustbeanalyzedandcorrected.Thismaygivetheimpressionthatinstruction is largely amatter of correcting errors. If thiswere the case, an effectivemachinewould, indeed, have to follow the student intomanyunprofitablepaths andtakeremedialaction.Butunderproperprogrammingnothingofthissortisrequired.Itistruethatarelativelyimportantfunctionoftheteacherwillbetofollowtheprogressofeachstudentandtosuggestcollateralmaterialwhichmaybeofinterest,aswellastooutline further studies, to recommend changes to programs of different levels ofdifficulty, and so on, and to this extent a student’s course of study will show“branching.”Butchangesinlevelofdifficultyorinthecharacterofthesubjectneednot be frequent and can bemade as the student moves from one set of material toanother.Teachingmachines based on the principle of “multiple choice” also often show a

misunderstandingofthelearningprocess.Whenmultiple-choiceapparatuseswerefirstused,theorganismwaslefttoproceedby“trialanderror.”Thetermdoesnotrefertoabehavioralprocessbutsimplytothefactthatcontingenciesofreinforcementwerelefttochance:someresponseshappenedtobesuccessfulandothersnot.Learningwasnotfacilitatedoracceleratedbyprocedureswhich increased theprobabilityof successfulresponses.Theresults,likethoseofmuchclassroominstruction,suggestedthaterrorswere essential to the learning process. But when material is carefully programmed,bothsubhumanandhumansubjectscanlearnwhilemakingfewerrorsorevennoneatall.RecentresearchbyHerbertS.Terrace, forexample,hasshownthatapigeoncanlearn todiscriminatecolorspracticallywithoutmakingmistakes.Thecontrolexertedby color may be passed, via a vanishing technique, to more difficult properties ofstimuli—againwithout error.Of coursewe learn something from ourmistakes—foronething,welearnnottomakethemagain—butweacquirebehaviorinotherways.The teachingmachines of S. J. Pressey, the first psychologist to see the “coming

industrialrevolutionineducation,”weremechanicalversionsofself-scoringtestforms,which Pressey and his students also pioneered. They were not designed forprogrammedinstructioninthepresentsense.Thestudentwaspresumedtohavestudieda subjectbeforecoming to themachine.By testinghimself,heconsolidatedwhathehadalreadypartiallylearned.Forthispurposeadevicewhichevaluatedthestudent’sselectionfromanarrayofmultiple-choiceitemswasappropriate.Forthesamepurposemultiple-choicematerialcan,ofcourse,beused inall themachinesdescribedabove.Butseveraladvantagesofprogrammedinstructionarelostwhensuchmaterialisusedinstraightforwardinstruction.

235

Inthefirstplace,thestudentshouldconstructratherthanselectaresponse,sincethisis thebehaviorhewill later finduseful.Secondly,he shouldadvance to the levelofbeingabletoemitaresponseratherthanmerelyrecognizeagivenresponseascorrect.Thisrepresentsamuchmoreconsiderableachievement,asthedifferencebetweenthesizesofreadingandwritingvocabulariesinaforeignlanguagedemonstrates.Thirdly,and more important, multiple-choice material violates a basic principle of goodprogramming by inducing the student to engage in erroneous behavior. Those whohave written multiple-choice tests know how much time, energy, and ingenuity areneededtoconstructplausiblewronganswers.(Theymustbeplausibleorthetestwillbeoflittlevalue.)Inamultiple-choicetest,theymaydonoharm,sinceastudentwhohasalreadylearnedtherightanswermayrejectwronganswerswitheaseandpossiblywithnoundesirable side-effects.Thestudentwho is learning,however, can scarcelyavoidtrouble.Tracesoferroneousresponsessurviveinspiteofthecorrectionoferrorsor the confirmation of a right answer. Inmultiple-choicematerial designed to teach“literary appreciation,” for example, the student is asked to consider three or fourplausibleparaphrasesofapassageinapoemandtoidentifythemostacceptable.Butasthestudentreadsandconsidersinacceptableparaphrases,theveryprocesseswhichthepoethimselfusedinmakinghispoemeffectiveareatworktodestroyit.Neithertherigorouscorrectionofwrongchoicesnortheconfirmationofarightchoicewillfreethestudentoftheverbalandnonverbalassociationsthusgenerated.Scientific subjects offer more specific examples. Consider an item such as the

following,whichmightbepartofacourseinhighschoolphysics:

Asthepressureofagasincreases,volumedecreases.Thisisbecause:(a)thespacebetweenthemoleculesgrowssmaller(b)themoleculesareflattened(c)etc….

Unless the student is as industrious and as ingenious as the multiple-choiceprogrammer,itwillprobablynothaveoccurredtohimthatmoleculesmaybeflattenedasagasiscompressed(withinthelimitsunderconsideration).Ifhechoosesitem(b)andiscorrectedbythemachine,wemaysaythathe“haslearnedthatitiswrong,”butthisdoesnotmeanthatthesentencewillneveroccurtohimagain.Andifheisunluckyenough to select the right answer first, his reading of the plausible but erroneousanswerwillbecorrectedonly“byimplication”—anequallyvagueandpresumablylesseffectiveprocess.Ineithercase,hemaylaterfindhimselfrecallingthat“somewherehehas read that molecules are flattened when a gas is compressed.” And, of course,somewherehehas.Multiple-choicetechniquesareappropriatewhenthestudent is to learntocompare

and choose. In forming a discrimination (aswith the device shown in Figure2), anorganismmust be exposed to at least two stimuli, one of whichmay be said to bewrong.Similarly,inlearningto“troubleshoot”equipmenttheremaybeseveralalmostequally plausible ways of correcting amalfunction. Games offer other examples. Agivenhandatbridgemayjustifyseveralbidsorplays,nooneofwhichiswhollyrightand all the others wrong. In such cases, the student is to learn the most expedientcoursetobetakenamonganaturalarrayofpossibilities.Thisisnottrueinthesimpleacquisition of knowledge—particularly verbal knowledge—where the task is onlyrarelytodiscriminateamongresponsesinanarray.Insolvinganequation,reportinga

236

factofhistory, restating themeaningofasentence,orengaging inalmostanyof theother behavior which is the main concern of education, the student is to generateresponses. He may generate and reject, but only rarely will he generate a set ofresponsesfromwhichhemustthenmakeachoice.Itmaybe argued thatmachineswhichprovide forbranching anddecision-making

aredesignedto teachmorethanverbalrepertoires—inparticular, that theywill teachthinking.There are strategies in choosing froman array, for example,which requirekindsofbehaviorbeyondthemereemissionofcorrectresponses.Wemayagreetothiswithoutquestioning thevalueofknowledge in the senseof averbal repertoire. (Thedistinction isnotbetween roteand insightful learning, forprogrammed instruction isespeciallyfreeofrotememorizingin theetymologicalsenseofwearingdownapaththroughrepetition.)Ifan“idea”or“proposition”isdefinedassomethingwhichcanbeexpressed inmanyways, then it may be taught by teachingmany of these “ways.”What is learned is more likely to generalize to comparable situations than a singlesyntactical form, and generalization is what distinguishes so-called deeperunderstanding.Butnotall thinking isverbal.Thereare, firstofall, alternative,parallelnonverbal

repertoires.Themathematicianbeginswith averbal problemandendswith averbalsolution,butmuchofhisinterveningbehaviormaybeofadifferentnature.Thestudentwho learns to followor construct aproof entirelybymanipulating symbolsmaynotengage in thiskindof thinking.Similarly, amerelyverbal knowledgeof physics, asoften seen in the student who has “memorized the text,” is of little interest to theserious educator. Laboratories and demonstrations sometimes supply contingencieswhichbuildsomenonverbalknowledgeofphysics.Specialkindsofteachingmachinescouldhelp,formachinesarenotonlynotconfinedtoverbalinstruction,theymaywellmakeitpossibletoreducetheemphasisonverbalcommunicationbetweenteacherandstudent.Amoreclear-cutexampleofthedistinctionbetweenverbalandnonverbalthinking

ismusicalcomposition.Thecomposerwho“thinksmusically”doesmorethanperformonaninstrumentorenjoymusic.Healsodoesmorethanusemusicalnotation.Insomesense he “thinks” pitches, intervals, melodies, harmonic progressions, and so on. Itshouldnotsurpriseus that individualsdiffergreatly in theirabilities todo this,sincethenecessarycontingenciesareinveryshortsupply.Onemightattacktheproblembysettingupanexplicitkinestheticrepertoireinwhich“thinkingapitch”takestheformof identifying a position on a keyboard. A device which arranges the necessarycontingenciesisunderdevelopment.Withitshelpwemaydiscovertheextenttowhichstudentscaningenerallearn(andatwhatagestheycanlearnmosteffectively)tostrikeakeywhichproducesatonewhichhasjustbeenheard.Similardevicesmightgenerateimportantformsofnonverbalmathematicalbehaviororthebehaviorexhibited,say,byaninventorconceivingofadeviceinthreedimensions,aswellascreativerepertoiresin other forms of art. Here is an extraordinary challenge to the technology ofinstrumentation.There is another sense in which the student must learn to think. Verbal and

nonverbal repertoires may prepare him to behave in effective ways, but he willinevitably facenovel situations inwhichhecannotat first respondappropriately.Hemaysolvesuchproblems,notbyexercisingsomementalability,butbyalteringeithertheexternalsituationortherelativeprobabilitiesofanadequateresponse.In this sense, thinking consists of a special repertoire which we may call self-

237

management.Forexample,thestudentmayaltertheextenttowhichtheenvironmentaffectshimby“attending”toitindifferentways.Asonestepinteachingthinkingwemustteacheffectiveattending.Thephrase“Payattention!”isascommononthelipsofteachersas“Open,please”onthoseofdentists—andformuchthesamereason:bothphrasessetupworkingconditions.Thestudentmaypayattentiontoavoidpunishmentand indoingsomany learn topayattention,butwhereaversivesanctionshavebeengivenup,teachershaveresortedtoattractingandholdingattention.Thetechniquesofthe publication and entertainment industries are extensively invoked. Primers areusually decorated with colored pictures, and high school textbooks are sometimesdesigned to resemble picture magazines. Films dramatize subject matters incompetitionwithnoneducationalfilmsandtelevision.Attention which is captured by attractive stimuli must be distinguished from

attentionwhich is “paid.”Only the lattermustbe learned.Lookingand listeningareformsofbehavior,andtheyarestrengthenedbyreinforcement.Apigeoncanlearntomatchcolors,forexample,onlyifit“paysattentiontothem.”Theexperimentermakessure that it does so, not by attracting its attention, but by reinforcing it for looking.Similarly, a well-taught student pays attention to sentences, diagrams, samples ofrecorded speech and music, and so on, not because they are attractive but becausesomethinginterestingoccasionallyhappensafterhehaspaidattention.Mostaudio-visualdevicesfail to teachattentionbecausetheystimulate thestudent

before he looks or listen closely. No matter how well a four-colored text or adramaticallyfilmedexperimentinphysicsattractsattention,itpreparesthestudentonlyfor comics, advertising, picture magazines, television programs, and other materialwhichisinterestingonitsface.Whatiswantedisanadultwho,uponseeingapageofblack-and-whitetext,willreaditbecauseitmayprove interesting.Unfortunately, thetechniquesassociatedwithcapturedandpaidattentionare incompatible.Wheneverateacherattractstheattentionofastudent,hedepriveshimofanopportunitytolearntopayattention.Teachingmachines,withtheircontrolovertheconsequencesofaction,canmakesurethatpayingattentionwillbeeffectivelyreinforced.Anotheractivityassociatedwiththinkingisstudying—notmerelylookingata text

andreadingitbutlookingandreadingforthesakeoffutureaction.Supposeweshowachild a picture and later, in the absence of the picture, reinforce himgenerously forcorrectanswers toquestionsabout it. Ifhehasdonenothing like thisbefore,hewillprobablynotbeverysuccessful.Ifwethenshowhimanotherpicture,hemaybegintobehave in a different way: he may engage in behavior which will increase theprobabilitythathewilllateranswerquestionscorrectly.Itwillbetohisadvantage(andtooursaseducators) if thiskindofbehavior is taught rather than left tochance.Weteachastudent“howtostudy”whenweteachhimtotakenotes,torehearsehisownbehavior, to test himself, to organize, outline, and analyze, to look for or constructmnemonicpatterns,andsoon.Someofthesebehaviorsareobvious,butothersareofmoresubtledimensionsandadmittedlyhardtoteach.Machineshaveanadvantageinmaintainingthecontingenciesrequiredforindirectormediatedreinforcement.Otheraspectsof thinking, includingthesolutionofpersonalproblems,canalsobe

analyzedanddirectlyprogrammed.Thisisnotcurrentpractice,however.Studentsaremostoften“taughttothink”simplybythrustingthemintosituationsinwhichalreadyestablished repertoires are inadequate. Some of them modify their behavior or thesituationeffectivelyandcomeupwithsolutions.Theymayhavelearned,buttheyhavenotnecessarilybeentaught,howtothink.

238

Logicians,mathematicians,andscientistshaveoftentriedtorecordandunderstandtheirownthinkingprocesses,butwearestillfarfromasatisfactoryformulationofallrelevantbehaviors.Muchremainstobelearnedabouthowaskillfulthinkerexaminesasituation, alters it, samples his own responseswith respect to it, carries out specificverbalmanipulations appropriate to it, and so on. It is quite possible thatwe cannotteachthinkingadequatelyuntilallthishasbeenanalyzed.Oncewehavespecifiedthebehavior,however,wehavenoreasontosupposethatitwillthenbeanylessadaptabletoprogrammedinstructionthansimpleverbalrepertoires.Teachingmachinesandtheassociatedpracticesofprogrammedinstructionwillhave

provedtoosuccessful if theirpracticalconsequencesareallowedtoovershadowtheirpromise for the future. We need teaching machines to help solve a very pressingproblem,butwealsoneedthemtoutilizeourbasicknowledgeofhumanbehaviorinthedesignofentirelyneweducationalpractices.Teachingmachinesareanexampleofthetechnologicalapplicationofbasicscience.

It is true that currentmachinesmight have been designed in the light of classroomexperienceandcommonsense,andthatexplanationsofwhytheyareeffectivecanbeparaphrasedintraditionalterms.Thefactremainsthatmorethanhalfacenturyoftheself-conscious examination of instructional processes had worked only moderatechanges in educational practices. The laboratory study of learning provided theconfidence,ifnotalltheknowledge,neededforasuccessfulinstrumentalattackonthestatusquo.Traditionalviewsmaynothavebeenactuallywrong,buttheywerevagueandwerenotentertainedwithsufficientcommitmenttoworksubstantialtechnologicalchanges.Asatechnology,however,educationisstillimmature,aswemayseefromthefact

that it defines its goals in terms of traditional achievements. Teachers are usuallyconcernedwith reproducing thecharacteristicsandachievementsofalreadyeducatedmen.Whenthenatureof thehumanorganism isbetterunderstood,wemaybegin toconsidernotonlywhatmanhasalreadyshownhimselftobe,butwhathemaybecomeundercarefullydesignedconditions.Thegoalofeducationshouldbenothingshortofthefullestpossibledevelopmentofthehumanorganism.Anexperimentalanalysisofbehavior, carried out under the advantageous conditions of the laboratory, willcontribute to progress toward that goal. So will practical experiments conducted inschoolsandcollegeswiththehelpofadequateinstrumentation.

FromHarvardEducationalReview,1961,31,377–398.

239

ReflectionsonaDecadeofTeachingMachines

Tothegeneralpublic,andtomanyeducatorsaswell,thenatureandscopeofteachingmachines arebynomeansclear.There is an extraordinaryneed formore andbetterteaching, and any enterprise which may help to meet it will not be left to developnormally. The demand for information about teachingmachines has been excessive.Articles and books have been published and lectures given; symposia have beenarranged, and conferences andworkshops have been held and courses taught.Thosewhohavehadanythingusefultosayhavesaiditfartoooften,andthosewhohavehadnothingtosayhavebeennomorereticent.Education is big business. Teaching machines were soon heralded as a growth

industry, and fantastic predictions of the sales of programed texts were circulated.Deviceshavebeen sold as teachingmachineswhichwerenotwell built ordesignedwith any understanding of their function or the practical exigencies of their use.Noauthorwasevermorewarmlyreceivedbyapublisherthantheauthorofaprogramedtext.Manyprograms,tobeusedeitherwithmachinesorintextbookform,havebeenmarketedwithoutadequateevaluation.

TeachersandDevices

The “mechanizing of education” has been taken literally in the sense of doing bymachine what was formerly done by people. Some of the so-called computer-basedteaching machines are designed simply to duplicate the behavior of teachers. Toautomate education with mechanical teachers is like automating banking withmechanicaltellersandbookkeepers.Whatisneededinbothcasesisananalysisofthefunctionstobeserved,followedbythedesignofappropriateequipment.Nothingwenowknowaboutthelearningprocesscallsforveryelaborateinstrumentation.Educational specialists have added to the confusion by trying to incorporate the

principlesuponwhichteachingmachinesarebasedintooldertheoriesoflearningandteaching.Inthebroadestsense,teachingmachinesaresimplydeviceswhichmakeitpossible

to apply our technical knowledge of human behavior to the practical field ofeducation.1Teachingistheexpeditingoflearning.Studentslearnwithoutteaching,buttheteacherarrangesconditionsunderwhichtheylearnmorerapidlyandeffectively.Inrecentyears,theexperimentalanalysisofbehaviorhasrevealedmanynewfactsaboutrelevantconditions.Thegrowingeffectivenessofanexperimentalanalysisisstillnotwidely recognized, even within the behavioral sciences themselves, but theimplicationsofsomeofitsachievementsforeducationcannolongerbeignored.An important condition is the relation between behavior and its consequences;

learning occurs when behavior is “reinforced.” The power of reinforcement is noteasilyappreciatedbythosewhohavenothadfirsthandexperienceinitsuseorhavenot

240

atleastseensomesortofexperimentaldemonstration.Extensivechangesinbehaviorcanbebroughtaboutbyarrangingso-calledcontingenciesof reinforcement.Variouskindsofcontingenciesareconcealedintheteacher’sdiscussionswithhisstudents,inthebookshegivesthemtoread,inthechartsandothermaterialsheshowsthem,inthequestions he asks them, and in the comments he makes on their answers. Anexperimentalanalysisclarifiesthesecontingenciesandsuggestsmanyimprovements.

ShapingbyProgram

Animportantcontributionhasbeentheso-called“programing”ofknowledgeandskills—the construction of carefully arranged sequences of contingencies leading to theterminal performances which are the object of education. The teacher begins withwhatever behavior the student brings to the instructional situation; by selectivereinforcement,hechangesthatbehaviorsothatagiventerminalperformanceismoreandmorecloselyapproximated.Evenwithlowerorganisms,quitecomplexbehaviorscanbe“shaped”inthiswaywithsurprisingspeed;thehumanorganismispresumablyfarmoresensitive.Soimportantistheprincipleofprogramingthatitisoftenregardedas the main contribution of the teaching-machine movement, but the experimentalanalysisofbehaviorhasmuchmoretocontributetoatechnologyofeducation.The direct contact which often exists between teacher and student favors the

constructionofprogramedsequences,andtheteacherwhounderstandstheprocesscanprofit from the opportunity to improvise programs as he goes. Programs can beconstructed inadvance,however,whichwill successfullyshape thebehaviorofmoststudentswithoutlocalmodifications,andmanyofthemcanconvenientlybemediatedbymechanicaldevices.Laboratorystudieshaveshownthatcontingenciesemphasizingsubtlepropertiesofbehaviorcanoftenbearrangedonlythroughinstrumentation.Thereare potentially as many different kinds of teaching machines as there are kinds ofcontingenciesofreinforcement.Teachingmachineswhichpresentmaterialtothestudentanddifferentiallyreinforce

his responses in well constructed programs differ in several ways from self-testingdevices and self-scoring test forms, aswell as from the training deviceswhichhavelongbeenusedbyindustryandthearmedservices.AsPresseypointedoutmanyyearsago,2astudentwilllearnwhiletakingamultiple-choicetestifheistoldimmediatelywhetherhisanswersarerightorwrong.Helearnsnot togivewrongsanswersagain,andhisrightanswersarestrengthened.Buttestinghastraditionallybeendistinguishedfrom teaching for good reason. Before using a self-testing device, the student mustalreadyhavestudiedthesubjectand,presumably,learnedmostofwhatheistolearnabout it. Tests usually occupy only a small part of his time. Their main effect ismotivational:Apoorscoreinduceshimtostudyharderandpossiblymoreeffectively.Materialsdesignedtobeusedinself-testingdeviceshaverecentlybeenprogramed,butthe contingencies which prevail during a test are not favorable to the shaping andmaintainingofbehavior.Conventionaltrainingdevicesarrangeconditionsunderwhichstudentslearn,usually

by simulating the conditions under which they eventually perform. Their originalpurposewastopreventinjuryorwasteduringearlystagesoflearning,butattentionhasrecentlybeengiventoprogramingtheactualbehaviorstheyaredesignedtoteach.To

241

the extent that they expedite learning, they are teaching machines. Terminalperformanceshaveusuallybeen selected forpractical reasons, but amorepromisingpossibilityistheanalysisandprogramingofbasicmotorandperceptualskills—agoalwhichshouldhaveanimportantplaceinanystatementofeducationalpolicy.In arranging contingencies of reinforcement, machines do many of the things

teachers do; in that sense, they teach. The resulting instruction is not impersonal,however.Amachinepresentsaprogramdesignedbysomeonewhoknewwhatwastobetaughtandcouldprepareanappropriateseriesofcontingencies.Itismosteffectiveifusedbyateacherwhoknowsthestudent,hasfollowedhisprogress,andcanadaptavailable machines and materials to his needs. Instrumentation simply makes itpossible for programer and teacher to provide conditionswhichmaximally expeditelearning.Instrumentationisthussecondary,butit isneverthelessinevitableifwhatisnowknownaboutbehavioristobeusedinaneffectivetechnology.

TheNewPedagogy

Anypracticalapplicationofbasicknowledgeaboutteachingandlearningis,ofcourse,pedagogy.IntheUnitedStatesatleast,thetermisnowdiscredited,butbyemphasizingananalysisof learningprocesses, teachingmachinesandprogramed instructionhavebeenresponsibleforsomeimprovementinitsstatus.Thesignificanceoftheteachingmachinemovementcanbe indicatedbynoting theastonishing lackof interestwhichotherproposalsfortheimprovementofeducationshowintheteachingprocess.Findbetterteachers.InhisTalkstoTeachers,WilliamJamesinsistedthattherewas

nothing wrong with the American school system which could not be corrected by“impregnatingitwithgeniuses.”3Itisanoldformula:Ifyoucannotsolveaproblem,findsomeonewhocan.Ifyoudonotknowhowtoteach,findsomeonewhoknowsorcan findout for himself.But geniuses are in short supply, andgood teachers donotcome ready-made. Education would no doubt be improved if, as Conant4 hasrepeatedlypointedout,goodteacherswhoknowandlikethesubjectstheyteachcouldbe attracted and retained.But somethingmore is needed. It is not true that “the twoessentials of a good teacher are (a) enthusiasm and (b) thorough knowledge of aninterestinhissubject.”5Athirdessentialisknowinghowtoteach.Emulatemodelschools.Rickover’scriticismofthepresentAmericanschool system

is well known.6 His only important positive suggestion is to set up model schools,staffedbymodelteachers.Theimplicationisthatwealreadyhave,oratleastcanhavefortheasking,schoolswhichneednoimprovementandwhosemethodscanbewidelycopied.Thisisadangerousassumptionifitdiscouragesfurtherinquiryintoinstruction.Simplifywhatistobelearned.Unsuccessfulinstructionisoftenblamedonrefractory

subject matters. Difficulties in teaching the verbal arts are often attributed to theinconsistencies and unnecessary complexities of a language. The pupil is taughtmanuscripthandwritingbecauseitmorecloselyresemblesprintedforms.Heistaughttospellonlythosewordsheislikelytouse.Phoneticalphabetsaredevisedtohelphimlearntoread.Itmaybeeasiertoteachsuchmaterials,butteachingitselfisnottherebyimproved. Effective teaching would correct these pessimistic estimates of availableinstructionalpower.Reorganize what is to be learned. The proper structuring of a subject matter is

242

perhapsapartofpedagogy,but it canalso serveasamodeofescape.Proposals forimprovingeducationbyreorganizingwhatistobelearnedusuallycontainanimplicitassumptionthatstudentswillautomaticallyperceiveandrememberanythingwhichhas“good form”—adoctrineprobably traceable toGestaltpsychology.Current revisionsofhighschoolcurriculaoftenseemtoleanheavilyonthebeliefthatifwhatthestudentis tobe taughthasbeen“structured,”hecannothelpunderstandingandrememberingit.7Otherpurposesofsuchrevisionscannotbequestioned:Materialsshouldbeuptodate and well organized. But a high school presentation acceptable to a currentphysicist is no more easily taught or easily remembered than the out-of-date anderroneousmaterialtobefoundintextsofadecadeormoreago.Similarly,theaccentofanativespeakerencounteredinalanguagelaboratoryisnomoreeasilylearnedthanabadaccent.Nomatterhowwellstructuredasubjectmattermaybe, itmuststillbetaught.Improvepresentation.Pedagogycanalsobeavoidedifwhatistobelearnedcanbe

madememorable.Audio-visualdevicesareoftenrecommendedforthispurpose.Manyof theirotherpurposesareeasilydefended. It isnotalwayseasy tobring thestudentinto contactwith the things he is to learn about.Words are easily imported into theclassroom,andbooks,lectures,anddiscussionsarethereforestaplesofeducation;butthis is often an unfortunate bias. Audio-visual devices can enlarge the student’snonverbalexperience.Theycanalsoservetopresentmaterialclearlyandconveniently.Theiruseinattractingandholdingthestudent’sattentionandindramatizingasubjectmatter insuchawaythat it isalmostautomaticallyrememberedmustbequestioned,however.Itisespeciallytemptingtoturntothemforthesepurposeswhentheteacherdoesnotusepunitivemethodsto“makestudentsstudy.”Buttheresultisnotthesame.When a student observes or attends to something in order to see it more clearly orrememberitmoreeffectively,hisbehaviormusthavebeenshapedandmaintainedbyreinforcement.Thetemporalorderwasimportant.Certainreinforcingeventsmusthaveoccurredafter thestudentlookedat,read,andperhapstestedhimselfonthematerial.Butwhen colored displays, attractive objects, filmed episodes, and other potentiallyreinforcingmaterialsareused toattractattention, theymustoccurbefore the studentengagesintheseactivities.Nothingcanreinforceastudentforpayingattentionifithasalready been used to attract his attention. Material which attracts attention fails toprepare the student to attend to material which is not interesting on its face, andmaterialwhich isnaturallymemorable fails topreparehimtostudyandrecall thingswhich are not, in themselves, unforgettable. Awell prepared instructional filmmayappear tobesuccessful inarousinginterest inagivensubject,andpartsof itmayberememberedwithouteffort,butithasnottaughtthestudentthatasubjectmaybecomeinterestingwhenmorecloselyexaminedorthatintensivestudyofsomethingwhichislikelytobeoverlookedmayhavereinforcingconsequences.Multiplycontactsbetween teacherandstudents. Audio-visual devices, particularly

whenadaptedtotelevision,arealsousedtoimproveeducationbybringingoneteacherintocontactwithanindefinitelylargenumberofstudents.Thiscanbedone,ofcourse,without analyzing how the teacher teaches, and it emphasizes a mode ofcommunicationwhichhastwoseriousdisadvantages:Theteachercannotseetheeffectheishavingonhisstudents,andlargenumbersofstudentsmustproceedatthesamepace.Contributionstopedagogymaybemadeindesigningprogramsforeducationaltelevision, but the mere multiplication of contacts is not itself an improvement in

243

teaching.Expandtheeducationalsystem.Inadequateeducationmaybecorrectedbybuilding

more schools and recruitingmore teachers so that the total quantity of education isincreased,eventhoughthereisnochangeinefficiency.Raise standards. Least effective in improving teaching are demands for higher

standards.Wemayagreethatstudentswillbebettereducatedwhenthey learnmore,buthowaretheytobeinducedtodoso?Demandsforhigherstandardsusuallycomefromcriticswhohaveleasttoofferinimprovingteachingitself.Themovementsymbolizedbytheteachingmachinediffersfromotherproposalsin

twoways. It emphasizes thedirect improvementof teachingon theprinciple thatnoenterprisecanimproveitselftothefullestextentwithoutexaminingitsbasicprocesses.In the second place, it emphasizes the implementation of basic knowledge. Ifinstructional practices violatemany basic principles, it is only in part because theseprinciplesarenotwidelyknown.Theteachercannotputwhatheknowsintopracticeintheclassroom.Teachingmachinesandprogramedinstructionconstituteadirectattackontheproblemofimplementation.Withappropriateadministrativechanges,theymaybridgethegapbetweenaneffectivepedagogicaltheoryandactualpractice.

EducationalGoals

An effective technology of teaching calls for a re-examination of educationalobjectives. What is the teacher’s actual assignment? Educational policy is usuallystated in traditional terms: The teacher is to “impart knowledge,” “improve skills,”“develop rational faculties,” and so on. That education is best, says Dr. Hutchins,8whichdevelops“intellectualpower.”Thetaskoftheteacheristochangecertaininnerprocessesorstates.Heistoimprovethemind.The role of the teacher in fosteringmental prowess has a certain prestige. It has

alwaysbeenheldsuperiortotheroleofthetrainerofmotorskills.Andithasthegreatadvantageofbeingalmost invulnerable tocriticism.Inreply to thecomplaint thathehasnotproducedobservableresults,theteacherofthemindcanlayclaimtoinvisibleachievements.His studentsmay not be able to read, but he has only been trying tomakesure theywanted to learn.Theymaynotbeable tosolveproblems,buthehasbeenteachingthemsimplytothinkcreatively.Theymaybeignorantofspecificfacts,buthehasbeenprimarilyconcernedwiththeirgeneralinterestinafield.Traditionalspecificationsofthegoalsofeducationhavenevertoldtheteacherwhat

todouponagivenoccasion.Nooneknowshowtoalteramentalprocessorstrengthenamentalpower,andnoonecanbesurethathehasdonesowhenhehastried.Therehavebeenmanygood teacherswhohave supposed themselves tobeworkingon themindsoftheirstudents,buttheiractualpracticesandtheresultsofthosepracticescanbe analyzed in other ways. The well educated student is distinguished by certaincharacteristics. What are they, and how can they be produced? Perhaps we couldanswer by redefining traditional goals: Instead of imparting knowledge, we couldundertake to bring about those changes in behavior which are said to be theconspicuousmanifestationsofknowledge,orwecouldsetupthebehaviorwhichisthemarkofamanpossessingwell-developedrationalpower.Butmentalisticformulationsarewarpedby irrelevanthistoricalaccidents.Thebehaviorof theeducatedstudent is

244

muchmoreeffectivelyanalyzeddirectlyassuch.Contrarytofrequentassertions,abehavioristicformulationofhumanbehaviorisnot

acrudepositivismwhichrejectsmentalprocessesbecausetheyarenotaccessibletothescientificpublic.9 Itdoesnotemphasizetherote learningofverbalresponses.Itdoesnot neglect the complex systems of verbal behavior which are said to show that astudenthashadan idea,ordevelopedaconcept,orentertainedaproposition. Itdoesnotignorethebehaviorinvolvedintheintellectualandethicalproblemsolvingcalled“thinking.” It does not overlook the value judgments said to be invoked when wedecide to teachone thing rather thananotherorwhenwedefend the timeandeffortgiventoeducation.It ismerelyaneffectiveformulationof thoseactivitiesof teacherandstudentwhichhavealwaysbeentheconcernofeducationalspecialists.10Not all behavioristic theories of learning are relevant, however. A distinction is

commonlydrawnbetweenlearningandperformance.Learningissaidtobeachangeinsomespecialpartof theorganism,possibly thenervoussystem,ofwhichbehavior ismerelytheexternalandoftenerraticsign.Withmoderntechniques,however,behaviorcanbemuchmoresuccessfullystudiedandmanipulated thananysuch innersystem,evenwhen inferences about the latter are drawn from the behaviorwith the help ofsophisticated statistics. An analysis of learning which concentrates on the behaviorapplies most directly to a technology, for the task of the teacher is to bring aboutchanges in the student’s behavior. His methods are equally conspicuous: He makeschanges in the environment. A teaching method is simply a way of arranging anenvironmentwhichexpediteslearning.

ManagingContingencies

Suchaformulationisnoteasilyassimilatedwiththetraditionalpsychologyoflearning.The teacher may arrange contingencies of reinforcement to set up new forms ofresponse,asinteachinghandwritingandspeechornonverbalformsofbehaviorinthearts, crafts, and sports.Hemay arrange contingencies to bring responses under newkindsofstimuluscontrol,as in teaching thestudent to readordrawfromcopy,or tobehave effectively upon other kinds of occasions. Current instructional programsdesigned to fulfill suchassignmentsaremainlyverbal,butcomparablecontingenciesgeneratenonverbalbehavior, includingperceptualandmotorskillsandvariouskindsofintellectualandethicalself-management.Asecondkindofprogramingmaintainsthestudent’sbehaviorinstrength.Theform

of the responseand thestimuluscontrolmaynotchange; thestudent issimplymorelikelytorespond.Somerelevantmethodsaretraditionallydiscussedundertheheadingofmotivation.Forexample,wecanstrengthenbehaviorbyintroducingnewreinforcersormakingoldonesmoreeffective,asingivingthestudentbetterreasonsforgettinganeducation. The experimental analysis of behavior suggests another importantpossibility: Schedule available reinforcers more effectively. Appropriate terminalschedulesof reinforcementwillmaintain thestudent’s interest,makehimindustriousand persevering, stimulate his curiosity, and so on; but less demanding schedules,carefully designed to maintain the behavior at every stage, must come first. Theprograming of schedules of reinforcement is a promising alternative to the aversivecontrolwhich,inspiteofrepeatedreforms,stillprevailsineducationalpractice.

245

In neglecting programing, teachingmethods havemerely followed the lead of theexperimentalpsychologyof learning,wherethealmostuniversalpracticehasbeentosubmitanorganismimmediatelytoterminalcontingenciesofreinforcement.11Amazeor a discrimination problem, for example, is learned only if the subject acquiresappropriatebehaviorbeforethebehaviorhebringstotheexperimenthasextinguished.Theintermediatecontingenciesarelargelyaccidental.Thedifferencesinbehaviorandinrateoflearningwhichappearundertheseconditionsareoftenattributedtoinheriteddifferencesinability.Inmaximizing the student’s success, programed instruction differs from so-called

trial-and-errorlearningwherethestudentissaidtolearnfromhismistakes.Atbest,helearnsnot tomakemistakesagain.Asuccessful responsemaysurvive,but trial-and-error teachingmakes littleprovision foractually strengthening it.Themethodseemsinevitablycommitted toaversivecontrol.For thesamereason,programed instructiondoesnotcloselyresembleteachingpatternedoneverydaycommunication.Itisusuallynotenoughsimplytotellthestudentsomethingorinducehimtoreadabook;hemustbe toldormustreadandthenbequestioned. In this“tell-and-test”pattern, the test isnotgiventomeasurewhathehaslearned,buttoshowhimwhathehasnotlearnedandthus induce him to listen and readmore carefully in the future. A similar basicallyaversivepatterniswidespreadatthecollegelevel,wheretheinstructorassignsmaterialand then examines on it. The studentmay learn to read carefully, tomake notes, todiscoverforhimselfhowtostudy,andsoon,becauseindoingsoheavoidsaversiveconsequences, but he has not necessarily been taught. Assigning-and-testing is notteaching.Theaversiveby-products,familiartoeveryoneinthefieldofeducation,canbeavoidedthroughtheuseofprogramedpositivereinforcement.Many facts and principles derived from the experimental analysis of behavior are

relevant to the construction of effective programs leading to terminal contingencies.The facts and principles are often difficult, but they make up an indispensablearmamentariumoftheeffectiveteacherandeducationalspecialist.Wehavelongsincepassed thepointatwhichourbasicknowledgeofhumanbehaviorcanbeapplied toeducationthroughtheuseofafewgeneralprinciples.

PrincipleandPractice

Thedifferencebetweengeneralprinciplesandaneffectivetechnologycanbeseenincertain efforts to assimilate the principles of programed instruction with earliertheories.Programedinstructionhas,forexample,beencalled“Socratic.”ItistruethatSocratesproceededbysmallstepsandoftenledhisstudentsthroughanargumentwithaseriesofverbalprompts,buttheexampleoftencitedtoillustratehismethodsuggeststhathewasunawareofanimportantdetail—namely,thatpromptsmusteventuallybe“vanished” inorder toput thestudentonhisown. In the famousscene in theMeno,Socrates demonstrates his theory that learning is simply recollection by leading anuneducatedslaveboy throughPythagoras’sGoldenTheorem.Theboyrespondswiththerathertimidcompliancetobeexpectedunderthecircumstancesandneverwithouthelp. Although Socrates himself and some of those among his listeners who werealreadyfamiliarwith thetheoremmayhaveunderstoodtheproofbetterat theendofthe scene, there is no evidence whatsoever that the boy understood it or could

246

reconstruct it. In this example of Socratic instruction, at least, the student almostcertainlylearnednothing.TheprogramoftheMenoepisodeconstructedbyCohen12isanimprovementinthatthestudentrespondswithlessprompting.Aseventeenth-centuryanticipationofprogramedinstructionhasalsobeenfoundin

theworkofComenius,whoadvocatedteachinginsmallsteps,nostepbeingtoogreatforthestudentwhowasabouttotakeit.Programingissometimesdescribedsimplyasbreakingmaterialintoalargenumberofsmallpieces,arrangedinaplausiblegeneticorder.Butsizeofstepisnotenough.Somethingmusthappentohelpthestudenttakeeachstep,andsomethingmusthappenashetakesit.Aneffectiveprogramisusuallycomposed of small steps, but the whole story is not to be found in Comenius’sphilosophyofeducation.Anothervenerableprincipleisthatthestudentshouldnotproceeduntilhehasfully

understood what he is to learn at a given stage. Several writers have quoted E. L.Thorndike13 to this effect, who wrote in 1912, “If, by a miracle of mechanicalingenuity, a book could be so arranged that only to him who had done what wasdirected on page one would page two become visible, and so on, much that nowrequires personal instruction could be managed by print.” In commenting on thispassage,FinnandPerrin14havewritten,“…Herearetheinsightsofagenius.Historycanveryoftenteachusalessoninhumility—anditdoeshere.Theinterestingquestionis:Whycouldn’twesee it then?”Wemightalsoask,whycouldn’tThorndikesee itthen?He remained active in education for at least 30 years, but he turned from thisextraordinarily promising principle to another and—as it proved—less profitableapproachtoeducationalpsychology.Itisalwaystemptingtoarguethatearlierideaswouldhavebeeneffectiveifpeople

had only paid attention to them. But a good ideamust bemore than right. It mustcommandattention;itmustmakeitsownwaybecauseofwhatitdoes.Educationdoesnotneedprincipleswhichwill improveeducationassoonaspeopleobserve them; itneedsatechnologysopowerfulthatitcannotbeignored.Nomatterhowinsightfultheanticipationofmodernprinciplesinearlierwritersmayseemtohavebeen,somethingwaslackingoreducationwouldbemuchfurtheradvanced.Weareonthethresholdofatechnologywhichwillbenotonlyrightbuteffective.15

CriteriaofResearch

A science of behaviormakes its principal contribution to a technology of educationthrough theanalysisofusefulcontingenciesof reinforcement. It also suggestsanewkind of educational research. Thorndike never realized the potentialities of his earlywork on learning because he turned to the measurement of mental abilities and tomatched-groupcomparisonsofteachingpractices.Hepioneeredinakindofresearchwhich,withtheencouragementofferedbypromisingnewstatisticaltechniques,wastodominateeducationalpsychologyfordecades.Itledtoaseriousneglectoftheprocessofinstruction.There are practical reasons why we want to know whether a given method of

instructionissuccessfulorwhetheritismoresuccessfulthananother.Wemaywanttoknowwhatchangesitbringsaboutinthestudent,possiblyinadditiontothoseitwasdesigned to effect. Themore reliable our answers to such questions, the better. But

247

reliability is not enough.Correlations between test scores and significant differencesbetweengroupmeanstelluslessaboutthebehaviorofthestudentintheactoflearningthan results obtainedwhen the investigator canmanipulatevariables and assess theireffects in a manner characteristic of laboratory research. The practices evaluated instudiesofgroupsofstudentshaveusuallynotbeensuggestedbyearlierresearchofasimilarnature,buthavebeendrawnfromtradition,fromtheimprovisationsofskillfulteachers,orfromsuggestionsmadebytheoristsworkingintuitivelyorwithotherkindsof facts. Nomatter howmuch theymay have stimulated the insightful or inventiveresearcher, the evaluations have seldom led directly to the design of improvedpractices.Thecontrastbetweenstatisticalevaluationandtheexperimentalanalysisofteaching

hasanilluminatingparallelinthefieldofmedicine.Variousdrugs,regimens,surgicalprocedures, and so on, must be examinedwith respect to a very practical question:Does thehealthof thepatient improve?But“health” isonlyageneraldescriptionofspecific physiological processes, and “improvement” is, so to speak, merely a by-productofthechangesintheseprocessesinducedbyagiventreatment.Medicinehasreachedthepointwhereresearchonspecificprocessesisamuchmorefertilesourceofnewkindsoftherapythanevaluationsintermsofimprovementinhealth.Similarly,ineducation,nomatterhowimportantimprovementinthestudent’sperformancemaybe,it remains a by-product of specific changes in behavior resulting from the specificchangesintheenvironmentwroughtbytheteacher.Educationalresearchpatternedonan experimental analysis of behavior leads to a much better understanding of thesebasicprocesses.Researchdirectedtowardthebehavioroftheindividualstudenthas,ofcourse, a long history, but it can still profit greatly from the support supplied by anexperimentalanalysisofbehavior.This distinction explains why those concerned with experimental analyses of

learning are not likely to takematched-group evaluations of teachingmachines andprogramed instruction very seriously. It is not possible, of course, to evaluate eithermachines or programs ingeneral because only specific instances can be tested, andavailableexamplesbynomeansrepresentallthepossibilities;buteventheevaluationof a givenmachine or program in the traditional manner may not give an accurateaccount of its effects. For example, thosewho are concernedwith improvement arelikely to test the student’scapacity togive rightanswers.Being righthas,ofcourse,practicalimportance,butitisonlyoneresultofinstruction.Itisadoubtfulmeasureof“knowledge”inanyusefulsense.Wesaythatastudent“knowstheanswer”ifhecanselect it fromanarrayofchoices,but thisdoesnotmean thathecouldhavegiven itwithout help. The right answer to one question does not imply right answers to allquestions said to show the “possession of the same fact.” Instructional programs areoftencriticizedasrepetitiousorredundantwhentheyareactuallydesignedtoputthestudent in possession of a number of different responses “expressing the sameproposition.”Whether such instruction is successful is not shown by any one rightanswer.

CorrectorEducated?

A preoccupation with correct answers has led to a common misunderstanding ofprogramed materials. Since a sentence with a blank to be filled in by the student

248

resemblesatestitem,itisoftensupposedthattheresponsedemandedbytheblankiswhat is learned. In that case, a student could not be learningmuch because hemayrespondcorrectlyin19outof20framesandmustthereforealreadyhaveknown95percentoftheanswers.Theinstructionwhichoccursashecompletesanitemcomesfromhavingrespondedtootherpartsofit.Theextentofthisinstructioncannotbeestimatedfromthefactthatheisright19outof20times,eitherwhilepursuingaprogramoronasubsequenttest.Norwillthisstatistictelluswhetherotherconditionsareimportant.Is it most profitable for the student to execute the response by writing it out, byspeaking it aloud,by speaking it silently,orby reading it in someotherway?Theseproceduresmayormaynothavedifferenteffectsonaselected“right-answer”statistic,butnoonestatisticwillcoveralltheireffects.Researchinteachingmustnot,ofcourse,losesightofitsmainobjective—tomake

educationmoreeffective.Butimprovementassuchisaquestionabledimensionofthebehavior of either teacher or student. Dimensions which are much more intimatelyrelatedtotheconditionstheteacherarrangestoexpeditelearningmustbestudiedeventhoughtheydonotcontributetoimprovementorcontributetoitinawaywhichisnotimmediatelyobvious.Thechangesinthebehavioroftheindividualstudentbroughtaboutbymanipulating

the environment are usually immediate and specific. The results of statisticalcomparisonsofgroupperformancesusuallyarenot.Fromhisstudyofthebehavioroftheindividualstudent, the investigatorgainsaspecialkindofconfidence.Heusuallyknowswhathehasdonetogetoneeffectandwhathemustdotogetanother.Confidence in education is another possible result of an effective technology of

teaching.Competitionbetweenthevariousculturesoftheworld,warlikeorfriendly,isnow an accepted fact, and the role played by education in strengthening andperpetuating a given way of life is clear. No field is in greater need of our mostpowerful intellectual resources. An effective educational technology based upon anexperimental analysiswill bring it support commensuratewith its importance in theworldtoday.

FromTeachersCollegeRecord,1963,65,168–177.1Skinner,B.F.TheScienceofLearningandtheArtofTeaching.HarvardEducationalReview,1954,24,86–97.[Thisvolume,pp.179–191.]

2Pressey,S.J.Asimpledeviceforteachingtesting,andresearchinlearning.Sch.&Soc.,1926,23,373–376.

3James,W.TalkstoTeachers.NewYork:Holt,1899.4Conant,J.B.TheEducationofAmericanTeachers.NewYork:McGraw-Hill,1963.5Helwig,J.Trainingofcollegeteachers.Science,1960,132,845.6Rickover,H.G.EducationandFreedom.NewYork:Dutton,1959.7Bruner,J.S.TheProcessofEducation.Cambridge:HarvardUniversityPress,1960.8Hutchins,R.M.OnEducation.SantaBarbara:CenterfortheStudyofDemocraticInstitutions,1963.

9Skinner,B.F.Behaviorismatfifty.Science,1963,140,951–958.10Skinner,B.F.Whyweneedteachingmachines.(Seepage217.)

249

11Skinner,B.F.Operantbehavior.AmericanPsychologist,1963,18,503–515.12Cohen,I.S.ProgramedlearningandtheSocraticdialogue.AmericanPsychologist,1962,17,772–775.

13Thorndike,E.L.Education.NewYork:Macmillan,1912.14Finn,J.D.,andPerrin,D.G.TeachingMachinesandProgrammedLearning:ASurveyoftheIndustry,1962.Washington,D.C.:U.S.OfficeofEducation,1962.

15Skinner,B.F.TheTechnologyofTeaching.NewYork:Appleton-Century-Crofts,1968.

250

TeachingScienceinHighSchool—WhatIsWrong?

Thescientificcommunityfacesaseriousproblem.Scienceandtechnologyaregrowingatanever-increasingrate,butthenumberofyoungmenandwomengoingintoscienceis not keeping pace. Only a fairly small percentage of high school students go tocollege expressing an interest in becoming scientists, and many of these eventuallyshift tootherfields.There isalreadyanacuteshortage,whichcouldprovedisastrousnot only for science itself but for a way of life which becomes more and moredependentonscienceastheyearspass.Apossibleexplanationisthatthelifeofthescientisthaslostsomeofitsglamour.It

may offer less chance for individual achievement, and its excitingmomentsmay bereservedonlyforthosewhohavehadaveryextensivepreparation.Evenso,themainfaultmustliewitheducation.Goodteachingshouldgiveanaccurateaccountofwhatscienceisanddoes,ofwhatasinglescientistmaycontributetotheworld,andofthegenuineexcitementofthosewhoenjoyscienceforwhatitis—thegreatartofthe20thcentury. Above all, education should recruit the scientists of the future, finding theright people, giving them the knowledge and skills they need, and providing thesatisfactionswhichwillmakethemcreativeanddedicatedmenandwomen.Onlyifitdoessocanwehopetofindthosewhowillpracticescienceinouruniversitiesandinindustrialandgovernmentallaboratories,andwhowillteachscienceinourschoolsandcollegestokeeptheenterprisegoing.Onlyeffectiveteachingwillcreatethatlargepoolfromwhich,ineachgeneration,afewgreatscientistsaredrawn.The problem has not gone unnoticed. For the past 10 or 15 years education as a

wholehasbeensharplycriticized,andmanyconstructivesuggestionshavebeenmade.We are all familiar with proposed remedies. Education needs support, and supportmeansmoney,and themoney is tobeused inavarietyofways.Weneedmoreandbetter schools. We need to recruit and hold better teachers, selecting them throughbettersystemsofqualificationandmakingthemmorecompetentinthefieldsinwhichtheyteach.Weneedtogiveallqualifiedstudentsachance,selectingthemimpartially,supporting them financially, and removing social and racial barriers.We needmoreand better capital equipment—texts, workbooks, films, and audiovisual devices,includingteachingmachinesandtelevision.Weneedtochangeourcurricula,makingasensibleselectionamongthethingstobetaughtandbringingwhatistaughtup-to-date.Highschoolscienceteachinghasbeensingledoutforspecialeffort,andthereisno

doubtthatimportantstepshavebeentaken,butthereisnotyetanygreatchange.Thecurveshowingthenumberofstudentsgoingintoscience,particularlyphysics,hasnotturnedsharplyupward.Possiblyit istoosoontoexpectresults.Educationalpracticeschange slowly, and we may yet see progress. But some possible reasons whyimprovementhasnotbeenmoredramaticmaybepointedout.Thereisacuriousomissioninthislistofeducationalneeds.Nothingissaidabouta

betterunderstandingoftheprocessesinvolvedinlearningandteaching.Nosuggestionismadethatweshouldlearnmoreaboutwhatishappeningwhenateacherteachesand

251

astudentlearns.Onthecontrary,theissueisavoidedinalmostallcurrentproposalsfortheimprovementofeducation.Pedagogyisadirtyword,andcoursesin“method”arediscounted, if not ridiculed. This is a serious mistake. As science itself has soabundantlydemonstrated,thepowerofanytechnologydependsuponanunderstandingofitsbasicprocesses.Wecannotreallyimproveteachinguntilweknowwhatitis.Themostcasualattitudetowardabetterunderstandingofinstructionisevidentatall

levels.Youwillnotfindanythinglikeamedicalschool,lawschool,orbusinessschoolfor those who want to be college teachers. No professional training is felt to benecessary. Preparation for grade and high school teaching is scarcelymore explicit.Schools of education no longer actively promote pedagogy ormethod as formalizedpractice. Instead, the beginning teacher serves an apprenticeship. He watches otherteachers and learns tobehaveas theybehave, andeventuallyhemayprofit fromhisownclassroomexperience.Inthelongrun,highschoolteachers,likecollegeteachers,teach as they themselves have been taught, as they have seen others teach, or asexperiencedictates.

ClassroomExperience

What is learned from classroom experience is perhaps likely to bemore useful thanformalizedrulesandprescriptions,buttheclassroomisneverthelessnotanidealsourceofeducationalwisdom.Onthecontrary,itcanbeseriouslymisleading.FrancisBacononce formulated his famous Idols—the false notions or fallacies which led to badthinking. I have suggested1 that we should add another to his list: the Idols of theSchool.TheIdolorFallacyoftheGoodTeacheristhebeliefthatwhatagoodteachercando,anyteachercando.Somepeoplearesociallyskillful;theyaregoodjudgesofcharacterandgetalongwellwithpeople.Theymakegoodteachers.Thetroubleis,wedonotknowwhy.Like theold-timedoctor, theypractice anartwhichhasnotbeenanalyzedandcanseldombecommunicated.Inthehandsofagoodteacheranewtext,anewsetofmaterials,oranewmethodmaybedramaticallysuccessful,butitdoesnotfollowthatitwillbesuccessfulinthehandsofteachersatlarge.ThecomplementaryFallacy of the Good Student is the belief that what a good student can learn, anystudentcanlearn.Somestudentsarehighlyintelligentandwellmotivated.Theyknowhow to study, and they learn without being taught or even when taught by a badteacher.Butatext,asetofmaterials,oramethodwhichworkswellwiththemwillnotnecessarilybeasuccesswithallstudents.Formanyyearseducational journals, schoolbulletins, and thepopularmediahave

reportedexamplesofeffective teaching.Theyhaveportrayed livelyclasses inwhichteachers and students work together in harmony and the students obviously learn agreatdeal.Everyoneispleased.Theteacherstakesatisfactioninwhattheyaredoing,the studentsenjoy themselvesandmakeprogress, andadministratorsandparentsaredelighted.But is itnot time toaskwhy theseexamplesarenotmorewidelycopied?Why,by this time, isnotall teachingequallypleasantandprofitable?Theanswer isprobably tobe found in the Idolsof theSchool.Weare lookingatgood teachersorgood students or both, but not at practices which have been analyzed or can becommunicated.Wecannotimproveeducationtoanygreatextentbyfindingmoregoodteachersandmoregoodstudents.Weneedtofindpracticeswhichpermitallteachersto

252

teachwellandunderwhichallstudentslearnasefficientlyastheirtalentspermit.Afirst step is to recognizehowmisleadingclassroomexperience isasasourceof

educationalwisdom.Itsoutstandingdefectisthattheteacherseldomseestheeffectsofwhathehasdone.Thesignificantresultsofteachinglieinthatdistantfutureinwhichstudentsmakeuseofwhat theyhave learned,and it is a futureusuallyclosed to theteacher.Heknowsnothingofwhathappenstomostofhisstudents.Heis influencedinsteadonly by short-term results, andmanyof these not only contribute nothing tolong-termgainsbutmayactuallyconflictwiththem.

ClassroomGames

No teacher enjoys studentswho are disorganized, inattentive, lethargic, or resentful.Butstudentsmaybelivelyandattentiveinwayswhichhavelittletodowithwhatorhowmuchtheyarelearning.Inafamiliar—perhapstoofamiliar—classroompractice,theteacherasksquestionsandthestudentsanswer.Thestudentsarerewardedforrightanswersandpunishedforwrong,andanythingastudentdoestobecalledonwhenheknows theansweroroverlookedwhenhedoesnotwillbe reinforced.The teacher isreinforcedeitherbyrightanswersiftheyshowthathehasbeenteachingsuccessfullyorbywrongifhemustcontroltheclassthroughathreatofpunishment,andanythinghemaydotogetarightanswerwhenhewantsarightanswerorawrongonewhenhewantsawrongwillbereinforced.Thesearetheessentialconditionsforacomplexgameinwhichteacherandstudents

attempttooutguesseachother.Thestudentwhoknowsananswerwaveshishand,anda teacherwhowantsa rightanswercallsonhim,buthecallson someoneelse ifhewantsawronganswer,andthestudentwhodoesnotknowtheanswerthenraiseshishandtoavoidbeingcalledonand thestudentwhoknowstheanswerkeepshishanddown, hoping to get a chance. The class is excited, the teacher is in control, andeveryonemaybehavingagood time.But thegame isquiteunrelated to the subjectbeing taught—it is the same for all subjects—and its educational value may bequestioned.Itmayinducesomestudentstoengageinmoreprofitableactivities,butitisnotcharacteristicofthoughtfuldiscussionorstudy,anditslong-termeffectsmaybenegligibleorevenharmful.Adull,lethargicclassisnodoubtthesignofabadteacher,butanexcitedclassisnotnecessarilythesignofagoodone.Hand-wavingmayseemtootrivialtomention,butthesamekindofgameisplayed

withverbalinterchanges.ThemodernSocrates,likehisfamouspredecessor,playscatand mouse with his students, pretending to misunderstand, constructing absurdparaphrases, making suggestions which lead his listeners into error, making ironiccommentswhichamusesomeofhislistenersattheexpenseofothers,andsoon.Ifheisskillful,hemayinducehisstudentstoprotest,disagree,insist,anddefendthemselvesinalivelyfashion.Allthisisvaluableinteachingstudentstoargueandingivingthemreasonsforacquiringfactstobeusedinanargumentbut,likethehand-wavinggame,itisunrelatedtosubjectmatteranditgivesthestudentawrongimpressionofscientificthinking.Itistruethatscientistsoccasionallydiscussthingsamongthemselves,butthecreativeinterchangesaremorelikelytobebetweenmenandthingsthanbetweenmenandmen.TheGreatConversationwhichhasbeengoingonformorethan2000yearshasnotbeennotablyproductiveofusefulinformationorwisdom.Tosuggesttohighschool students that science is a kind of runningdebate is to risk selectingpotential

253

debatersratherthanpotentialscientists.Bothteacherandstudentcanbesimilarlymisledbypracticesdesignedprimarilyto

make science interesting. Studentswho take an interest in things are likely to learnsomethingaboutthem,andmakingasubjectinterestingisnodoubtworthwhile,butitisamistaketoconfusearousinginterestwithteaching.Inarecentreviewofabookonthemathematicscurriculum2 the reviewer insisted that remarkson thepsychologyofteaching should “confine themselves [my italics] to observing that mathematicsteaching(indeed,allteaching)mustmakethesubjectmatterattractive.”Andhowoftendowehear it said that the good teacher is simply onewhoknows his field and canmake it interesting!But teaching ismuchmore than arousing interest, andmaterialsandtechniquesdesignedtogenerateinterestmayconflictwithgoodteaching.

Attention

A student who is not paying attention is obviously not learning, and the teacher istherefore reinforced when he behaves in ways which attract attention. Audiovisualmaterials, texts with colored pictures and charts, animated films, and demonstrationexperiments full of surprises are often used for this reason. Advertisers and theentertainment industry face a similar problem and solve it in similar ways. But toattractattentionis todeprive thestudentof thechanceto learn topay attention.Theimportant thing is for the student to discover that interesting things happenwhenheattends to something which, on its face, is not interesting at all. We do not wantstudentswhoreadbooksonlywhentheyareprintedinfourcolors,orwhowatchfilmsor demonstrations only when something interesting is always happening. We wantstudentswhoreadblack-and-whitepagesbecausesomethinginterestinghappenswhentheydo,andwhowatchfilmsanddemonstrationswhichseemnomoreinterestingthannature itself,until closeobservationshowshowfascinating they reallyare.Materialsmiscarry in the samewaywhen they are designed to appeal to a student’s interestsoutside the classroom—thephysics of the tennis court, the chemistry of the kitchen.Faradaybecame interested inelectricitywhenhe readanarticle in theencyclopedia,anditwasnotentitled“ElectricityforyoungBritons.”Iamnot saying thata student shouldnotbe interested inwhathe isdoingor that

interestingaspectsofasubjectshouldnotbepointedout,butinrelyingtooheavilyontheattractionsofsciencewegivethestudentawrongimpressionofwhatheistofindwhenhepursuessciencefurther,andweshouldnotbesurprisedthathedropsoutwhenhediscoverstheactualstateofaffairs.Thethingswhichcommitthematurescientisttoalifetimeofdedicatedresearcharenotthekindsofthingswhichinterestthelaymanorthebeginningstudent. It ischaracteristicof thesuccessfulscientist, forexample, thathecontinuestoworkforlongperiodswhennothinginterestingishappening.Thatkindof dedication can be instilled in the student, as we shall see, but not by making asubjectinteresting.

Discovery

Another practice which has the effect of immediately rewarding the teacher eventhough the ultimate consequences are questionable is letting the student discover

254

scienceforhimself.ThiswasthegreatprinciplewhichRousseaudevelopedinhisbookEmile.Letthestudentlearnfromnature,notfromwhatothershavesaidaboutnature.Lethimgodirectlytothefacts,tothings,whichaloneareincorruptible.Theprincipleis supported by Pascal’s earlier observation that the arguments we discover forourselves are better understood and remembered than thoseweget fromothers.Theprinciple seems particularly appropriate in teaching science, where the greatachievementstaketheformofdiscoveries.Thescientistworksinordertodiscover,andhecontinuestoworksolongashehasachancetodiscover.Whyshouldthestudentnothavethesamemotivation?Wecannotmean,however,thatthestudentistodiscoverallofscienceforhimself,

orevenanyappreciablepartofit.Scienceisavastaccumulationofthediscoveriesofagreatmanymen.Itmustbetransmittedfromonegenerationtoanother—eitherintheform of books, charts, tables, and so on, or in the form of behavior taught to newmembersofaculture.Educationischargedwiththetransmissionofknowledgeinthesecond sense, and it cannot possibly fulfill its obligation simply by arranging forrediscovery.Whetherwelikeitornot,agreatdealofsciencemustbetaught.Weraisea serious obstacle to teaching whenwe suggest to the student that it is beneath hisdignity to learn what someone else already knows. How much of science is to betaught,howmuchissimplytobemadeavailableinrecordedform,andhowmuchistobe left for rediscovery are questions concerning the available time and energy ofteachersandstudents.Theanswersmust take intoaccount theefficiencyof teachingmethods.The problem is particularly difficult because scientific knowledge changes so

rapidly. Textbooks and other records go out-of-date, and so do the behavioralrepertoiresimpartedthroughinstruction,butwecannotsolvethatproblembyrefusingtowritebooksortoteach.Wemustbepreparedtochangeourbooksandtoteachinsuchawaythatthebehaviorofourstudentscanchangeasoccasiondemands.Itisnosolutiontothisproblemtoletthestudentdiscoverthingsforhimself,becausewhathediscoverswillalsosoonbeout-of-date.Of coursewewant to encourage students to inquire, explore, and discover things,

and we want to teach them to do so efficiently. We must teach a wide range ofscientificmethodsaswellasfacts.Manyoftheverbalpracticesofsciencehavebeencarefully formulated bymathematicians, logicians, statisticians, and others, and theyare usually part of a science curriculum. The nonverbal day-to-day behavior of thescientist in his laboratory has in contrast been sadly neglected, and it is here thattechniquesofdiscoveryaremorelikelytoberelevant.Wenodoubtneedtoknowmoreaboutthemifwearetoteachthemwell,butevensothereisnoreasonwhytheyshouldbetaughtbythediscoverymethod.Indeed, it is not likely that they are taught well by that method. The guided

discoveries of the classroom bear only a vague resemblance to genuine scientificdiscoveries.ThearchetypalpatternofthiskindofteachingisthesceneinPlato’sMenoinwhichSocrates leads the slave boy throughPythagoras’ theorem for doubling thesquare.Thisisstillhailedasagreateducationalinnovation,butthefactisthattheslaveboylearnednothing.TherewasnottheremotestchancethathecouldgothroughtheproofhimselfwhenSocrateshadfinishedwithhim,andevenifhecouldhavedoneso,his behavior in assenting to Socrates’ suggestions almost certainly had nothing incommonwiththestepswhichledPythagorastohisdiscoveryofthetheorem.Polya3

255

has published a delightful account of how one might tease out the formula for thediagonal of a parallelepiped from a class of high school students, but the hints,suggestions, corrections, and heuristic exhortations he uses do not give a veryconvincingpictureoftheconditionsunderwhichtheoriginaldiscoverymusthavebeenmade.A few students no doubt benefit from this kind of teaching in the hands of agoodteacher.Theyexperiencesomeofthedelightofmakingadiscovery,whichmaysustain them in further work. Even so, they are not necessarily thenmore likely tomakeotherdiscoveriesbythemselves,andmeanwhilealltheotherstudentsintheclasshavereceivedaparticularlyconfusingpresentation.Althoughthemomentofdiscoveryisimportantinthelifeofascientistandmayexplainhisdedication,itisnecessarilyarareeventandcannotexplainthequalityornatureofmostofhisbehavior.

AversiveControl

These, then, are a few examples of classroompracticeswhich flourish because theirimmediateeffectsarereinforcingtostudentsandteachersinspiteofthefactthatlong-termeffectsmaybeweak, lacking,oractuallyundesirable.Therearenodoubtotherreasonswhy thepractices flourish.Education is in transition. It is a transition in therightdirection,butithasalongwaytogo.Weareintheprocessofrejectingmethodswhich have long dominated the field, in which students study primarily to avoidpunishment and which impose upon the teacher the necessity of maintaining asustainedthreat.Adictatorial,despoticteacher—an“authority”inapoliticalaswellasascholarlysense—isoutofplaceinmodernlife.Wewantlearningtomeanmorethanpractice,drill,orrotememorizing,whicharethecommonestproductsofsuchasystem.It is not surprising, therefore, thatwe should turn first tomaking science attractive,engagingthestudentindiscussion,givinghimmaterialswhicharousehisinterest,andlettinghimdiscoverthingsforhimself.Butasenjoyableasthesepracticesmaybe—forteacher and student alike—the fact remains that they are not really effectivealternatives.Theproofisthattheteacherisforcedbackagainandagainupontheoldcoercivepattern.Inspiteofallourefforts, it isstill truethatstudentslearnmainlytoavoidtheconsequencesofnotlearning.Thecommonestpracticeinhighschoolaswellascollegeisstill“assignandtest.”Wetellthestudentwhatheistolearnandholdhimresponsible for learning it bymaking a variety of unhappy consequences contingentuponhis failure. In doing sowemaygive him some reason to learn, butwe do notteach.Our failure is clear in the frequencywithwhicheducators conclude that a teacher

cannotreallyteachbutcanonlyhelpthestudentlearn.Thisisadisastrousphilosophy.It canbe asserted, of course, onlyofmethodswhichhave actually been tried, but ittendstobeusedasanargumentagainsttryingnewones.Itisnotonlyaconfessionoffailure but a form of exculpation. By admitting that we cannot teach, we avoidconfessingthatwehavefailedtodoso,andwethuscontinuetomaintain,asteachershavemaintainedforcenturies, that it isalways thestudentwhofails,not the teacher.Wecandiscardcoercivepracticesonlywhenwehavefoundsatisfactoryreplacements,andthepresentstateofeducationisproofthatwehavenotyetbeensuccessful.

WhatDoesTeachingMean?

256

Animportantfirststepinsearchingforbetterwaysofteachingistodefineourterms.What is happening when a student learns? Traditional theories of education almostalwaysanswer thatquestion inmentalisticways.The student is said tobeginwith adesire to learn, a natural curiosity, of which the teacher must take advantage. Theteachermustexercisethestudent’sfaculties,strengthenhisreasoningpowers,develophiscognitivestylesandskills,lethimdiscoverstrategiesofinquiry.Thestudentmustacquire concepts, come to see relations, and have ideas. He must take in and storeinformationinsuchaformthatitcanbequicklyretrievedwhenneeded.Statementsofeducationalpolicy are repletewith expressionsof this sort. Itwouldbe amistake tounderestimate their power, for they are supported by ancient systems of psychologyimbedded in our language and by vestigial cognitive theories. It is therefore hard torealizethattheyareeithermetaphorswhichinadequatelyrepresentthechangestakingplace in the student’s behavior or explanatory fictionswhich really explain nothing.Theirmostseriousshortcomingisthattheydonottelltheteacherwhattodoinordertobring about changes in his students or give him any satisfactory way of knowingwhetherhehasdoneso.Iftheseareindeedthetasksoftheteacher,wemustagreethathecannotreallyteach.Itisevendoubtfulwhetherhecanhelpthestudentlearn.Amuchmorepromisingapproachistolookatthestudent’sbehavior—thebehavior

from which mentalistic states and processes are inferred and which they soinadequatelydescribeandexplain.Thebasicquestion,initscrudestform,isthis:whatdowewantthestudenttodoastheresultofhavingbeentaught? (It isnoanswertocite theexaminationsheis topass,for theyareonlysamplesofhisbehavior,andnomatterhowreliabletheymaybe,theyare,wehope,verysmallsamplesindeedofwhathewillactuallylearn.)Tosaythatwewantthestudentto“behavelikeascientist”isontherighttrack,butitisonlyastart.Forhowdoesascientistbehave?Theanswerwillbenothinglessthananepistemology,atheoryofscientificknowledge.Itmustinfactbemore:weneedanempiricaldescriptionofthebehaviorofthescientistatwork,inallitsmyriadforms.Such a description is not to be had for the asking. Scientific thinking is an

extraordinarily difficult field, and we have not advanced very far in analyzing it,possiblyjustbecausewehavesooftenbeenseducedbymetaphor.Ifweannouncethatweareinterestedingivingthestudentathoroughknowledgeofascience,agraspofitsstructure, an understanding of its basic relations, we shall be endlessly admired. If,instead,we specify the thingswewant him todo,verballyandnonverbally,we riskbeingcalledmechanicalandshallow,eventhoughthethingswelistarepreciselythethingsfromwhichanunderstandingorgraspofthestructureofthescienceisinferred.Thereisnothingaboutbehaviorwhichevokesthemysterywhichhasalwaysattachedtomind,butitisimportanttorememberthatwestandinaweofmindjustbecausewehavebeenabletodosolittleaboutit.

ProgrammedInstruction

Toremovethemystery,wemustdefineourgoalsinthemostexplicitway.Andwecanthenbegintoteach.Havingspecifiedtheterminalbehaviorourstudentsaretoexhibit,we can proceed to generate it. One way is through programmed instruction, acontribution to education which has been widely misunderstood. Many educationaltheoristshave insisted that it isnothingnewandhave tried to assimilate it to earlier

257

theoriesandpractices.Wearetoldthatitissimplyamatterofbreakingthematerialtobelearnedintoeasysteps,arrangingstepsinalogicalorderwithnogaps,makingsurethestudentunderstandsonestepbeforemovingon toanother,and thus, incidentally,makingsurethatheisfrequentlysuccessful.Allthesethingsaredoneinconstructingagoodprogram,butthecentralpointhasstillnotbeenreached.Programmed instruction is primarily a way of using recent advances in our

understandingofhumanbehavior.Wewanttostrengthencertainkindsofbehaviorinour students and so far aswe know, there is only onewayof doing so.Behavior isstrengthenedwhenitisfollowedbycertainkindsofconsequences.Tobemoreprecise,a responsewhichproduces a so-calledpositive reinforceror terminates anegative ismore likely to occur again under similar circumstances. We use this principle of“operantconditioning” to strengthenbehaviorbyarranging reinforcingconsequences—by making available reinforcers contingent on behavior. This is often said to benothingmorethanrewardandpunishment,andthereiscertainlyaconnection.Butthetraditional concepts of reward and punishment are about as close to operantconditioning as traditional concepts of heat, space, or matter are to contemporaryscientific treatments. Only a detailed experimental analysis of contingencies ofreinforcementwillsupplytheprinciplesweneedinthedesignofeffectiveinstructionalpractices.Teaching is the arrangement of contingencies of reinforcement which expedite

learning. Learning occurs without teaching, fortunately, but improved contingenciesspeed the process and may even generate behavior which would otherwise neverappear. Programmed instruction is designed to solve a special problem.We cannotsimplywaitforourstudenttobehaveinagivenway,particularlyinthecomplexwayscharacteristicofascientist,inordertoreinforcehim.Somehoworotherwemustgethimtobehave.Ourculturehasdevisedrelevanttechniquesforotherthaneducationalpurposes.Weresorttoverbalinstruction,forexample,whenwesimplytellthestudentwhattodo,orweshowhimwhattodoandlethimimitateus.Ifweinducethestudenttoengage in terminalbehavior in thatway,however,hewillbemuch toodependentuponbeingshownortold.Hewillnothavelearned.Webegininsteadwithwhateverbehaviorthestudenthasavailable—withbehaviorwhichdoesnotcallformuchhelp.Weselectivelyreinforceanypartwhichcontributestotheterminalpatternormakesitmore likely that the student will behave in other ways which contribute to it. Thedevicesweusetoevokethebehaviorcanthenbeeasilywithdrawn,sothattheterminalbehaviorappearsuponappropriateoccasionswithouthelp.Ahighdegreeoftechnicalknowledgeisneededtodothis.Manyinstructionalprogramshavebeenwrittenbythosewhodonotunderstandthe

basicprinciple,anditisanunhappyreflectiononthestateofeducationtodaythattheyare still probably better than unprogrammed materials, but they give a wrongimpression. Even a good program may be misleading to anyone who is alreadyproficient in a field because he cannot easily appreciate its effect on a new learner.Anyonewhowantstogetthefeelofprogrammedinstructionshouldtryhishandatagoodprograminanunfamiliarsubject.Acolleaguewhoseworkhadbeguntomoveinthedirectionofbiochemistryworkedthroughanexcellentprograminthatfield.“In3days,”hetoldme,“Iknewbiochemistry!”Hewasexaggerating,ofcourse,aswebothknew,buthewasexpressingverywellthealmostmiraculouseffectofagoodprogram.A furthermisunderstanding has arisen from the fact that industry and theArmed

Serviceshave takenupprogrammed instructionmuchmore rapidly thanschoolsand

258

colleges.Therearesomeobviousreasons.Foronething,teachingtechniquesintheseorganizationscanbeeasilychanged.Foranother,therearepeopleinindustryandtheArmed Services whose job it is to see that no possible improvement in teaching isoverlooked. Unfortunately they have no counterparts in school and collegeadministrations.Explanationsofthissorthavenotpreventedtheerroneousconclusionthat there is another reasonwhy instruction is particularly suited to industry and theServices.Instructionthereissaidtobeofaspecialnature,amatterof trainingratherthan teaching. This is a very dubious distinction. Training once meant nonverbalinstruction, usually through the use of training devices, but that is no longer true.Industry and the Services teach many of the things taught in schools and colleges,although the terminalbehavioradmittedlycomes in smallerpackages.The importantthingisthatitcanbemoreeasilyspecified.Thetraditionaldistinctioncomesdowntothis:whenweknowwhatwearedoing,wearetraining;whenwedonotknowwhatwearedoing,weareteaching.Oncewehavetakentheimportantfirststepandspecifiedwhatwewant the student todoas the resultofhavingbeen taught,wecanbegin toteachinwayswithrespecttowhichthisoutworndistinctionismeaningless.In doing soweneednot abandon anyof our goals.Wemust simplydefine them.

Anybehaviorwhich canbe specified canbeprogrammed.Anexperimental analysishasmuchmoretoofferinthisdirectionthanisgenerallyrealized.Itisfarfromacrudestimulus-responsetheoryandisnotcommittedtorotememorizingortheimpartingofmonolithic,unchangingtruth.Ithasasmuchtosayaboutsolvingproblems,inductiveordeductive reasoning,andcreative insightasabout learningfacts.Wehaveonly todefinethesetermsandatechnologyofteachingbecomesapplicable.Specification,ofcourse,isonlythefirststep.Goodprogramsmustbeconstructed.Atthemomentonlya few people have the necessary competence, but this is one of the points atwhicheducational reform should start. Scientists, as subjectmatter specialists,must play amajorrole.

ClassroomManagement

Another important application is in classroom management. The teacher whounderstandsreinforcementandisawareofthereinforcingeffectsofhisownbehaviorcancontrolhisclass.Thosewhoareinterestedintheintellectualsideofeducationhavetended to neglect classroom discipline, but at great cost. Much of the time of bothstudentandteacherisnowspentinwayswhichcontributelittletoeducation.Studentswhoareparticularlyhardtomanageareoftenineffectabandoned,althoughthereareprobablygeniusesamongthem.Itisherethatthetransitionfromolderaversivepracticesismostconspicuous.Many

educationalreformers—AdmiralRickoveramongthem,forexample—lookwithenvyon the disciplined classroom of European schools. It appears to be a backgroundagainstwhichthestudentuseshistimemostprofitably.Butpunitivetechniqueshaveobjectionable by-products, and we are led to explore the possibility of creating anequallyfavorablebackgroundinotherways.Specialskillsonthepartoftheteacherareneeded, not only in maintaining discipline but in teaching the kinds of nonverbalbehaviorwhichfiguresoprominentlyinsuchfieldsaslaboratoryexperimentation.Itisaparticularlydifficultproblembecausewemustcompetewithothercontingenciesinthe student’sdaily life involving sex, aggression, competitive sports, and soon.Too

259

oftenthegoodstudentissimplyonewhoisunsuccessfulinotherways.Herespondstoour instructional contingencies only because he has not come under the control ofothers.Theresult,ofcourse,ispoorselection.Weneedtorecruitscientistsfromthosewho could be successful in any walk of life. To do so wemust take the design ofclassroombehaviorseriously.Effective instructionalcontingencies in theclassroomaremoredifficult toarrange

thanthoseinprogrammedinstruction.Curiouslyenough,thenatureoftheenterpriseisclearest with respect to a more difficult kind of student. Institutions for the care ofautisticorretardedchildrenandtrainingschoolsforjuveniledelinquentshavebeguntomake effective use of operant conditioning. Because of either their heredity or theirearly environments, certain people do not respond well to normal contingencies ofreinforcement. A special environment must be constructed. Ogden R. Lindsley hascalled itaprostheticenvironment.Eyeglassesandhearingaidsareprostheticdeviceswhich compensate for defective sense organs, as crutches and artificial limbscompensate for defective organs of response.A prosthetic environment compensatesfor adefective sensitivity to contingenciesof reinforcement. In such an environmentreinforcers may be clarified; many institutions reinforce students with tokens,exchangeable for other reinforcers such as sweets or privileges,which can bemadecontingent on behavior in conspicuous ways. Many of these defective people willalwaysrequireaprostheticenvironment,butotherscanbebroughtunderthecontrolofthereinforcersindailylife,suchaspersonalapprovalorthesuccessfulmanipulationofthephysicalenvironment,andcanthusbepreparedforlifeoutsideaninstitution.Contrived reinforcers intended to have a similar effect are by no means new in

education.Marks, grades, diplomas, honors, and prizes, not tomention the teacher’spersonalapproval,areseldomthenaturalconsequencesofthestudent’sbehavior.Theyare used on the assumption that natural consequenceswill not induce the student tolearn.Severalobjectionsmaybeleveledagainstthem.Inthefirstplace,asconditionedreinforcerstheyarelikelytolosetheirpower.Thisiseventrueofpersonalreinforcersif theyarenotgenuine.Whenour telephonesays tous,“I’msorry.Thenumberyouhavereachedisnotinserviceatthistime,”wemayrespondatfirsttothe“I’msorry”asifitwerespoken,say,byafriend.Eventually,wemaystoptoask,“Whoissorry?”andlookforwardtothedaywhenmachineswillbepermittedtobehavelikemachines.The computers used in computer-aided instruction are particularly likely to “getpersonal”inthisway.Theycalltheyoungstudentbynameandtypeoutexclamationsofdelightathisprogress.Butthenaturalconsequenceswhichmadetheseexpressionsreinforcing in the firstplacearenot forthcoming,and theeffectsextinguish.What isnot so obvious is that personal approval may be equally spurious. George BernardShawisresponsibleforaprinciplewhichmaybestatedinthisway:neverstrikeachildexcept in anger.Acomplementaryprinciple in the classroom is this: never admire astudentexceptwhenheisbehavingadmirably.Contrivedadmirationisself-defeating.Buttheobjectiontogrades,prizes,andsyntheticpersonalapprovalisnotthatthey

are contrived, but that the contingencies in which they are used are bad. Anexperimentalanalysisismostvaluableatjustthispoint.Tobringaclassundercontrol,the teacher must begin by making available reinforcers explicitly contingent on thedesired behavior. Some students may need reinforcers as conspicuous as tokens orpointsexchangeableforgoodsorprivileges.Moneyisatokenreinforcerwhichshouldnot be ruled out of account. (It could solve the high school dropout problem if thecontingencies were right.) But once a classroom has been brought under control, a

260

teachermustmovetomoresubtlecontingenciesandeventuallytothoseinherentintheeverydayphysicalandsocialenvironmentofthestudent.Techniques of reinforcement are now available which can replace the aversive

techniques which have dominated education for thousands of years. We can havestudentswhopayattentionnotbecausetheyareafraidoftheconsequencesiftheydonot, or because they are attracted by fascinating if often meretricious features, butbecausepayingattentionhasprovedtobeworthwhile.Wecanhavestudentswhoareinterested in their work not because work has been chosen which is interesting orbecause its relation to interesting things has been stressed, but because the complexbehavior we call taking an interest has been abundantly reinforced. We can havestudentswholearnnotbecausetheywillbepunishedfornotlearning,butbecausetheyhavebeguntofeelthenaturaladvantagesofknowledgeoverignorance.Wecanhavestudentswhowill continue tobehaveeffectivelyafter instructionhasceasedbecausethecontingencieswhichhavebeenusedbytheirteachersfindcounterpartsindailylife.Above all, we can have dedicated students who will become dedicated men and

women.Many interesting aspects of human behavior, often attributed to somethingcalled motivation, are the results of various schedules of reinforcement4 to whichalmost no attention has been given in educational theory. A common criticism ofprogrammedinstruction,forexample,isthatfrequentreinforcementleavesthestudentunpreparedforaworldinwhichreinforcersmaybescarce,andthiswouldbetrueifthepossibilitywereneglected.Butprogrammingtechniquesareavailablewhichpermitustosustainthebehaviorofthestudentevenwhenreinforcersareveryrareindeed.Oneofthemostpowerfulschedules, theso-calledvariable-ratioschedule, ischaracteristicof all gambling systems. The gambler cannot be sure the next play will win, but acertainmeanratioofplaystowinsismaintainedbythesystem.Ahighratiowillnottake control if it is encounteredwithout preparation, because any available behaviorwill extinguish during a long run, but a low ratio will be effective and can be“stretched”as thebehaviorbuildsup.This is thewayadishonestgamblerhookshisvictim.Atfirstthevictimispermittedtowinfairlyoften,butastheprobabilitythathewillcontinuetoplayincreases, theratioisincreased.Eventuallyhecontinuestoplaywhen he is not winning at all. The power of the schedule is most obvious when itproduces a pathological gambler, but pigeons, rats, monkeys, and other lowlyorganismshavebecomepathologicalgamblersonthesameschedule.And so have scientists. The prospector, the explorer, the investigator, the

experimenter—all meet with success on a variable-ratio schedule. The dedicatedscientistcontinuestoworkeventhoughtheratioofresponsestoreinforcementisveryhigh,buthewouldnothavebecomeadedicatedscientistifhehadstartedatthatratio.Itwouldnotbecorrecttosaythatwecanalwaysarrangeaprogramwhichstartswithfrequentsuccessesandleadsinevitablytoahighratio,butatleastweknowthekindofschedule needed. In any case, the extraordinary effects of scheduled reinforcementsshouldnotbeoverlooked.Indesigningalaboratorycourse,forexample,ifwekeepaneyeonthestudent’ssuccessesandparticularlyonthewayinwhichtheyarespaced,wearemorelikelytoproduceastudentwhonotonlyknowshowtoconductexperimentsbutshowsanuncontrollableenthusiasmfordoingso.The newmaterials which have been made available for teaching science in high

school are genuinely exciting, but the fact remains that classroom practice has notreallychangedverymuch.Theforceswhichmakepracticestraditionalmakethemeasy

261

totransmittonewteachers.Therelationsbetweenstudentandteachersdemandedbysuch practices arouse no anxiety. The practices can be justified to parents, policy-makers, supportersofeducation,andstudents themselves.Theycall fornoextensivechanges in administration. And of course they have their occasional successes—particularlywith good students or in the hands of good teachers.All this favors thestatusquo.Thechangewhichisneededmustovercomemanyhandicaps.Muchmoreisknown

about thebasicprocessesof learningand teaching than isgenerally realized,butweneedtoknowstillmore.What isknownhasnotyetbeenput touseveryeffectively.Thedesignand constructionofmethods andmaterials is a difficult enterprisewhichdemands a kind of specialistwho is, at themoment, in short supply.New practicesneed to be thoroughly tested. And when, at last, we have devised more effectivemethods, we must convince educators that they should be used. Extensiveadministrative changes must be made. (The changes required simply to permit theindividual student to progress at his own rate are prodigious.) Teachers need to beretrained as skillful behavioral engineers.The commoncomplaint that newmaterialsdonotworkbecausetheteachersareincompetentisnotonlyunfair,itshowsafailureto recognize another point at which the improvement of teaching might begin.Materialsaregoodonlyif theycanbeusedbyavailableteachers.It isquitepossiblethatmaterialscanbedesignedwhichwillpermitteacherstoteachwelleveninfieldsinwhichtheyhavenospecialcompetence.

TheImprovementofTeaching

Scientists are wary of being asked about their “values.” They hesitate to speak ofprogressbecausetheyarelikelytobeasked,“Progresstowardwhat?”Theyareuneasyinsuggestingimprovements.“Improvementsinwhatsense?”Thecurrentfashionistospeakonlyofeducationalinnovation.Allthatisclaimedforanewpracticeisthatitisnew.We need amuchmore positive attitude. The efficiency of current methods ofteaching is deplorably low. The change which occurs in a student as the result ofspending one day in high school is discouragingly small. We need to improveeducationinthesimplesenseofmakingitpossibletoteachmoreinthesametimeandwith thesameefforton thepartof teacherandstudent. It isadifficultassignment—possiblyasdifficult,say,asthecontrolofpopulationorresolvingthethreatofnuclearwar,butthereisnomoreimportantproblemfacingAmericatodaybecauseitssolutionwilladvanceallothersolutions.It is the sortofchallenge that scientistsareaccustomed toaccept.They,aboveall

others,shouldappreciatetheneedtodefineobjectives—toknow,inthisinstance,whatitmeanstoteachscience.Theyshouldbequicktorecognizetheweaknessesofcasualexperienceandoffolkwisdombasedonthatexperience.They,aboveallothers,shouldknow that no enterprise can improve itself to any great extentwithout analyzing itsbasic processes. They should be best able to gage the importance of science in theimmediateanddistantfutureandthereforetheextentofthedisasterwhichwillfollowifwefailtorecruitforsciencelargenumbersofourmostintelligentanddedicatedmenandwomen.Itisnotimeforhalf-heartedmeasures.Theimprovementofteachingcallsforthemostpowerfulmethodswhichsciencehastooffer.

262

FromScience,1968,159,704–710.1InTheTechnologyofTeaching.2Hilton,P.J.Science,147(1965),138.3Polya,G.HowtoSolveIt.Princeton:PrincetonUniversityPress,1945.4See,forexample,pages132–134.

263

ContingencyManagementintheClassroom

Whydostudentsgotoschool?Whydotheybehavethemselvesinclass?Whydotheystudy and learn and remember? These are important questions, but they are seldomasked—possiblybecausewearenotproudoftheanswers.Whetherwelikeitornot,moststudentsstillcometoschool,behavethemselves,andstudyinordertoavoidtheconsequences of not doing so. True, most teachers have abandoned the birch rod(though its return is called for in some quarters), but there aremany ingenious, lessviolent replacements. Violent or not, punitive methods have serious consequences,among them truancy, apathy, resentment, vandalism, and ultimately an anti-intellectualismwhich includes an unwillingness to support education. These are thegreatproblemsoftheeducationalestablishment,andtheycanbetracedinlargeparttothetechniquesoftheestablishmentitself.Few teachers are happy about punitive methods (most of them would like to be

friends with their students), but alternatives have seldom proved fruitful. Simply toabandonpunishmentandallowstudentstodoastheypleaseistoabandonthegoalsofeducation.A“freeschool”wasrecentlydescribedinanewspaperarticleasfollows:

ThemiddleschoolclassroomIsawwasfullofchildrenworkinginanendlessvarietyofsubjects, the life cycle of the beetle, action painting, physical properties of water,mathematics (by choice), making dressing-up clothes, writing poetry. Some of themwanderedupandstartedaconversation.Theywereconfidentandarticulate.Iwasaskedtojoin various games, give an honest opinion on a painting, listen to poetry. Ten year oldMichaeliswritingpoetrynearlyallthetimenow….Anotherchildiscoaxingawoodwormoutofapieceofrottingwood.

Itisnodoubtanattractivepicture—untilwestarttothinkaboutwhataschoolisfor.Menhavebeendreamingof thepermissiveorfreeschoolforat least twohundred

years.The idea first appeared incloseassociationwith the ideaofpolitical freedom,andoneman—JeanJacquesRousseau—waslargelyresponsibleforboth.HehasbeencreditedwithinspiringnotonlytheFrenchRevolutionbut,inhisgreatworkÉmile, arevolution of perhaps comparable magnitude in education. He was interested, quitejustly, inabolishing thepunitivemethodsofhis time,andsowere thediscipleswhofollowedhim—Pestalozzi,Froebelandhiskindergarten,Montessori,JohnDewey,and(adabsurdum)NeillwithhisSummerhill.With Rousseau the proposal was clearly a dream, for Émile was an imaginary

studentwith, aswe now know, imaginary learning processes.When Pestalozzi triedRousseau’s principles on his own child, he came to grief. And, sooner or later, thedreamisalmostalwaysfollowedbyarudeawakening.Secondaryschoolsarefoundedbywell-intentioned peoplewhowant their students to be free, but the schools growsteadilymore disciplined as the exigencies of teachingmake themselves felt.Whenprospectiveparentsbegintoask“Howmanyofyourstudentsgoontocollege?”and“Whatcollegesdotheygoto?”thegoalofthefreestudentisabandoned.Coursesshow

264

the samepattern.Language instruction begins painlesslywith the directmethod, butsoonerorlaterthestudentwillbefoundmemorizingvocabularylistsandgrammaticalparadigms.Andoneof the freedomsenjoyedby the students inSummerhillwas thefreedomtotreattheirfellowspunitively.Occasionally thedreamcomes true. In anygeneration there are a fewoutstanding

teachers, just as there are a few outstanding artists, writers, executives, andpersonalities in films and television. There are also many exceptional students—studentswhoscarcelyneedtobetaughtatall.Anoutstandingteacherandafewgoodstudentscomposeapicturethatweshouldallliketocopy,butitisnotamodelfortheteachingofordinarystudentsbyordinaryteachers.Nor can we replace punishment simply by telling our students about long-term

advantages.Wemakeagreatdealofthe“dollarvalue”ofaneducation(convenientlyoverlookingthefactthattruckdriversandcarpentersmakeasmuchasmostteachers),but the ultimate consequences of an education are too remote to have any importanteffect on the student as he reads a textbook or listens to a lecture. The gold stars,marks, grades, honors, promotion, and prizes which we think of as alternatives topunitivesanctionsalsolackanecessaryimmediacy.Norcanwesolvetheproblembybringingreallifeintotheclassroomsothatstudentswillcomeintocontactwiththingswhich are naturally rewarding, for we cannot find interesting things relevant toeverythingwewanttoteach.“Reallife”philosophiesofeducationhavealsomeanttheabandonmentofimportantgoals.All thesemeasures fail because they do not give the student adequate reasons for

studyingandlearning.Punishmentgavehimareason(wecansaythatforit),butifweare to avoid unwanted by-products, we must find nonpunitive forms. It is not animpossible assignment. The “reasons” why men behave are to be found among theconsequencesoftheirbehavior—what, toputitroughly,they“getoutofbehavingingiven ways.” And these have been carefully studied. Behavior which acts upon theenvironmenttoproduceconsequences—“operant”behavior—hasbeenexperimentallyanalyzedingreatdetail.Certainkindsofconsequencescalledreinforcers(amongthemthe things the layman calls rewards) aremade contingent uponwhat an organism isdoinganduponthecircumstancesunderwhichitisdoingit.Changesinbehaviorarethenobserved.Thecontingencies, rather than the reinforcers,are the important things. Ithas long

been obvious that men act to achieve pleasure and avoid pain (at least most of thetime),butthefacttobeemphasizediswhattheyaredoingatthemomenttheyachievethese results. Special equipment is used to arrange so-called “contingencies ofreinforcement”(andifteachingcanbedefinedastheexpeditingoflearning,thenthisequipment is a kind of teaching machine). The complexity of the equipment to befound in hundreds of laboratories throughout theworld is not a bad indicator of thecomplexityofthecontingenciesnowunderinvestigation.Fewpeopleoutsidethefieldareawareofhowfartheanalysishasgone.Asmoreandmorecomplexcontingencieshave been arranged, it has been possible to studymore andmore complex kinds ofbehavior,includingbehavioronceattributedtohighermentalprocesses.Anapplicationtoeducationwasinevitable,butithasnotbeenunopposed.Thefact

thatmuchof theearlywork involved thebehaviorof loweranimalssuchas ratsandpigeons has often been held against it. But man is an animal, although anextraordinarily complex one, and sharesmany basic behavioral processeswith otherspecies. Human behavior must nevertheless be studied in its own right, and human

265

subjectsare in factnowcommonlyused inexperimentalanalyses.Whencomparablecontingenciesofreinforcementcanbearranged,theyyieldcomparableresults;butthecontingencies towhich the human organism can adjust are extraordinarily complex.Efforts currently under analysis have the subtlety, variety, and intricacy whichcharacterizehumanbehaviorintheworldatlarge.Thatthemethodsofanexperimentalanalysisofoperantbehaviorareappropriateto

humansubjectsisconfirmedbythesuccesswithwhichtheyhavebeenputtoworkinpractical ways. Psychotherapy, for example, has undergone an important change. Arecent book by Ayllon and Azrin, The Token Economy, shows how a hospital forpsychotics canbe converted into a community inwhichpatients care for themselvesand their possessions, avoid trouble with their associates, and (within the limitsimposed by their illness) enjoy life. Such an arrangement of contingencies ofreinforcement has been called a “prosthetic” environment. Like eye glasses, hearingaids,andartificiallimbs,itpermitspeopletobehavesuccessfullyinspiteofdefects.Inthepsychoticthedefectisoftenaninsensitivitytocontingenciesofreinforcements.The principles of operant conditioning were first applied to education in

programmed instruction. The step-by-step shaping of complex behavior was firstdemonstratedinanexperimentalanalysis,andthetechniqueisprobablystillbestseeninexperimentswithanimals.Ahungrypigeon, forexample, canbe induced throughreinforcement with food to respond in specified ways. Quite complex forms ofbehavior can be generated, often with surprising speed, through a series of stagesleadingtotheterminalspecifications.Oneactually“seeslearningtakeplace,”andthevisibility is important. When a teacher can bring about conspicuous changes inbehavior,changeswhichdonotneedtobeconfirmedbyastatisticaltreatmentoftestscores,heknowsimmediatelywhathehasdone,andheisthenmostlikelytolearntoteach effectively. Traditional research in learning has seldom been very useful ineducation,and inpartbecause ithasneglected theprocessofshaping.Subjectshavebeenplungedintoterminalcontingenciesandlefttostruggletowardadequateformsofbehavior through “trial and error.” (Although shaping is important, it is not alwaysnecessary. There are effective ways of evoking complex behavior so that it can bedirectly reinforced, and there is often a great gain in efficiency.Relevant techniquescanalsobeattributedtotheexperimentalanalysisofbehavior.)Programmed instruction has been largely responsible for the current emphasis on

behavioralspecifications.Aprogramcanbewrittenonlywhencertainbasicquestionshavebeenanswered.Whatisthestudenttodoastheresultofhavingbeentaught?Tosaythataprogramisto“impartknowledge,”“trainrationalpowers,”or“makestudentscreative” is not to identify the changes which are actually to be brought about.Somethingmorespecificisneededtodesigneffectiveprogrammedcontingencies(asitisneededinordertoteachwellintheclassroom).Wedonotteachtheskillsstudentsaresaidtodisplaywhentheybehaveskillfully,weteachskillfulbehavior.Wedonotimpart knowledge,we generate behavior said to show the possession of knowledge.Wedonotimproveabilitiesorstrengthenrationalpowers;wemakeitmorelikelythatthestudentwillshowthebehaviorfromwhichabilitiesandpowersareinferred.Whengoalsareproperlyspecified,theteacherknowswhatheistodoand,later,whetherhehas done it.Behavioral objectives removemuchof themystery from education, andteachersmay feeldemeanedwhen their task is reduced to lessawesomedimensions.Butthelossismorethanoffsetbyagreatersenseofachievement.Many early programs were constructed by writers who missed some of the

266

implications of the basic analysis. They were encouraged to do so by educationalphilosopherswho tried to assimilate programming to traditional theories of learning.Programmingwassaidtobesimplyamatterofproceedinginsmallsteps,ofaskingthestudenttomasteronestepbeforemovingontothenext,ofarrangingstepsinalogicalsequencewithnogaps, and soon.Thiswas true enough, andprogramsdesignedonthese principles were better than no programs at all, but other points need to beconsidered.Animportantexamplehastodowith“motivation.”Studies of operant reinforcement differ from earlier studies of learning by

emphasizingthemaintenanceaswellastheacquisitionofbehavior.Acquisitionistheconspicuouschangebroughtaboutbyreinforcement,butthemaintenanceofbehaviorinagivenstateofstrengthisanequallyimportanteffect.Agoodprogramreinforcesthe student abundantly and at just the right times. It shapes new forms of behaviorunderthecontrolofappropriatestimuli,buttheimportantthingisthatitmaintainsthestudent’sbehavior.Itholdshisattention;itkeepshimatwork.Traditional studies of learninghave paid little attention towhy the student learns,

andthishasencouragedthebeliefthatmenhaveanaturalcuriosityorloveoflearning,orthattheynaturallywanttolearn.Wedonotsaythataboutapigeon;wesayonlythatundertheconditionswehavearranged,apigeonlearns.Weshouldsaythesamethingabout human students. Given the right conditionsmenwill learn—not because theywant to, but because, as the result of the genetic endowment of the species,contingenciesbringaboutchangesinbehavior.Oneofthemaindifferencesbetweenatextbookandaprogramisthatatextbookteachesonlywhenstudentshavebeengivensome extraneous reason for studying it. A program contains its own reasons.Fortunatelyforusall,thehumanorganismisreinforcedbymanythings.Successisoneof them.A baby shakes a rattle because the production of noise is reinforcing, andadults put jigsawpuzzles together andwork crossword puzzles for nomore obviousreasonthanthattheycomeoutright.Inagoodprogramthestudentmakesthingscomeoutright;hemakesthingswork;hebringsorderoutofchaos.Agoodprogramhelpshimdo so. Itmakes right responses highly probable—just short of telling himwhatthey are.Again themotivational issuemaybemissed.Manypeople resistmaking astudent’staskeasy,andthebeginningprogrammermayfindhimselfunwillingto“givea response away.”As a teacher hehas felt the need to keep students under aversivecontrol,andhemaynotyetbefullyawareofhispowertocontroltheminotherways.Aprogramisalsoreinforcingbecauseitclarifiesprogress.Ithasadefinitesize.The

studentknowswhenheishalf-waythroughandwhenhehasfinished.Becauseofallthis a good program pulls the student forward.Hemay feel exhaustedwhen he hasfinished,buthedoesnotneedtoforcehimselftowork.Thereisanotherproblemineducationwhichoperantreinforcementhelpstosolve.In

primaryandsecondaryschools(andtosomeextentatotherlevels)ateachernotonlyteaches, he has custody of his students for an appreciable part of the day. Theirbehavior in the classroom, quite apart from what they are learning, is part of hisassignment. Coming to class, behaving well toward other students, attending to theteacher,enteringintodiscussions,studying—theseareasessentialtoeducationaswhatis being learned, and here the teacher plays a different role. He is not a source ofknowledgeoranevaluatorofwhatastudentknows;heisinasensethegovernorofacommunity.It should be a community in which learning takes place expeditiously, and the

teachercanmeetthatassignmentifheknowshowtousereinforcement.Buthemust

267

first answeran importantquestion:what reinforcersareavailable?Toput it roughly,whatdoeshepossessthathisstudentswant?Itisoftenanembarrassingquestion,butalmostneverwhollyunanswerable.Thebuilt-in reinforcersofprogrammedmaterialswillnotsuffice,butotherthingsareavailable.Thephysicalaspectofaschoolmayormaynotbereinforcing,andthiswillhavea

bearing on what happens when a student turns a corner and comes in sight of theschool.Ifthebuildingisnotattractive,hewillbelesslikelytoturnthatcorneragainandmaygoinsomeotherdirection.Theappearanceofabuildingisusuallybeyondtheteacher’s control, but reinforcing features of a classroom may not be. Businessenterprises understand the principle. A well-run store smells good; it is tastefullydecoratedandpleasantlylighted;theremaybemusicinthebackground.Thebehaviorofentering thestore is thereforereinforced,andcustomersaremore likely toenter itagain.To“reduceabsenteeism”theteachershouldtakesimilarstepstomakesurethathisstudentsarereinforcedwhentheyenterhisclassroom.Whatgoesonintheroomisalsorelevant.Theaversivetechniquesofthebirchrod

orcanearenotlikelytoreinforcecomingtoschool,andstudentssotreatedarelikelytoplaytruantorbecomedrop-outswhentheycanlegallydoso.Socialcontingenciesareimportant.Achildismorelikelytocometoschoolifhegetsalongwithhispeersandhis teacher; he is not likely to come if he is frequently criticized, attacked, orostracized.Unfortunately, social contingencies are often hard to arrange. To induce the

members of a classroom community to behave well with respect to each other,additional reinforcersmaybeneeded.The teachermayhavesomecontroloverwhatfoodchildreneatatlunchtime,whatsuppliestheyarepermittedtouse,whatprivilegesthey can enjoy (such as access to play areas),whom theymay associatewith,whenthey may turn to preferred activities, and what field trips they may take. Personalcommendation is often a powerful reinforcer, but a merely synthetic approval oraffectionhasitsdangers.Themainproblemis tomakethesereinforcerscontingentonthedesiredbehavior.

They are often not available on the spur of the moment. The teacher cannotconvenientlyreinforceachildwhenhesitsquietlybysendinghimoffonafieldtrip,orwhenhestopsfightingbyhandinghimanicecreamcone.A“generalizedreinforcer”isneeded—somethingwhich is exchangeable for reinforcing things.Money shows thearchetypalpattern.Wepaypeopleeventhoughatthetimetheyreceiveourmoneytheyarenothungryforthefoodtheywillbuywithitorinthemoodforthefilmtheywilluseittogotosee.Creditpointsortokenscanbeusedasmoneyintheclassroom.Theyarerelativelyindependentofthedeprivationswhichmakethemreinforcingandofthecircumstancesunderwhichthethingstheyareexchangedforwillbeconsumed.In one procedure the behavior of the students is sampled from time to time. A

student ischosenwithsomemechanicalsystemsuchasspinningadialordrawinganamefromabowl,andhisbehaviorissampledfor,say,20or30seconds.Heisthentoldthathehasbeenobservedandthathehasorhasnotreceivedatokenorcredit.Adayortwoofthisisoftenenoughtomakeagreatchange:theroomgrowsquietasthestudents go to work. Sampling can then become less frequent. Eventually, as thestudentsbegintobereinforcedinotherwayswhentheyfindthemselvesworkingmoreeffectively inaquiet room, theywillconstruct theirownsocialcontingencies,whichmayeventuallyreplacethosearrangedbytheteacher.No one procedure will work well in every classroom, and a certain ingenuity is

268

needed todevise the right system in the rightplace,but theprincipleofcontingencymanagement is sound and it is proving effective in a rapidly increasing number ofexperiments. Research conducted in a classroom is not always impressive“statistically,”butenoughhasbeendonetowarrantfurtherexperimentationonabroadscale.Thereareobjections,however,andsomeofthemcallforcomment.Reinforcement

is sometimes called bribery. (To say this is tomake a confession: a bribe is paid toinduce a person to do something he is for some reason inclined not to do, and it istragicthatwearesoreadytoseeschoolworkinthatlight.)Thepointofabribeisanimpliedcontract(“DothisandIwillgiveyouthat”),butacontracttendstodestroytheeffectofareinforcer.Contingenciesofreinforcementaremosteffectivewhenthereisnoprioragreementastoterms.Amorevalidobjectionisthatcontingenciesofthissortareartificial.Inreallifeone

does not sit quietly in order to take a field trip to the zoo or stop annoying one’sneighborinordertogetanicecreamcone.Theconnectionbetweenthebehavioranditsconsequence iscontrived. (It iscurious thatnooneraises thesameobjectionwithrespecttopunishment,forthereisnonaturalconnectionbetweensolvingaprobleminarithmeticandavoidingthecane.Andgoodmarks,promotion,honors,andprizesarenot only artificial reinforcers, they are artificially and ineffectively contingent onbehavior.)But artificiality is not the issue.Weuse contrivedcontingencies to set upbehaviorwhichwill,wehope,bereinforcednaturallyunderthecontingenciesofdailylife.Theproblemistomakesurethatthebehaviorwesetupwillindeedbeeffectiveintheworldatlarge.Therehaveoftenbeengreatdiscrepanciesbetweenwhatistaughtandwhatstudents

eventuallyuse.Verbalmaterialsareeasilyimportedintotheclassroom(intheformofdiscussions, lectures, and testbooks), and they have often been overemphasized.Students spend a great deal of time answering questions, but answering questions isonly a small part of daily life.Nonverbal behavior also needs to be taught.But thisdoesnotmeanthatweshouldgetridofverbalteachingaltogether.Thevalueofverbalprogramsinsuchafieldasmedicalschoolanatomymaywellbequestioned.Nothingbutacadaverwillteachthewould-bedoctorwhatthehumanbodyislikeorpermithimtoacquirethespecialbehaviorsheneeds.Onewouldcertainlynotwanttobeoperatedupon by a surgeon who had merely worked through a programmed text in humananatomy.Butthereisagreatdealtobesaidforprogrammedinstructionbeforeturningto a cadaver. What one learns in verbal or pictorial form facilitates learning aboutthingsthemselves.Thereisnothingunrealaboutverbalmaterial.Anotherobjectionisthatreinforcersindailylifearenotalwaysimmediate,andthat

thestudentmustbepreparedtobehaveforthesakeofremoteconsequences.Nooneiseveractually reinforcedby remoteconsequences,but ratherbymediating reinforcerswhich have acquired their power through some connection with them. Mediatingreinforcers can be set up, however, and the student can be taught with availableprinciplesandtechniquestofindorconstructthemforhimself.Arathersimilarobjectionisthatindailylifeastudentisnotalwaysreinforcedwhen

hebehaves,andthatheshouldbecomeaccustomedtononreinforcement.Butthisisasubject which has been studied with particular care. High levels of activity can besustained by intermittent reinforcement, particularly if the schedule of reinforcementhasbeensuitablyprogrammed.Agamblerisreinforcedonwhatiscalleda“variable-ratioschedule.”Itmaysustainhisbehaviortothepointatwhichhelosesallhismoney,

269

butitwillnothavethiseffectunlessthemeanratioofresponsestoreinforcementshasbeenextendedgradually.Studentsreinforcedonavariable-ratioschedulewillshowafantasticdedicationiftheschedulehasbeenproperlyprogrammed.Theywillworkforlongperiodsoftimewithnoreinforcementwhatsoever,andarethuswellpreparedforaworldinwhichreinforcementsmayindeedberare.Current applications of operant conditioning to education are no doubt crude, but

theyareabeginning,andabeginningmustbemade.Thetaskisparticularlydifficultbecause we must contend with theories and practices which are deeply entrenched.Thereisnothingverynewinprevailingeducationaltheories,anditwillbealongtimebeforewecanproperlyestimate theharmtheyhavedone.Most teachers todayteachessentiallyas teachershave taught forcenturies.Thebestof themare simplypeoplewhohaveaknackingettingalongwithothers.All thismustchange,andthechangewill take time. But we are on the verge of a new educational “method”—a newpedagogy—inwhichtheteacherwillemergeasaskilledbehavioralengineer.Hewillbeabletoanalyzethecontingencieswhichariseinhisclasses,anddesignandsetupimprovedversions.Hewillknowwhatistobedoneandwillhavethesatisfactionofknowingthathehasdoneit.Thetrainingofateachershouldbeginwithbasicprinciples.Everyonewhointends

tobea teachershouldhaveachancetosee learningtakeplaceor,better, toproducevisible learning himself, as by shaping the behavior of a rat or a pigeon. It is aheartening experience to discover that one can produce behavior of specifiedtopographyandbringitunderthecontrolofspecifiedstimuli.Somesuchexperienceisparticularly valuable because the effects of positive reinforcement are somewhatdelayed incontrastwithpunishment,which tends tobeused inpart justbecause theresults arequick.Laboratoryor classroompractice inoperant conditioninggives theteacher the confidence he needs to change behavior in less immediate but moreeffectiveways.It also clarifies the mistakes teachers make when they are careless about

reinforcement. Many problems in classroom management arise because the teacherreinforcesstudentswhentheybehaveinobjectionableways.Forexample,theteachermaypayspecialattentionwhenthestudentusesobscenitiesormovesaboutortalksatinappropriate times.The teacher tends todoso“naturally,”andhewillbedissuadedfromdoingsoonlywhentheeffectsofreinforcementhavebeenmadecleartohim.An example of the misuse of operant reinforcement in the classroom has been

analyzedelsewhere.Nomatterhowbadateachermaybe,hehasatleastoneavailablereinforcer—dismissing his class. If, near the end of a period, he is free to tell hisstudents that theymay leave (if there isno routinesignalsuchasabell),hecanusedismissalasapowerfulreinforcer.Heshouldwaituntilthebehavioroftheclassisasacceptableas it is likely tobeand thendismiss.Butalmost inevitablyhewilldo thewrongthing:hewilltendtodismisstheclasswhentroubleisbrewing.Asurreptitiousfightisstartinginthebackoftheroom,andsohesays“That’senoughfortoday.”Indoing so he gets out of today’s trouble, but a fight will be more likely to starttomorrow.Anothernaturalmistakeistoshifttoamoreinterestingtopicwhenadiscussionor

lectureappearstobeboringthelistener.Amoreinterestingtopicisareinforcer,andbyshifting to it we reinforce expressions of boredom. Another common mistake is todistracttheattentionofalikelytroublemaker.Adistractionisbydefinitionreinforcing,and it reinforces what the student is doing when we distract him—namely, making

270

trouble.Wemakemistakesofthissortuntilagreaterfamiliaritywiththeprinciplesofreinforcementinducesustostop.In England a “black paper” recently criticized the educational establishment. It

performeda servicebybringing into theopenagrowingdissatisfactionwith currentmethods.Wehavebeen too ready to assume that the student is a free agent, thathewants to learn, thatheknowsbestwhathe should learn, thathis attitudes and tastesshoulddeterminewhathelearns,andthatheshoulddiscoverthingsforhimselfratherthan learnwhat others have already discovered. These principles are allwrong, andtheyareresponsibleformuchofourcurrenttrouble.Educationisprimarilyconcernedwith the transmission of a culture—with teaching new members what others havealreadylearned—anditisdangeroustoignorethisfunction.Buttheblackpapertookthewronglinebysuggestingthatwereturntowhatareessentiallypunitivetechniques.Theteachermustregaincontrol,buthemustdosoinwayswhicharenotonlymoreefficient but free of the undesirable by-products of older practices. Progressiveeducationmadeanhonesteffort todispensewithpunishment,but itnever found thealternativesitneeded.Effectivealternativesarenowavailable.Theclassroomisakindofcommunity,withacultureofitsown,andwecandesign

suchaculturewhilerespectingthestandardsofdignityandfreedomwhichwevalueintheworldatlarge.Theassignmentisimportantbecauseinthelongruneducationmusttakeitsplaceasthemethodofchoiceinallformsofsocialcontrol.Itmustreplacetheaversive sanctions of government, both international and domestic, and the undulycompellingeconomic sanctionsofbusinessand industry.Theby-productsareall toovisibletoday,inpartbecauseoftheviolencewithwhichtheyareattacked.Thesoonerwe find effectivemeans of social control, the sooner we shall produce a culture inwhichman’s potential is fully realized. Thosewho are genuinely trying to improveeducation have, therefore, a frightening responsibility, but they face a tremendousopportunity.

FromEducation,1969,90,93–100.

271

PARTIV

TheAnalysisandManagementofNeurotic,Psychotic,andRetardedBehavior

ACritiqueofPsychoanalyticConceptsandTheoriesPsychologyintheUnderstandingofMentalDiseaseWhatisPsychoticBehavior?SomeRelationsBetweenBehaviorModificationandBasicResearchCompassionandEthicsintheCareoftheRetardate

272

ACritiqueofPsychoanalyticConceptsandTheories

AttheannualmeetingoftheAmericanAssociationfortheAdvancementofSciencein1953,theInstitutefortheUnityofScience, thePhilosophyofScienceAssociation,and Section L of AAAS sponsored a series of papers subsequently published in TheScientificMonthlyandlaterundertheeditorshipofPhilippG.FrankinTheValidationof Scientific Theories (Boston, 1956). The present paper appeared in The ScientificMonthlyinNovember,1954,andisreprintedherebypermissionoftheeditor.

Freud’sgreatcontributiontoWesternthoughthasbeendescribedastheapplicationofthe principle of cause and effect to human behavior. Freud demonstrated thatmanyfeaturesofbehaviorhithertounexplained—andoftendismissedashopelesslycomplexorobscure—couldbe shown tobe theproductofcircumstances in thehistoryof theindividual.Manyofthecausalrelationshipshesoconvincinglydemonstratedhadbeenwholly unsuspected—unsuspected, in particular, by the very individuals whosebehavior theycontrolled.Freudgreatly reduced thesphereofaccidentandcaprice inourconsiderationsofhumanconduct.Hisachievement in this respectappearsall themore impressivewhenwe recall thathewasnever able to appeal to thequantitativeproofscharacteristicofothersciences.Hecarriedthedaywithsheerpersuasion—withthe massing of instances and the delineation of surprising parallels and analogiesamongseeminglydiversematerials.Thiswasnot,however,Freud’sownviewofthematter.Attheageof70hesummed

uphisachievementinthisway:“Mylifehasbeenaimedatonegoalonly:toinferorguesshowthementalapparatusisconstructedandwhatforcesinterplayandcounteractinit.”1 It isdifficult todescribe thementalapparatusherefers to innoncontroversialterms,partlybecauseFreud’sconceptionchangedfromtimetotimeandpartlybecauseits very nature encouragedmisinterpretation andmisunderstanding.But it is perhapsnottoowideofthemarktoindicateitsprincipalfeaturesasfollows:Freudconceivedof some realm of themind, not necessarily having physical extent, but neverthelesscapableoftopographicdescriptionandofsubdivisionintoregionsoftheconscious,co-conscious,andunconscious.Withinthisspace,variousmentalevents—ideas,wishes,memories, emotions, instinctive tendencies, and so on—interacted and combined inmanycomplexways.Systemsofthesementaleventscametobeconceivedofalmostas subsidiary personalities and were given proper names: the Id, the Ego, and theSuperego.Thesesystemsdividedamongthemselvesalimitedstoreofpsychicenergy.Therewere,ofcourse,manyotherdetails.Nomatter what logiciansmay eventually make of this mental apparatus, there is

littledoubtthatFreudaccepteditasrealratherthanasascientificconstructortheory.One does not at the age of 70 define the goal of one’s life as the exploration of anexplanatory fiction. Freud did not use his “mental apparatus” as a postulate systemfromwhichhededucedtheoremstobesubmittedtoempiricalcheck.Iftherewasanyinteractionbetweenthementalapparatusandempiricalobservations,ittooktheform

273

ofmodifyingtheapparatustoaccountfornewlydiscoveredfacts.TomanyfollowersofFreudthementalapparatusappearstobeequallyasreal,andtheexplorationofsuchan apparatus is similarly accepted as the goal of a science of behavior. There is analternative view, however, which holds that Freud did not discover the mentalapparatus but rather invented it, borrowing part of its structure from a traditionalphilosophyofhumanconductbutaddingmanynovelfeaturesofhisowndevising.There are those who will concede that Freud’s mental apparatus was a scientific

constructratherthananobservableempiricalsystembutwho,nevertheless,attempttojustify it in the light of scientific method. One may take the line that metaphoricaldevicesareinevitableintheearlystagesofanyscienceandthatalthoughwemaylookwith amusement today upon the “essences,” “forces,” “phlogistons,” and “ethers” ofthescienceofyesterday, theseneverthelesswereessential to thehistoricalprocess. Itwould be difficult to prove or disprove this. However, if we have learned anythingabout the nature of scientific thinking, if mathematical and logical researches haveimprovedour capacity to represent andanalyze empirical data, it is possible thatwecanavoidsomeofthemistakesofadolescence.WhetherFreudcouldhavedonesoispastdemonstrating,butwhetherweneedsimilarconstructsinthefutureprosecutionofascienceofbehaviorisaquestionworthconsidering.Constructs are convenient and perhaps even necessary in dealing with certain

complicated subject matters. As Frenkel-Brunswik shows,2 Freud was aware of theproblemsofscientificmethodologyandevenofthemetaphoricalnatureofsomeofhisownconstructs.Whenthiswasthecase,hejustifiedtheconstructsasnecessaryoratleasthighlyconvenient.Butawarenessofthenatureofthemetaphorisnodefenseofit,and if modern science is still occasionally metaphorical, we must remember thattheorywise it is also still in trouble. The point is not that metaphor or construct isobjectionablebutthatparticularmetaphorsandconstructshavecausedtroubleandarecontinuing to do so. Freud recognized the damageworked by his ownmetaphoricalthinking,buthefelt that itcouldnotbeavoidedandthat thedamagemustbeputupwith.Thereisreasontodisagreewithhimonthispoint.Freud’sexplanatoryschemefollowedatraditionalpatternoflookingforacauseof

humanbehaviorinsidetheorganism.Hismedicaltrainingsuppliedhimwithpowerfulsupportinganalogies.Theparallelbetween theexcisionofa tumor, forexample,andthereleaseofarepressedwishfromtheunconsciousisquitecompellingandmusthaveaffectedFreud’sthinking.Now,thepatternofaninnerexplanationofbehaviorisbestexemplifiedbydoctrinesofanimism,whichareprimarilyconcernedwithexplainingthe spontaneity and evident capriciousness of behavior. The living organism is anextremelycomplicatedsystembehavinginanextremelycomplicatedway.Muchofitsbehaviorappearsatfirstblushtobeabsolutelyunpredictable.Thetraditionalprocedurehad been to invent an inner determiner, a “demon,” “spirit,” “homunculus,” or“personality”capableofspontaneouschangeofcourseoroforiginationofaction.Suchaninnerdetermineroffersonlyamomentaryexplanationofthebehavioroftheouterorganism,becauseitmust,ofcourse,beaccountedforalso,butitiscommonlyusedtoputthematterbeyondfurtherinquiryandtobringthestudyofacausalseriesofeventstoadeadend.Freud, himself, however, did not appeal to the inner apparatus to account for

spontaneityor capricebecausehewas a thoroughgoingdeterminist.He accepted theresponsibilityofexplaining,inturn, thebehavioroftheinnerdeterminer.Hedidthis

274

by pointing to hitherto unnoticed external causes in the environmental and genetichistoryoftheindividual.Hedidnot,therefore,needthetraditionalexplanatorysystemfortraditionalpurposes;buthewasunabletoeliminatethepatternfromhisthinking.Itledhim to representeachof thecausal relationshipshehaddiscoveredasaseriesofthree events. Some environmental condition, very often in the early life of theindividual,leavesaneffectupontheinnermentalapparatus,andthisinturnproducesthe behavioral manifestation or symptom. Environmental event, mental state orprocess, behavioral symptom—these are the three links in Freud’s causal chain. Hemadenoappealtothemiddlelinktoexplainspontaneityorcaprice.Insteadheusedittobridge thegap inspaceand timebetween theeventshehadproved tobecausallyrelated.Apossible alternative,whichwouldhavehadnoquarrelwith established science,

wouldhavebeentoarguethattheenvironmentalvariablesleavephysiological effectswhichmay be inferred from the behavior of the individual, perhaps at amuch laterdate.Inonesense,toolittleisknownatthemomentaboutphysiologicalprocessestomakethemusefulinalegitimatewayforthispurpose.Ontheotherhand,toomuchisknownofthem,atleastinanegativeway.Enoughisknownofthenervoussystemtoplacecertaindimensionallimitsuponspeculationandtoclipthewingsofexplanatoryfictions.Freudaccepted,therefore,thetraditionalfictionofamentallife,avoidinganout-and-out dualism by arguing that eventually physiological counterparts would bediscovered. Quite apart from the question of the existence of mental events, let usobservethedamagewhichresultedfromthismaneuver.We may touch only briefly upon two classical problems which arise once the

conceptionofamentallifehasbeenadopted.Thefirstoftheseistoexplainhowsuchalife is tobeobserved.The introspectivepsychologistshadalready tried to solve thisproblemby arguing that introspection is only a special case of the observationuponwhichallsciencerestsandthatman’sexperiencenecessarilystandsbetweenhimandthephysicalworldwithwhichsciencepurportstodeal.ButitwasFreudhimselfwhopointedoutthatnotallofone’smentallifewasaccessibletodirectobservation—thatmanyeventsinthementalapparatuswerenecessarilyinferred.Greatasthisdiscoverywas,itwouldhavebeenstillgreaterifFreudhadtakenthenextstep,advocatedalittlelater by theAmericanmovement calledBehaviorism, and insisted that conscious, aswell as unconscious, events were inferences from the facts. By arguing that theindividual organism simply reacts to its environment, rather than to some innerexperienceofthatenvironment,thebifurcationofnatureintophysicalandpsychiccanbeavoided.3Asecondclassicalproblemishowthementallifecanbemanipulated.Intheprocess

oftherapy,theanalystnecessarilyactsuponthepatientonlythroughphysicalmeans.HemanipulatesvariablesoccupyingapositioninthefirstlinkofFreud’scausalchain.Nevertheless, it is commonly assumed that the mental apparatus is being directlymanipulated.Sometimes it isargued thatprocessesare initiatedwithin the individualhimself, suchas thoseof freeassociationand transference,and that these in turnactdirectlyupon themental apparatus. But how are thesemental processes initiated byphysicalmeans?Theclarificationofsuchacausalconnectionplacesaheavyandoftenunwelcomeburdenofproofupontheshouldersofthedualist.TheimportantdisadvantagesofFreud’sconceptionofmental lifecanbedescribed

somewhatmorespecifically.Thefirstoftheseconcernstheenvironmentalvariablesto

275

whichFreudsoconvincinglypointed.Thecogencyof thesevariableswas frequentlymissed because the variableswere transformed and obscured in the course of beingrepresented in mental life. The physical world of the organism was converted intoconsciousandunconsciousexperience,andtheseexperienceswerefurthertransmutedastheycombinedandchangedinmentalprocesses.Forexample,earlypunishmentofsexual behavior is an observable fact which undoubtedly leaves behind a changedorganism.Butwhenthischangeisrepresentedasastateofconsciousorunconsciousanxiety or guilt, specific details of the punishment are lost. When, in turn, someunusual characteristic of the sexual behavior of the adult individual is related to thesupposedguilt,manyspecificfeaturesoftherelationshipmaybemissedwhichwouldhavebeenobviousif thesamefeaturesofbehaviorhadbeenrelatedtothepunishingepisode.InsofarasthementallifeoftheindividualisusedasFreudusedittorepresentandtocarryanenvironmentalhistory,itisinadequateandmisleading.Freud’s theory of the mental apparatus had an equally damaging effect upon his

studyofbehaviorasadependentvariable.Inevitably,itstoletheshow.Littleattentionwaslefttobehaviorperse.Behaviorwasrelegatedtothepositionofameremodeofexpressionoftheactivitiesofthementalapparatusorthesymptomsofanunderlyingdisturbance. Among the problems not specifically treated in the manner which wastheirdue,wemaynotefive.1. The nature of the act as a unit of behavior was never clarified. The simple

occurrence of behaviorwas neverwell represented. “Thoughts” could “occur” to anindividual; he could “have” ideas according to the traditional model; but he could“have”behavioronly ingivingexpression to these inner events.Wearemuchmorelikelytosaythat“thethoughtoccurredtometoaskhimhisname”thanthat“theactofaskinghimhisnameoccurredtome.”It is thenatureof thoughtsandideasthat theyoccurtopeople,butwehavenevercometobeathomeindescribingtheemissionofbehavior in a comparableway.This is especially trueofverbalbehavior. In spiteofFreud’svaluableanalysisofverbalslipsandofthetechniquesofwitandverbalart,herejectedthepossibilityofananalysisofverbalbehaviorinitsownrightratherthanasthe expression of ideas, feelings, or other inner events, and therefore missed theimportanceofthisfieldfortheanalysisofunitsofbehaviorandtheconditionsoftheiroccurrence.Thebehavioralnatureofperceptionwasalsoslighted.Toseeanobjectasanobject

isnotmerepassivesensing;itisanact,andsomethingverymuchlikeitoccurswhenweseeanobjectalthoughnoobjectispresent.FantasyanddreamswereforFreudnottheperceptualbehavioroftheindividualbutpicturespaintedbyaninnerartistinsomeatelierofthemindwhichtheindividualthencontemplatedandperhapsthenreported.This division of labor is not essential when the behavioral component of the act ofseeingisemphasized.2. The dimensions of behavior, particularly its dynamic properties, were never

adequatelyrepresented.Weareallfamiliarwiththefactthatsomeofouractsaremorelikely to occur upon a given occasion than others. But this likelihood is hard torepresentandhardertoevaluate.Thedynamicchangesinbehaviorwhicharethefirstconcernofthepsychoanalystareprimarilychangesinprobabilityofaction.ButFreudchose to dealwith this aspect of behavior in other terms—as a questionof “libido,”“cathexis,” “volume of excitation,” “instinctive or emotional tendencies,” “availablequantitiesofpsychicenergy,”andsoon.Thedelicatequestionofhowprobabilityofaction is to be quantified was never answered, because these constructs suggested

276

dimensions to which the quantitative practices of science in general could not beapplied.3. In his emphasis upon the genesis of behavior, Freud made extensive use of

processes of learning.Thesewere never treated operationally in termsof changes inbehavior but rather as the acquisition of ideas, feelings, and emotions later to beexpressed by, or manifested in, behavior. Consider, for example, Freud’s ownsuggestionthatsiblingrivalryinhisownearlyhistoryplayedanimportantpartinhistheoreticalconsiderationsaswellasinhispersonalrelationshipsasanadult.AninfantbrotherdiedwhenFreudhimselfwasonly1½yearsold,andasayoung

child Freud played with a boy somewhat older than himself and presumably morepowerful, yet who was, strangely enough, in the nominally subordinate position ofbeinghisnephew.Toclassifysuchasetofcircumstancesassiblingrivalryobscures,aswehaveseen,themanyspecificpropertiesofthecircumstancesthemselvesregardedasindependentvariablesinascienceofbehavior.Toarguethatwhatwaslearnedwastheeffectofthesecircumstancesuponunconsciousorconsciousaggressivetendenciesorfeelings of guilt works a similar misrepresentation of the dependent variable. Anemphasis upon behavior would lead us to inquire into the specific acts plausiblyassumed to be engendered by these childhood episodes. In very specific terms, howwasthebehaviorof theyoungFreudshapedbythespecialreinforcingcontingenciesarisingfromthepresenceofayoungerchildinthefamily,bythedeathofthatchild,and by later association with an older playmate who nevertheless occupied asubordinatefamilyposition?WhatdidtheyoungFreudlearntodotoachieveparentalattention under these difficult circumstances? How did he avoid aversiveconsequences? Did he exaggerate any illness? Did he feign illness? Did hemake aconspicuousdisplayofbehaviorwhichbroughtcommendation?Wassuchbehaviortobe found in the field of physical prowess or intellectual endeavor? Did he learn toengage in behaviorwhichwould in turn increase the repertoires available to him toachievecommendation?Didhestrikeorotherwiseinjureyoungchildren?Didhelearntoinjurethemverballybyteasing?Washepunishedforthis,andifso,didhediscoverother forms of behavior which had the same damaging effect but were immune topunishment?Wecannot,ofcourse,adequatelyanswerquestionsofthissortatsolateadate,but

theysuggestthekindofinquirywhichwouldbepromptedbyaconcernfortheexplicitshapingofbehavioral repertoiresunderchildhoodcircumstances.Whathas survivedthrough theyears is not aggressionandguilt, later tobemanifested inbehavior, butratherpatternsofbehavior themselves. It isnotenough to say that this is “all that ismeant” by sibling rivalry or by its effects upon the mental apparatus. Such anexpression obscures, rather than illuminates, the nature of the behavioral changestaking place in the childhood learning process.A similar analysis could bemade ofprocessesinthefieldsofmotivationandemotion.4.An explicit treatment of behavior as a datum, of probability of response as the

principal quantifiable property of behavior, and of learning and other processes intermsofchangesofprobability isusuallyenough toavoidanotherpitfall intowhichFreud,incommonwithhiscontemporaries,fell.Therearemanywordsinthelayman’svocabulary which suggest the activity of an organism yet are not descriptive ofbehavior in the narrower sense. Freud usedmany of these freely—for example, theindividual is said to discriminate, remember, infer, repress, decide, and so on. Suchterms do not refer to specific acts. We say that a man discriminates between two

277

objects when he behaves differently with respect to them; but discriminating is notitselfbehavior.Wesay thathe repressesbehaviorwhichhasbeenpunishedwhenheengages in other behavior just because it displaces the punished behavior; butrepressingisnotaction.Wesaythathedecidesuponacourseofconducteitherwhenheentersupononecourse to theexclusionofanother,orwhenhealterssomeof thevariablesaffectinghisownbehaviorinordertobringthisabout;butthereisnoother“actofdeciding.”Thedifficultyisthatwhenoneusestermswhichsuggestanactivity,one feels it necessary to invent an actor, and the subordinate personalities in theFreudianmentalapparatusdo,indeed,participateinjusttheseactivitiesratherthaninthemorespecificbehavioroftheobservableorganism.Among these activities are conspicuous instances involving the process of self-

control—the so-called “Freudian mechanisms.” These need not be regarded asactivitiesoftheindividualoranysubdivisionthereof—theyarenot,forexample,whathappens when a skillful wish evades a censor—but simply as ways of representingrelationships among responses and controlling variables. I have tried to demonstratethis by restating the Freudian mechanisms without reference to Freudian theory [inScienceandHumanBehavior].5.SinceFreudneverdevelopedaclearconceptionofthebehavioroftheorganism

andneverapproachedmanyofthescientificproblemspeculiartothatsubjectmatter,itis not surprising that he misinterpreted the nature of the observation of one’s ownbehavior.Thisisadmittedlyadelicatesubject,whichpresentsproblemswhichnoone,perhaps, has adequately solved. But the act of self-observation can be representedwithin the framework of physical science. This involves questioning the reality ofsensations,ideas,feelings,andotherstatesofconsciousnesswhichmanypeopleregardasamongthemostimmediateexperiencesoftheirlife.Freudhimselfpreparedusforthischange.Thereis,perhaps,noexperiencemorepowerfulthanthatwhichthemysticreports of his awareness of the presence ofGod.The psychoanalyst explains this inotherways.He himself, however,may insist upon the reality of certain experienceswhichotherswishtoquestion.Thereareotherwaysofdescribingwhatisactuallyseenorfeltundersuchcircumstances.Eachofusisinparticularlyclosecontactwithasmallpartoftheuniverseenclosed

withinhisownskin.Undercertainlimitedcircumstances,wemaycometoreacttothatpartoftheuniverseinunusualways.Butitdoesnotfollowthatthatparticularparthasanyspecialphysicalornon-physicalpropertiesor thatourobservationsof itdiffer inanyfundamentalrespectfromourobservationsoftherestoftheworld.Ihavetriedtoshowelsewhere4 how self-knowledge of this sort arises and why it is likely to besubjecttolimitationswhicharetroublesomefromthepointofviewofphysicalscience.Freud’s representations of these events was a particular personal contributioninfluencedbyhisownculturalhistory.Itispossiblethatsciencecannowmoveontoadifferentdescriptionofthem.Ifitisimpossibletobewhollynonmetaphorical,atleastwemayimproveuponourmetaphors.The crucial issue here is the Freudian distinction between the conscious and

unconsciousmind.Freud’scontributionhasbeenwidelymisunderstood.Theimportantpointwasnotthattheindividualwasoftenunabletodescribeimportantaspectsofhisownbehaviororidentifyimportantcausalrelationshipsbutthathisabilitytodescribethemwas irrelevant to the occurrence of the behavior or to the effectiveness of thecauses.Webeginbyattributingthebehavioroftheindividualtoeventsinhisgenetic

278

andenvironmentalhistory.Wethennotethatbecauseofcertainculturalpractices,theindividual may come to describe some of that behavior and some of those causalrelationships. We may say that he is conscious of the parts he can describe andunconsciousoftherest.Buttheactofself-description,asofself-observation,playsnopartinthedeterminationofaction.Itissuperimposeduponbehavior.Freud’sargumentthatweneednotbeawareofimportantcausesofconductleadsnaturallytothebroaderconclusionthatawarenessofcausehasnothingtodowithcausaleffectiveness.

In addition to these specific consequences of Freud’smental apparatus in obscuringimportant details among the variables ofwhichhumanbehavior is a function and inleadingto theneglectof importantproblemsin theanalysisofbehaviorasaprimarydatum, we have to note the most unfortunate effect of all. Freud’s methodologicalstrategy has prevented the incorporation of psychoanalysis into the body of scienceproper.Itwasinherentinthenatureofsuchanexplanatorysystemthatitskeyentitieswould be unquantifiable in the sense in which entities in science are generallyquantifiable,butthespatialandtemporaldimensionsoftheseentitieshavecausedotherkindsoftrouble.Onecansenseacertainembarrassmentamongpsychoanalyticwriterswithrespectto

the primary entities of themental apparatus.There is a predilection for termswhichavoid the embarrassing question of the spatial dimensions, physical or otherwise, ofterms at the primary level. Although it is occasionally necessary to refer to mentaleventsandtheirqualitiesandtostatesofconsciousness,theanalystusuallymovesonin some haste to less committal terms such as forces, processes, organizations,tensions, systems, andmechanisms. But all these imply terms at a lower level. Thenotionofaconsciousorunconscious“force”maybeausefulmetaphor,butifthisisanalogous to force in physics, what is the analogous mass which is analogouslyaccelerated?Humanbehavior is ina stateof fluxandundergoingchangeswhichwecall “processes,” but what is changing in what direction when we speak of, forexample, an affective process? Psychological “organizations,” “mental systems,”“motivational interaction”—these all imply arrangements or relationships amongthings, but what are the things so related or arranged? Until this question has beenanswered the problem of the dimensions of the mental apparatus can scarcely beapproached.Itisnotlikelythattheproblemcanbesolvedbyworkingoutindependentunitsappropriate to thementalapparatus,althoughithasbeenproposedtoundertakesuchastepinattemptingtoplacepsychoanalysisonascientificfooting.Beforeoneattemptstoworkoutunitsoftransferenceorscalesofanxiety,orsystems

ofmensurationappropriatetotheregionsofconsciousness,itisworthaskingwhetherthere isnotanalternativeprogramforarapprochementwithphysical sciencewhichwouldmake such a task unnecessary. Freud could hope for an eventual union withphysicsorphysiologyonly through thediscoveryofneurologicalmechanismswhichwouldbe theanalogsof,orpossiblyonlyotheraspectsof, thefeaturesofhismentalapparatus. Since this depended upon the prosecution of a science of neurology farbeyond its current state of knowledge, itwas not an attractive future. Freud appearsnever to have considered the possibility of bringing the concepts and theories of apsychologicalscienceintocontactwith therestofphysicalandbiologicalsciencebythesimpleexpedientofanoperationaldefinitionofterms.Thiswouldhaveplacedthementalapparatusinjeopardyasalifegoal,butitwouldhavebroughthimbacktotheobservable,manipulable,andpreeminentlyphysicalvariableswithwhichhewasinthe

279

lastanalysisdealing.

1Jones,E.LifeandWorkofSigmundFreud.NewYork:BasicBooks,1953,Vol.1.2Frenkel-Brunswik,E.P.ScientificMonthly,1954,79,293.3AlthoughitwasFreudhimselfwhotaughtustodoubtthefacevalueofintrospection,heappearstohavebeenresponsiblefortheviewthatanothersortofdirectexperienceisrequiredifcertainactivitiesinthementalapparatusaretobecomprehended.Sucharequirementisimpliedinthemodernassertionthatonlythosewhohavebeenpsychoanalyzedcanfullyunderstandthemeaningoftransferenceorthereleaseofarepressedfear.

4Seepage416below.

280

PsychologyintheUnderstandingofMentalDisease

Twoconferencesonthe“EtiologyofMentalDisease”wereheldin1953and1954under the auspices of theCommittee onPublicHealth of theNewYorkAcademyofMedicine. The present paper was given at the second conference. It appears inIntegratingtheApproachestoMentalDisease(NewYork,PaulB.Hoeber,Inc.,1957)andisreprintedherebypermission.Thatvolumealsoreportsmuchofthediscussionwhich followed the papers. My own scattered comments reveal a concern with theoperationaldefinitionoftermsinthefieldofmentaldiseaselaterclarifiedinthepaperonpage303.Asampleisincludedattheendofthefollowingpaper.Anysurveyof thecontributionswhichpsychologycanmaketoourunderstandingofmentaldiseasewilldependuponhowpsychologyisdefined.Inpractice,themethodsand concepts of all four of the disciplines represented at this Conference overlapextensively.Narrowlyconsidered,however,thespecialprovinceofpsychologymaybetaken to be the description of the behavior of the individual as a whole and theexplanation of that behavior in terms of environmental factors and conditions.Morespecifically, psychology is concernedwith recording andmeasuringhumanbehavioranditsvariousaspects,andwithrelatingthequantitiessomeasuredtovariablesinthepastandcurrentenvironment.Manypsychologists,ofcourse,havebroaderinterests.Inadditiontoforceswhicharecurrentlyactingupontheorganism,orhaveacteduponitinthepast,theymaybeconcernedwithvariablesinitsgenetichistory,thephysiologyof its parts, or, at the other extreme, its social environment or cultural history. Anarrowerdelineationofthefieldis,thougharbitrary,desirableforourpresentpurposes.Mental disease appears to refer to modes of behavior which are troublesome or

dangerouseithertotheindividualhimselfortoothers.Behaviormaybetroublesomeordangerousbyitsverynatureorbecauseofthecircumstancesunderwhichitoccurs.Itis not strictly correct to describe such behavior as “atypical,” since extreme orunrepresentativevaluesofmanypropertiesofbehaviordonotalwayspresentproblemsappropriately described as the result of disease. Genius is atypical but, presumably,healthy. It is probably also not of any great value to characterize troublesome ordangerousbehavioras“nonadaptive,”orasviolatingsomeprincipleof“homeostasis”or “equilibrium.” The problems of mental disease arise when an individual showsbehaviorwhich,becauseofitscharacterorthecircumstancesunderwhichitappears,causestrouble.Oneproblemistoexplainthisbehavior,andanotheristochangeit.One contribution which experimental psychology has to offer, by virtue of its

methods and concepts, is a precise description of the behavior under examination.Psychologicaltechniquesarepeculiarlydesignedtoprovidetheclearestpossiblerecordof behavioral manifestations, together with a rigorous demonstration of relations tocausalfactors.Thebehaviorofthementallydiseasedisoftensoobviouslytroublesomeordangerousthatprecisemeasurementisfelttobeunnecessary,butitispossiblethatsuch behavior differs from what might be called normal merely by occupying anextremepositiononacontinuum.Lesserconditions,notsoeasilydetected,mayoffera

281

clue to the causal factors involved. Further advances in the study of the mentallydiseasedmaycreateademandforthetypeofdescriptionwhichcanbeestablishedonlybymorecarefulmeasurement.Inotherwords,thefirstquestiontowhichpsychologymayaddressitselfis,Whatis

neurotic, psychotic, or defective behavior? We cannot answer this with a logicaldefinitionof terms.Thequestion ismorespecific.Givenanagreed-uponexampleofneurotic, psychotic, or defective behavior, what are its significant properties? Thecommoner manifestations of mental disease may not at the present time requiredetaileddescription,butaneventualaccountofthesephenomenawillalmostcertainlyneed to lean upon themethods and terms of a science primarily concernedwith thebehaviorofthewholeorganism.Amongthespecialtopicstowhichpsychologyhasaddresseditself,forexample,is

the sensory control of behavior. Gross instances of hallucinations, anesthesias,confusions,ordefectivecategorizationscanoftenbeacceptedassymptomsofmentaldisease without further inquiry, but a comprehensive account of such phenomena,whichwouldrelatethemto“normal”manifestations,requiresthetechniquesofsensorypsychology.Psychologyhasalsobeenespeciallyconcernedwithmotorbehavior.Thestudyof

gross instances of paralysis, ataxia, loss of skill, or confusion may not require theprecise techniques of the laboratory, but here again a full description of these“symptoms”may.Emotionalbehavior is also a subject towhich psychology has given considerable

attention.Theemotionalbehaviorofthementallydiseasedis,again,oftensogrossthatno appeal is made to the methods of the laboratory. But as further precision incharacterizingabnormalbehaviorisrequired,andastheexperimentalstudyofemotionprogresses,thepsychologicalcharacterizationoftheemotionalpatternofthepsychoticshouldbecomemoreimportant.Motivationhasonlyrecentlybeenstudiedonasubstantialscaleinthecaseofhuman

subjects, but much has been learned from the study of animals with respect to theeffects of deprivation, satiation, and aversive stimulation. Here again the extremeconditionsmanifested by thementally diseased—behaviorswhich suggest excessivedeprivation, or complete satiation, or a failure to avoid or escape from powerfullyaversive conditions—may seem to make the precise methods of experimentalpsychology irrelevant, but an improved account of the psychotic condition musteventuallybebasedupontherelationshipsdemonstratedinsuchascience.Lastly, in the field of learning, quantitative properties of processes have been

demonstrated against which the unusual conditions of the mentally diseased musteventuallybe evaluated.Manydeficiencies in the fieldof learningare at thepresenttime measured indirectly through intelligence tests, but methods which are moreappropriate to a laboratory science are available. The speed with which behavior isacquiredinconditioning,thecomplexityofthebehaviorwhichmaybesoacquired,therate at which such behavior will be extinguished, and the precision with whichbehavior may be brought under stimulus control, all enter into many of thecharacteristicsofmentaldisease.Althoughagrosseffect,suchasaconspicuouslossofmemory,may not at themoment require precisemeasurement, an eventual detailedaccountofthenatureofsuchadifficultymayneedtoappealtothemethodsandresultsoftheexperimentallaboratory.Of special importance in the field ofmental disease aremany forms of behavior

282

resulting from theuseofpunishment as ameasureof control. It isnowclear that inpunishinga response,wedonot simply lower theprobability that itwillbeemitted.Punishmentactsbylettingupcertainaversiveconditionsfromwhichtheorganismmayescape,orwhichitmayavoid,throughmanydifferenttypesofbehavior.Amongsuchavoidance or escape responses are frequently found the troublesome or dangerousresponsescharacteristicofmentaldisease.Themethodsandconceptsofexperimentalpsychologywhicharelikelytobeuseful

in the precise description of the behavior characteristic ofmental disease cannot bedescribed indetail.Thepresentpoint is simply that, among the fourdisciplineshererepresented,1 it isprimarilyexperimentalpsychologywhichhasconcerneditselfwiththe problem of describing and explaining the behavior of the intact organism in theabovesense.Theadvantagewhichpsychologyhasgainedfromthelaboratorycontrolofthevariablesithasstudiedshouldbeemphasized,eventhoughitmayreasonablybeobjected that the variables so far manipulated are by no means comprehensive orentirelyrepresentative.Thesearenot,strictlyspeaking,contributions to theetiologyofmentaldiseasebut

merely to the investigation of that etiology. In exploring the causes of the behaviorcharacteristic of mental disease, the technical contributions of psychology arefrequentlyjoinedwiththemethodsandconceptsofotherdisciplines.Forexample,toinvestigatehereditaryfactorsinthecausationofmentaldisease,wemustcombinethemethodsandconceptsofgeneticswiththoseofexperimentalpsychology.Similarly,ifweare to investigateorganiccauses,weneed to combine themethods and termsofphysiologyandexperimentalpsychology.Muchthesamerelationprevailsinthestudyof endocrinological and pharmacological effects. In each case the psychologist isreachingbeyondthevariableswhichareusuallytakentobecharacteristicofhisspecialfield.Hispeculiar contribution is toprovide, so to speak, abaselineuponwhich theeffectofgenetic,organic,andothervariablesmaybeobserved.Causalfactorsimportantinunderstandingmentaldiseaseare,however,tobefound

amongtheindependentvariablestowhichthepsychologistcharacteristicallyturns.Anexcessive emotional condition, a dangerous mode of escape from anxiety, atroublesomepreoccupationwithsex,oranexcessiveenthusiasmforgamblingmaybenothing more than extreme cases of the effects of environmental conditions. Theseaspects of the personal history and the current environment of the individual arecommonly taken to be in the realm of psychology and within reach of thepsychologist’stechniques.Modesofbehaviorcharacteristicofmentaldiseasemaybesimplytheresultofahistoryofreinforcement,anunusualconditionofdeprivationorsatiation, or an emotionally exciting circumstance. Except for the fact that they aretroublesome or dangerous, they may not be distinguishable from the rest of thebehavioroftheindividual.Insofarasthisisthecase,theetiologyofmentaldiseaseandthepossibilityofanalysisandtherapyliewithinthefieldofpsychologyproper.(Atthispoint an overlap with psychodynamics is obvious. The distinction between thepsychologicalandpsychodynamicviewisnotbasicallyadistinctioninsubjectmatterorintherangeoffactorsstudied.Thedistinctionisprimarilyoneofmethod,anditispossible that these two fields will eventually fuse or at least become very closelyassociated.)Recent work in the field of learning has enabled the psychologist to achieve an

extensivecontroloverthebehaviorofanorganism,andtobringthisbehaviorunderthe

283

control of complex environmental conditions. By manipulating the event called areinforcement, it is possible not only to shapeupmanynovel formsof behavior butalsotosustainalmostanygivenlevelofactivityforlongperiodsoftime.InthefieldofPavlovianconditioning,comparableadvanceshavebeenmadeintheunderstandingoftheoriginsofemotionalpatterns.Itisreasonabletosupposethatsuchanexperimentalsciencewilleventuallyproduceatechnologycapableofmodifyingandsustaininganygivenpatternofbehavioralmostatwill.Supposeitcanbeshownthatanorganismwithagivengenetichistoryandagivenorganicconditioncanbeinducedtoengageinthekinds of behavior characteristic of mental disease through the manipulation ofenvironmental variables. Then in order to demonstrate one type of cause of mentaldisease, it will remain only to show that comparable environmental variables couldhave been operative upon a given person showing neurotic, psychotic, or defectivecharacteristics.Anothertechnologicalextensionofsuchanexperimentalsciencewouldbetochangethebehaviorofthementallydiseased,inthedirectionusuallyreferredtoastherapy,throughthearrangementofenvironmentalvariables.Ifexperimentalpsychologycontinues in itscurrentdirection, itmayreasonablybe

expectedtoshowhowsomeofthebehavioralcharacteristicsofmentaldiseasecanbegeneratedandhowtheycanbecorrected.Byappealingonlytoenvironmentalvariables(whileassumingagivensetofgeneticandorganicvariables),psychologymaymakethis contribution entirely within its own traditional field. That there are etiologicalfactors lyingbeyond this field isdoubtless true.Herepsychologycanmakeonly thekindofco-operativecontributionpreviouslydescribed.Acertainpracticalhierarchyofcauses may, however, be pointed out. Although genetic and organic factors can beefficiently evaluated only by holding environmental factors constant, and althoughenvironmental factors can be correctly evaluated only against a stable genetic andorganiccondition,itisprobablyausefulpracticetoexploreenvironmentalfactorsfirstto seewhether any behavioralmanifestations remain to be attributed to genetic andorganiccauses.

Summary

In a narrow, though traditional sense, psychology is concerned with describing andmeasuring the behavior of the individual and with relating that behavior toenvironmental factors. The techniques and concepts which it has developed for thispurposemaycontribute to thestudyof theetiologyofmentaldiseasebyprovidingamoreprecise characterizationof the formsofbehavior at issue.Thecontributionsofpsychologyinthisrespectwillbecomemoreimportantwhenthegrossdisorderswhichfirstattractourattentionarenolongeradequateincharacterizingadiseasedcondition.Themethodsandconceptsofexperimentalpsychologyarerequiredintheinvestigationof genetic, physiological, endocrinological, and pharmacological factors in thecausationofneuroticandpsychoticbehavior,althoughthemethodsandtechniquesofotherdisciplinesareherealsorequired.Someetiologicalfactorsaretobefoundwithinthe narrower province of psychology itself. Some of the emotional andmotivationalconditionswhich are taken to be symptoms ofmental diseasemay be nothingmorethanextremevaluesoftheeffectsofvariablesencounteredintheanalysisandcontrolofthenormalorganism.Behaviorwhichissotroublesomeordangerousastobesaidtocharacterizementaldiseasemayalsosimplybelearned—thatis,itmaybetheproduct

284

of reinforcing contingencies which affect the organism according to the learningprocesses encountered in the behavior of the normal individual. It is a reasonableexpectation that a developing experimental psychology will find itself increasinglymore effective in producingbehaviorwhichwould be said to reflectmental disease,andinchangingthebehavioroftheactuallymentallydiseasedinthedirectionknownas therapy—and all of this by manipulating environmental variables traditionallyassigned to the field of psychology. In this latter approach to the problem of thecausationofmentaldisease, it is clear thatpsychologyandpsychodynamicsoverlap,thesetwofieldsbeingdistinguishednotintermsofsubjectmatterorthecausalfactorstowhichappealismade,butonlyintechnique—adistinctionwhichmaybelostasthetwosciencesarefurtherdeveloped.

The following comment on some remarks of Dr. Franz Alexander was made at theconferenceatwhichtheprecedingpaperwasgiven.Dr. Alexander has developed a theory of knowledge with the intention, I think, ofjustifyinganalyticpracticesand theories. It isa theoryofknowledgewhichhasverydeephistoricalandtraditionalphilosophicalroots.Heseemstobewillingtobasethemattereventuallyonapragmatictest.Wecannotaskanymorethanthat.Iamperfectlywillingtolethimoranyonewhosodesiresusethatconceptionofknowledgetoshowhow useful it is. However, he does seem to imply that it is the only theory orinterpretation of human knowledge, and that alternative views will necessarily misssomethingwhichisfundamentalandparticularlyimportant.Thenotionthatknowledgeconsistsofsenseimpressionsandconceptsderivedfrom

sense impressionswas,ofcourse, theviewofBritishempiricismand is stillheldbymanypeople.Butothers,includingmyself,believethatitisincapableofrepresentinghuman knowledge adequately. Even a simple idea is not, as Locke supposed, anassemblageof sensorymaterials in response to stimulation.Tosuppose thatphysicalknowledgeexistsinthemindofaphysicistaspsychicormentalmaterial—asthewayhe looks at theworld—seems tome quite absurd.At no time is a physical theory apsychiceventinthesenseofanimageorsensation.Tosay thatphysicsalwaysgetsback tosense impression is simply tosay that the

organism is in contact with the environment only through its sense organs—a veryobvious axiom.But the organismdoesmore than soakup the environment. It reactswith respect to the environment, and throughout its lifetime it learnsmore andmorevariedways of reacting.An alternative conception of knowledge,whichmanyof ushold, is that knowledge is action rather than sensing, and that a formulation ofknowledgeshouldbeintermsofbehavior.Itistruethatwemaybeawareofourownbehaviorinthesensethatwecanseeourselvesseeingsomething.Someofthetimeweknowwhenwearereactingtotheworldaroundus,but,asFreudpointedout,atothertimeswedonotknow.Whenwedoknow,wearereactingtoourselvesaswereacttotheworldaboutus.The notion of knowledge as response is useful in enabling us to formulate the

assembledknowledgeofphysicsintermsoftheequationsandlawswhichcomprisetheverbalbehaviorofthescientist,ratherthanhismentalstates.Italsogivesamuchmoreplausibleviewoftheinsightwhichisapparentlysousefultotheanalyst.Weacquirethe vocabularywhich describes our own behavior under great difficulty. The verbalcommunitywhichcan easily teacha child todistinguish colors, for example, cannotwiththesametechniqueteachhimtodistinguishaches,pains,feelings,andemotions.

285

As physical states in the individual, these are a part of the physical world, but theindividualhimselfhasa special connectionwith them.Myaching tooth ismine inaveryrealsensebecausenoneofyoucanpossiblygetnervesintoit,but thatdoesnotmakeitdifferentinnaturefromtheceilinglightwhichweallreacttoinmoreorlessthesameway.Asaresultofthisphysicalprivacy,thesubjectivevocabulary,suchaswasusedby introspectivepsychologyandseems tobeused todaybysomeanalysts,haslimitationsinprecisionwhichnoonehasbeenabletosurmount.Whentheanalyststatesthathecanpredictwhatapatientisgoingtodo,Iacceptthe

factthathecanmakethatprediction.ButIchallengethestatementthathedoesitbyexperiencing the same sensations or feelings. I do not see any distinction betweenpredictingwhatanindividualisgoingtodoandpredictingwhat,letussay,asailboatisgoingtodo.Apersonwhoisfamiliarwithasailboatandknowshowitwill takethewaveshasahighpredictiveknowledgeofhowtohandlethetiller.Thatseemstometobeparalleltowhattheskilledanalystdoesinknowingwhattoexpectandwhattodonextinthecourseofananalysis.Wedonotneedtoassumethattheskillfulsailormustbeexperiencingsomethingwhichtheboatisexperiencing.Itisobviousthatheisnot.Theanalystshavetendedtoconceptualizecertainwaysofdealingwiththebehaviorofthemselvesandtheirpatientswhichmaybeconvenientor inconvenient,butneednotbeevaluatedaccordingly.Thoseofuswhoworkwithverbalbehaviorperse,ratherthanwiththeexpressionof

meanings,becomesuspiciousofeffortstoimprovevocabulariesfortalkingaboutwhatis going on inside the individual. It will be recalled that the old introspectivepsychologist had a trained observer, but it is now clear that in training him thepsychologistwasdefinitely shaping theway inwhich theobserver reacted to stimuliand was to some extent creating the very data which he was trying to collect.Something of the same sort may very well happen in the training analysis. It isnecessary to be on guard lest the interchange between the patient and the therapistresult in building up a very particular vocabulary which will lead to quite spuriousresults.Severalexperimentsarenowintheliteratureinwhichaninterviewerhasskillfully

shaped, by very slight reinforcements and punishments, the verbal behavior of theperson being interviewed. It is fairly easy to get another person onto an arbitrarilychosen topic during a conversation by showing attention when the topic is beingapproached and inattentionwhen it is being left.The sameprocess could generate amisleadingcommunityofvocabularyinthepersonalinteractionbetweentwopeople.

1Theconferenceatwhichthispaperwasgivenbeganwithstatementsfromfourpointsofview:organic,psychological,psychodynamic,andpsychosocial.

286

WhatIsPsychoticBehavior?

Ascientificprogramon“NewerAspectsoftheTheory,Etiology,andTreatmentofthePsychoses”marked theopeningof theRenardHospital,apsychiatricunitof theWashingtonUniversitySchoolofMedicineandtheBarnesandAffiliatedHospitals,inSt. Louis on October 10, 1955. As part of that program this paper was addressedprimarily to psychiatrists and others concerned with mental health. The analysis,particularlyasrepresentedbythefourfiguresandtheaccompanyingtext,hasprovedusefulinabroadercontext.Incharacterizingthetraditionalmethodofdescribingandexplaining behavior and suggesting a more profitable operational definition ofcommon psychological terms, the paper amplifies points made in “The operationalanalysisofpsychologicalterms”(page416).Sincemyfieldofspecializationliessomedistancefrompsychiatry,itmaybewelltobeginwithcredentials.Thefirstwillbenegative.Inthesenseinwhichmytitleismostlikelytobeunderstood,Iamwhollyunqualifiedtodiscussthequestionbeforeus.ThenumberofhoursIhavespentinthepresenceofpsychoticpeople(assumingthatIammyselfsane)isnegligiblecomparedwithwhatmanyofyoumightclaim,andthetimeIhave spent in relevant reading and discussion would suffer equally from the samecomparison.Iamcurrentlyinterestedinsomeresearchonpsychoticsubjects,towhichIshallreferagainlater,butmyassociationwiththatprograminnowayqualifiesmeasaspecialist.Fortunately, I am not here to answer the question in that sense at all. A more

accurate titlewouldhavebeen “What isbehavior?—with an occasional reference topsychiatry.”HereIwilllistsuchpositivecredentialsasseemappropriate.Ihavespenta good share ofmy professional life in the experimental analysis of the behavior oforganisms.Almostallmysubjectshavebeenbelowthehumanlevel(mostofthemratsor pigeons) and all, so far as I know, have been sane. My research has not beendesignedtotestanytheoryofbehavior,andtheresultscannotbeevaluatedintermsofthe statistical significance of such proofs. The object has been to discover thefunctionalrelationswhichprevailbetweenmeasurableaspectsofbehaviorandvariousconditions and events in the life of the organism. The success of such a venture isgaugedbytheextenttowhichbehaviorcan,asaresultoftherelationshipsdiscovered,actuallybepredictedandcontrolled.Herewehave, I think,been fortunate.Withinalimitedexperimentalarrangement,mycolleaguesandIhavebeenabletodemonstratealawfulnessinbehaviorwhichseemstousquiteremarkable.Inmorerecentresearchithasbeenpossible tomaintain—actually, to sharpen—thisdegreeof lawfulnesswhileslowlyincreasingthecomplexityofthebehaviorstudied.Theextentofthepredictionandcontrolwhichhavebeenachieved isevidentnotonly in“smoothnessofcurves”anduniformityofresultsfromindividualtoindividualorevenspeciestospecies,butinthe practical useswhich are already beingmade of the techniques—for example, inprovidingbaselinesforthestudyofpharmacologicalandneurologicalvariables,orinconvertingalowerorganismintoasensitivepsychophysicalobserver.

287

Althoughresearchdesignedinthiswayhasanimmediatepracticalusefulness,it isnotindependentofonesortoftheory.Aprimaryconcernhasbeentoisolateausefulandexpedientmeasure.Ofallthemyriadaspectsofbehaviorwhichpresentthemselvesto observation, which are worth watching? Which will prove most useful inestablishing functional relations?From time to timemanydifferentcharacteristicsofbehavior have seemed important. Students of the subject have asked how wellorganized behavior is, howwell adapted it is to the environment, how sensitively itmaintainsahomeostaticequilibrium,howpurposefulitis,orhowsuccessfullyitsolvespracticalproblemsoradjuststodailylife.Manyhavebeenespeciallyinterestedinhowan individual compares with others of the same species or with members of otherspecies in some arbitrary measure of the scope, complexity, speed, consistency, orotherpropertyofbehavior.Alltheseaspectsmaybequantified,atleastinaroughway,andanyonemayserveasadependentvariableinascientificanalysis.Buttheyarenotall equally productive. In research which emphasizes prediction and control, thetopography of behavior must be carefully specified. Precisely what is the organismdoing? The most important aspect of behavior so described is its probability ofemission.Howlikelyisitthatanorganismwillengageinbehaviorofagivensort,andwhat conditionsor events change this likelihood?Althoughprobabilityof actionhasonlyrecentlybeenexplicitlyrecognizedinbehaviortheory,itisakeyconcepttowhichmanyclassicalnotions,fromreactiontendenciestotheFreudianwish,maybereduced.Experimentallywedealwithitasthefrequencywithwhichanorganismbehavesinagivenwayunderspecifiedcircumstances,andourmethodsaredesignedtosatisfythisrequirement.Frequencyofresponsehasprovedtobearemarkablysensitivevariable,andwithitsaidtheexplorationofcausalfactorshasbeengratifyinglyprofitable.Onedoesnotengageinworkofthissortforthesheerloveofratsorpigeons.Asthe

medical sciences illustrate, the study of animals below the level of man is dictatedmainlybyconvenienceandsafety.Buttheprimaryobjectofinterestisalwaysaman.SuchqualificationsasIhavetoofferinapproachingthepresentquestionspringaboutequallyfromtheexperimentalworkjustmentionedandfromaparallelpreoccupationwithhumanbehavior,inwhichtheprinciplesemergingfromtheexperimentalanalysishavebeen testedandput towork in the interpretationofempirical facts.The formaldisciplinesofgovernment,education,economics,religion,andpsychotherapy,amongothers,togetherwithoureverydayexperiencewithmen,overwhelmuswithafloodoffacts. To interpret these facts with the formulation which emerges from anexperimentalanalysishasprovedtobestrenuousbuthealthfulexercise.Inparticular,the nature and function of verbal behavior have taken on surprisingly fresh andpromisingaspectswhenreformulatedunderthestricturesofsuchaframework.Inthelongrun,ofcourse,mereinterpretationisnotenough.Ifwehaveachieveda

true scientific understanding of man, we should be able to prove this in the actualprediction and control of his behavior. The experimental practices and the conceptsemerging from our research on lower organisms have already been extended in thisdirection,notonlyintheexperimentsonpsychoticsubjectsalreadymentioned,butinother promising areas. The details would take us too far afield, but perhaps I canindicatemy faith in the possibilities in a single instance by hazarding the predictionthatweareonthethresholdofarevolutionarychangeinmethodsofeducation,basednot only upon a better understanding of learning processes, but upon a workableconceptionofknowledgeitself.Whetherornotthisbriefpersonalhistoryseemstoyoutoqualifymetodiscussthe

288

questionbeforeus,thereisnodoubtthatithascreatedahighprobabilitythatIwilldoso, as shown by the fact that I am here. What I have to say is admittedlymethodological.Icanunderstandacertainimpatiencewithsuchdiscussionparticularlywhen, as in the field of psychiatry, many pressing problems call for action. Thescientistwhotakes timeout toconsiderhumannaturewhensomanypractical thingsneedtobedoneforhumanwelfare is likelytobecast in theroleofaNero,fiddlingwhileRomeburns. (It isquitepossible that the fiddling referred to in thisarchetypalmythwasalaterinventionofthehistorians,andthatinactualfactNerohadcalledinhis philosophers and scientists and was discussing “the fundamental nature ofcombustion” or “the epidemiology of conflagration.”) But I should not be here if IbelievedthatwhatIhavetosayisremotefrompracticalconsequences.Ifwearenowenteringaneraofresearchinpsychiatrywhichistobeasextensiveandasproductiveas other types of medical research, then a certain detachment from immediateproblems, a fresh look at human behavior in general, a survey of applicableformulations, and a consideration of relevant methods may prove to be effectivepracticalstepswithsurprisinglyimmediateconsequences.Thestudyofhumanbehavioris,ofcourse,stillinitsinfancy,anditwouldberashto

suppose that anyone can foresee the structure of a well-developed and successfulscience. Certainly no current formulation will seem right fifty years hence. Butalthoughwecannotforeseethefutureclearly,itisnotimpossibletodiscoverinwhatdirectionwearelikelytochange.Thereareobviouslygreatdeficienciesinourpresentwaysofthinkingaboutmen;otherwiseweshouldbemoresuccessful.Whatarethey,andhowaretheytoberemedied?WhatIhavetosayrestsupontheassumptionthatthebehaviorofthepsychoticissimplypartandparcelofhumanbehavior,andthatcertainconsiderations which have been emphasized by the experimental and theoreticalanalysisofbehavioringeneralareworthdiscussinginthisspecialapplication.It is important to remember that I am speaking as an experimental scientist. A

conceptionofhumanbehaviorbasedprimarilyonclinicalinformationandpracticewillundoubtedly differ from a conception emanating from the laboratory. This does notmeanthateitherissuperiortotheother,orthateventuallyacommonformulationwillnotproveusefultoboth.Itispossiblethatquestionswhichhavebeensuggestedbytheexigencies of an experimental analysismaynot seemof first importance to those ofyou who are primarily concerned with human behavior under therapy. But aspsychiatry moves more rapidly into experimental research and as laboratory resultstake on a greater clinical significance, certain problems in the analysis of behaviorshould become common to researcher and therapist alike, and should eventually begivencommonandco-operativesolutions.

Thestudyofbehavior,psychoticorotherwise,remainssecurelyinthecompanyofthenaturalsciencessolongaswetakeasoursubjectmattertheobservableactivityoftheorganism, as it moves about, stands still, seizes objects, pushes and pulls, makessounds,gestures,andsoon.Suitableinstrumentswillpermitustoamplifysmall-scaleactivitiesaspartofthesamesubjectmatter.Watchingapersonbehaveinthiswayislikewatchinganyphysicalorbiologicalsystem.Wealsoremainwithintheframeworkofthenaturalsciencesinexplainingtheseobservationsintermsofexternalforcesandeventswhichactupon theorganism.Someof theseare tobe foundinthehereditaryhistoryof the individual, includinghismembership in agiven species aswell ashispersonalendowment.Othersarisefromthephysicalenvironment,pastorpresent.We

289

mayrepresentthesituationasinFigure1.Ourorganismemitsthebehaviorwearetoaccountfor,asourdependentvariable,attheright.Toexplainthis,weappealtocertainexternal,generallyobservable,andpossiblycontrollablehereditaryandenvironmentalconditions, as indicated at the left. These are the independent variables of whichbehavioristobeexpressedasafunction.Bothinputandoutputofsuchasystemmaybetreatedwiththeaccepteddimensionalsystemsofphysicsandbiology.Acompleteset of such relations would permit us to predict and, insofar as the independentvariablesareunderourcontrol, tomodifyorgeneratebehavioratwill. Itwouldalsopermitus to interpret given instances of behavior by inferring plausible variables ofwhichwe lack direct information. Admittedly the data are subtle and complex, andmanyrelevantconditionsarehardtogetat,but theprogramassuchisanacceptableone from the point of view of scientificmethod.We have no reason to suppose inadvancethatacompleteaccountcannotbesogiven.Wehaveonlytotryandsee.

FIG.1

It is not, however, the subtlety or complexity of this subject matter which isresponsiblefortherelativelyundevelopedstateofsuchascience.Behaviorhasseldombeenanalyzed in thismanner. Instead,attentionhasbeendiverted toactivitieswhichare said to take place within the organism. All sciences tend to fill in causalrelationships,especiallywhentherelatedeventsareseparatedbytimeandspace.Ifamagnetaffectsacompassneedlesomedistanceaway,thescientistattributesthistoa“field”setupbythemagnetandreachingtothecompassneedle.Ifabrickfallsfromachimney,releasingenergywhichwasstoredthere,say,ahundredyearsagowhenthechimneywas built, the result is explained by saying that the brick has all this timepossessed a certain amount of “potential energy.” In order to fill such spatial andtemporalgapsbetweencauseandeffect,naturehas fromtime to timebeenendowedwithmanyweirdproperties,spirits,andessences.Somehaveprovedhelpfulandhavebecome part of the subjectmatter of science, especiallywhen identifiedwith eventsobserved in other ways. Others have proved dangerous and damaging to scientificprogress.Sophisticatedscientistshaveusuallybeenawareof thepracticeandalert toitsdangers.Such inner forceswere, indeed, thehypotheseswhichNewtonrefused tomake.Among the conditionswhich affect behavior, hereditary factors occupy a primary

position,atleastchronologically.Differencesbetweenmembersofdifferentspeciesareseldom, if ever, disputed, but differences between members of the same species,possiblyduetosimilarhereditaryfactors,aresocloselytiedupwithsocialandethicalproblemsthattheyhavebeenthesubjectofseeminglyendlessdebate.Inanyevent,the

290

newly conceived organism begins at once to be influenced by its environment; andwhenitcomesintofullcontactwiththeexternalworld,environmentalforcesassumeamajorrole.Theyaretheonlyconditionswhichcanbechangedsofarastheindividualisconcerned.Among theseare theeventswecall“stimuli,” thevarious interchangesbetween organism and environment such as occur in breathing or eating, the eventswhich generate the changes in behavior we call emotional, and the coincidencesbetweenstimuliorbetweenstimuli andbehavior responsible for thechangeswecalllearning. The effects may be felt immediately or only after the passage of time—perhapsofmanyyears.Sucharethe“causes”—theindependentvariables—intermsofwhichwemayhopetoexplainbehaviorwithintheframeworkofanaturalscience.In many discussions of human behavior, however, these variables are seldom

explicitlymentioned.Theirplaceistakenbyeventsorconditionswithintheorganismforwhichtheyaresaidtoberesponsible(seeFigure2).Thus,thespeciesstatusoftheindividual is dealt with as a set of instincts, not simply as patterns of behaviorcharacteristicofthespecies,butasbiologicaldrives.Asonetextputsit,“instinctsareinnatebiological forces, urges, or impulsionsdriving theorganism to a certain end.”The individual genetic endowment, if not carried by body type or other observablephysicalcharacteristic,isrepresentedintheformofinheritedtraitsorabilities,suchastemperament or intelligence.As to the environmental variables, episodes in the pasthistoryoftheindividualaredealtwithasmemoriesandhabits,whilecertainconditionsofinterchangebetweenorganismandenvironmentarerepresentedasneedsorwants.Certainincitingepisodesaredealtwithasemotions,againinthesensenotofpatternsbut of active causes of behavior. Even the present environment as it affects theorganism is transmuted into “experience,” aswe turn fromwhat is the case towhat“seemstobe”thecasetotheindividual.

FIG.2

291

FIG.3

Thesamecentripetalmovementmaybeobservedon theothersideof thediagram(seeFigure3).Itisraretofindbehaviordealtwithasasubjectmatterinitsownright.Insteaditisregardedasevidenceforamentallife,whichisthentakenastheprimaryobjectofinquiry.Whattheindividualdoes—thetopographyofhisbehavior—istreatedasthefunctioningofoneormorepersonalities.Itisclear,especiallywhenpersonalitiesaremultiple,thattheycannotbeidentifiedwiththebiologicalorganismassuch,butareconceivedof, rather,as innerbehaversofdoubtfulstatusanddimensions.Theactofbehaving in a given instance is neglected in favor of an impulse orwish,while theprobability of such an act is represented as an excitatory tendency or in terms ofpsychic energy.Most important of all, the changes in behavior which represent thefundamental behavioral processes are characterized as mental activities—such asthinking,learning,discriminating,reasoning,symbolizing,projecting,identifying,andrepressing.Therelativelysimpleschemeshowninthefirstfiguredoesnot,therefore,represent

the conception of human behavior characteristic of most current theory. The greatmajorityofstudentsofhumanbehaviorassumethattheyareconcernedwithaseriesofevents indicated in the expanded diagram of Figure 4. Here the hereditary andenvironmental conditions are assumed to generate instincts, needs, emotions,memories, habits, and soon,which in turn lead thepersonality to engage invariousactivities characteristic of the mental apparatus, and these in turn generate theobservable behavior of the organism.All four stages in the diagram are accepted as

292

properobjectsofinquiry.Indeed,farfromleavingtheinnereventstootherspecialistswhileconfiningthemselvestotheendterms,manypsychologistsandpsychiatriststakethementalapparatusastheirprimarysubjectmatter.

FIG.4

Perhaps the point of my title is now becoming clearer. Is the scientific study ofbehavior—whether normal or psychotic—concerned with the behavior of theobservableorganismunderthecontrolofhereditaryandenvironmentalfactors,orwiththefunctioningofoneormorepersonalitiesengagedinavarietyofmentalprocessesunder the promptings of instincts, needs, emotions, memories, and habits? I do notwant to raise the question of the supposed nature of these inner entities. A certainkinship between such an explanatory system and primitive animism can scarcely bemissed,butwhateverthehistoricalsourcesoftheseconcepts,wemayassumethattheyhavebeenpurgedofdualisticconnotations.Ifthisisnotthecase,iftherearethosewhofeel thatpsychiatry isconcernedwithaworldbeyondthatof thepsychobiologicalorbiophysical organism, that consciousor unconsciousmind lacksphysical extent, andthatmentalprocessesdonotaffecttheworldaccordingtothelawsofphysics,thenthefollowing arguments should be all themore cogent. But the issue is not one of thenatureoftheseevents,butoftheirusefulnessandexpedienceinascientificdescription.ItcanscarcelybedeniedthattheexpansionofsubjectmatterrepresentedbyFigure4

hastheunfortunateeffectofalossofphysicalstatus.Thisismorethanaquestionofprestigeor“face.”Asubjectmatterwhichisunquestionablypartofthefieldofphysicsandbiologyhasbeenrelinquishedforoneofdoubtfulcharacteristics.Thiscannotbecorrected merely by asserting our faith in the ultimately physical nature of innerprocesses.Toprotestthattheactivitiesoftheconsciousandunconsciousmindareonlyinsomesenseanaspectofthebiologicalfunctioningoftheorganismwillnotanswerthepracticalquestion.Inabandoningthedimensionalsystemsofphysicsandbiology,we abandon the techniques of measurement which would otherwise be a naturalheritage from earlier achievements in other sciences. This is possibly an irreparableloss.Ifwecomeoutflatlyfortheexistenceofinstincts,needs,memories,andsoon,ontheonehand,andthementalprocessesandfunctionsof thepersonalityon theother,thenwemust accept the responsibility of devisingmethods of observing these inner

293

events and of discovering dimensional systems according to which they can bemeasured. The loss of the opportunity to measure and manipulate in the mannercharacteristic of the physical sciences would be offset only by some extraordinaryadvantagegainedbyturningtoinnerstatesorconditions.It is possible, however, to argue that these inner events are merely ways of

representingtheouter.Manytheoristswillcontendthatahabitisonlyasortofnotationuseful in reporting a bit of the history of the individual, just as so-called “mentalprocesses”arewaysoftalkingaboutchangesinbehavior.Thisisatemptingposition,forwemaytheninsistthattheonlydimensionalsystemsrequiredarethoseappropriatetotheterminalevents.Butifwearetotakethatline,agreatdealstillneedstobedoneto put our house in scientific order. The concepts which one encounters in currentbehaviortheoryrepresenttheobservableeventsinanextremelyconfusingway.Mostof them have arisen from theoretical or practical considerations which have littlereferencetotheirvalidityorusefulnessasscientificconstructs,andtheybearthescarsofsuchahistory.Forexample,Freudpointed to important relationshipsbetween thebehaviorofanadultandcertainepisodesinearlychildhood,buthechosetobridgetheveryconsiderablegapbetweencauseandeffectwithactivitiesorstatesofthementalapparatus. Conscious or unconscious wishes or emotions in the adult represent theearlierepisodesandaresaid tobedirectlyresponsiblefor theireffectuponbehavior.Theadultissaid,forexample,tobesufferingfromconsciousorunconsciousanxietygeneratedwhenasachildhewaspunishedforaggressivebehavior towardasibling.But many details of the early episode are glossed over (and may, as a result, beneglected)inattributingthedisturbancesinhisbehaviortoacurrentanxietyratherthantotheearlierpunishment.Thenumberofreferencestoanxietyintreatisesonbehaviormustgreatlyexceedthenumberofreferencestopunishingepisodes,yetwemustturntothelatterforfulldetails.Ifthedetailsarenotavailable,nothingcantaketheirplace.Otherkindsofindependentvariablesprovidesimilarexamples.Everyoneisfamiliar

with the fact that, in general, organisms eat or do not eat depending upon a recenthistory of deprivation or ingestion. If we can establish that a child does not eat hisdinner because he has recently eaten other food, there may seem to be no harm inexpressingthisbysayingthat“heisnothungry,”providedweexplainthisinturnbypointing to the history of ingestion. But the concept of hunger represents quiteinadequately themany features of schedules of deprivation and other conditions andevents which alter the behavior of eating. In the same way the inner surrogates ofhereditary variables function beyond the line of duty. We often have no otherexplanation of a given bit of behavior than that, like other features of anatomy andphysiology,itischaracteristicofaspecies;butwhenwechooseinsteadtoattributethisbehavior to a set of instincts,we obscure the negative nature of our knowledge andsuggestmoreactivecausesthanmerespeciesstatuswarrants.Similarly,weacceptthefact that individualsdiffer in their behavior, andwemay, in some instances, showarelation between aspects of the behavior of successive generations, but thesedifferences and relationships are optimistically misrepresented when we speak ofhereditary traits and abilities. Again, the term experience incorrectly represents ourinformation about a stimulating field. It has often been observed, for example, thatsometrivialincidentgeneratesareactionaltogetheroutofproportiontoitsmagnitude.Apersonseemstobereacting,nottothephysicalworldassuch,buttowhattheworld“meanstohim.”Eventually,ofcourse,theeffectmustbeexplained—forexample,bypointing to some earlier connection with more important events. But whatever the

294

explanation, it is almost certainly not adequately expressed by the notion of amomentary experience. There are obvious difficulties involved in representing aphysicalenvironmentplusapersonalhistory asacurrentpsychologicalenvironmentalone.So far as our independent variables are concerned, then, the practice we are

examiningtendstoglossovermanyimportantdetailsandcomplexities.Theconceptualstructureconceals fromus the inadequacyofourpresentknowledge.Much thesamedifficulty is encountered with respect to the dependent variable, when observablebehaviortakessecondplacetomentalfunctioningsofapersonality.Justasthephysicalenvironmentistransmutedintoexperience,sophysicalbehaviorcomestobedescribedintermsofitspurposeormeaning.Amanmaywalkdownthestreetinpreciselythesamewayupontwooccasions,althoughinone instancehe isoutforexerciseand inanotherheisgoingtomailaletter.Andsoitisthoughtnecessarytoconsider,notthebehavior itself, but “what it means” to the behaving individual. But the additionalinformationwearetryingtoconveyisnotapropertyofbehaviorbutofanindependentvariable.Thebehaviorweobserveinthetwocasesisthesame.Inreadingmeaningorintentionintoit,wearespeculatingaboutsomeofitscauses.Totakeanotherexample,itiscommonlysaidthatwecan“see”aggression.Butwe“see”itintwosteps:(1)weobserve the behavior of an organism, and (2) we relate it to observed or inferredvariableshavingtodowithinjuriousconsequencesandwiththekindsofcircumstanceswhich make such behavior probable. No behavior is itself aggressive by nature,althoughsomeformsofbehavioraresooftenafunctionofvariableswhichmakethemaggressivethatweareinclinedtooverlooktheinferencesinvolved.Similarly,whenweobserve twoormorebehavioralsystems in thesame individualandattribute themtodifferent personalities, we gain a considerable advantage for certain descriptivepurposes.Forexample,wecanthendescribeoppositionsbetweensuchsystemsaswewouldbetweendifferentpersons.Butwehavealmostcertainlysuggestedaunitywhichis not justified by the observed systems of behavior, andwe have probablymade itmoredifficult to represent the actual extentof anyconflicts aswell as to explain itsorigins. And when we observe that the behavior of a person is characterized by acertain responsiveness or probability of responding and speak instead of a givenamountofpsychicenergy,weneglectmanydetailsof theactual factsanddodge theresponsibility of finding a dimensional system. Lastly, mental processes are almostalwaysconceivedofassimplerandmoreorderlythantheratherchaoticmaterialfromwhichtheyareinferredandwhichtheyareusedtoexplain.The“learningprocess”inexperimental psychology, for example, does not give us an accurate account ofmeasuredchangesinbehavior.Welookinsidetheorganismforasimplersystem,inwhichthecausesofbehavior

arelesscomplexthantheactualhereditaryandenvironmentaleventsandinwhichthebehaviorofapersonalityismoremeaningfulandorderlythantheday-to-dayactivityoftheorganism.Allthevarietyandcomplexityoftheinputinourdiagramseemstobereduced to a few relatively amorphous states, which in turn generate relativelyamorphous functions of the personality, which then suddenly explode into theextraordinaryvarietyandcomplexityofbehavior.But the simplificationachievedbysuchapracticeis,ofcourse,illusory,foritfollowsonlyfromthefactthataone-to-onecorrespondencebetweeninnerandoutereventshasnotbeendemanded.It isjustthislack of correspondence which makes such an inner system unsuitable in theexperimental analysis of behavior. If “hunger” is something which is produced by

295

certainschedulesofdeprivation,certaindrugs,certainstatesofhealth,andsoon,andifin turn it produces changes in theprobability of a great variety of responses, then itmust have very complex properties. It cannot be any simpler than its causes or itseffects. If thebehaviorweobservesimplyexpresses thefunctioningofapersonality,the personality cannot be any simpler than the behavior. If some common learningprocessisresponsibleforthechangesobservedinanumberofdifferentsituations,thenit cannot be any simpler than these changes. The apparent simplicity of the innersystemexplainstheeagernesswithwhichweturntoit,butfromthepointofviewofscientific method it must be regarded as a spurious simplicity, which foreshadowsultimatefailureofsuchanexplanatoryscheme.There is another objection. Although speculation about what goes on within the

organismseemstoshowaconcernforcompletingacausalchain,inpracticeittendstohavetheoppositeeffect.Chainsareleftincomplete.Thelaymancommonlyfeelsthathehasexplainedbehaviorwhenhehasattributedittosomethingintheorganism—asinsaying“Hewentbecausehewantedtogo,”or“Hecouldnotworkbecausehewasworriedabouthishealth.”Suchstatementsmayhavevalueinsuggestingtherelevanceofonesetofcausesasagainstanother,buttheydonotgiveafullexplanationuntilitisexplained why the person wanted to go, or why he was worried. Frequently thisadditionalstepistaken,butperhapsjustasoftentheseincompleteexplanationsbringinquirytoadeadstop.No matter how we may wish to represent such a sequence of causal events, we

cannot satisfy the requirements of interpretation, prediction, or control unlesswe goback toeventsactingupon theorganismfromwithout—events,moreover,whichareobserved as any event is observed in the physical and biological sciences. It is onlycommon sense, therefore, as well as good scientific practice, to make sure that theconceptswhichenter into a theoryofbehavior are explicitly andcarefully related tosuch events.What is needed is an operational definition of terms. Thismeansmorethansimpletranslation.Theoperationalmethodiscommonlymisusedtopatchupandpreserve conceptswhich are cherished for extraneous and irrelevant reasons.Thus itmight be possible to set up acceptable definitions of instincts, needs, emotions,memories,psychicenergy,andsoon,inwhicheachtermwouldbecarefullyrelatedtocertain behavioral and environmental facts. But we have no guarantee that theseconcepts will be themost useful when the actual functional relationships are betterunderstood. A more reasonable program at this stage is to attempt to account forbehavior without appeal to inner explanatory entities. We can do this within theacceptedframeworkofbiology,gainingtherebynotonlyacertainpersonalreassurancefrom the prestige of a well-developed science but an extensive set of experimentalpractices and dimensional systems.We shall be prevented fromoversimplifying andmisrepresentingtheavailablefactsbecauseweshallnottransmuteourdescriptionsintoother terms.Thepracticalcriteriaofpredictionandcontrolwill forceus to take intoaccountthecompletecausalchainineveryinstance.Suchaprogramisnotconcernedwith establishing the existenceof inferred events, butwith assessing the stateofourknowledge.Thisdoesnotmean,ofcourse,thattheorganismisconceivedofasactuallyempty,

or that continuity between input and output will not eventually be established. Thegeneticdevelopmentoftheorganismandthecomplexinterchangesbetweenorganismandenvironmentarethesubjectmattersofappropriatedisciplines.Somedayweshallknow, for example,what happenswhen a stimulus impinges upon the surface of an

296

organism,andwhathappensinsidetheorganismafterthat,inaseriesofstagesthelastofwhich is thepoint atwhich theorganismactsupon the environment andpossiblychangesit.Atthatpointweloseinterestinthiscausalchain.Someday,too,weshallknowhowtheingestionoffoodsetsupaseriesofevents,thelastofwhichtoengageourattentionisareductionintheprobabilityofallbehaviorpreviouslyreinforcedwithsimilar food. Some day we may even know how to bridge the gap between thebehavioralcharacteristicscommontoparentsandoffspring.Butalltheseinnereventswillbeaccountedforwithtechniquesofobservationandmeasurementappropriate tothephysiologyofthevariouspartsoftheorganism,andtheaccountwillbeexpressedintermsappropriatetothatsubjectmatter.Itwouldbearemarkablecoincidenceiftheconcepts now used to refer inferentially to inner eventswere to find a place in thataccount. The task of physiology is not to find hungers, fears, habits, instincts,personalities,psychicenergy,oractsofwilling,attending,repressing,andsoon.Noristhat task to find entities or processes of which all these could be said to be otheraspects.Itstaskistoaccountforthecausalrelationsbetweeninputandoutputwhicharethespecialconcernofascienceofbehavior.Physiologyshouldbeleftfreetodothisinitsownway.Justtotheextentthatcurrentconceptualsystemsfailtorepresenttherelationshipsbetweenterminaleventscorrectly,theymisrepresentthetaskoftheseother disciplines. A comprehensive set of causal relations stated with the greatestpossibleprecisionisthebestcontributionwhichwe,asstudentsofbehavior,canmakein the co-operative venture of giving a full account of the organism as a biologicalsystem.But are we not overlooking one important source of knowledge?What about the

directobservationofmentalactivity?Thebeliefthatthementalapparatusisavailableto direct inspection anticipated the scientific analysis of human behavior by manyhundredsofyears. Itwas refinedby the introspectivepsychologistsat theendof thenineteenthcenturyintoaspecialtheoryofknowledgewhichseemedtoplacethenewlycreated science of consciousness on a par with natural science by arguing that allscientistsnecessarilybeginandendwiththeirownsensationsandthatthepsychologistmerelydealswiththeseinadifferentwayfordifferentpurposes.Thenotionhasbeenrevivedinrecenttheoriesofperception,inwhichithasbeensuggestedthatthestudyofwhat used to be called “optical illusions,” for example,will supply principleswhichhelp inunderstanding the limitsofscientificknowledge. Ithasalsobeenargued thatthe especially intimate empathic understanding which frequently occurs inpsychotherapy supplies a kind of direct knowledge of themental processes of otherpeople.FranzAlexanderandLawrenceKubiehavearguedinthismannerindefenseofpsychoanalytic practices. Among clinical psychologists Carl Rogers has activelydefendedasimilarview.Somethingofthesamenotionmayunderliethebeliefthatthepsychiatristmaybetterunderstandthepsychoticif,throughtheuseoflysergicacid,forexample,hemaytemporarilyexperiencesimilarmentalconditions.Whether theapproach tohumanbehaviorwhich Ihave justoutlined ignores some

basic fact, whether it is unable to take into account the “stubborn fact ofconsciousness,” is part of a venerable dispute which will not be settled here. Twopointsmaybemade,however,inevaluatingtheevidencefromdirect“introspection”ofthementalapparatus.Knowledgeisnottobeidentifiedwithhowthingslooktous,butratherwithwhatwedoaboutthem.Knowledgeispowerbecauseitisaction.Howthesurroundingworldsoaksintothesurfaceofourbodyismerelythefirstchapterofthestory and would be meaningless were it not for the parts which follow. These are

297

concernedwith behavior.Astronomy is not how the heavens look to an astronomer.Atomicphysicsisnotthephysicist’sperceptionofeventswithintheatom,orevenofthemacroscopiceventsfromwhichtheatomicworldisinferred.Scientificknowledgeiswhatpeopledoinpredictingandcontrollingnature.Thesecondpoint is thatknowledgedependsuponapersonalhistory.Philosophers

haveofteninsistedthatwearenotawareofadifferenceuntilitmakesadifference,andexperimentalevidenceisbeginningtoaccumulateinsupportoftheviewthatweshouldprobablynotknowanythingatall ifwewerenotforcedtodoso.Thediscriminativebehavior called knowledge arises only in the presence of certain reinforcingcontingencies among the things known. Thus, we should probably remain blind ifvisual stimuli were never of any importance to us, just as we do not hear all theseparate instruments in a symphony or see all the colors in a painting until it isworthwhileforustodoso.Someinterestingconsequencesfollowwhenthesetwopointsaremadewithrespect

to our knowledge of events within ourselves. That a small part of the universe isenclosedwithintheskinofeachofus,andthatthisconstitutesaprivateworldtowhicheach of us has a special kind of access can scarcely be denied. But the world withwhich we are in contact does not for that reason have any special physical ormetaphysical status.Now, it is presumably necessary to learn to observe or “know”eventswithinthisprivateworldjustaswelearntoobserveor“know”externalevents,andourknowledgewillconsistofdoingsomethingaboutthem.Butthesocietyfromwhichweacquiresuchbehaviorisataspecialdisadvantage.Itiseasytoteachachildto distinguish between colors by presenting different colors and reinforcing hisresponsesasrightorwrongaccordingly,butitismuchmoredifficulttoteachhimtodistinguish between different aches or pains, since the information as towhether hisresponsesarerightorwrongismuchlessreliable.Itisthislimitedaccessibilityoftheworldwithin theskin, rather than itsnature,whichhasbeenresponsibleforsomuchmetaphysicalspeculation.Terms which refer to private events tend to be used inexactly.Most of them are

borrowed in the first place from descriptions of external events. (Almost all thevocabularyofemotion,forexample,hasbeenshowntobemetaphoricalinorigin.)Theconsequences arewell known. The testimony of the individual regarding hismentalprocesses, feelings, needs, and so on, is, as the psychiatrist above all others hasinsisted, unreliable. Technical systems of terms referring to private events seldomresemble each other.Different schools of introspective psychology have emphasizeddifferent features of experience, and the vocabulary of one may occasionally beunintelligibletoanother.Thisisalsotrueofdifferentdynamictheoriesofmentallife.Theexponentofa“system”mayshowextraordinaryconvictioninhisuseoftermsandin his defense of a given set of explanatory entities, but it is usually easy to findsomeone else showing the same conviction and defending a different and possiblyincompatiblesystem.Justasintrospectivepsychologyoncefounditexpedienttotrainobservers in the use of terms referring to mental events, so the education ofexperimental psychologists, educators, applied psychologists, psychotherapists, andmanyothersconcernedwithhumanbehaviorisnotalwaysfreefromacertainelementof indoctrination. Only in this way has it been possible to make sure that mentalprocesseswillbedescribedbytwoormorepeoplewithanyconsistency.Psychiatry itself is responsible for the notion that one need not be aware of the

feelings, thoughts, andsoon,whichare said toaffectbehavior.The individualoften

298

behavesasifhewerethinkingorfeelinginagivenwayalthoughhecannothimselfsaythatheisdoingso.Mentalprocesseswhichdonothavethesupportof thetestimonysupplied by introspection are necessarily defined in terms of, and measured as, thebehavioral facts from which they are inferred. Unfortunately, the notion of mentalactivitywaspreserved in the faceofsuchevidencewith thehelpof thenotionofanunconsciousmind.Itmighthavebeenbettertodismisstheconceptofmindaltogetherasanexplanatoryfictionwhichhadnotsurvivedacrucialtest.Themodesofinferencewith which we arrive at knowledge of the unconscious need to be examined withrespect to the consciousmind as well. Both are conceptual entities, the relations ofwhichtoobserveddataneedtobecarefullyreexamined.In the long run the point will not be established by argument, but by the

effectivenessofagivenformulationinthedesignofproductiveresearch.Anexampleofresearchonpsychoticsubjectswhichemphasizestheendtermsinourdiagramistheprojectalreadymentioned.Thisisnottheplacefortechnicaldetails,1buttherationaleof this researchmaybe relevant. In these experiments a patient spends one ormorehoursdaily,alone,inasmallpleasantroom.Heisnevercoercedintogoingthereandisfree to leave at any time.The room is furnishedwith a chair, and contains a devicesimilartoavendingmachine,whichcanbeoperatedbypushingabuttonorpullingaplunger. The machine delivers candies, cigarettes, or substantial food, or projectscoloredpicturesonatranslucentscreen.Mostpatientseventuallyoperatethemachine,are “reinforced” by what it delivers, and then continue to operate it daily for longperiodsoftime—possiblyayearormore.Duringthistimethebehaviorisreinforcedon various “schedules”—for example, once every minute or once for every thirtyresponses—inrelationtovariousstimuli.Thebehaviorisrecordedinanotherroominacontinuouscurvewhichisreadsomewhat in themannerofanelectrocardiogramandwhichpermitsareadyinspectionandmeasurementoftherateofresponding.Theisolationofthissmalllivingspaceis,ofcourse,notcomplete.Thepatientdoes

notleavehispersonalhistorybehindasheenterstheroom,andtosomeextentwhathedoesthereresembleswhathedoesorhasdoneelsewhere.Nevertheless,astimegoeson,theconditionsarrangedbytheexperimentbegintocompose,sotospeak,aspecialpersonal history, the important details of which are known. Within this small andadmittedly artificial life space, we can watch the patient’s behavior change as wechange conditions of reinforcement, motivation, and to some extent emotion. Withrespect to these variables the behavior becomes more and more predictable andcontrollable or—as characteristic of the psychotic subject—fails to do so in specificways.The behavior of the patientmay resemble that of a normal human or infrahuman

subject in response to similar experimental conditions, or it may differ in a simplequantitativeway—forexample,therecordmaybenormalexceptforalowerover-allrate.Ontheotherhand,aperformancemaybebrokenbybriefpsychoticepisodes.Theexperimental control is interrupted momentarily by the intrusion of extraneousbehavior. In somecases it hasbeenpossible to reduceor increase the time takenbythese interruptions, and todeterminewhereduring the session theywilloccur.As insimilar work with other organisms, this quantitative and continuous account of thebehavior of the individual under experimental control provides a highly sensitivebaselinefortheobservationoftheeffectsofdrugsandofvariousformsoftherapy.Forourpresentpurposes,however,theimportantthingisthatitpermitsustoapplytothe

299

psychotica fairly rigorous formulationofbehaviorbaseduponmuchmoreextensiveworkunderthemuchmorepropitiouscontrolofconditionsobtainedwithotherspecies.Thisformulationisexpressedin termsof inputandoutputwithoutreferenceto innerstates.The objection is sometimes raised that research of this sort reduces the human

subject to the status of a research animal. Increasing evidence of the lawfulness ofhuman behavior only seems to make the objection all the more cogent. Medicalresearchhasmetthisproblembefore,andhasfoundananswerwhichisavailablehere.Thanks to parallel work on animals, it has been possible, in some cases at least, togeneratehealthierbehavior inmen,even thoughat thisstagewemaynotbedirectlyconcernedwithsucharesult.Another common objection is that we obtain our results only through an

oversimplificationofconditions,andthattheyarethereforenotapplicabletodailylife.But one always simplifies at the start of an experiment.We have already begun tomake our conditions more complex and will proceed to do so as rapidly as theuniformityofresultspermits.Itispossibletocomplicatethetaskofthepatientwithoutlimit,andtoconstructnotonlycomplexintellectualtasksbutsuchinteractionsbeweensystemsofbehaviorasareseenintheFreudiandynamisms.Onesimplificationsometimescomplainedofistheabsenceofotherhumanbeingsin

this small life space.Thiswas, of course, a deliberate preliminarymeasure, for it ismuchmoredifficult tocontrol social thanmechanical stimulationand reinforcement.Butwearenowmovingontosituationsinwhichonepatientobservesthebehaviorofanother working on a similar device, or observes that the other patient receives areinforcementwheneverheachievesonehimself,andsoon.Inanothercasethepatientis reinforced only when his behavior corresponds in some way to the behavior ofanother.Techniquesforachievingextraordinarilyprecisecompetitionandco-operationbetweentwoormoreindividualshavealreadybeenworkedoutwithlowerorganisms,andareapplicabletothepresentcircumstances.Thisprojecthas,ofcourse,barelyscratchedthesurfaceofthesubjectofpsychotic

behavior.Butso faras ithasgone, it seems tous tohavedemonstrated thevalueofholdingto theobservabledata.Whetherornotyouwillall find themsignificant, thedatawereporthaveaspecialkindofsimpleobjectivity.Atleastwecansaythatthisiswhatapsychoticsubjectdidunderthesecircumstances,andthatthisiswhathefailedtodoundercircumstanceswhichwouldhaveahadadifferenteffecthadhenotbeenpsychotic.Althoughwehavebeenabletodescribeandinterpretthebehaviorobservedinthese

experiments without reference to inner events, such references, are of course, notinterdicted.Othersmayprefer tosaythatwhatweareactuallydoingismanipulatinghabits,needs,andsoon,andobservingchangesinthestructureofthepersonality, inthestrengthoftheego,intheamountofpsychicenergyavailable,andsoon.Buttheadvantage of this over a more parsimonious description becomes more difficult todemonstrateasevidenceoftheeffectivenessofanobjectiveformulationaccumulates.In that bright future to which research in psychiatry is now pointing, we must beprepared for the possibility that increasing emphasis will be placed on immediatelyobservable data and that theories of human behavior will have to adjust themselvesaccordingly.Itisnotinconceivablethatthementalapparatusandallthatitimplieswillbeforgotten.Itwillthenbemorethanamereworkinghypothesistosay—toreturnatlonglasttomytitle—thatpsychoticbehavior,likeallbehavior,ispartoftheworldof

300

observableevents towhich thepowerfulmethodsofnaturalscienceapplyand to theunderstandingofwhichtheywillproveadequate.

1Dr.HarrySolomonoftheBostonPsychopathicHospitalhasservedasco-directoroftheproject,althoughtheprecedingargumentsdonotnecessarilyrepresenthisviews.Dr.OgdenR.Lindsleyisinimmediatechargeandresponsibleformuchoftheoverallexperimentaldesignaswellasfortheactualday-to-dayconductoftheexperiments.SupporthasbeenprovidedbytheOfficeofNavalResearchandbytheNationalInstituteofMentalHealth.TheworkisbeingcarriedoutattheMetropolitanStateHospitalinWaltham,Massachusetts,withthecooperationofDr.WilliamMcLaughlin,Superintendent,andDr.MeyerAsakoff,DirectorofResearch.[Seereferenceonpage158.]

FromTheoryandTreatmentofthePsychoses:SomeNewerAspects.F.Glidea,editor.WashingtonUniversityStudies,1956.TheresearchdescribedattheendofthisarticlewascarriedoutattheMetropolitanStateHospitalatWaltham,Massachusetts.Dr.HarrySolomon,thenoftheBostonPsychopathicHospital,collaboratedwiththeauthorinsettingupthelaboratory.Dr.OgdenR.Lindsleytookimmediatechargeandwasresponsibleformuchoftheoverallexperimentaldesignandtheday-to-dayconductoftheexperiments.

301

SomeRelationsBetweenBehaviorModificationandBasicResearch

Thefirstscientificlawswereprobablytherulesofcraftsmen.Inotherwords,scienceseemstohaveemergedfromeffortstosolvepracticalproblems.Thereareexamplesofthis in psychology. Binet’s invention of a way to place children in schools led tointelligence testing, and the problem of selecting personnel in industry led to othermeasures in what later became the psychology of individual differences. ClinicalproblemsledFreudtoinventthementalapparatusofpsychoanalytictheory.Eventhephysiological psychology of Fechner andWundt arose from efforts to solve certainphilosophicalproblemsconcerning thenatureofknowledgeand therelationbetweenmindandbody.Asscienceadvances,however,thedirectionchanges.Subjectmattersbecome toocomplex toyield to laywisdomor rulesof thumb,and it is the scientistwhoseestheusefulimplicationsofhisdiscoveries.Mostimprovementsintechnologynow come from what is essentially basic research. Behavior modification is anexample.Itsoriginslayinarelatively“pure”experimentalanalysis.Theshapingofbehaviorthroughastep-by-stepapproximationtoafinaltopography

isacaseinpoint.Asadistinctionbegantobemadebetweenoperantandrespondentconditioning, or between emitted and elicited behavior, the role of the controllingstimuluschanged,andproblemsarose.Theexperimenterhadnocontroloverthefirstinstanceofanoperantresponse;hecouldonlywaitforittoappear.Itseemedtofollowthatcomplexformsofbehaviorwhichseldomifeveroccurredwereessentiallyoutofreachofoperantconditioning.Butthepossibilityremainedthatthecomponentpartsofacomplexpatternmightoccurseparatelyandthatlargerpatternsmightbeconstructedbitbybit.Anearlyefforttoshapecomplexbehaviorwassuggestedbyanexperimentinwhich

achimpanzeeusedpokerchipstooperateasortofvendingmachine.1Itwasimpliedbysomewriters that indoingso thechimpanzeedemonstratedcertainhighermentalprocesses—inwhich,forexample,thepokerchipservedasymbolicfunction.Toshowthatoperantconditioningcouldexplainthebehavior,Iconditionedarattobehaveinasimilarway.AsoriginallyreportedinTheBehaviorofOrganisms,2“everystepintheprocesshadtobeworkedoutthroughaseriesofapproximations,sincethecomponentresponseswerenotintheoriginalrepertoireoftherat.”Followingthepracticeinearlyexperimentsonoperantconditioning,andinotherexperimentsonlearningatthattime,I made reinforcers contingent on behavior by constructing mechanical systems. Tobeginwith,anumberofmarbleswerestrewnontheflooroftherat’scage.Thesidesofthecagewereclearofthefloorsothatamarblecouldrolloffanyedge.Wheneveramarbledroppedoffthefloor,ittrippedaswitchandoperatedafooddispenser.Whenresponseswhich causedmarbles to roll about the floor becamemore frequent, threeedges of the floor were blocked so that amarble could fall off only one edge, andmovementswhich causedmarbles tomove in other directions underwent extinction.

302

Later,onlyasmallsectionofoneedgewasleftclear,andtheedgewasthenslightlyraised.Bythis timetheratwaspushingthemarbleswithitsforepawsornose,andastrong directed response would force a marble up and over the edge and would bereinforced.Eventuallytheratgraspedandliftedthemarbleovertheedge.Anewproblem then arose.Evidently it is hard for a rat to acquire theoperant of

“letting go.” It lets go of an object when it is either punished or not reinforced forholdingon,butapparentlyitlearnsonlywithdifficultytodoso“toproduceaneffect.”Theratwouldholdthemarbleovertheedge,showingaconsiderabletremorbecauseofthe unnatural posture, and eventually themarblewould fall from its hands. (At onestage, it learned to knock themarble out of its pawswith its nose, often shooting itsomedistance,andasortofbackstophadtobebuilttomakesurethemarblewouldnotmiss the triggeringplatformbelow.)Eventuallyanappropriate responseof lettinggoappeared.Theedgeandthebackstopwerethenchangedtoformaslot,whichgrewsmallerin

cross section and taller until itwas eventually a tube less than one inch in diameterextending two inches above the floor.A rackwas then added fromwhich amarblecouldbeobtainedbypullingachain thathung from the topof thecage. In the finalperformancetheratwouldpullthechain,pickupthemarblethusdischarged,walkonitshindlegsacrossthecage,liftthemarbletremblinglytothetopofthetube,anddropitin.Thecuriousthingaboutthisexperiment,seeninretrospect,isthateverystepinthe

shapingprocesswastakenbychangingtheequipment.Mechanicalcontingencieshavetheadvantagethattheyworkintheabsenceoftheexperimenter,andinspiteofmuchevidencetothecontraryImayhavebeenundertheinfluenceofearlylearningtheoriesinwhichitwasassumedthatagreatmanytrialswereneededtoteacharatanything.(Imadethesamemistakewithmystudents;Iassumedthattheywouldalwaysneedtogoovermaterialmore than once, and the first teachingmachine designed for their useallowedforthepresentationofeachitemuntiltwocorrectresponseshadbeenmade.3Iwellremember thedaywhenNormanGuttman,KellerBreland,andIdiscovered

howwrongall thiswasbydispensingwithmechanicalcontingenciesandreinforcingsuccessive approximations to a complex response by hand. By operating a fooddispenserwithahandswitchwetaughtapigeontostrikeawoodenballwithaswipingmotionofitsbeakandtoknocktheballaboutitscage,ratherinthemannerofasquashplayer.Shapingcomplexbehaviorthroughaprogrammedsequenceofcontingenciesisnowsocommonplacethatitishardtounderstandwhyweshouldhavebeenamazedatthespeedwithwhichthiswasdone.

Anotherusefulprinciplediscoveredinbasicresearchhastodowiththemaintenanceofbehavior.Early experiments on learning concentratedon certain conspicuous facts—the acquisition of new forms of behavior (in learning) and the disappearance of oldones (in forgetting). “Overlearning”was studied,but therewasno recognitionof thefact that behavior could be maintained in various states of strength by appropriateschedules of reinforcement. Schedules are now widely used to solve what areessentially practical problems—for example, in generating baselines to show drugeffectsandemotionalchangessuchastheconditionedsuppressioncalled“anxiety”—buttheimplicationsforbehaviormodificationhavenotbeenfullyexplored.Althoughthemaintenanceofavailablebehaviorinstrengththroughintermittentreinforcementis

303

animportantpracticalproblem,theemphasisisstillontheproductionofnewbehavior.But this will change as practical implications are discovered. The behaviormodification of the future will also make a far more extensive use of the controlexertedbythecurrentenvironment,ofdeprivationandsatiation,oftheconditioningofnew reinforcers, and of the explicit design of instructional repertoires (of whichimitationisanexample).Not all innovations in behavior modification are traceable to a basic analysis, of

course,andtraditionalrulesofthumbforshapingandmaintainingbehaviorwhichusethe same principles arose long before any research had been done. The analysis isneverthelessimportantininterpretingandexplainingtheeffectofamethod,whateverits provenance. It is hard to see the contingencies of reinforcementwhich prevail indaily life and hence to understand the behavior they generate.4 Laboratory researchtellsuswhattolookforand,equallyimportant,whattoignore,andindoingsoitleadstotheimprovementofpracticalcontingencies.It isnotalwayseasy tomaintainadistinctionbetweenbasicandapplied research,

sincebothmayusethesamemethodsandequipmentandcomeupwiththesamekindsof results,butan importantdifference lies in the reasonswhy research isundertakenand supported. The applied researcher is under the influence of a special kind ofconsequence. He carries on, in part, because he will make someone healthier orwealthierratherthansimplywiser.Thisisnotareflectiononhis“purity”oranyothertrait of character. It is simply a fact which bears on certain characteristics of hispractice,andforthatreasonsomefurtherrelationsbetweenappliedandbasicresearchareworthconsidering.Oneobviousvalueofpracticalresultsisillustratedbythehistoryoftheexperimental

analysis of behavior. Operant research was not at first generously supported bydepartmentsofpsychology,andmanyyoungresearcherstookjobsinlaboratories—inpharmaceutical companies, for example—whichwere supported because of practicalconsequences. Itwasnot alwaysnecessary todesign researchunder the influenceofsuchconsequences,however,andmanyimportantbasicexperimentswerecarriedout.Evenwhen thepractical consequenceswere stressed, therewasoften again inbasicknowledge.Behaviormodificationsuppliesanotherexample.Itspracticalimplicationshaveled

to thesupportofworkfromwhichcontributions tobasicknowledgearebeingmade.Practical consequences often force the scientist to deal with variables he wouldotherwise put off for later consideration. To improve the lives of institutionalizedretardatesorpsychotics,forexample,itisusuallynecessarytodealwithmorevariablesthan convenience dictates, and the simplifying practices of the laboratory are notfeasible.Asaresult,discoveriesaremadewhichstandardexperimentalpracticewould,inasense,putoutofreach.Acontributionintheotherdirection—frombasictoapplied—wouldtraditionallybe

describedas the“confidence”withwhichcontingenciesarenowdesigned in solvingpractical problems. Laboratory successes generalize to daily life. The effects ofreinforcementareoftendeferredandneedtobemediated,andthisisparticularlytruewhen reinforcement is used in place of punishment because the latter has quickereffects. The amenable conditions of the laboratory are likely to bring the researcherunderthecontrolofdeferredconsequencesandtomaintainhisbehaviorwhenitisonlyintermittently reinforced.Apracticalmethodmaycontinue tobeusedbecauseof its

304

success, but help from basic research is often needed in its early development.Techniques of behavior modification often seem, after the fact, like the plainest ofcommon sense, butwe should remember that they remainedundiscoveredorunusedforalongtimeandthatthesame“commonsense”stillleadstomanyviolationsofthebasicprinciples.The theorywhich accompanies an experimental analysis is particularly helpful in

justifyingpracticebecausebehaviormodificationoftenmeansavastchangeinthewayinwhichwedealwithpeople.Itisvastnotonlyinscope(touchingfieldsasdiverseaseducation, psychotherapy, economics, and government) but in its very nature, as thestatesofmind,feelings,andotherattributesoftheinnermanwhofiguresintraditionalexplanationsofhumanbehaviorarerejectedinfavorofantecedentcircumstancesinaperson’sgeneticandindividualhistories.Thegenetichistoryisatthemomentbeyondcontrol, but the environmental history, past and present, can be supplemented andchanged, and that is what is done in a genuine technology of behavior. Behaviormodificationisenvironmentmodification,butthisisnotwidelyrecognized.Verylittlecurrent “behavioral science” is really behavioral, because prescientific modes ofexplanationstillflourish,butbehaviormodificationisanoutstandingexception.One practical contribution of an experimental analysis is easily overlooked. It is

derived from the scientific methodology upon which the analysis is based. Appliedpsychologists have usually found themselves in a subordinate position. They haveserved as fact-finders rather than decision-makers. More than any other science,psychologyhashadtomoveagainstaweightoffolklore,superstition,anderror,anditisnot surprising thatpsychologistshaveputapremiumon the factualandobjective.Theyhave struggled to escape from the limitationsofpersonal experience.Toknowwhatamanactuallyhearsorseestheyhavecontrolledthestimulatingenvironment.Toknowwhathedoesorsaystheyhaverecordedhisbehavioraspreciselyaspossibleandquantifieditwithinventories,questionnaires,andtests.Todiscoverwhatheisinclinedtodoorsaytheyhavesampledhisopinionsandbeliefs.This“objectivity”hasnodoubtbeenvaluable,but ithascast thepsychologist ina

merely supportive role. Clinical psychologists and psychometricians find themselvestestingpatientsandsupplying the information to thepsychiatristwhocarriesonwiththerapy.Theschoolpsychologistreportstoateacheroradministrator,whoistheonewho takes appropriate action. The results of opinion polls are used by statesmen orpoliticians,notbythepollsters,anditisthedirectorswhoplanthefutureofacompanyin the lightofamarketanalysis.Whenapsychologistmoves intoadecision-makingspot,itisgenerallyfeltthatheisnolongeractingasapsychologist.Somethingofthesamesortholdsforall thebehavioralsciences,and therelationshiphas, infact,beengivensemi-official status.An“AdvisoryCommitteeonGovernmentPrograms in theBehavioral Sciences,” appointed by the National Academy of Sciences and theNational Research Council in 1965, has been careful to point out that “there is noassumption… that knowledge is a substitute for wisdom or common sense or fordecision-making.”5The wisdom, common sense, or decision-making faculty thus denied to social

scientistsisprobablyaremnantofaprescientificautonomousman.Itisnotsupportedby the conceptionofhumanbehaviorwhich emerges froman experimental analysis,and this fact explains the unique characteristic of behavior modification that it isdirectly concerned with decision-making and control. When we have specified the

305

goalstowhich“wisdomandcommonsense”aretobeapplied,wemaygostraighttothe design of relevant contingencies. The experimental analysis of behavior ismorethanmeasurement.Itismorethantestinghypotheses.Itisanempiricalattackuponthemanipulable variables of which behavior is a function. As a result the behaviormodifierplaysanexceptionalrole.Evenso,hemayfinditdifficulttomoveintoapositionwhereimportantdecisions

aremade.Changesinpracticedonotusuallycomeaboutinthecenterofpower.Itisnot the administrator or teacher who is likely to make a major contribution toinstruction,butratherthosewhoanalyzeteachingandlearningexperimentally.Itisnotthe employer orworkerwho is likely to propose better incentive systems but rathersomeonewhohasexamined theeffectsof reinforcersonhumanbehavior ingeneral.Wehavebecomeusedtothisfactinmedicine,wheremajorimprovementscomefromsciencesperipheraltomedicineitselfratherthanfrompracticingphysicians.Butunlikethe physician, who has learned to accept new practices originating elsewhere, theadministrator,teacher,psychiatrist,orindustrialistmayfeelthathehasaspecialkindofwisdomarising from familiaritywithhis field andmay resist the applicationof abasic science or a new method closely associated with such a science. The currentcondition of American elementary and high school education is a tragic example.Prevailingpracticesarederivedfromunscientific“philosophiesofeducation”andfromthepersonalexperiencesofadministratorsandteachers,andtheresultsareparticularlydisturbing to thosewhoknowwhatmightbedone instead.The first behaviorwhichneedstobemodifiedisobviouslythatoftheteacher,administrator,orphilosopherofeducation.Another practical consequence of basic research remains to be emphasized. Our

culture has made us all sensitive to the good of others, and we are generouslyreinforcedwhenweactfortheirgood,butthedisplayofgratitudewhichreinforcestheteacher or therapist who is in immediate contact with another person is oftendangerous.Thosewhoareespeciallysensitivetothegoodofothersareofteninducedtogointoteachingortherapyratherthanbasicresearch.Progresswouldbemorerapidif the same kind of reinforcement could be brought to bear on the researcher, if hecouldbeappropriatelyaffectedbytheextraordinaryextent towhichhe isalsoactingfor good of others. The basic researcher has, in fact, a tremendous advantage. Anyslight advance in our understanding of human behavior which leads to improvedpractices in behavior modification will eventually work for the good of billions ofpeople.

FromCumulativeRecord(3rded.).NewYork:Appleton-Century-Crofts,1972,pp.276–282.

1Cowles,J.T.Comp.Psychol.Monogr.,1937,14,No.5.2Skinner,B.F.BehaviorofOrganisms.NewYork:Appleton-Century-Crofts,1938.3Skinner,B.F.TheTechnologyofTeaching.NewYork:Appleton-Century-Crofts,1968.Seepage36,Figure4.

4Skinner,B.F.ContingenciesofReinforcement:ATheoreticalAnalysis.NewYork:Appleton-Century-Crofts,1969.

5QuotedinareviewbyAlbertD.Biderman(Science,1970,169,1064–1067)ofthe

306

reportoftheYoungCommittee,appointedbyNAS-NRCin1965as“AdvisoryCommitteeonGovernmentProgramsintheBehaviorSciences.”ThereportwaspublishedunderthetitlePoliticsofSocialResearch,byRalphL.Beals(Chicago:Aldine,1969).

307

CompassionandEthicsintheCareoftheRetardate

Weallknowwhatitmeanstofeelthatsomethingisrightorwrong,orgoodorbad,butdowebehaveethicallybecausewedowhatwefeelisrightorgood,oriswhatwefeelamereby-product?Ifourfeelingsdonotexplainourbehavior,someotherexplanationforrightorwrongorgoodorbadconductmustbefound.Ethicalandmoralprinciplesmustthenbederivedfromthereasonswhypeoplebehave(andonlyincidentallyfeel)astheydo.Thepointis importantbecausewecanchangereasonsmoredirectlythanfeelings,andareformulationmaythereforeimproveourchancesofinducingpeopletobehaveethically.Prominent among the reasons why a person behaves in a given way are the

consequences which follow when he does so. Relevant consequences are usuallydescribedintermsoffeelings.Thus,apersonissaidtobehaveinagivenwaybecausethe effect is “satisfying” or “pleasant” or because he thus avoids an effectwhich is“annoying”or“unpleasant.”(Itisclearthatwearealreadyclosetoaquestionofethics,becausewecallonekindofconsequence“good”andsaythatbehaviorfollowedbyitis“right,”andanotherkind“bad”and thatbehavior followedby it “wrong.”Theseareelemental value judgments.)The laymen calls such consequences rewarding, but theword“reinforcing”isanimprovement.Itemphasizesthefactthatreinforcedbehaviorisstrengthenedinthesensethatitbecomesmorelikelytoberepeated.Thereinforcingorstrengtheningeffectshavebeenexhaustivelystudiedintheexperimentalanalysisofoperantbehavior.We use the terms good and bad, or right and wrong, even when speaking of a

solitarypersonwhosebehaviordoesnotaffectothers.Therearemanynaturalwaysinwhichthehumanorganismisreinforced:foodreinforcesthehungry,waterthethirsty,sexualcontactthesexuallydeprived,andescapefrompainthosewhoareinpain.Whythis isso isaquestionabout theevolutionof thespecies;organismshavebeenmorelikely tosurvivewhensuchconsequenceshavestrengthened thebehaviors theywerecontingent upon.But these natural values do not explainmuch of the behavior of acivilizedperson.Nonsocialreinforcementproducesatbestonlyaferalchild—achildwho has grown up untouched by human care—or Rousseau’s noble savage,uncorruptedbycivilization,perhaps,butnotreallyverynobleandstillasavage.Peoplelivingtogetheringroupsaresensitivetootherkindsofvaluesbecausethey

are subject tootherkindsof consequences.Theirbehavior is shapedandmaintainedbecauseitseffectsaregoodorbadforothers.Wecall thebehaviorofanotherpersongoodorbadaccordingtoitseffectuponus,andwealsotakemorepositiveaction;wereinforcesuchbehaviorwhenitsconsequencesarereinforcingtous(thusinducingthepersontoactinthesamewayagain)andpunishitwhenitispunishingtous(sothatthe personwill henceforth avoid acting in thatway again). Thewords “Good!” and“Bad!” eventually become social reinforcers in their own right. Comparable socialcontingencies are implied by the concepts of duty and obligation.We are likely tospeak of a “sense of duty” or a “feeling of obligation,” but the basic facts concern

308

socialreinforcement.Thetermsrefertowhatisduetoothers(whomaydemandtheirdue)andtowhatothersobligeustodo.Abehavioristicreformulationdoesnotignorefeelings;itmerelyshiftstheemphasis

fromthefeeling towhat is felt.Apersonresponds to thephysicalworldaroundhimand,witharatherdifferentsetofnerves,tothenolessphysicalworldwithinhisskin.Whathefeelsishisownbody,andamongthethingshefeelsishisownbehaviorasithasbeenaffectedby itsconsequences.Hemayfeel itas rightorwrong.Whenhe iswalkingfromoneplacetoanotherinacity,hemayfeelthatheisorisnotontherightstreet.Hetakesonestreetoranotherdependinguponhowwellheknowsthecityor,inother words, how well his behavior has been shaped and maintained by its earlierconsequences.Hedoesnottakeaparticularstreetbecausehefeelsitisright;hefeelsitis right because he is inclined to take it, and he is so inclined because he has beenreinforcedfordoingso.Similarly,indealingwithotherpeoplehefeelsthatheisorisnotdoingtherightthing,buthedoesnotdoitbecausehefeelsitisright;hefeelsitisright because it is what he tends to do, and he tends to do it because it has hadconsequenceswhichwerereinforcingfirsttoothersandtheninturntohim.Whensomeonemistreatsus,wemayfeelangryorenraged,butifwecallhimbad

andhisbehaviorwrongortakemoreeffectiveaction,itisnotbecauseofourfeelingsbut because of the mistreatment. But why do we protest mistreatment when weourselvesarenotdirectlyinvolved?Again,thetraditionalexplanationissoughtintheworldoffeeling:wearesaidtorespondtothemistreatmentofothersoutofsympathyorcompassion,intheetymologicalsenseofsufferingwiththemistreatedperson.Ifthatweretrue,wecouldscarcelydenythatfeelingscanserveasthecausesofaction.Butthe explanation is not satisfactory. Compassion will not explain action until thecompassion has in turn been explained.Why we act in the interest of others is nohardertoexplainthanwhywefeeltheirfeelings.Tosaythatitishumannaturetobecompassionateistoappealtothehumangenetic

endowment,whichmusteventuallybeexplainedbyshowing that ithashadsurvivalvalue for the species. A tendency to feel compassionate would contribute to thesurvivalofthespeciesifitinducedpeopletoprotectandhelpeachother,butitisthebehavior of protecting and helping others which is selected by the contingencies ofsurvival.Wedonotsaythatafemalerathasaninnatetendencytofeelcompassionateandthatthefeelingtheninduceshertocareforherhelplessyoung;insteadwespeakofthesurvivalvalueofmaternalbehavior.Thehumanmotherwillfeeltheconditionofherbodyasshecaresforherchild(ifherculturehas inducedher tofeel it),andshemaycallitcompassion(ifherculturehastaughthertodoso),butwhatshefeelsisaby-product,notthecauseofherbehavior.Whenpeoplecallitwrongtomistreatananimalorchildorhelplesspersonorwhen

theytakemorepositiveaction,theymayfeelaccompanyingconditionsoftheirbodies,which they have learned to call, say, anger, but it is their behavior which is to beexplained.Theprocessofgeneralizationmayberelevant.Whatweseewhenapersonismistreatingothers resembleswhatwe seewhenhe ismistreatingus.Thephysicalaspects ofmistreatmentbecome aversive, and we may act to escape or attack eventhough the original reasons no longer prevail.We kill themosquito on our arm andthrough generalization on the arm of a friend. We attack those who attack us andthrough generalization those who attack others. Verbal practices offer support. Aculture teaches itsmembers to call thosewhomistreatothersbadand their behaviorwrong. It establishes rules fordealingwith themistreatmentofothers andmaintains

309

appropriatesanctionstoinducepeopletofollowthem.Thereareotherreasonswhyweprotestmistreatment,andtheyexplainaswelltheconditionwefeelascompassionorsympathywhenwedoso.Such an explanationmay not bewholly convincing to thosewho insist upon the

priorityoffeelings,butitmustberememberedthatitwouldbeatleastequallydifficulttoexplainthefeelings.Moreover,noverypowerfulexplanationisneeded,becausethebehavioritselfislikelytobeweak.Theactionwetakewithrespecttothemistreatmentofothersislessvigorousthanwhenweourselvesaremistreated,andourfeelingsarecorrespondingly weaker. The point has an important bearing on some classicalexamplesofmistreatment inwhichanycountercontrollingactionmustbe takenbyathirdpartywhoisnotdirectlyaffectedandisthereforelessinclinedtoactandtohavethefeelingsassociatedwithaction.Consider, first, the care of small children in orphanages or in schools to which

children are sent by people who soon lose contact with them. Charles Dickensdescribed typical examples of the abuse which has been characteristic of suchinstitutionsthroughouttheirlonghistory,andhewasresponsibleforacertainmeasureofreform,butthereisstillmuchtobedone.Thetroublearisesbecausethosewhoexertcontrol are subject to littleornocountercontrol.Children are too small andweak toprotest and can only occasionally turn to that archetypal response to mistreatment,escape. In the absenceof effective countercontrol those in chargebegin to change—from being perhapsmerely careless to being callous or even cruel. In the end theyalmostnecessarilybehaveinwayswecallethicallywrong.Theydonotdosobecausethey lack compassion, but because reciprocal action has beenweak or absent.Whattheymayatfirsthavefeltascompassiongrowsweakorvanishesaltogether.The care of the chronically ill or the aged is subject to the same deterioration.

Nursing homes and homes for the elderly are often supported by the state or byrelativeswho take little further interest inwhathappens in them.The illor theagedcannot protest effectively, nor can they escape, and little or no countercontrol isthereforeexerted.Undertheseconditionsthoseinchargebegintobehaveinwayswecallethicallywrong.Theydosonotbecausetheylackcompassionbutbecausethereisnoadequatecountercontrol,andastheirbehaviordeteriorates,nothingislefttobefeltascompassion.Prisons offer a rather different example. A prison is itself a device for the

countercontrol of unlawful behavior, and its practices are at the start almost alwayspunitive.Thosewhoestablishprisonsandincarceratepeopleinthemusuallytakelittlefurther interest in what happens, and prisons have therefore been notorious forinhumane treatment. The prisoner escapes if he can, but no other action is possibleshortofextremeviolence,whichtendstobesuppressedbyevenmoreviolentmeans.Itisusuallyobviousthatthemistreatmentofaprisonerisduetoinabilitytoprotest,andwearenotlikelytolooktothecompassionofprisonauthoritiesforreform.The care of psychotics has also shown a long history of inhumane treatment. It

wouldbedifficultforanysanepersontoavoidbecomingpsychoticundermanyformsofinstitutionalcare.Theinstitutionsareusuallyrunbythestate,andthosewhocommitpatients to them often lose interest or accept the care they provide as somehowinevitable. The psychotic cannot protest effectively and, if he escapes, he is usuallyeasily caught. Counter-control is negligible, and the behavior of those in chargethereforetendstobecomecareless,callous,orcruel.Feelingschangeaccordingly.Weareespeciallyconcernedherewitha fifth field inwhich therehasalsobeena

310

longhistoryofmistreatment,thecareoftheretardate.Itwouldbedifficultforanormalchildtodevelopnormallyundermanyformsofinstitutionalcare.Likethepsychotic,theretardatecannotescapeorprotesteffectively.Thoseinchargeeasilyfallintomodesofconductwecallethicallywrong,andtheirfeelingspresumablychangeaccordingly.Fortunately,therearethosewhoareinclinedtodosomethingaboutthemistreatment

ofchildren,theaged,prisoners,psychotics,andretardates.Wesaythattheycare,butitis important to make clear that caring is first of all a matter of acting and onlysecondarilyamatteroffeeling.Fromtimetotime,energeticactionhasbeentaken,butreformhasneverthelessremainedepisodic.Programsdesignedtocorrectmistreatmentwaxandwane,andcompassionwaxesandwanestoo.Thetroubleliesintheweaknessofthecontingencieswhichinduceathirdpartytoundertakereform.Professionalethics(asignificantexpression!)showstheproblem.Acodeofethicsstatesthatitiswrongforaprofessionalperson tobehave incertainwaysand that thosewhodosowillbecensuredorstrippedofprofessionalprivileges.Thepeopleprotectedbysuchacodedonotexertcountercontrolbecausetheydonotknowthattheyhavebeeninjured.Athirdpartyentersbecausethegoodoftheprofessionsuffers;butthegoodoftheprofessionisaremotegain,andonlywell-organizedprofessionsareabletobringittobearontheconduct of their members.More remote andmore difficult tomake effective is the“good” which induces a third party to act in the interests of children, the aged,prisoners,psychotics,andretardates.Onestepwhichmightbetakentocorrect thecontingenciesunderwhichtreatment

tends to deteriorate is to make sure that the care of others is adequately supported.Mistreatment is often amatter of poor conditions and of underpaid and overworkedpersonnel. Butmaking good care as easy as possiblewill not redress the imbalancewhich is causing trouble. Affluence alone has never made people particularlyconcernedforthewelfareofothers.Anotherpossibilityistorecruitpeoplewhoalreadytendtotreatotherpeoplewell.

They do so, and presumably also feel compassionate, not because they arecompassionate people but because they have been exposed to countercontrollingcontingencieselsewhere.Werecruittheminthehopethattheywillcontinuetobehavewellevenundertheeffectivecontingenciesofacustodialinstitution.Butunfortunatelytheeffectsofcontingenciesarenotpermanent.Whenthereisnocurrentcountercontrol,compassionatepeopleoftenfindthemselvesbecomingcarelessorcruel,oftentotheirdismay. Good behavior borrowed from other environments will not last; we needsustainedcontingencieswhichcontinuetoinducepeopletobehavewell.Anobviousstepistocorrecttheimbalancebyexertingcountercontrolonbehalfof

thosewhocannotexertitthemselves.Wemayacttowardthoseinchargeofchildren,oldpeople,prisoners,psychotics,andretardatesasifwewereourselvesaggrievedandthussubjectthemtonormalcountercontrol.Butpreciselywhatarewetodo?Aversionofthegoldenrulemayberelevant:wemayinsistthatthosewhoareinchargeofothersdountothemasweshouldhavethemdountous.Buttheruleisnotinfallible;itdoesnot,forexample,specifythedegreeofcountercontroltobesupplied.Whatistokeepthebalancefromswingingtoofartheotherway?Givenunlimitedpoweritispossiblethatweshouldallbecomeselfishmonsters.Themaintroublewiththerule,however,isthatwedonotalwaysknowwhatweshouldwantifwewerechildren,old,inprison,or, inparticular,psychoticorretarded.“Sharedfeelings”willnothelp.Therulemayservetopreventgrossmistreatment,butitwillnotprescribeadetailedplanofaction.What,afterall,isthegoodlifeforsuchpeople?

311

In using our feelings—or, technically speaking, our own susceptibilities toreinforcement—asaguide,wemakeamistakewhichisinherentintheverynotionofcaring—ofcaritasorcharityintheBiblicalsense.Wefeedthehungry,healthesick,andvisitthelonely—andpossiblybecausethesearethingswelikeotherstodotous.Butthepatternwillnotsufficeforthedesignofabetterworld.Somethingismissinginthe formula “to each according to his need.” It ismissing in both an affluent and awelfare state. We imagine, wish for, and pray for a world—or heaven—in whichreinforcers are available without effort, and we do so because any behavior whichbringsusintosuchaworldisabundantlyreinforced.Butwhatistobedonewhenweareoncethere?Allthatremainsistheconsumptionofgratifiers.Theimportantthingisnotwhatapersongets,butwhatheisdoingwhenhegetsit,andthisisalmostalwaysneglectedbythosewhowishforabetterwayoflife.Thosewhoargueagainstcaringforpeoplebyfulfillingtheirneedsarelikelytofall

undersuspicion.Theyseemtobesayingthatpeople“ought”todeserveorappreciatewhattheygetor“oughtnot”tospongeonothers.Thatisoftentheviewofthosefromwhomgoodshavebeentaken—whohavebeenexploitedinthepursuitofaffluenceortaxed to build a welfare state. But in improving custodial care, we must take intoaccount the good of those who receive rather than those who give. The humanorganism has evolved under conditions in which great effort has been needed forsurvival, and a person is in a very real sense less than human when he is merelyconsumingthingssuppliedtohimgratis.We see the problem in the world at large. We have developed more and more

efficient ways of getting the things we need, and in doing so we have deprivedourselvesofsomepowerful reinforcers.Wehavebuiltaworld inwhichwe lessandlessoftenengageinstronglyreinforcedbehavior.Wefeeltheresultingconditionandcallitboredom.Itistheproblemofleisure.Whenwedonot“have”todoanything,weseemtobefreetoengageinart,music,literature,andscience,butwearemuchmorelikelytofallunderthecontrolofotherkindsofweakreinforcerswhichgettheirchancewhenthepowerfulreinforcershavebeenattenuated.Ourforefathersputitintheformofanaphorism:“Thedevilalwayshassomethingforidlehandstodo.”Leisureclasseshave characteristically gambled, used alcohol and drugs, and enjoyed themselves bywatching other people behave seriously. The problem is particularly acute in aninstitutional environment, where those who do not “have” to do anything (in partbecausetheyarebeingcaredfor)havenothingtodoandarethereforelikelytobehaveinparticularly troublesomeways.Thecharitable fulfillmentofneedscanbe justifiedonly for thosewhocannot fulfill theirneeds themselves.Help ischarityonly for thehelpless.Wedonothelpthosewhocanhelpthemselveswhenwemakeitunnecessaryforthemtodoso.Instead,wedeprivethemofthechancetobehaveinwayswhicharesaidtoshowaninterestinlifeorpossiblyevenexcitementorenthusiasm.Sympatheticunderstandingmaysuggestthedesignofareinforcingenvironment,but

itwillnotspecifydetails.Whatisneededistechnicalknowledgeoftheeffectsoftheenvironment on human behavior. Fortunately, the ways in which reinforcers can bemade contingent upon behavior have been extensively studied, and the results havebeenputtouseinatechnologyofbehaviormodification,whichisconcernedwiththedesignofeffectivecontingenciesofreinforcement.That technology has much to contribute to custodial care, but it will make its

contribution only when certain ethical questions have been answered. One of thesebringsusbacktotheproblemwehavejustconsidered.Increatingabetterworldfor

312

the institutionalized retardate should we be limited to the reinforcers which remainwhenbasicneedshavebeensatisfied?Astrongpositionwastakenbyacommitteeofpsychiatristswhoissuedsomeprinciplesgoverningtheuseoftokeneconomiesinstatehospitalsforpsychotics:“Deprivationisnevertobeused.Nopatientistobedeprivedof expected goods and services or ordinary rights… that he had before the programstarted.Inaddition,deficitrewardingmustbeavoided;thatis,rewardsmustnotconsistoftherestorationofobjectsorprivilegesthatweretakenawayfromthepatientorthatheshouldhavehadtobeginwith.”Butwhatshouldhehavehadtobeginwith?A program of contingency management which begins by taking something away

from people suggests a lack of compassion if compassion is associated with givingpeoplewhattheyneed.Butalthoughastateinwhichallneedsaresatisfiedmayappealto those who have struggled to satisfy needs, it is in a curious sense a state ofdeprivation,astateinwhichpeoplearedeprivedofreinforcementswhichinducethemto behave productively. We honor and admire normal people who do things forthemselves.Psychoticsandretardateshavefailedtodosointhepast,butitisamistaketo suppose that they cannot do so in the future. Through the proper use of what isknownaboutcontingenciesofreinforcement,wecanconstructsimplifiedenvironmentsinwhichdefectivepeoplecantosomeextenttakecareofthemselves.Indoingso,wecouldbesaidtogivethemsomemeasureofself-relianceanddignity.Other “values”whichmay help us decidewhat the retardate should be doing can

alsobederivedfrompracticesconcerningnormalpeople.Whatapersondoeswhenhedoes not “have” to do anything may be judged in terms of the extent to which heexhibitsall thebehaviorofwhichhe iscapable.Thebehaviorsmostcharacteristicofleisuretendtoberepetitious(asingambling),stupefying(as intheuseofalcoholorother drugs), and passive (as in being a mere spectator). They are characteristic ofleisurejustbecausetheydemandverylittleofaperson.Apersonmayspendalifetimedoingthesethingsandbeverylittlechangedattheendofit.Boththeindividualandhis culture suffer. Other activities characteristic of leisure, however, lead to thedevelopment of extensive, varied, and subtle repertoires—among them the arts andcrafts,sports,scholarship,scientificinvestigation,exploration,andthemaintenanceofgood relations with others who are also changing. When activities of that sort areencouraged,theretardateleadsthefullestlifeofwhichheiscapable.The design of such aworld needsmore than compassion or sympathy; it needs a

highlydevelopedscienceofbehavior.Wedonotyethaveasciencewhichcansolveallourproblems,andthatfactpointstosomethingelsewhichtheretardatecouldbedoingwhich raises an ethical question. A science of behavior needs research whichnecessarilyinvolvespeople.Butunderwhatconditionsisitethicaltoforceorinducepeopletoserveassubjects?It isgenerallyheldthatparticipantsmustbeawareofallpossibleconsequencesandmustgivetheirconsent,but this isobviouslynotpossibleforpsychoticsandretardates.Theconsentofthosetowhomtheyhavebeenentrustedforcustodialcareisrequired.Andunfortunatelyitisoftenhardtoget.Administrativeproblemsarebadenoughwithoutmaking furtheradjustments in supportof research.Nevertheless, anyonewho is concernedwith theultimatewelfareof thosewhomustliveinthecareofothersshouldinsistthatsuchpeoplecontributetoresearch.Personalsafetyshould,ofcourse,beprotected,butguidelinesarenothardtosetup.Psychoticsand retardates have an opportunity to contribute to a world in which others likethemselveswillleadbetterlives.Theymaynotknowthattheyaredoingso,butthosewhodoknowshould insist that theybegiven thechance. In thatway theycanshare

313

one other commonly admired feature of normal behavior: they can be controlled byremoteconsequenceshavingtodowiththegoodofothers.We shall not bring aboutmajor changes in custodialmanagementby appealing to

compassion, sympathy, or ethical principles. The contingencies affecting those whomanagecustodialinstitutionsmustbechanged.Environmentscanbedesignedinwhichthosewhoare in thecareofotherswill leadbetter lives,and itmay justhappen thatthosewhoconstructandmaintainthemwillasaresultfeelcompassionateandfollowethicalprinciples.

TobepublishedbytheJosephP.Kennedy,Jr.,FoundationforMentalRetardation.[FromCumulativeRecord(3rded.).Wehavenoevidencetoshowwhetherthepublicationdescribedinthefootnoteaboveevertookplace.]

314

PARTV

ForExperimentalPsychologistsOnly

CurrentTrendsinExperimentalPsychologyTheFlightfromtheLaboratory

315

CurrentTrendsinExperimentalPsychology

This lecturewas given inMarch, 1947, at the first annual conference onCurrentTrendsinPsychologyattheUniversityofPittsburgh.ItappearedinCurrentTrendsinPsychology (Pittsburgh, University of Pittsburgh Press, 1947) and is reprinted bypermissionoftheUniversityofPittsburghPress.There isa familiarcaricatureof theexperimentalpsychologistwhichrunssomethinglike this: He is first of all an apparatusman, who spends a good share of his timetinkeringwithsundrypiecesofequipmentwhichneverquiteworktohissatisfaction.Heinvestigatesonlyproblemswhichhecallsappropriatetothelaboratory.Hecannotstudylearningaspartofthecomplexandsubtleinterplayofbehaviorandenvironmentin everyday life, so he confines himself to the memorizing of meaningless wordspresentedwithclocklikeregularityinastandardaperture.Hecannotbringloveorhateor envy into the laboratory, so he investigates reactions to garter snakes and pistolshots.Theonlystrongmotivesheknowsarehisown,forhissubjectsperformmerelytoobligehimorbecausetheyarerequiredtodosoaspartofacoursetheyaretaking.(In an exceptional case, if he “has a grant,” theymay be paid seventy-five cents anhour.)Heremainsanexperimentalpsychologistonlysolongashisproblemshavenopracticalvalue;thatishowhestayspure.Ifhisfieldsuddenlybecomesimportantforindustryorthepublicweal,thenhebecomesanindustrialorappliedpsychologistanddoes the whole thing over again in no time at all with better and more expensiveapparatus. He works only with amenable subjects—that is to say, with subjects inwhomnooneisreallyinterested:whiteratsordogsorhumanbeingswhohavesteppedoutoftheirnormallivesandintoalaboratoryframeasstandardorganisms.Thepictureisnotasamusingasitmayseem.Partsofitareperhapstooclosetothe

truthtobefunny,andwhetherjustifiedornotthegeneraltoneisdisturbing.Itsupportsa conviction,whichmost of us have reached on the strength of other evidence, thatexperimentalpsychologyispassingthroughacriticalphaseinitshistory,andthatitisunder close and not always sympathetic scrutiny. Psychologists who take a broaderinterestintheaffairsofmenhavegrownimpatientwiththeirexperimentalcolleagues,if not openly critical. They often appear to resent the historical seniority of theexperimentalfieldandtheprestigewhichseniorityhasengendered.Theexperimentalpsychologists themselves have grown uncertain of their scientific position. Theirconfidence has been shaken, and desertions from the ranks occur more and morefrequently.Thisdoesnotmean thatagreatdealofwhatmayproperlybecalledexperimental

psychology is not still going on, or that the results are not duly published in thejournals. It would be possible to write a paper on “Current Trends in ExperimentalPsychology” by describing the latest improvements in techniques, by reporting themostimportantrecentadvances,andsoon,andsuchapaperwouldnotsufferfromanyshortageofmaterial.Buttheimportantissueisthesurvivalofthefielditself,oratleastitsultimatepositionwithrespecttootherbranchesofthescience.Thisoughttohave

316

firstclaimuponourattention.The very definition of experimental psychology is in doubt. It is always easy to

overemphasizesomeincidentalorsuperficialfeature.Forexample,thereisnoreasonwhy we should suppose that experimental psychology is concerned with a specialsubdivisionofhumanbehavior.Atonetime,itistrue,experimentalistswerededicatedtoalimitedsubjectmatter,particularlythefieldsofsensoryprocesses,reactiontimes,andcertainlimitedlearningsituations,butthisisnolongerso.Experimentationisnowcommon inevery fieldofhumanbehavior.Nor is theexperimentalpsychologist anylongerdistinguishedbythefactthatheusesapparatus.Itischaracteristicofhimthatheisnotsatisfiedtoobservebehaviorwithhiseyesandearsalone,butmustconnecthissubjects toamplifiersandrecordersofonesortoranother.Characteristically, too,hedoesnot take the environment simply as he finds it, butmodifies it in variouswayswith various ingenious devices. He was once almost alone in these practices, butvirtually every sort of investigator now adopts them from time to time. The use ofapparatus may improve an experiment, but it must not be confused withexperimentationitself.Itispossibletobeanexperimentalistwithoutusingapparatusatall. It is alsonot true thatexperimentalpsychologynecessarilydealswith somethinglessthanthewholemaninsomethinglessthantherealworld.Tosimplifythematerialof a science is oneof the purposes of a laboratory, and simplification isworthwhilewheneveritdoesnotactuallyfalsify.Buttheexperimentalpsychologisthasnocorneron simplification.Thepsychoanalytic couch is a simplifiedworld, and so is any testsituation.Sincetheexperimentalpsychologistisnolongerdistinguishedbyaspecialfieldof

research, or by his technical equipment, or by laboratory simplification, still anotherhistoricaldistinctionmustbegivenup:hisstatementsarenotnecessarilymorereliablethanthoseofanyoneelse.Atonetimethismighthavebeenregardedastheessentialdifference.Experimentalpsychologystoodforprecisionversuscasualobservation,forexperimental validation versus general impression, for fact versus opinion.Most ofwhatwassaidabouthumanbehaviorineducation,publicaffairs,industry,letters,andsoon,wasontheotherside.Theexperimentalpsychologistwasdistinguishedbythefact that one could trust his statements, no matter how limited their application.Elsewhereoneexpectednothingmorethancasualorphilosophicaldiscourse.Butthisisno longer true.Statementsofcomparablevalidityarecharacteristicofmostof thefieldsrepresentedinthisconferenceandmaybefoundinotherandstilllargerspheresofhumanbehavior.Rigorousdefinition,carefulmeasurement,andvalidationnolongercomprise a sufficient criterion; and even the consolation that the experimentalpsychologist was at least first to take these matters seriously will not suffice for acurrentdefinition.Wecanmakesomeprogress towarddelimitinga fieldofexperimentalpsychology

which is not merely an historical accident by looking more closely at the wordexperimental.Inpsychology,asinanyscience,theheartoftheexperimentalmethodisthedirectcontrolof the thingstudied.Whenwesay,“Letus tryanexperiment,”wemean, “Let us do something and seewhat happens.” The order is important:we dosomethingfirstandthenseewhathappens.Inmoreformaltermswemanipulatecertain“independent variables” and observe the effect upon a “dependent variable.” Inpsychology the dependent variable, towhichwe look for an effect, is behavior.Weacquire control over it through the independent variables. The latter, the variableswhichwemanipulate, are found in the environment.Wemanipulate themwhenwe

317

stimulateanorganism,whenwealterconditionsofmotivationorlearning,andsoon.The greatmajority of psychological experiments can be reduced to this form.Theremaybevariationsonthetheme:insensorypsychology,forexample,wemaywishtoseehowfarwecanchange theenvironmentwithout changingbehavior, aswhenwestudydifferencelimens.Butthebasicpatternofcontrolremainsthesame.This is a narrow definition of an experimental science. It does not identify

“experimental” with “scientific.” Physics, chemistry, physiology, and genetics areexperimental sciences in this sense. Astronomy, geology, and taxonomical biologywould not generally qualify. This is no reflection upon the latter. We are merelyclassifyingthemaccordingtomethodology.Theclassificationisworthmakingbecausethe psychologist is more likely to find common problems and common solutionsamongscienceswhichhavethesameformalstructure.Oneinterestingconsequenceofdefiningexperimentalpsychologyasabranchofthe

science in which we control the variables which govern behavior is that we thusexcludemost investigationsusingcorrelationalmethods. Itmaybepossible toprovethe existence of a functional relation of the sort here in question by running acorrelationbetweensomeaspectofbehaviorandsomeaspectoftheenvironment,butif we are able to manipulate the aspect of the environment, letting it take differentvaluesatdifferenttimes,wecangetamuchmorecompleteaccountoftherelation.Theexperimentalcontroloreliminationofavariable is theheartofa laboratory science,and,ingeneral,itistobepreferredtomanipulationthroughstatisticaltreatment.Itisnotaquestionofachoiceofmethods,however.Thetwoapproachesrepresentdifferentscientific plans and lead to different results. It is curious that our definition shouldsingleoutthekindofresultwhichhasbeentraditionallyacceptedascharacteristicofthe field of experimental psychology.A possible explanation ofwhy it does sowillappearlater.Alinedrawnbetweenfunctionalandcorrelationalanalyseswillrunapproximately

alongtheacceptedboundarybetweenpureandappliedpsychology.Ifthiswerenotanaccidentwemightseizeuponit inordertoreplacethedistinctionbetweentheusefulandtheuseless—adistinctionwhichisnotexactlyflatteringtothepurescientist.Buttheagreementisroughandaccidental.Correlationaltechniqueshavebeenextensivelyusedinpureresearch,andthereasontheyhavedominatedthescienceofpsychologyinits application to education, industry, public affairs, and elsewhere is not that theprocesses to be dealt with in those fields are of any special nature, but that it hasgenerallybeenimpossibletogiveanaccountofrelevantfactorsinanyotherway.The special problem of the applied psychologist is a practical one.Hemust gain

controlofcertainrelativelycomplexmaterial—ifnotdirectly,asinthelaboratory,thenindirectly and frequently after-the-fact through statistical procedures. He is notconfrontedwith any special sort of psychological fact forwhich a specialmethod isrequired. The preference for correlational techniques in applied psychology maythereforechange.Ithasbeentrueoftechnologyingeneralthatasthebasicengineeringproblemissolved,astheappliedscientistgainscontrolofhismaterial,theconnectionwithpureorlaboratoryscienceisstrengthened.Commonmethodsandcommontermscanbeadopted.Somethingofthissortmaybeexpectedinpsychologyasengineeringcontrolisimproved.It is a familiar complaint that the kind of control possible in the laboratory is

impossible in the world at large. The argument is that we cannot modify a naturalenvironmentinsubtlewaysormeasurenormalunhamperedbehaviortothousandthsof

318

a second. The complaint is especially loud with respect to the laboratory study ofanimalbehavior.Thefactthatscienceslikephysiology,embryology,andgeneticsarevery largely concerned with the study of animals and yet yield results constantlyappliedtomenisdismissedasbesidethepoint.Eventhoughbehavioralprocessesmaybe essentially similar in man and rat, it is argued that men cannot be similarlycontrolled,and that the resultsof theanimal laboratoryare thereforeworthlesswhenappliedtothelargerproblemsofhumanbehavior.Thispositionisboundtogrowweakerastheappliedsciencesgrowstronger.Itisnot

true that human behavior is not controlled. At least we cannot proceed very far asscientistsonthatassumption.Tohaveascienceofpsychologyatall,wemustadoptthefundamentalpostulatethathumanbehaviorisalawfuldatum,thatitisundisturbedbythecapriciousactsofanyfreeagent—inotherwords,thatitiscompletelydetermined.Thegeneticconstitutionoftheindividualandhispersonalhistorytodateplayapartinthis determination. Beyond that, the control rests with the environment. The moreimportantforces,moreover,areinthesocialenvironment,whichisman-made.Humanbehavioristhereforelargelyunderhumancontrol.Exceptforthetrivialcaseofphysicalrestraintorcoercion,thecontrolis,ofcourse,

indirect.Itfollowsthegeneralpatternofalteringadependentvariablebymanipulatingthe independent variables. Now, there are many cases in which the independentvariables are freelymanipulable with respect to human behavior. In the nursery, incertaintypesofschools,incorrectiveandpenalinstitutionsthedegreeofcontrolmaybe very great. Although there are certain legal and ethical restrictions, the kind ofmanipulationcharacteristicofthelaboratoryisquitefeasible.Elsewhere—ineducation,industry,law,publicaffairs,andgovernment—thecontrolisnotsolikelytobelodgedinasinglepersonoragency.Here,thebasicengineeringproblemistoacquirecontrol.Butwemustrememberthattheproblemhasfrequentlybeensolved—perhapsasoftenasnottooursorrow.Sincehumanbehavioriscontrolled—andcontrolled,moreover,bymen—thepattern

ofanexperimentalscienceisnotrestrictedinanyway.Itisnotamatterofbringingtheworldintothelaboratory,butofextendingthepracticesofanexperimentalsciencetothe world at large. We can do this as soon as we wish to do it. At the momentpsychologists are curiously diffident in assuming control where it is available or indevelopingitwhereitisnot.Inmostclinicstheemphasisisstilluponpsychometrics,andthisisinpartduetoanunwillingnesstoassumetheresponsibilityofcontrolwhichis implied inguidanceandcounseling.Mostpersonnelpsychologistsstillobtainmenwith desired capacities or personalities by selecting them from a larger populationratherthanbycreatingthemthroughtrainingandguidance.Ineducationwedesignandre-design our curricula in a desperate attempt to provide a liberal education whilesteadfastly refusing to employ available engineering techniques which wouldefficiently build the interests and instill the knowledge which are the goals ofeducation. In some curious way, we feel compelled to leave the active control ofhuman behavior to those who grasp it for selfish purposes: to advertisers,propagandists,demagogues,andthelike.This diffidence in accepting control has had far-reaching consequences. It is

doubtlesstosomeextentresponsibleforthecontinuedefforttoanalyzebehaviorintotraits,abilities,factors,andsoon.Theendresultofsuchaprogramisadescriptionofbehavior in terms of aspect rather than process. It is a static rather than a dynamicdescription,andagainitisprimarilycorrelationalratherthanfunctional.Noonedoubts

319

the value of investigating relations between ability and age, intellect andsocioeconomic status, emotionality and body type, and so on. The resultsmay haveimportantengineeringapplications.Butsofarasthesingleindividualisconcerned,wedonot thenproceedtoalterage,orbody type,orsocioeconomicstatus.Relationsofthissortmaymakeusmoreskillfulinusingtheinstrumentsofcontrolalreadyinourpossession, but they do not help us to acquire new instruments. No matter howsatisfactorilywemaydemonstratetherealityofabilities,traits,factors,andsoon,wemustadmitthatthereislittlewecandoaboutthem.Theygiveusanaspectdescriptionofbehaviorwhichmayhaveapracticalvalueinclassifyingorselectingthemembersofagroup,buttheydonotcarryusveryfartowardthecontrolofthebehavioroftheindividual. That control requires techniques which are peculiarly experimental innature,accordingtothepresentlimiteddefinition,andwemaythereforeanticipatethatassoonasappliedpsychologyemphasizesactivecontrol,theexperimentalpatternwillemerge.Ourdefinitionoftheexperimentalfieldisthereforenotyetcomplete,sinceitdoes

notexcludetheappliedinterestinfunctionalcontrol.Butafinaldistinctioncannowbemade. It concerns the use to which the control is put. What the experimentalpsychologist is up towhenhe is being essentiallyexperimental isdisinguished fromotherfieldsofpsychologybythefactthathehasaspecialgoal.Weneednotblushtoexpress this in rather general terms.The experimental psychologist is fundamentallyinterestedinaccountingforbehavior,orexplainingbehavior,orinaverybroadsenseunderstanding behavior. If these are synonymous expressions I have been redundantandIapologize.Ifeachcarriesitsownspecialshadeofmeaning,thenallthree,takentogether,willcomenearer toanadequatestatement. Inanyevent,wemust try tobemoreprecise.Wedonotunderstandathingsimplybybecomingfamiliarwithit.Norisitenough

tobe abledescribe it, nomatterhowspecificor subtleour termsmaybe.Wemakesomeprogress towardunderstandinganythingwhenwediscoverhow it is related tootherthings,especiallytoantecedentevents.Thisiswhatthelaymanmeansbycauseandeffect,andthesatisfactionwhichhefeelswhenhediscoversthecauseofaneventis probably not to be distinguished from the satisfactionwhich the scientist takes indemonstratingafunctionalrelationship.Thediscoverythattheenvironment,inactingupontheorganism,couldberegardedasacausalagentinthedirectionandcontrolofbehavior, and the realization that itwas thereforepossible todispensewith fictitiousinner controlsmarked thebeginningof a scienceofbehavior.Thiswas asmuch thespirit of the sensory analysis ofmind begun by theBritishEmpiricists as itwas thespiritofDescartesandthelateranalystsofaction.But the cataloguing of functional relationships is not enough. These are the basic

factsofascience,buttheaccumulationoffactsisnotscienceitself.Therearescientifichandbooks containing hundreds of thousands of tabulated facts—perhaps the mostconcentratedknowledgeinexistence—butthesearenotscience.Physicsismorethanacollection of physical constants, just as chemistry is more than a statement of thepropertiesofelementsandcompounds.Thereisnobetterproofofthisthanthefailureofsimple fact-collecting to inspire thescientificworker.Mostof the factsentered inourscientifichandbooksarevirtuallyhackwork.Somewerecollectedinthecourseofmorerewardingscientificpursuits,butthetablesarefilledoutonlybythetypeofmanwhomightotherwisebefoundcollectingstampsoroldcoins.Thereisnomorepatheticfigureinpsychologytodaythanthemerecollectoroffacts,whooperates,orthinkshe

320

operates, with no basis for selecting one fact as against another. In the end, he isusuallytobefounddoingsomethingelse,orperhapsnothingatall.Behavior can only be satisfactorily understood by going beyond the facts

themselves.Whatisneededisatheoryofbehavior,butthetermtheoryisinsuchbadrepute that Ihasten toexplain.Psychologyhashadnoworse theories thananyotherscience,but ithashadthemmorerecently,andtheyhavesufferedinthelightofourimprovedunderstandingofscientificmethod.Noone todayseriouslyusesa fictionalexplanation as a theory, but all sciences have done so at one time or another. Thatmercurystandsatacertainheight inabarometerbecausenatureabhorsavacuumtoexactly that extent, or that certain bodiesmove because they are possessed by a visviva,orthatasubstanceburnsbygivingoffphlogistonarethekindsoftheorieswhosedemisemarkstheprogressofascience.TheyarethesortofhypotheseswhichNewtonrefusedtomake,andmostscientistshavefollowedhisexample.ButNewtonhimselfdemonstratedthevalueofaproperscientifictheory.Atheory,asIshallusethetermhere,hasnothingtodowiththepresenceorabsence

of experimental confirmation. Facts and theories do not stand in opposition to eachother.The relation, rather, is this: theories are basedupon facts; they are statementsaboutorganizationsoffacts.Theatomictheory,thekinetictheoryofgases,thetheoryofevolution,andthetheoryofthegeneareexamplesofreputableandusefulscientifictheories.Theyareallstatementsaboutfacts,andwithproperoperationalcaretheyneedbenothingmorethanthat.Buttheyhaveageneralitywhichtranscendsparticularfactsandgivesthemawiderusefulness.Everyscienceeventuallyreachesthestageoftheoryinthissense.Whether particular experimental psychologists like it or not, experimental

psychology is properly and inevitably committed to the construction of a theory ofbehavior.Atheoryisessentialtothescientificunderstandingofbehaviorasasubjectmatter.Butifwearetoconsiderthecurrentstatusofexperimentalpsychologyratherthan its destiny, wemust admit that it is at themoment in themidst of theoreticalchaos. This is, in fact, the explanation of the present crisis. Many experimentalistsobviouslylackmotivationanddirectionandfinditdifficulttoimparteitheronetotheirstudents.Manyofthemhavelostinterestandareturningtootherfields.Thisisnotdueto any lack of financial support.Our universities can stillwin out against industrialoffers when that is the only thing at issue. Nor is it a question of the support ofresearch,althoughmanyuniversitieshavenot fullyunderstood their responsibility ingeneratingaswellasimpartingknowledge.Therealdifficultyisthattheexperimentalpsychologistisunabletodoanythingwiththefactshehasaccumulated,andheseesnoreasontoaccumulateanymore.Helacksaprofessionalgoal.Partofthisdifficultycanbetracedtothefactthatthetwogreatexplanatorysystems

whichhaveheldthepsychologicalfieldforahundredyearsarenolongerpayingtheirway.Theyhavelosttheirpowertointegrateandilluminatethefactsofthescienceandtoinspireandmotivatethescientificworker.Theonlyresearchtowhichtheynowleadisasortofdesperatepatchworktokeepthetheoriesintact,andthisisunsatisfying.One of these explanatory theories is the notion of a controlling mind. From our

modernvantagepointtheessentiallyfictionalnatureofthisexplanationisclear.Itisonaparwiththeabhoredvacuumorthevisvivaorphlogiston.Mostofusliketofeelthattheghosthasbeenlaid,andthatwearefreeofmentalisticexplanations.Buttheinnerman,constructedofsuchstuffasdreamsaremadeon,stillflourishes.Atleasthalfthetextbooks in psychology still talk about mental life, and few are successful in

321

convincingthestudent that thiscanbereducedtothestuffwhichisdealtwithinthephysicalsciences.Inpsychiatrythescorewouldbealmostahundredtooneinfavorofanappealtopsychicdeterminersofbehavior.Psychoanalysishasassignednamestoatleastthreeoftheseinnermen,anditistheexceptionalpsychoanalystwhoiswillingtoregardthemasphysicalentities.We cannot break away from these hoary practices simply by resolving to avoid

theory altogether.We need a better theory. But this will be of a different sort andcannotbereachedbypatchingupanoldmodel.Onecurrentpractice,forexample,istomake the inner man more respectable by stripping him of what we may call hispersonification.Henolongerexistsasacompleteperson,butonlyassmallfractionsofhisoldself—aswants,drives,attitudes,interests,andsoon.Itistheexceptionalwriterwho convincingly defines termsof this sort in a nonmentalisticway; and even if anoperational re-definition is successful, the old theory may leave its mark in thestructuresurviving.Theothercurrentexplanatorytheoryflourisheswithgreaterprestigeandpresumably

inmorerobusthealth.This is thephysiological theoryofbehavior.The innermanisgivenneurologicalproperties,withagreatgaininscientificrespectability.Psychiatrybecomes neuropsychiatry, and psychology the study of the nervous system. It isdifficult to attack this theory without seeming to criticize the physiologicalpsychologist,butnocriticismisinvolved.Therearemanyprecedentsinthehistoryofscience for borderline disciplines. To integrate the facts of two sciences is aninteresting and profitable endeavor. Eventually, we may assume, the facts andprinciples of psychology will be reducible not only to physiology but throughbiochemistry and chemistry to physics and subatomic physics. But this reduction isundoubtedlyalongwayoff.Thecurrenttheoreticalpracticewhichisobjectionableistheuseofahypotheticalneuralstructure,theconceptualnervoussystem,asatheoryofbehavior.Theneurologicalreferencesintroducedintosuchatheory,likereferencestomentalstates,interferewithfreetheorybuilding,andtheyproduceastructurewhichisnotoptimalfortheorganizationofbehavioralfacts.The traditional physiological theory, too, eventually fails to motivate the

psychologist.Validneurologicalexplanationsofimportantpsychologicallawsarenotarrivedatwithaveryrewardingfrequency,andtheinvestigationswhichtheyinspirehaveatendencytoleadtosuchajumbleofdetailsthattheoriginalplanislostsightof.Weareallfamiliarwiththetypeofgraduatestudentwhocomestostudypsychologyfullofenthusiasmforascienceofbehavior,whoclimbsthephysiologicalfamilytreethroughBerkeley,Hume,Wundt, and themoderns, and findshimself studying somedetailedphysiologicalmechanism.Hismotivationeventually flagswhenhe sees thathiscurrentactivitieshaveonlythemosttenuousconnectionwithhisoriginalinterestinhuman behavior. Such a case history is only a scale model of the history ofexperimental psychology. No matter how critically we may now view the originalprogramofascienceofmind,wemustadmitthatagreatdrivingforcewaslostwhenthenervoussystemhadtobebroughtin.Insteadofthebasicpsychophysicalrelation,the object of research became the operation of specific physiological mechanisms.Generalized brain theories of the Gestalt variety and dimensional analyses ofconsciousnessareefforts tobring togetheragain the fragmentsofa scienceofmind,and to add something of theoretical interest to the study of the physiology of end-organs.Butthesparkhasbeenlost.Ifwetrytoputthesetwogreatexplanatorysystemsingoodscientificorderthrough

322

operational re-definition,we only succeed in dealing the coupdegrâce.We can, ofcourse, define “mind” in behavioral terms, andwe can set up a conceptual nervoussystemfortherepresentationofbehavioralfacts,leavingthespecificationoftheactualneuralpropertiesuntilsomelaterdate.Butinthiswayweeliminatealltheexplanatoryforceofthetheories.Anoperationaldefinitionispossibleineverycase,butitdoesnotnecessarily lead to a satisfactory theoretical construct.Whatever its success, it spoilstheexplanatoryfun.The appeal to what wemay call naive physiologizing, like the appeal to psychic

determiners,ismadeinanattempttoexplainbehaviorbyshiftingtoadifferentlevelofobservation.Theseare“outside”theories,whichaccountforonethingbypointingtosomethingwhich is goingon somewhere else at the same time.For this reason theycannot fill the need for a theory of behavior, nomatter how carefully theymay beextendedorrepaired.Whatisemerginginpsychology,asithasemergedatsomepointin the history ofmost sciences, is a theorywhich refers to facts at a single levelofobservation.The logicof this issimpleenough.Webeginwithbehaviorasasubjectmatteranddeviseanappropriatevocabulary.Weexpressthebasicprotocolfactsofthescienceinthetermsofthisvocabulary.Inthecourseofconstructingatheorywemayinventnewterms,buttheywillnotbeinventedtodescribeanynewsortoffact.Atnotimewillthetheorygeneratetermswhichrefertoadifferentsubjectmatter—tomentalstates, for example, or neurones. It is not the purpose of such a theory to explainbehaviorbyturningto“outside”determiners.Therealachievementofsuchtheorybuildingisnoteasytodemonstratebecauseof

thepresentconfusedconditioninpsychology.Thereisnogenerallyacceptedtheoryofbehaviorwhichwill serve as an example. But the situation is not quite hopeless.Ascientific theory isneverfullysubscribed tobyall thepractitionersofascience; if itwere,therewouldbenofurtherneedforscientificeffort.Andwhilenoexplicittheoryin experimental psychology today hasmore than a handful of adherents, in practicemost psychologists respect certain underlying assumptions which constitute thebeginningofanimplicittheory.Werealizehowextensivethisimplicittheoryiswhenweobservenon-psychologistsdealingwiththesamesubjectmatterandseehowtheyrepeatedlyviolateourassumptions.Wehave,then,somethingtobeginwithbywayofactualtheoreticalpractice.Wemayalsogetaplausibleglimpseofthefuture,forsomeofthefeaturesofaneffectivetheorycanbeinferredfromthenatureofbehaviorasasubjectmatterandfromcomparabletheoriesinotherfields.Itshould,therefore,alsobepossible to evaluate the present status of psychology with respect to theoryconstruction.The first step in building a theory is to identify the basic data. Itmay be easy or

difficult, depending upon the science. It was relatively easy, for example, to decidewhateventsweretobetakenintoaccountintheCopernicantheoryofthesolarsystem.Astronomershadobserved thepositionsof theplanetsatgiven times; the theoreticalproblemwastorelatethesefacts,nottoidentifythem.Ingenetics,ontheotherhand,itisrelativelydifficulttodiscoverwhatcharacteristicsofanorganismarevalidgeneticunits.Psychologyfacesanevenmoredifficultproblem:whatarethepartsofbehaviorandenvironmentbetweenwhichorderlyrelationsmaybedemonstrated?Thelaymanhaslittledifficultyinanalyzingthebehaviorofhimselfandhisfellow

men. He breaks it into discrete acts. He may report, for example, that someone“watchedacaruntilitpassedoutofsight.”Thestatementconveysusefulinformationatthelevelofcasualdiscourse,butisitnecessarilyavalidscientificdescription?The

323

language of the kitchenmay be of no use to the chemist, though the cook finds itmeaningful enough. Anyone who has tried to analyze pursuit behavior knows theprobleminvolvedinorientingtheeyestowardamovingobject,andverymuchmorethanthat iscoveredby thewordwatch.Andwhenthe layman,withwhatseemslikebreathlessdaring,reportsthatsomeone“chosetoremainsilent,”hestakesoutafieldwhichmightsufficeforalifetimeofresearch.Thestatementmaybequiteeffectiveforpractical purposes, but itwill not necessarily suffice for a scientific description. Forwhatis“choice”?Eventhebehaviorinvolvedinchoosingbetweensimpleobjectslikecigarettesornecktiesiscomplexenough.Butwhatishappeningwhenone“choosestoremainsilent”?Andinwhatsenseisremainingsilenttoberegardedasbehavioratall?Inpracticepsychologistsdefine“response” inmanyways—frommuscle twitch to

teliceffect.Inthelattercasetheypresentthephysiologistwiththebafflingproblemofhowtworesponsesexecutedbydifferentpartsofthebodycanbemutuallyreplaceableinalawfulphysiologicaltrainofevents.Itisacommoncurrentpracticetododgetheproblembyacceptingsomepracticalmeasureofbehavior,oftenlimitedtoaparticularmeasuring device, such as “maze performance” or some arbitrary criterion of“success.”Thephysiologisthasalsobeenappealedto,butinvain,sinceanindicationofthepresenceorabsenceofactivityinaparticulareffectorisoflittlehelp.Wecannotcontinuetoleavetheproblemunsolvedifwearetoconstructaneffective

theory. It may be that the notion of a unit of response is at fault and that a finalstatementwill reflect the fluidity and continuity of behavior as awhole.Thiswouldrequiremorepowerfulanalyticaltools,butitmaybenecessary.Afurtherrequirementmust also be recognized; it is not themere formof behaviorwhichweundertake topredict but rather its occurrence.Expressions like “reaction tendency” or “excitatorypotential” have attempted to take account of this fact. The end term in a theory ofbehavior,inshort,istheprobabilityofaction.In the companion problem of environment, the layman again shows an enviable

talent, for he describes and analyzes the environmentwith no hesitationwhatsoever.Theworldtohimissimplyacollectionofthings.Buthissuccessgivesthecaseaway.Hehasanalyzedtheenvironmentintermsofitspracticalimportance.Thisisjustifiablefor his purposes; and in so far as various aspects of the environment have commonpractical consequences for everyone, the lay vocabulary might even be adopted forscientific use. But a complete scientific account must go back to properties of theenvironmentwhichareeffectivebeforeanyconsequenceshavebeenfelt,anditmustaccount for the process by which consequences alter the effectiveness of theseproperties.Currentpracticesareagaindiverse.Somepsychologists, as inpsychophysics,deal

withstimulionedimensionata time.Others,at theotherextreme,refer to the“totalsituation”—anexpressionwhichseemssafebecauseitcanscarcelyoverlookanything,but which is unpleasantly vague. Our present knowledge of the physiology of thereceptorsofferslittleifanyhelpindecidinguponaneffectivepractice.Sincewehavenotclearlyidentifiedthesignificantdataofascienceofbehavior,we

donotarrivewellpreparedatthesecondstageoftheorybuilding,atwhichwearetoexpressrelationsamongdata.Observedrelationsofthissortarethefactsofascience—or,when a sufficient degree of generality has been reached, its laws.The generalformoflawsofbehaviorcanbeinferredfromthenatureofourprogram,butexamplesarenotveryabundantamongtheachievementstodate.Aweaknessatthefirststageoftheoryconstructioncannotbecorrectedatthesecond.Inpsychophysicsthestimulusis

324

definedrigorously,ifnotverycomprehensively,andanarbitrarydefinitionofresponseseemstosuffice.Consequently,somegeneralityatthesecondstagehasbeenachieved.In the field of learning, on the other hand,we have collected thousands of separatelearningcurves,buttheyrepresentchangesinhundredsofdifferentaspectsofbehaviorinhundredsofdifferentsituations.Asa result,wehavenovalidgeneralexpressionsfor learning processes. This is characteristic of most of the facts of experimentalpsychology,and thenextstep in theconstructionofasatisfactory theory is thereforeverydifficult.This step—at the third stage in theory building—can be exemplified by a simple

example from the science of mechanics. Galileo, with the help of his predecessors,began by restricting himself to a limited set of data. He proposed to deal with thepositionsofbodiesatgiventimes,ratherthanwiththeircolororhardnessorsize.Thisdecision, characteristic of the first stage in building a theory, was not so easy as itseemstoustoday.Galileothenproceededtodemonstratearelationbetweenpositionandtime—thepositionofaballonaninclinedplaneandthetimewhichhadelapsedsince its release.Somethingelse thenemerged—namely, theconceptofacceleration.Later, as other facts were added, other concepts appeared—mass, force, and so on.Third-stageconceptsofthissortaresomethingmorethanthesecond-stagelawsfromwhich theyarederived.Theyarepeculiarly theproductof theory-making in thebestsense,andtheycannotbearrivedatthroughanyotherprocess.There are few, if any, clear-cut examples of comparable third-stage concepts in

psychology, and the crystal ball grows cloudy. But the importance of the stage isindicatedbythefactthattermslikewants, faculties,attitudes,drives, ideas, interests,andcapacities properly belong there. When it is possible to complete a theoreticalanalysisatthisstage,conceptsofthissortwillbeputingoodscientificorder.Thiswillhave theeffectofestablishing themin theirownright.Atpresent theyneedexternalsupport.Someofthem,likewantsandattitudes,cometoustrailingcloudsofpsychicglory,andawispor twoof thepsychiccanusuallybedetectedwhen theyareused.Other concepts, like drives and motives, borrow physiological support in certainfavorable cases. Still others, like abilities and traits, have been made respectablethroughcorrelationalanalyses,whichgivethemthestatusof“individualdifferences.”Althoughmostpsychologiststhinkofanabilityassomethingwhichhasmeaninginthebehaviorofasingleindividual,currenttechniquesofmeasurementfinditnecessarytomakeuseofthepositionoftheindividualinapopulation.Magnitudesareassignedtotheabilitiesandtraitsoftheindividualintermsofhisrelationtothegroupratherthanthrough direct measurement. A proper theory at this stage would characterize thebehaviorofanindividualinsuchawaythatmeasurementwouldbefeasibleifheweretheonlyindividualonearth.Thiswouldbedonebydeterminingthevaluesofcertainconstantsinequationsdescribinghisbehavior—clearlyathird-stageenterprise.Fromallofthisshouldemergeanewconceptionoftheindividualasthelocusofa

system of variables. Fortunately for psychology it has been possible to deal withbehaviorwithoutaclearunderstandingofwhoorwhatisbehaving,justasitseemstobepossibletodealwithpersonalitywithoutdefining“person.”Theintegrityorunityoftheindividualhasbeenassumed,perhapsbecausetheorganismisabiologicalunit.Butitisquiteclearthatmorethanoneperson,inthesenseofanintegratedandorganizedsystem of responses, exists within one skin. The individual proves to be no moreundividable than theatomwasuncuttable.Manysortsofmetaphoricalschemeshavebeendevised to represent this fact.A single personalitymaybe regarded asmoving

325

about from one level of consciousness to another, or personalities may be franklymultiple. A proper theory must be able to represent the multiplicity of responsesystems.Itmustdosomethingmore:itmustabolishtheconceptionoftheindividualasa doer, as an originator of action. This is a difficult task. The simple fact is thatpsychologistshavenevermadea thoroughgoingrenunciationof theinnerman.Heissurreptitiouslyappealedtofromtimetotimeinallourthinking,especiallywhenwearefacedwithabitofbehaviorwhichisdifficulttoexplainotherwise.Eventually we may expect the main features of a behavioral theory to have

physiologicalsignificance.Asthescienceofphysiologyadvances, itwillpresumablybepossibletoshowwhatishappeninginvariousstructureswithintheorganismduringparticularbehavioralevents,andthetheoreticalsystemsofthetwosciencesmayalsobeseentocorrespond.Anexampleofthisrapprochementisthewayinwhichfactsandprinciples of genetics arrived at from the study of the characteristics of parents andoffspringareseentocorrespondtofactsandprinciplesofcellstructure.Thescienceofgenetics has already reached the stage at which it is profitable to investigate bothsubjectmattersat thesametime.Termswhichoriginallydescribedrelationsbetweenthe characteristics of parents and offspring may now carry additional cytologicalreferences.A similar day may come in psychology. That is up to the physiologist and the

physiologicalpsychologist.Buttheeventualcorrespondenceshouldnotbeallowedtoobscure the present need for a behavioral theory. The hypothetical physiologicalmechanisms which inspire so much research in psychology are not acceptable assubstitutes for a behavioral theory. On the contrary, because they introduce manyirrelevant matters, they stand in the way of effective theory building. There is atendencyinsomequarterstoadmitthiswhileinsistinguponcompensatingadvantages.It isargued that thesolidityof thenervoussystemgives it thestrength todispossesspsychicfictionswhichapurelybehavioral theorymaylack.It isalsothoughttobeanecessary intellectualcrutch—anever-presenthelp in timeof theoreticalneed.Manypeople cannot think of the origination of an actwithout thinking of amotor center.Theycannotconceiveoflearningwithoutthinkingofchangesinsynapticresistanceorsomeotherprotoplasmicchange.Theycannotcontemplateaderangementofbehaviorwithoutthinkingofdamagedtissue.Moreover,itisoftenpointedoutthatthehistoriesofothersciencesshowmanyexamplesof theorieswhich,underaproperoperationalanalysis,wouldhavebeen found tocontainunwarranted references tootherkindsofdata but which made it possible to think more effectively about relevant data thanwould have been possible with a purely conceptual scheme. But this remains to beproved. It is not necessarily true that physiological theories have in the long rundirectedtheenergiesofpsychologistsintothemostprofitablechannels.Anenlightenedscientificmethodologyshouldenableustoimproveuponthepracticesexemplifiedbythe history of science. In any event an independent theory of behavior is not onlypossible, it is highly desirable, and such a theory is in no sense opposed tophysiologicalspeculationorresearch.Becauseoftheunhappyfateofsomanypsychologicaltheoriesofthepast,asound

theoryofbehaviormustworkitselfoutagainstaweightofindifferenceandevenactiveopposition.Veryfewpsychologistsunderstandthenatureofsuchatheoryorareawarethatithasacounterpartinmostestablishedsciences.Manyofthemdenythepossibilityof a respectable theory. It is encouraging to recall, however, that a good tentativetheoryhas usually proved to be autocatalytic; a demonstration ofwhat can be done,

326

even within a limited sphere, draws attention to theory-building, and the process isaccelerated.There isusuallynoneed to justify a theoryofbehaviorwhen its potentialities are

made clear, for these are very great. Consider the case of the social sciences, forexample. The current practice of the sociologist is either to express his facts andtheorieswithoutreferringtoindividualbehavioratall,ortoconstructapsychologyofhis own—devoting at least an introductory chapter (if not an entire treatise) to themotivesandhabitswhichleadmentolivetogetherandbehavetogetherastheydo.Thesociologistmayormaynotagreethatthebehaviorofthegroupistobepredictedfroma study of the psychology of the individual, but he has no hesitation in using thebehavior of the individual to expound, if not to explain, sociological facts. Theeconomist,whetherprofessionalorprofessorial,facesthesamealternatives.Eitherhemuststatelawsandmakepredictionswithoutmentioninghumanbehavior,orhemustdevise a special psychology to explain the activities of the great-grandchildren ofAdamSmith’s“economicman.”Itistheexceptionaleconomistwhodoesnotaccountfor facts about goods or money or labor or capital by pointing to what men willtypicallydoundercertaincircumstances.Similarly, thepolitical scientist,whetherornothehopes toderive theprinciplesofgovernmentor thecharacteristicsofpoliticalstruggles frompsychology,usuallycontinues to talkaboutsomespeciesof“politicalman,” towhomhe assigns just themotives and capacities needed to account for hispoliticalfacts.Whatever his field, the social scientist does not currently find in the science of

psychology a conceptual scheme with which he can talk about human behaviorconsistentlyandeffectively.Economicman,politicalman,thegroupmind—thesearecrude explanatory fictionswhich need to be replaced by a sound behavioral theory.That such a theory need not be essential to a true social science is beside the point.Thereisnoquestionthatitwouldbeenormouslyhelpful.Thereisagreaterneedforsuchatheoryinthosebroadfieldsofhumanendeavorin

which rigorous scientific practices are not yet feasible. For example, a widespreadcritical examination of our educational practices is currently in progress. This isbasicallyaprogramofpsychologicalengineering.Yetitisbeingprojectedandcarriedthroughwithaquiteunrealisticconceptionofhumanbehavior.Ancienttheoriesofthenatureofman recuragainandagainwith their familiarcant—“an integratedviewoflife,”“asenseofpersonalresponsibility,”“acapacitytoexperienceandunderstandlifeas a related whole,” “the development of the mind,” and so on. Educators are notwholly to blame, for we have not yet put forth a workable conception of humanbehaviorsuitablefortheirpurposes.Ourlegalsystem,totakeanotherexample,isbaseduponanevenolderformofthe

traditional theory. It is becoming more and more difficult to reconcile our modernconceptionofmanandsocietywiththelegalnotionofpersonalresponsibility,ofawillcapableofconsciousmotionanddominatedfromtimetotimebyideas,feelings,andinfluences.Butanalternativetheoryisapparentlynotyetinworkableform.Thelackofanadequateunderstandingofhumanbehaviorismostcruellyfeltinthe

fieldofgovernmentandworldaffairs.Wearefacedwiththedishearteningspectacleofhundredsofmenofgoodwilldrawingupblueprintsfortheworldofthefuture,whilemakingassumptions abouthumannaturewhichmostofusknow tobe invalid.Twoworldwarshavenotbeenfoughtoveranythingassimpleasworldtradeorboundaries.We are in transition from one conception of man to another and to an effective

327

understanding of the possible relationshipswhichmay exist betweenmen.We havepaida terribleprice forknowledgewhichcouldconceivablybeacquired through thepeacefulandprofitablemethodsofscience,andasyetwehavelittletoshowforit.Agreatdealmaydependuponwhetherwecanreachinthenearfutureaworkabletheoryofhumanbehavior.Oneimportantroleofascientifictheoryofbehavior,then,istoreplacethetheories

which now pervade our thinking, which are part of our everyday speech, whichinfluenceallourdealingswithourfellowmen,andwhichstandinthewayofapplyingthe methods of science to human affairs. As everyone knows, many technicalprocedureswhichwould improveourpractices in education, law,politics, and soonare now available. The contributions which the science of psychology can make inthesemattersisverygreat.Psychologistshavebeenpowerfuladvocatesofanobjectiveattitudeandwillundoubtedlycontinuetoinsistthatthemethodsofsciencebeappliedto human behavior and human society wherever possible. If we are to talk aboutbehavior,letusbeprecise.Ifwearetoinsistthattwofactsarerelated,letusprovetherelation.Psychologycanofferbetterwaysofdescribingandmeasuringbehavior,bettermethodsofguaranteeingthevalidityofstatements,andsoon.Butnothingofthissortisanylongerexclusivelyapsychologicalcontribution.Themaintaskistomakethesetechnicalcontributionsfelt, toput theminto thehandsof thepeoplewhoneed them;andwecando thisonlywhenwemake it clear that a science ismore thanmethod,morethanfacts.Themostimportantcontributionthatpsychologycanmaketodayisaworkable theory of behavior in the present sense—a conception ofmanwhich is inaccordwithallthefactsofhumanbehaviorandwhichhasbeencruciallytestedintheexperimental laboratory. Only an effective and progressive theory of behavior canbring about the proper change in attitude which will make it possible to apply themethodsofsciencetohumanaffairsineveryfield.Thesurvivalofthetraditionalconceptionofmanasafreeandresponsibleagentis

anexcellentexampleofthegeneralprinciplethatatheoryisneveroverthrownbyfacts,but only by another theory. There are facts which have been well established forcenturieswhichareincompatiblewiththetraditionaltheoriesofhumanbehavior,andthesetheoriesmoveaboutinthemodernworldinawelterofcontradiction.Buttheirproponents work busily to patch them up, and somehow they survive. A newinterpretationhere,aconspiracyofsilencethere,andthetrickisturned;andthiswillcontinuetobesountilanewandeffectivetheoryisworkedout.Wecannotremedythesituationbymeredialectic.Weneedtoarriveatatheoryof

human behavior which is not only plausible, not only sufficiently convincing to be“sold” to the public at large, but a theory which has proved its worth in scientificproductivity.Itmustenableus,notonlytotalkabouttheproblemsoftheworld,buttodosomethingabout them, toachieve thesortofcontrolwhich it is thebusinessofascienceofbehaviortoinvestigate.Thesuperiorityofsuchatheorywill thenbeclearandweshallnotneedtoworryaboutitsacceptance.Theimportanttrendinexperimentalpsychology,then,istowardasatisfactorytheory

ofbehavior.Perhapsweshouldnotbesurprisedatthis,sincethefieldwasdefinedinsuchawaythatitwouldnecessarilybetrue.Butthefieldhadtobedefinedinthatway.Experimental psychology ismore than a tradition; it ismore than an assemblage ofpracticesandinterestspassedalongfromgenerationtogenerationwithoutrespecttoachangingworld.A traditionneeds tobe reviewedand justified,and this isespeciallytrue inexperimentalpsychology,where ithasbeeneasy to lose themain theme.The

328

trend,then,istowardaclarificationofthistheme,towardasortofself-realization.Theexperimentalpsychologistisnotusingmethodformethod’ssake.Heisnotfollowinganinteresttowhichhehasbeenledbyindulginginoneidlecuriosityafteranother.Hedoes not seize upon a field of research because the practical-minded have left ituntouched.Insofarasheisbehavingasanexperimentalpsychologist,heistryingtounderstandbehavior. In this work he must discover and collect facts, and he mustconstructanadequatetheory.A clear realization of this aim should be helpful. There is nothing wrong with

experimentalpsychologywhichaclear-cutobjectivewillnotcure.Thedevelopmentofaneffective theoryofbehavior is ideal for thispurpose.Thescienceofexperimentalpsychologywill presumably remain in the hands of the professors.Critical issues inappliedfieldsmayleadtoimportantcontributionstotheory;methodswillbedevisedandfactsdiscoveredinindustry,education,theclinic,andsoon,whicharerelevanttoacentralscience.Butthehusbandingoffacts,thesiftingofinformationfromallfieldsofhumanbehavior,thespecialstudyofquestionswhicharetheoreticallycrucial,andthe working out of a satisfactory conceptual system will presumably remain thefunction of the psychologists in our universities. This is still so, at least, in olderscienceswithmuchmore extensive technological applications. It is appropriate, too,that a concern for theory in this sense should remain closely associated withinstruction.But the academic psychologist is limited in the time and facilities available for

research, and at themoment hemay be rather bewildered by, if not envious of, theglittering technical advantages of his erstwhile colleagues. In theory-construction,however,hefindsafieldwhichisnotonlyexclusivelyhisown,butoneinwhichhecanexperimenteffectivelyand tosomepurposewith relatively limitedresources.Hewillnotneedtoconfinehimselftofactswhichhavebeenneglectedbythosewhocanexperimentmore efficiently.Hewill be able to explorekeypositionsof thegreatestimportance.Theexperimentumcrucisishisfield,andinithemayusuallyrestcontentwith one subject for every hundred studied by his applied colleagues and with onechronoscopeorpursuitmeterorcathode-rayoscillographinplaceofdozens.This isnotagestureofescape. It isnotaconclusion that thegrapesaresour.The

experimentalpsychologistisaboveallascientist,andthisistheproperfieldofscience—the discovery and ordering and understanding of nature. This is Faraday andMaxwellratherthanthelaboratoriesofGeneralElectricorWestinghouse.ItisMendelandT.H.Morganrather thananagriculturalbreedingstation. It isPasteurandKochratherthanresearchlaboratoriesofgreatpharmaceuticalhouses.Thisisgoodcompany.Tounderstandhumanbehaviorinthesenseinwhichanypartofnatureisunderstoodbyscienceistrulyanexcitingandsatisfyinggoal.

329

TheFlightfromtheLaboratory

The circumstances under which this paper was written are described in the firstparagraphs. It was given at the University of Pittsburgh in January, 1958, and isreprintedherebypermissionoftheUniversityofPittsburghPress.

Anexperimentalpsychologistsometimesinvitesamanintoa laboratory,askshimtomemorizealistofnonsensesyllablesorlearntokeepapointeronamovingtarget,andsends him on hisway quite unaware that hewill be asked to come back later for asecondseriesofobservations.Theexperimentwillnotsucceedunlessheisignorantofthe future test. I do not know whether the originator of these conferences wasconductingsuchanexperimenttenyearsago,butIcannowreportwhatitfeelsliketobeinvitedbackforthesecondsession.Itismainlyafeelingofregret.If,whenIwaspreparing my earlier paper, I had known that I would be asked to compare mypredictionoftrendsinexperimentalpsychologywithadecadeofactualhistoricalfact,Ishouldhaveconfinedmyselftostatementswhichcouldhavebeenmoreeasilytwistedtoaccommodatetheeventualities.Ishouldhavepreparedamuchmorepalatabledishofhumblepie.Itisobviousnow,afterthefact,thatthetrendsIdescribedwerescarcelymorethan

myhopes for the futureof experimentalpsychology.Possiblymybehavior couldbedefended as a gesture appropriate to the intellectual climate of 1947. Experimentalpsychologywas thenat thenadirof itspopularity.Graduatestudentswere turning tosocial, personal, clinical, and applied psychology in ever-increasing numbers, anddefectionsfromtheranksamongoldermenwerecommon.Thepracticalcontributionswhich experimental psychologists had made duringWorld War II had not offset agrowingimpatiencewiththeirstubborndedicationtoseeminglyunimportantaspectsofhumanbehavior.Butwastherenotabrightspotonthismurkyhorizon?Ifthehistoryofsciencewereanyguide,aneffectivepsychologywouldeventuallydevelopacentralconceptionofhumanbehaviorwhichnotonlywouldbefundamentally“right” in thesense of enabling us to understand behavior, whatever that might mean, but wouldgeneratepowerful techniques having important applications in every field of humanaffairs.Notheoryofbehaviorhadyetcomeclosetothatachievement.Psychoanalysiswas the only disciplinewhich had spread beyond its original boundaries, and it hadgonenofurtherthansomeofthesocialsciencesandliterarycriticism.Elsewhere—ingovernment, economics, religion, education, and all the natural sciences—provincialtheories of human behavior were eked out by the tattered theory which had beenbequeathedtotheEnglishlanguagebyalonglineofoutmodedphilosophies.Itwasasif each of the technologies of physical science had its own scientific conception ofnature—as if specialists in synthetic fibers used one theory of molecular structure,pharmacologists another, and biochemists still another, while the layman carried onwith a commonsense view of the structure of matter untouched by any of thesetechnicaltreatments.Suchastateofaffairswasfarfromsatisfactory.Afterall,itwas

330

the samemanwhowas of interest to psychologists, political scientists, theologians,psychotherapists,economists,educators,literarycritics,andscientificmethodologists.Whyshouldtherebeadifferenttheoryofhumanbehaviorineachcase?Intothispowervacuum,itseemedtome,experimentalpsychologymusteventually

move. A general theory of human behavior was needed, and only an experimentalscience could supply it. Separate technologies of behavior could temporize withparticular theories, but the special control of variables attainable only in laboratoryexperimentationwould ultimately supply the accountwhich, being in closest accordwiththeactualpropertiesofthehumanorganism,wouldbemostusefulineveryfieldofhumanaffairs.Theclosecheckwith realitycharacteristicofexperimentalanalysiswouldbemostlikelytoexposethefictionalentitieswhichhadplayedsodevastatingaroleinwhatpassedforpsychologicalexplanationandwouldpermitustoescapefromthe inaccessible, hypothetical constructs emerging from statistical analyses. Thisextrapolation of the history of science was intended to give the experimentalpsychologistabroaderhorizon.Inpointingoutthepotentialsignificanceofaneffectivetheoryofhumanbehaviorandthespecialplaceofalaboratoryscienceindevelopingsuchatheory,Iwastryingtoalterthecontingenciesofreinforcementofmycolleaguesinthehopeofstemmingwhatseemedtobeaperpetuallyebbingtide.It is temptingtoarguethat thisproved, indeed, tobeanactual trend.It ispossible

thattheoriesofbehaviorderivedfromtheclinicorfromfieldstudies,ratherthanfromthelaboratory,areonthewane.AstrictFreudianpsychology,forexample,isnolongerstoutly defended. Certain general points have beenmade—in some sensewe are allFreudians—but the facts and principles which have been salvaged can be stated inrelativelynon-technicallanguage.Eventhepatientundertherapyisnolongerlikelytobe burdened with technical references to the structure and function of the psyche.Experimental psychologists are not responsible for this change, but if the commonheritage of psychoanalysis is to be put in good scientific order, if an effectivetechnologyistobemorethanageneralunderstandingofthemotivesandemotionsofoneselfandone’sfellowmen,experimentalpsychologistswillplayanimportantrole.TheFreudiandynamismscanbesubjectedtoexperimentalanalysis,andtheresultingchanges in definition reveal the experimental method at work.1 The Freudianexplanatory system seldom traces the causal linkage far enough. We do not reallyexplain“disturbedbehavior”byattributingitto“anxiety”untilwehavealsoexplainedtheanxiety.Theextrasteprequired is in thespiritofanexperimentalscience: it isasearchforamanipulablevariableratherthanaverbalexplanation.Psychoanalysisitselfhas identified some of the conditions which must be changed in order to modifybehavior in psychotherapy, and to bring about other behavioral effects, but itsmethodology is not adapted to themanipulation and control of these conditions. Incontrast, experimental psychology is becomingmore andmore successful in dealingwiththevariablestowhichonemusteventuallyturnforafullaccountofbehaviorandforeffectivecontrol.There are other signs of a change. The layman’s way of talking about behavior,

deeplyentrenchedinoureverydayvocabularythoughitis,haslostground.Itisviewedwith greater uneasiness by those who use it. Ten years ago the physiologist,neurologist,orpharmacologistwhoseresearchinvolvedbehaviorwaslikelytosetuphisownexperimentsandtodescribehisresultsinnon-technicalterms.Henowacceptstheexperimentalpsychologistasaspecialisttowhomhemustturnforhelp.Totakea

331

verydifferentexample,thelayterminologyisnowmoreoftenusedwithapologies(orin quotation marks) by political scientists. The ultimate danger of arguing fromhistoricalanalogy,andofpredictingor recommendingcoursesofactionbydeducingtheoremsfromaxiomaticprinciplesorgovernmentalstereotypes, ismorelikelytoberecognized. The ideological use of the work of Pavlov by Soviet propagandists haslittle to recommend it,butweprobablymake the samemistakewhenwecounterbyexpressing contempt for techniques of government based on conditioned reflexes. Inthe long run all thiswill have a salutary effect if it leads us to askwhether amoreadequate science of behavior may not be relevant to the design of governmentalpractices.Aconceptionofhumanbehaviorwilleventuallyproveworkable,notbecauseitfitsamomentarypredilectionforaphilosophyofgovernment,butbecauseitsurvivesthetestofexperimentalanalysis.Somewherebetweentheextremesofphysiologyandgovernment liesa thirdbitof

evidence for a possible trend.Educational psychologists have long been devotees ofresearch,but thepatternofa laboratorysciencehasnotbeenclosely followed.Theirexperimentshaveseldomcometogripswiththebehavioroftheindividualstudentintheactoflearning.Ontheotherhand,theexperimentalpsychologyoflearning,thoughonceastapleintextbooksoneducation,hasbeenreceivinglessandlessattention.Butwehavelearnedagreatdealaboutlearninginthepastdecade.Aproposaltoputthistouseineducationwasmadeatanearlierconferenceinthisseries.Theprinciplesofanexperimentalanalysisarenowbeingextendedtothefieldofverbalbehavior,anditisinconceivablethattheresultswillnotbeusedtoimproveinstructionalprocedures.Andwithfabulousresults.Enoughhasalreadybeendonetojustifythepredictionthatwhatisnowlearnedbytheaveragecollegestudentwillsomedaybelearnedinhalfthetimewithhalftheeffort.Thereis,then,evidenceofarenaissanceinexperimentalpsychologywhichmightbe

attributed in part to a realization of the potential contribution of the experimentalmethod.But itdoesnotwarrant theclaim that Icorrectlypredictedamajor trend.Ageneraltheoryofhumanbehaviorinthissensehasappealedtoonlya“happyfew.”Asone can easily discover by glancing at the tables of contents of our journals,experimentalpsychologyasawholehasnot shownmuchchange.Very littlecurrentresearchisreportedintheframeofreferenceofacomprehensivetheory.Norhasthepointofviewofanexperimentalanalysisyetreachedfarafield.Manysocialsciencesremainuntouched,andamongnaturalscientiststhereisalmostcompleteignoranceofthepromiseandachievementofthescientificstudyofbehavior.Dr.NeilsBohr,oneofthemostdistinguishedlivingphysicists,recentlydiscussedcertainissuesinpsychologyasfollows:

Quiteapartfromtheextenttowhichtheuseofwordslike“instinct”and“reason”inthedescriptionofanimalbehaviorisnecessaryandjustifiable,theword“consciousness,”appliedtooneselfaswellastoothers,isindispensablewhendescribingthehumansituation….Theuseofwordslike“thought”and“feeling”doesnotrefertoafirmlyconnectedcausalchain,but to experienceswhich exclude each other because of different distinctions between theconscious content and the background which we loosely term ourselves…. We mustrecognize thatpsychicalexperiencecannotbesubjected tophysicalmeasurementsand thattheveryconceptofvolitiondoesnotrefertoageneralizationofadeterministicdescription,but from the outset points to characteristics of human life.Without entering into the oldphilosophical discussion of freedom of the will, I shall onlymention that in an objective

332

description of our situation the use of the word “volition” corresponds closely to that ofwords like “hope” and “responsibility,” which are equally indispensable to human

communications.2

These terms and issueswould have been at home in psychological discussions fiftyyearsago.(Indeed,onecommentatormentionedthesimilarityofDr.Bohr’sviewstothoseofWilliamJames.)How shocked Dr. Bohr would be if a distinguished psychologist were to discuss

modernproblemsinphysicalscienceintermswhichwerecurrentatthebeginningofthecentury!Psychologyingeneral,andexperimentalpsychologyinparticular,isstillalongwayfromprovidingaconceptionofhumanbehaviorwhichisasreadilyacceptedby thosewhodealwithmenas theviewsofphysicsareacceptedby thosewhodealwiththephysicalworld.Andpsychologiststhemselvesarenotdoingmuchaboutit.I therefore return to the attack. (In doing so I assertmymembership in a species

distinguishedbythefact that,at leastwhenpsychotic, itsmemberssometimesfail toshowextinction.)3But I shall not doggedly repeatmy exhortations or promises of adecade ago. It is evidently not enough to strengthen the scientific behavior ofpsychologists by giving them a glimpse of an exciting future. Fortunately, as oneachievementoftheinterveningdecade,theproblemcannowbeattackedwithabetterbrand of behavioral engineering. I propose to analyze the behavior of psychologists.Whyaretheynotcurrentlydevelopingthepurescienceofhumanbehaviorfromwhichsuchtremendoustechnologicaladvanceswouldcertainlyflow?Howarewetoexplainthe continuing flight from the experimental field? Where have the experimentalpsychologistsgone,andwhataretheydoinginstead?Andwhy?And,aboveall,whatstepscanbetakentoremedythesituation?SuchquestionsclarifytheengineeringtaskwhichfacesusifwearetoproducethetrendinexperimentalpsychologywhichIinsistuponpredicting.Sostated,theproblemhasananalogyinatypeofexperimentwhichisgrowingin

importance in the experimental analysis of behavior. When we have studied theperformancesgeneratedbyvariouscontingenciesofreinforcementinasinglearbitraryresponse,wecanmoveontotwoormoreconcurrentresponses.Insteadofonelevertobepressedbyaratoronekeytobepeckedbyapigeon,ourexperimentalspacenowfrequently contains two or three levers or keys, eachwith its own set of reinforcingcontingencies. In the present experiment, we are to account for the fact thatpsychologists have stopped pressing the experimental lever and have turned to otheravailable manipulanda. To explain this two questions must be asked: (1)What hashappened to the reinforcing contingencies on the experimental lever? and (2)Whatcontingencies compete so effectively elsewhere? Once these questions have beenanswered, we can proceed to the engineering task of increasing the relativeeffectivenessoftheexperimentalcontingencies.Itwouldprobablybeunfairtodothisby attacking competing conditions, for any source of scientific zeal should berespected,butitispossiblethatsomeofthereinforcementsresponsibleforactivityonother levers can be made contingent upon the response in which we are primarilyinterested.Somedeficienciesintherewardsoftheexperimentalpsychologistwereanalyzedin

myearlierpaper.Allsciencesundergochangesinfashion.Problemsloseinteresteventhoughtheyremainunsolved.Inpsychologymanygreenpastureshavebeenglimpsed

333

ontheothersideoftheexperimentalfence.Theverysuccessofasciencemayforceittobecomepreoccupiedwith smaller and smaller details,which cannot competewithbroadnewissues.Thephilosophicalmotivationofthepioneersofa“mentalscience”hasbeenlost.Althoughidealismisevidentlystillafightingwordinsomepartsoftheworld, dualism is no longer a challenging issue in American psychology. Classicalresearchontherelationbetweenthepsychicandthephysicalhasbeentransmutedintothestudyof thephysiologicalandphysicalactionsofend-organs.This isascientificstepforward,butanimportantsourceofinspirationhasbeenleftbehind.Someofthemosteffectiverewardscontingentuponexperimentalpracticehavebeen

inadvertentlydestroyed inanotherway.Weowemostofourscientificknowledge tomethods of inquiry which have never been formally analyzed or expressed innormativerules.Formorethanageneration,however,ourgraduateschoolshavebeenbuilding psychologists on a different pattern of Man Thinking. They have taughtstatistics in lieu of scientific method. Unfortunately, the statistical pattern isincompatiblewithsomemajorfeaturesoflaboratoryresearch.Asnowtaught,statisticsplays down the direct manipulation of variables and emphasizes the treatment ofvariationafterthefact.Ifthegraduatestudent’sfirstresultisnotsignificant,statisticstellshimtoincreasethesizeofhissample;itdoesnottellhim(and,becauseofself-imposedrestrictionsonmethod,itcannottellhim)howtoachievethesameresultbyimprovinghisinstrumentsandhismethodsofobservation.Biggersamplesmeanmorework,thebruntofwhichtheyoungpsychologistmayhavetobear.Whenhegetshisdegree (and a grant), hemay pass the labor on to someone else, but in doing so hehimselflosescontactwiththeexperimentalorganismheisstudying.Whatstatisticianscall experimentaldesign (I havepointedout elsewhere that thismeansdesignwhichyieldsdatatowhichthemethodsofstatisticsareappropriate)usuallygeneratesamuchmoreintimateacquaintancewithacalculatingmachinethanwithabehavingorganism.Oneresult isadamagingdelay in reinforcement.Anexperimentmay“payoff”onlyafter weeks of routine computation. A graduate student who designs an experimentaccording to accepted statistical methods may survive the ordeal of the calculatingroombyvirtueofhisyouthfulzeal,buthisultimatereinforcementasascientistmaybesolongdeferredthathewillneverbeginanotherexperiment.Otherleversthenbeckon.The psychologist who adopts the commoner statistical methods has at best an

indirect acquaintance with the “facts” he discovers—with the vectors, factors, andhypotheticalprocessessecretedbythestatisticalmachine.Heisinclinedtorestcontentwithroughmeasuresofbehaviorbecausestatisticsshowshimhowto“dosomethingabout them.” He is likely to continue with fundamentally unproductive methods,becausesqueezingsomethingofsignificanceoutofquestionabledatadiscouragesthepossiblymoreprofitablestepofscrappingtheexperimentandstartingagain.Statistics offers its own brand of reinforcement, of course, but this is often not

contingentuponbehaviorwhichismostproductiveinthelaboratory.Onedestructiveeffect is tosupplyasortofbusyworkfor thecompulsive. In theearlystagesofanyinquiry the investigatoroftenhas toweather aperiodof ignorance andchaosduringwhichapparentprogressisslight,ifnotlackingaltogether.Thisissomethinghemustbetaughttoendure.Hemustacquireakindoffaithintheultimatevalueofostensiblyundirectedexploration.Hemustalso learn tobe indifferent to thecriticismthathe isnotgettinganywhere.Ifhehasacceptedfundsinsupportofhisresearch,hemustlearnto tolerate a gnawing anxiety about the annual report.At such times statistics offersconsolingcomfortand,whatisworse,anall-too-convenientescape-hatch.Howsimple

334

it is to match groups of subjects, devise a crude measure of the behavior at issue,arrangeforteststobeadministered,andpunchthescoresintoIBMcards!Nomatterwhatcomesofitall,noonecansaythatworkhasnotbeendone.Statisticswillevenseetoitthattheresultwillbe“significant”evenifitisprovedtomeannothing.The intention of the statistician is honorable and generous. He wants the

experimental scientist tobe sureofhis results and toget themostoutof them.But,whetherornotheunderstandstheessenceoflaboratorypractice,hisrecommendationsareofteninimicaltoit.Perhapsagainsthiswill,hehasmadecertainessentialactivitiesin good laboratory research no longer respectable.The very instrumentwhichmighthavemadeanexperimentalsciencemorerewardinghas,instead,allbutdestroyeditsbasicfeatures.Inthelongrunthepsychologisthasbeendeprivedofsomeofhismostprofitable,andhenceeventuallymostreinforcing,achievements.The resulting flight from the laboratory can be stopped by pointing to alternative

methodsofresearch.Ifallpsychologistsaretoberequiredtotakecoursesinstatistics,they should also bemade familiarwith laboratory practices andgiven the chance tobehaveasscientistsratherthanastherobotsdescribedbyscientificmethodologists.Inparticular,youngpsychologistsshouldlearnhowtoworkwithsingleorganismsratherthanwithlargegroups.Possiblywiththatonestepalonewecouldrestoreexperimentalpsychologytothevigoroushealthitdeserves.But it will be worthwhile to examine the competing contingencies. Psychologists

havefledfromthelaboratory,andperhapsforgoodreason.Wherehavetheygone?

TheFlighttoRealPeople

Laboratories can be dull places, and not only when furnished with calculatingmachines. It is not surprising that psychologists have been attracted by the humaninterestof real life.Theexperimental subject in the laboratory isonlypartofaman,andfrequentlyanuninterestingpart,whilethewholeindividualisafascinatingsourceof reinforcement.Literature flourishes for that reason.Psychologists have long sincelearnedtoborrowfromtheliterarydomain.Ifalectureflags,orachapterseemsdull,one has only to bring in a case history and everything literally “comes to life.”Therecipe is so foolproof that the lecture or text which consists of nothing but casehistorieshasbeencloselyapproximated.Butinresortingtothisdeviceforpedagogicalor therapeutic effect psychologists have themselves been influenced by thesereinforcers; their courses of action as scientists have been deflected. They oftenrecognizethisandfromtimetotimehavefelttheneedforaspecialtheoryofscientificknowledge (based, forexample,onempathyor intuition) to justify themselves.Theyseldom seem to feel secure, however, in the belief that they have regained fullcitizenshipinthescientificcommonwealth.The reinforcementswhich flow from real people are not all related to, on the one

hand, an intellectual conviction that the proper study ofmankind isman or, on theother, the insatiable curiosity of a Paul Pry. In a world in which ethical training iswidespread,mostmenarereinforcedwhentheysucceedinreinforcingothers.Insuchaworldpersonalgratitudeisapowerfulgeneralizedreinforcer.Wecanscarcelyholditagainstpsychologiststhat,likeothermenofgoodwill,theywanttohelptheirfellowmen—eitheronebyoneintheclinicornationbynationin,say,studiesofinternationalgoodwill.Wemayagree that theworldwouldbeabetterplace ifmoremenwould

335

concernthemselveswithpersonalandpoliticalproblems.Butwemustnotforgetthatthe remedial step is necessarily a short-termmeasure and that it is not the only stepleadingtothesamegoal.Thelivelyprosecutionofascienceofbehavior,appliedtothebroadproblemofculturaldesign,couldhavemoresweepingconsequences.Ifsuchapromisingalternativeisactuallyfeasible,anyonewhoiscapableofmakingalong-termcontributionmaywiselyresisttheeffectofotherconsequenceswhich,nomatterhowimportant theymaybe to himpersonally, are irrelevant to the scientific process andconfinehim to short-term remedial action.Aclassical example fromanother field isAlbert Schweitzer. Here is a brilliant man who, for reasons we need not examine,dedicatedhislifetohelpinghisfellowmen—onebyone.Hehasearnedthegratitudeof thousands, but we must not forget what he might have done instead. If he hadworked as energetically for as many years in a laboratory of tropical medicine, hewouldalmostcertainlyhavemadediscoverieswhichinthelongrunwouldhelp—notthousands—butliterallybillionsofpeople.WedonotknowenoughaboutSchweitzerto saywhy he took the short-term course. Could he not resist the blandishments ofgratitude?Was he freeing himself from feelings of guilt?Whatever his reasons, hisstorywarnsusofthedangerofaculturaldesignwhichdoesnotharnesssomepersonalreinforcement in the interests of pure science. The young psychologist who wantsabove all to help his fellow men should be made to see the tremendous potentialconsequences of even a small contribution to the scientific understanding of humanbehavior. It is possibly this understanding alone,with the improved cultural patternswhichwill flowfromit,whichwilleventuallyalleviate theanxietiesandmiseriesofmankind.

TheFlighttoMathematicalModels

Theflightfromtheexperimentalmethodhassometimesgoneintheotherdirection.Ifthehumanbeingstudied in the laboratoryhasbeen toodrabandunreal forsome,hehasbeenjust theoppositeforothers. Inspiteofourvauntedcontrolofvariables, theexperimentalsubjecttoooftenremainscapricious.Sometimesheisnotonlywarmbut,asbaseballplayerssay,toohottohandle.Eventhe“averageman,”whencapturedinthestatisticalnet,maybeunpleasantlyrefractory.Somepsychologistshave thereforefledtoanivoryimageoftheirownsculpturing,mountedonamathematicalpedestal.ThesePygmalionshaveconstructedaGalateawhoalwaysbehavesassheissupposedtobehave,whoseprocessesareorderlyand relatively simple, and towhosebehaviorthemostelegantofmathematicalproceduresmaybeapplied.Sheisacreaturewhoseslightest blemish can be erased by the simple expedient of changing an assumption.JustaspoliticalscientistsusedtosimplifytheirproblemsbytalkingaboutanabstractPolitical Man, and the economists theirs by talking about Economic Man, sopsychologistshavebuilttheidealexperimentalorganism—theMathematicalModel.The effect of this practice on so-called learning theory has been pointed out

elsewhere (page69). Early techniques available for the study of learning—from thenonsense syllables of Ebbinghaus, through the problem boxes of Thorndike and themazes ofWatson, to the discrimination apparatuses ofYerkes and Lashley—alwaysyielded learning curves of disturbing irregularity. In experiments with theseinstrumentsanorderlychangeinthebehaviorofasingleorganismwasseldomseen.Orderlyprocesseshadtobegeneratedbyaveragingdata,eitherformanytrialsormany

336

organisms. Even so, the resulting “learning curves” varied in a disturbingway fromexperimenttoexperiment.Thetheoreticalsolutiontothisproblemwastoassumethatan orderly learning process,which always had the same properties regardless of theparticularfeaturesofagivenexperiment,tookplacesomewhereinsidetheorganism.Agiven result was accounted for by making a distinction between learning andperformance. Though the performance might be chaotic, the psychologist couldcontinue to cherish the belief that learning was always orderly. Indeed, themathematical organism seemed so orderly that model builders remained faithful totechniqueswhichconsistentlyyieldeddisorderlydata.Anexaminationofmathematicalmodelsinlearningtheorywillshowthatnodegreeofdisorderinthefactshasplacedanyrestrictionontheeleganceofthemathematicaltreatment.Thepropertieswhich(todroptoatwo-dimensionalfigureofspeech)makeapaper

doll4 more amenable than a living organism are crucial in a scientific account ofbehavior.Nomatterhowmanyoftheformulationsderivedfromthestudyofamodeleventually prove useful in describing reality (remember wave-mechanics!), thequestions to which answers are most urgently needed concern the correspondencebetween the two realms.How canwe be sure that amodel is amodel ofbehavior?Whatisbehavior, andhow is it tobeanalyzedandmeasured?Whatare the relevantfeaturesoftheenvironment,andhowaretheytobemeasuredandcontrolled?Howarethesetwosetsofvariablesrelated?Theanswerstothesequestionscannotbefoundbyconstructingmodels. (Nor is amodel likely tobehelpful in furthering thenecessaryempiricalinquiry.Itisoftenarguedthatsomemodel,hypothesis,ortheoryisessentialbecausethescientistcannototherwisechooseamongthefactstobestudied.Butthereare presumably as many models, hypotheses, or theories as facts. If the scientificmethodologistwill explain howhe proposes to choose among them, his answerwillserveaswelltoexplainhowonemaychooseamongempiricalfacts.)What sort of behavioral engineering will reduce the rate of responding to the

mathematical lever and induce distinguished psychologists to get back to thelaboratory?Twostepsseemtobeneeded.First,itmustbemadeclearthattheformalpropertiesofasystemofvariablescanbeprofitablytreatedonlyafterthedimensionalproblems have been solved. The detached and essentially tautological nature ofmathematical models is usually frankly admitted by their authors, particularly thosewho come into experimental psychology frommathematics, but for the psychologistthesedisclaimersareoftenlostamongtheintegralsigns.Secondly,theopportunitytobemathematical indealingwith factualmaterial shouldbeclarified.Toreturn to theexample of learning theory, the psychologist should recognize that with propertechniquesonecanseelearningtakeplace,notinsomeinnerrecessfarremovedfromtheobservableperformanceofanorganism,butasachangeinthatperformanceitself.Techniquesarenowavailable for theexperimentalanalysisofverysubtlebehavioralprocesses,andthisworkisreadyforthekindofmathematicaltheorywhichhasalwaysbeenproductiveat theproperstageinthehistoryofscience.What isneededisnotamathematicalmodel,constructedwithlittleregardforthefundamentaldimensionsofbehavior,butamathematical treatmentofexperimentaldata.Mathematicswill comeintoitsownintheanalysisofbehaviorwhenappropriatemethodsyielddatawhicharesoorderlythatthereisnolongeranyneedtoescapetoadreamworld.

TheFlighttotheInnerMan

337

Experimentalpsychologyhassufferedperhaps itsgreatest lossofmanpowerbecausecompetentinvestigators,beginningwithadescriptiveinterestinbehavior,havepassedalmostimmediatelytoanexplanatorypreoccupationwithwhat isgoingoninside theorganism.Indiscussing this flight to the innermanIshould like tobelieve that Iamwhipping a dead horse, but the fact remains that human behavior is still mostcommonly discussed in terms of psychic or physiological processes. A dualisticphilosophy isnotnecessarily implied ineithercase for itmaybeargued,on theonehand,thatthedataofphysicsreduceatlasttothedirectexperienceofthephysicistor,ontheother,thatbehaviorisonlyahighlyorganizedsetofbiologicalfacts.Thenatureof any real or fancied inner cause of behavior is not at issue; investigative practicessufferthesamedamageinanycase.Sometimes,especiallyamongpsychoanalysts,theinnermenaresaidtobeorganized

personalitieswhoseactivitiesleadatlasttothebehavioroftheorganismweobserve.Thecommonerpracticeistodissecttheinnermananddealseparatelywithhistraits,perceptions, experiences, habits, ideas, and so on. In thisway an observable subjectmatter is abandoned in favor of an inferred. It was Freud himself who insisted thatmental processes could occur without “conscious participation” and that, since theycould not always be directly observed, our knowledge of themmust be inferential.Muchofthemachineryofpsychoanalysisisconcernedwiththeprocessofinference.In the analysis of behavior we may deal with all mental processes as inferences,whether or not they are said to be conscious. The resulting re-definition (call itoperational if you like) conveniently omits the mentalistic dimension. At the sametime, however, the explanatory force is lost. Inner entities or events do not “cause”behavior,nordoesbehavior“express”them.Atbesttheyaremediators,butthecausalrelationsbetweentheterminaleventswhicharemediatedareinadequatelyrepresentedbytraditionaldevices.Mentalisticconceptsmayhavehadsomeheuristicvalueatonestageintheanalysisofbehavior,butithaslongsincebeenmoreprofitabletoabandonthem. In an acceptable explanatory scheme the ultimate causes of behaviormust befoundoutsidetheorganism.The physiological inner man is, of course, no longer wholly inferential. New

methodsandinstrumentshavebroughtthenervoussystemandothermechanismsunderdirect observation. The new data have their own dimensions and require their ownformulations.Thebehavioralfactsinthefieldoflearning,forexample,aredealtwithintermsappropriatetobehavior,whileelectricalorchemicalactivitiesoccurringatthesame time demand a different conceptual framework. Similarly, the effects ofdeprivation and satiation on behavior are not the same as the events seen through agastric fistula.Nor isemotion, studiedasbehavioralpredisposition,capableofbeinganalyzed in termsappropriate topneumographsandelectrocardiographs.Bothsetsoffacts, and their appropriateconcepts, are important—but theyareequally important,not dependent one upon the other. Under the influence of a contrary philosophy ofexplanation,whichinsistsuponthereductivepriorityoftheinnerevent,manybrilliantmenwhobeganwithaninterestinbehavior,andmighthaveadvancedourknowledgeofthatfieldinmanyways,haveturnedinsteadtothestudyofphysiology.Wecannotdispute the importance of their contributions,we can only imaginewith regretwhattheymighthavedoneinstead.Ifwearetomakeastudyofbehaviorsufficientlyreinforcingtoholdtheinterestof

youngmenincompetitionwithinnermechanisms,wemustmakeclearthatbehaviorisanacceptablesubjectmatterinitsownright,andthatitcanbestudiedwithacceptable

338

methodsandwithoutaneyetoreductiveexplanation.Theresponsesofanorganismtoagivenenvironmentarephysicalevents.Modernmethodsofanalysisrevealadegreeof order in such a subject matter which compares favorably with that of anyphenomena of comparable complexity. Behavior is not simply the result of morefundamentalactivities,towhichourresearchmustthereforebeaddressed,butanendinitself,thesubstanceandimportanceofwhicharedemonstratedinthepracticalresultsofanexperimentalanalysis.Wecanpredictandcontrolbehavior,wecanmodifyit,wecanconstructitaccordingtospecifications—andallwithoutansweringtheexplanatoryquestionswhichhavedriveninvestigatorsintothestudyoftheinnerman.Theyoungpsychologistmaycontemplateatruescienceofbehaviorwithoutanxiety.

TheFlighttoLaymanship

Experimentalpsychologyhasalsohadtocontendwithwhatisinessencearejectionofthe whole scientific enterprise. In a recent review of a study of the psychologicalproblemsofaging,thereviewercommentsupon“atendencyinpsychologicalthoughtwhichisreturningtoprominenceaftersomeyearsofrelativedisfavor.Thestatementshave a certain refreshing directness and ‘elegance’ in their approach to the study ofhumanbehavior.The sterile arguments of so-called ‘learning theory,’ thedoctrinairehalf-truthsofthe‘schools,’thepanaceatreatmentsof‘systems,’andthehigh-sounding,emptytechnical termsoftenfoundinpsychologicalwritingsareconspicuousbytheirabsence.”Noonewillwanttodefend“sterilearguments,”“half-truths,”“panaceas,”or“emptytechnicalterms,”nomatterwhattheirsources,buttheforceofthepassageismorethanthis.Theauthorisrejectingalleffortstoimproveuponthepsychologyofthelaymaninapproachingtheproblemsoftheaged.Andmanypsychologistsagreewithhim.“Enoughof thelingoof the laboratory!” theargumentruns.“Enoughofclinicaljargon!Enoughoffrighteningequations!Aplagueonallyourhouses!Letusgobacktocommonsense!Letussaywhatwewanttosayabouthumanbehaviorinthewell-wornbutstillusefulvocabularyofthelayman!”Whetherthisisagestureoffatigueorimpatience,ortheexpressionofadesiretogetonwithpracticalmattersattheexpenseofabasicunderstanding,itmustbeansweredbyanyonewhodefendsapurescience.Itwouldbeeasiertofindtheanswerifexperimentalpsychologyhadmovedmorerapidlytowardahelpfulconceptionofhumanbehavior.Someprogresshasbeenmadeinprovingthesuperiorityofscientificconceptsover

thoseof traditionalusage.Consider, forexample, twopsychologicalaccountswritteninthevulgartongue.First,asampleinthefieldofemotionalbehavior:

Theemotionaltemperofthetypeofjuveniledelinquentjustmentionedisasextraordinaryas it iswell-known.Farfrombeingnaturallypeaceful,sympathetic,orgenerous,menwhoareexcluded from the societyof their fellowmenbecome savage, cruel, andmorose.Thewanton destructiveness of the delinquent is not due to sudden bursts of fury, but to adeliberateandbroodingresolvetowagewaroneverything.

Thesecondhastodowithintellect.Itisanexplanationofhowachildlearnstoopenadoorbydepressingathumb-latchandpushingagainstthedoorwithhislegs.

Ofcoursethechildmayhaveobservedthatdoorsareopenedbygrownupsplacingtheir

339

handsonthehandles,andhavingobservedthisthechildmayactbywhatistermedimitation.But theprocess as awhole is somethingmore than imitative.Observationalonewouldbescarcely enough to enable the child to discover that the essential thing is not to grasp thehandlebut todepressthelatch.Moreover, thechildcertainlyneversawanygrownuppushthedoorwithhislegsasitisnecessaryforthechildtodo.Thispushingactionmustbeduetoan originally deliberate intention to open the door, not to accidentally having found thisactiontohavethiseffect.

Bothpassagesmakeintelligiblepointsandwouldconceivablybehelpfulindiscussingjuveniledelinquencyortheteachingofchildren.Butthereisatrap.Actuallytheheroesofthesepieceswerenothumanatall.ThequotationsareslightlyalteredpassagesfromRomanes’ Animal Intelligence, published about seventy-five years ago. The firstdescribes the behavior of the prototype of all delinquents—the rogue elephant. The“child”ofthesecondwasacat—possiblytheverycatwhichsetThorndiketoworktodiscoverhowanimalsdo,indeed,learntopresslatches.Theexperimentalanalysisofbehaviorhasclearlyshownthepracticalandtheoretical

value of abandoning a commonsense way of talking about behavior and hasdemonstrated the advantages of an alternative account of emotion and intelligence.That is tosay, ithasdone this forcats, rats,pigeons,andmonkeys. Itssuccessesareonly slowly reaching into the field of human behavior—not becausewe any longerassumethatmanisfundamentallydifferentbutinpartbecauseanalternativemethodofanalysis is felt to be available because of the scientist’s membership in the humanspecies. But the special knowledge resulting from self-observation can be given aformulation which preserves intact the notion of the continuity of species.ExperimentalmethodscanbeappliedfirsttothebehavioroftheOtherOne,andonlylatertotheanalysisofthebehaviorofthescientisthimself.Thevalueofthispracticeisdemonstrated in the consistencyof the resulting account and the effectivenessof theresultingtechnologicalcontrol.Itisnotdifficulttoexplainthestrengthoftraditionalconcepts.Manyofthosewho

discuss human behavior are speaking to laymen andmust adapt their terms to theiraudience.Theimmediateeffectofthelayvocabularyalsogainsstrengthfromitsdeepintrenchmentinthelanguage.Ourlegalsystemisbasedonit,andtheliteratureofideasis couched in it. Moreover, from time to time efforts are made to rejuvenate thephilosophical systems fromwhich it came.Aristotle, through ThomasAquinas, stillspeakstosomestudentsofbehavior.TheveryfactthatAristotle’spsychology,scarcelymodified,canbeseriouslychampionedinbehavioralsciencetodayshowshowlittleithas done to advance our understanding.Aristotelian physics, chemistry, and biologyhaveenjoyednosuchlongevity.WemaylookforwardtotheearlydemiseofthissolesurvivorofGreekscience.Areturntothelayvocabularyofbehaviorcannotbejustified.Themoveisamatter

ofmotivation,competence,ortheaccessibilityofgoals.Theseareallirrelevanttothelong-term achievement of a scientific account of behavior.Nodoubt,many pressingneedscanstillbemostreadilysatisfiedbycasualdiscussion.Inthelongrun,however,weshallneedaneffectiveunderstandingofhumanbehavior—sothat, intheexamplecited,weshallknow thenatureof thechangeswhich takeplaceasmenandwomengrowoldandshall,therefore,beinthemostfavorablepositiontodosomethingaboutthem.To reach thatunderstandingwemust recognize the limitationsof the remedialpatchworkwhichemergesfromcommonsensediscussionandmustbewillingtoresort

340

toexperimentswhichquitepossiblyinvolvecomplicatedtechniquesandtotheoreticaltreatmentsquitepossiblyexpressedindifficultterms.

Conclusion

Wehaveglancedbrieflyatfourdivertissements in thegrowthofascienceofhumanbehavior.RealMen,MathematicalMen,InnerMen,andEverydayMen—itwouldbeamistaketounderestimatetheirseductivepower.Togethertheyconstituteaformidablearrayofrivalsuitors,andtogroomtheExperimentalOrganismforthisracemayseemahopelessenterprise.Buthehasachance,forinthelongrunheoffersthegreatestnetreinforcement to the scientist engaged in the studyofbehavior. I doubtwhether thisfactwillaffectmanyofthosewhohavealreadyflownfromthelaboratory,butIamnotspeaking to them. A story aboutWilliam James is appropriate. James wasmuch indemandasalecturerandonedaydiscoveredthathewasscheduledtoaddressaladies’literarysocietyinasuburbofBoston.Hesetofftokeephisappointmentafterhavingpicked up from his desk the first lecture which came to hand. It happened to be alecturehehadpreparedforoneofhisRadcliffeclasses.Hisaudience,incontrast,wascomposed of elderly New England matrons. James was reading his paper, possiblythinkingofother things,when tohishorrorheheardhimselfsaying,“…andso,myfairyoungfriends….”Helookedoutuponaseaofstartledfacesand—failingutterlyinthis pragmatic test of a psychologist—blurted out, “I should explain that this lecturewaswrittenforaverydifferentaudience.”IwishIcouldsay,andalsowithmoretact,whataudiencethislecturewasprepared

for.Nomatterhowstrongmyconviction thatweareclose toaneffectivescienceofhumanbehavior,withallwhichsuchascienceimplies,Idonotexpecttorecapturetheinterest and enthusiasm of those who have fled from the laboratory to pleasurabledalliance elsewhere. But some of you, I hope, are not yet committed. For you thepossibility of an adequate theory of behavior, in the sense in which any empiricalscience leads eventually to a theoretical formulation, together with its enormoustechnicalpotential,maybeenough to tip thebalance.And if suchofyou therebe, Ilook to you to restore to experimental psychology the energy, enthusiasm, andproductivitywhichcharacterizeditinanearlierepoch.We are living in an age in which science fiction is coming true. The thrilling

spectacleofman-madesatelliteshasturnedoureyestowardouterspace.Whatweshallfind there only timewill tell.Meanwhile, we are confronted by far more importantproblemsonthesurfaceoftheearth.Apossiblesolutionisinthespiritofanotherkindofsciencefiction:theeighteenth-centuryutopiandreamofPerfectionismwithitsbasiccontentionthat,ifhumannatureisdeterminedbyenvironmentandifenvironmentcanbechanged,humannaturecanbechanged.Likeanartificialsatelliteorarockettothemoon,thiswasonceafoolishdream.Butsciencemovesforwardatabreathlesspace.Wemayshortlybedesigningtheworldinwhichmenwillhenceforthlive.Buthowisittobedesigned,andtowhatend?Thesearedifficultquestions,towhichnothingshortof an effective scienceofmanwill provide the answers.Themethods of sciencenolongerneedverbaldefense;onecannotthrowamoonaroundtheearthwithdialectic.Applied to human behavior, the same methods promise even more thrillingachievements.That prospectwill, I still believe, determine the trend in experimentalpsychologyintheyearstocome.

341

FromCurrentTrendsinPsychologicalTheory.Pittsburgh:UniversityofPittsburghPress,1961.

1SeeScience&HumanBehavior,Chapter24.2Bohr,N.Atomicphysicsandhumanknowledge.NewYork,1958.3Seereferencepage158.4Thereference,ofcourse,istothewell-knownsongbyJohnnyS.Black,inwhichthelyricistexpresseshispreferencefor“apaperdolltocallhisown”ratherthana“fickle-mindedreallivegirl.”

342

PARTVI

CreativeBehavior

CreatingtheCreativeArtistALectureon“Having”aPoem

343

CreatingtheCreativeArtist

Therearemanyreasonswhywemaywant togiveartamore importantplace inourculture.Perhapswe simplywantmore art tobe available tobe enjoyed.Perhapswebelieve thataculture inwhichart flourishes is strongerbecause it attractsandholdspeoplewhocansolvepracticalproblems.Butourreasons,whatevertheymaybe,arebynomeansasimportantasourprospects.Whatcanbedonetofurthertheproductionandconsumptionofart?

ArtandLeisure

A relation between art and leisure has long been recognized. Earlyman had to freehimself fromaconstantpreoccupationwith food, shelter, andsafetybeforehecouldbegintodecoratehisclothing,hisdwelling,hisweapons,andhisbody,andeventuallycreatethingswithnootherfunctionthantobedecorative.Whencivilizationsreachthestage at which many people enjoy leisure, great periods of art often begin. Weacknowledge the connection when we encourage artists by giving them leisure—through patronage, fellowships, grants, or sinecures. The archetypal connectionbetweenartandbohemianismorbetweenartandlifeinagarretisinthesamepattern,for these are devices through which the artist gives himself leisure, by avoidingcommitmentsandlivingcheaply.The relation is easilymisunderstood. Certainly not everyone becomes an artist as

soonasheisfreetodoso.Andtheartistwillbethefirsttoinsistthatjustbecausehehasmanagedtodisposeofthethingswesayhe“has”todo,heisnotinanyrealsensefree. The serious, dedicated artist must do what he does as earnestly and ascompellinglyasothermenstruggleforfood,shelter,orsafety.Thedifferenceismerelyintheconspicuousnessofthecauses.Weusuallyknowwhypeoplebehaveastheydowhenthey“have”todoso,butlesscompellingreasonsareusuallylessobvious.Theyexist,however,andifwearegoingtoencouragepeopletobeartists(or,forthatmatter,consumersofart),weoughttoknowwhattheyare.

TheReasonsforArt

Why, indeed, do artists paint pictures? The traditional answers are not very helpful.Theyrefer toeventssupposedly takingplace inside theartisthimself.Theartistwhohasbeenfreedfromthepressuresoftheworldaroundhimissaidtobeabletoexpresshisindividuality,hiscreativeimpulses,hisloveofbeauty,theagonyandecstasyofhisinner struggles. These are engaging theories. They represent the artist as a complexpersonlivingadramaticlife,andtheygivehimexclusivecreditforthebeautifulthingshecreates.Butwehavenot really explained the artist’s achievement in termsofhisinnerlifeifwehavelearnedaboutthatlifeonlyfromhisachievements.

344

It is true that artists talk about themselves and very often about their inner lives.Theyarenodoubt inanexcellentposition toobserve thebehaviorofartists,butwemustaccepttheiraccountswithcaution.Whentheytalkabouttheiremotions,thoughts,ideas, and impulses, they necessarily use a vocabulary that they have learned frompeoplewho have had no contactwith these things andwho, therefore, cannot teachthemtodescribewhattheyobserveaccurately.Asaresult,everyartistgiveshisownidiosyncraticaccount,andhisanswertothequestion“Whydoestheartistpaintashedoes?”isprobablynomorehelpfulthanthatofanyoneelse.Nordoesthetraditionalviewhelpusinfurtheringtheproductionandenjoymentof

art.Wehavenodirectcontactwiththemindoremotionsoftheartist.Onlyindirectly,ifatall,canweinducehimtohavestrongfeelingsororiginalideas.Ifartspringsfromaninnerlifewhichistrulyoriginal,inthesensethatitbeginswiththeartist,thenthereis nothing to be done beyond giving the artist an opportunity. It is much morepromising,however, toargue that theachievementsof theartist canbe traced to theworld in which he lives, for we can then begin to examine that world not only toexplaintheachievementsbutalsotofindthemeansoftakingpracticalsteps.Thecolloquial“whatfor”isoftenausefulsynonymfor“why.”Whatdoartistspaint

picturesfor?Whatdopeople lookatpictures for?The termpoints to the future, andemphasizesanimportantfact.Artistspaintpicturesbecauseoftheconsequences, andpeople look at pictures because of the consequences. In traditional terms, theconsequencesmaybesaidtodefinethepurposesofart,butthereisamoreimportantimplication. The relation between behavior and its consequences has recently beenstudied in considerabledetail. Inhundredsof laboratories throughout theworld, in aspecialscientificdisciplinecalledtheexperimentalanalysisofbehavior,variouskindsofconsequencesaremadecontingentuponbehaviorincomplexways,andtheeffectsobserved.Certainkindsofconsequencesaresaidto“reinforce”behaviorinthesensethattheymakeitmorelikelytooccur.Weneednotgointodetailhere,butifwecandiscoverthereinforcerswhicharecontingentupontheartist’sbehaviorwhenhepaintsapicture,anduponthebehaviorofotherswhentheylookatapicture,wecannotonlyexplaintheirbehaviorbutalsouseourknowledgetogiveartamoreimportantroleinourculture.The word “reinforcing,” though technical, is useful as a rough synonym for

“interesting,” “attractive,” “pleasing,” and “satisfying,” and all these terms arecommonly applied to pictures. For our present purposes it is particularly useful as asynonym for “beautiful.”Pictures arebydefinition reinforcing in the sense that theyare responsible for the fact that artists paint them and people look at them. It is amistaketosupposethattheydothisbecauseofhowpeoplefeelaboutthem.Feelingsaremereby-products;theimportantthingiswhatapicturedoestobehavior.Theartistputspaintoncanvasandisorisnotreinforcedbytheresult.Ifheisreinforced,hegoesonpainting.Otherslookatthepictureandareorarenotreinforcedwhentheydoso.Iftheyarereinforced,theycontinuetolookandtoseekotherpicturestolookat.Tosomeextentwearereinforcedbypicturesforidiosyncraticreasons.Consumers

ofarthavedifferentpreferencesandsohaveartists.Theintegrityofanartist’sworkisinpartamatterofwhatfeatureshavereinforcedhim.Ifheisundulyreinforcedwhenhispicturessell,hemaybegintopaintpictureswhicharelikelytosell.Thereinforcingeffectofapictureis,however,significantforanotherreason.

345

TheCompetitorsofArt

Letuslookagainat therelationbetweenartandleisure.Thingswe“havetodo”areunderthecontrolofpowerfulreinforcers.Whenhungry,wearedominatedbybehaviorwhichhasbeenreinforcedbyfood.Whenunderathreat,weareabsorbedinavoidanceor escape.Butwhen free of powerful reinforcers,weare simplymore vulnerable toweakones.Leisurebringstheartistunderthecontrolofinconspicuousreinforcers.But works of art are not the only reinforcers which take over when serious

consequenceshavebeeneliminated.Membersoftheleisureclassarenotallartists.Onthecontrary, theleisuretobeobservedineitheranaffluentsocietyorawelfarestatehasavarietyofquitedifferenteffects.Itmaylead,forexample,toplay—tobehaviorresemblingseriousbehaviorbutexhibitedfor less thanseriousreasons(as inhuntingandfishing,whenwhatiscaught,captured,orkilledisnoteaten)ortoplayinggamesinwhichtrivialconsequenceshavebeenmadecritical(willtheballfallintothehole?).Gambling ischaracteristicof leisure,andallgamblingsystemsaredesigned tomaketheir consequences particularly effective bymaking them contingent on behavior inunpredictableways.Sexualbehaviorischaracteristicofleisurebecauseitisconcernedwith thesurvivalof thespeciesrather than the individual,and isnotsubject to long-termsatiation;inasenseitcannotbemadeunimportant.Syntheticreinforcers,suchasalcohol or marijuana, are among the nonessentials which take over in leisure.Spectatorship is a common feature, and recent technological advances havemade itpossibleforvastnumbersofpeopletowatchothersengageintheseriousbusinessoflife, as they undergo crucial personal experiences in drama, are hurt in football andothergames,orruntherisksofexplorationinspace.Thesearethenaturalcompetitorsofart.Howcanwegiveartachanceagainstthem?

Ifartneedsleisure,thenleisureneedsart.Inessence,howcanweinducetheartisttopaintmore pictures and everyone else to look at them?The obvious answer is this:make sure thatpainting and lookingare abundantly reinforced.Andherewe see theadvantageinsubstituting“reinforcing”for“beautiful.”Nomatterwhatotheraspectsofbeauty thecriticmaywish toemphasize,wearefor themoment interestedsimply inthereinforcingeffectofapicture.Whatisitaboutapicturethatmakestheartistmoreinclinedtocontinuepaintingitandtheviewertolookatitandtocontinuetodoso?

WhyaPictureIsReinforcing

From time to time answers have been sought in pictures themselves. Beauty, it hasbeen said, is in theobject, andwehaveonly to analyze a largenumberof beautifulobjects to discover it.But different people in different ages and in different cultureshavefounddifferentthingsbeautiful,andifthereareindeedobjectivepropertieswhichmakeathingbeautifulorreinforcing,theymustbelessimportantthanotherreasons.We say that a picture is reinforcing because of its content, but this is not a full

explanationofwhywelookatit.Itisreinforcingbecauseitresemblesrealthings,butthese are reinforcing for other reasons. An exhaustive account would explain all ofhuman behavior. Contingencies of survival in the evolution of the human race havemade the human form an important visual reinforcer, and the human figure is, ofcourse, common subject matter in the art of many cultures. Foodstuffs become

346

reinforcingforotherbiologicalreasons,anditshouldoccasionnosurprisethatpeoplehave, from time to time,hungpicturesof fruit, fish, andgame in their living rooms.Portraitsofpeopleweloveoradmirepermitustolookatthepeopleinabsentia,andwhentheyoungloverkisses theportraitofhisbeloved,he isonlyexemplifying inaconspicuousway thekindof thingweall dowhenwe lookat aportrait:webehavetoward the portrait to some slight extent as we should behave toward the personportrayed.Theartistwhoreliesmainlyoncontent tomakehispicturesreinforcing isusuallyhelddeficientinotherrespects.We turn from content to abstract form in a search for other kinds of reinforcing

thingswhicharelessobviouslytraceabletothehistoryoftheindividualorspecies.Butwemustnot forget thatwhatwedo about an abstraction is still the important thing.DenmanRosspointedthisoutin1907inhisTheoryofPureDesign,1 abookwhich,though not particularly influential, could be regarded as the manifesto ofabstractionism.Adesigninducestheviewertolookatitinaparticularway.Whathedoesmaybecommontosomanyvisualpresentationsthatwedivorceitfromcontentaltogether and regard it as possibly close to an essential artistic process. But if anabstract picture induces a painter to finish painting it and viewers to look at it, it isbecause ithas reinforcingproperties. (The lossofcontent is tosomeextenta lossofpower,andsomecontemporaryartistsoffsetthiswithauseofabstractmaterialwhichisasexaggeratedasthemostmaudlinsentimentalityinrepresentationalart.)Otherissuescommonlyencounteredindiscussionsofartcanberestatedintermsof

reinforcement. The history of art is to a large extent the history ofwhat artists andviewershavefoundreinforcing.Universalityistheuniversalityofreinforcingeffects.Changes in fashioncomeaboutas some reinforcers losepowerandothersgain.Theemphasisisimportantinitsbearinguponthepracticalproblemofimprovingtheplaceofartinaculture.Thereisnotmuchthenonartistcandotomakeartitselfmorereinforcing.Thatisthe

artist’sownfield,anditmustbelefttohim.Itishewhomustdiscovernewkindsofreinforcersinthesenseofnewformsofbeauty.Butthereareotherthingstobedone.There areways inwhich thenonartist canmake itmore likely that the artistwill bereinforcedbythepictureshepaints.Tobecomeanartist(or,inthecourseofacareer,tobecomeadifferentkindofartist)isaformoflearning.The“instruction”responsibleforitmaybeentirelyaccidental.Canitbedeliberate?Canweteachapersontobeanartist? So far as technique goes, the answer is yes. And that fact is relevant to thepresentissue.Themorecompetenttheartist,themorereinforcinghisworkislikelytobe—otherthingsbeingequal.Butwhataboutthe“otherthings”?Canweactuallyteachanartisthowtodiscoveror inventnewformsofbeautyin thesenseofnewkindsofreinforcers?The easy answer is “no,” and it is usually given by thosewho continue to regard

artisticachievementastheexpressionofaninnerlife.Suchalifeisnotdirectlywithinreachofateacher;geniusmustbelefttoworkitswayout.Andifthatisthenatureofart,then,apartfromtechnique,theteachercannotteachbutonlyhelptheartistlearn.Moreover,hemustthinktwiceaboutteachingtechniquelestheinterferewithuntaughtcreative expression. The position has the support of many educational philosophiesoutside the field of art where subjectmatter is abandoned in order to strengthen aninquiring spirit. The position is in essence a renunciation of teaching: the student issomehowbetteroffifheislefttodiscoverthingsforhimself.

347

The “discovery method” is particularly attractive to those who are interested inproducingoriginalartists,butweshouldfirstbesurethat itworks.It isbydefinitionnot amethod of instruction at all. It arose, in fact, from a concernwithmotivation.Whenteachersabandonedolderformsofdiscipline,theylostcontrol,andtotheextentthattheyhavenotfoundsuitablesubstitutes,itisquitecorrecttosaythattheycannolonger teach. And they have, therefore, been tempted to let students discoverknowledgeforthemselves.Butwedonotneedtoabandonsubjectmatterinordertoteachdiscovery.It isnot

truethatifwefillthestudent’sheadwithfactshewillbeunabletothinkforhimself.He is not damaged by facts but only by the ways in which facts have been taught.Thereisnoreasonwhymethodsofdiscoverymustbetaughtbythediscoverymethod.Learningthetechniquesofothersdoesnotinterferewiththediscoveryoftechniquesofone’sown.Onthecontrary,theartistwhohasacquiredavarietyoftechniquesfromhispredecessorsisinthebestpossiblepositiontomaketrulyoriginaldiscoveries.Andheismostlikelytobeoriginalifhehasbeentaughthowtodoso.Theveryassignmentofproducingacreativeartistmayseemcontradictory.Howcan

behavior be original or creative if it has been “produced”? Production presupposessomeformofexternalcontrol,butcreativity,takenliterally,deniessuchcontrol.Thatiswhywetendtoassociateitwithaninnerlife.ArthurKoestlerhastakenthislineinhisbook,TheActofCreation.ForKoestlerabehavioralanalysisofcreativity isnotonlyimpossiblebutludicrous,sincenoveltycannotariseina“mechanistic”system.Acreativemindmustbeatwork.Butacreativemindexplainsnothing.Itisanappealtothemiraculous:mindisbroughtintodowhatthebodycannotdo.Butwemustthenexplainhowtheminddoes it,and ifweaccept thatassignment,wediscover thatwehavemerelyrestatedouroriginalprobleminmuchmoredifficultterms.Novelty or originality can occur in a wholly deterministic system. A convenient

archetypal pattern is the theory of evolution. The living forms on the earth show avarietyfarbeyondthatofworksofart.ThediversitywasonceattributedtothewhimsandvagariesofacreativeMind,butDarwinproposedanalternativeexplanation.Theword“origin”inTheOriginofSpeciesisimportant,forthebookisessentiallyastudyoforiginality.Themultiplicityof living forms is accounted for in termsofmutationandselection,withoutappealingtoanypriordesign.Therearecomparableelementsinthebehavioroftheartistwhoproducesoriginalworks.Theartistfacingacleancanvasisinmuchthesamepositionasthewriterfacinga

cleansheetofpaper.Whatistobeputonit,andwhereisittocomefrom?(Thosewhoinsist that artists, like writers, must first have ideas must rephrase the questionaccordingly:Whatideasaretobeputonthecleanslateofthemind,andwheredotheycomefrom?)Therearesomesimpleanswers.Iftheartisthasalreadysuccessfullyputpaintoncanvas,heislikelytodothesamethingagain.Ifhehaslearnedtocopythingswhich are reinforcing, he can convert his clean canvas into a reinforcing object bycopyingsomethingwhichhasprovedreinforcingelsewhere.Itistempting,ofcourse,tocopyotherpictures,butwhentheotherpictureshavebeenpaintedbyotherartists,thecopieswillbethesourceoflittlesatisfactionorapproval.Itislegitimate,however,forartiststocopythemselves.OnlythefirstPicassowasnotderivative:alltheotherswerederivedfromearlierPicassos.Whatwecallanoriginalorcreativepaintingmustariseforotherreasons.Wemust

look for “mutations.”Many of these are accidental in the sense that they arise fromconditionswhichwecannotnowidentifyinthegeneticandenvironmentalhistoriesof

348

theartist and fromunpredictabledetails of hisworkingmethods and conditions.Wemay not like to credit any aspect of a successful painting to chance, but, if we arewillingtoadmitthatchancedoesmakeacontribution,wecantakestepstoimprovethechances.Mutationsmaybemademoreprobablebymaking thecontrolofamediumlesspreciseorbyencouragingdisturbances.DuringtheSecondWorldWarnewtypesof electronic equipment were used which could not be made wholly reliable in theavailable time, and an element was therefore sometimes introduced to keep theequipment in constantvibration. If a relay stuck, itwouldbe instantly shaken loose.Thevibrationwascalled“dither.”Theartistintroducesasourceofditherwhenheaddsanextralengthofhandletohisbrush,orpaintswithbitsofspongeinsteadofabrush,or pours paint on a horizontal canvas. He can generate mutations by changing hisworkingconditions,byworkingwhenheistired,cold,discouraged,ordrunk.Hecangenerateotherkindsofmutationsbydeliberatelydoingwhathehasbeentaughtnottodo;hecanviolatestandards,conventions,andtaboos,asamathematiciandeniesself-evidentaxiomsorasacomposerusespreviouslyforbiddenharmonies.Randomnessismostobviouslydeliberatewhentheartistspinsadial,throwsdice,orconsultsatableofrandomnumbersandputspaintoncanvasastheresultsdictate.Mutationmust,however,befollowedbyselection.Noteveryproductofcarelessness

—a cold studio, the deliberate rejection of a convention, or the roll of dice—is art.Putting paint on canvas is no more important than letting it stand, changing it, orscraping it off. The picture eventually left on the canvas is only one product of thecombinedprocessesofmutationandselection.Anartistwhowillhenceforthpaintinadifferent way is another. The selective side of the artist’s role emphasizes hisuniquenessand thealmost infinitevarietyof thecircumstancesunderwhichhe livesandpaints.Butselectionisalsolearnedandcanpresumablybetaught.Theyoungartistmaybetaught,forexample,totolerateeffectsheoncerejected,topermitsomefeaturestostandforthesakeofothers,tostoppaintingintime,andsoon.

TheRoleoftheConsumer

Apart from the idiosyncratic factorswhich play a part inmutation and selection, anartististosomeextentreinforcedwhenothersenjoyhiswork.Wecanhelpbymakingsurethathisworkwillbeenjoyed.Weshouldbeexplicitaboutthis:wewanttoinducemore people to look at more pictures and for longer periods of time—to seek outpicturestolookat,asbygoingtomuseums,andtobuypicturesinorderthattheymaybelookedat.Peopledoallthiswhenpicturesarereinforcing.Howcantheybemadeso?Theartisthimselfisconcernedwithproducingconsumers.Whetherornothepaints

primarilybecauseofeffectsuponhimself,heconstructspictureswhicharereinforcingtootherswhentheylookat them.Asnonartistswecanhelp.Wewantpicturestobemore valuable to those who look at them in the sense of being more effective ininducingthemtocontinuetolook.Thatisoneofthefunctionsofarteducation:peopleare taught to look at pictures in ways which aremore likely to be reinforced. “Artappreciation”isanaptexpression,foritisamatterofincreasingthereinforcingvalueofart,anditisappropriatetousethesametermtospeakofanappreciationinprice.Wecanscarcelybeproudofwhatisnowbeingdone,however.Onlyasmallpartof

theaverageschoolcurriculumisdevotedtotheenjoymentofart—or,forthatmatter,

349

musicor literature.It is themisfortuneofall threefieldstobetaughtasmysteries towhicheffectivemethodsofinstructionareheldnottoapply.Verylittlebeyondsimplecommunication and conspicuous enjoyment is attempted. The emphasis is on thefeelingsengenderedbybooks,music,andpictures.Studentsmaynotreadbooks,listentomusic,orlookatpictureswithoutfeelings,butanincreasedlikelihoodthattheywilldothesethingsshouldbethegoalofeducation.An important opportunity to encourage creative art by multiplying consumers is

often overlooked. A comparisonwithmusic is instructive. A generation or two agoveryfewpeoplecouldheargoodmusic.Foreveryonewhoheardasymphonyorchestraoropera,thousandsheardnomorethanthebrassbandonthevillagegreenortheparlorpiano. The phonograph and radio made a prodigious difference. Vast numbers ofpeople now hear music of unlimited variety and excellent quality. A recordedperformanceisoftenbetterthananactualsymphonyheardfrommanyoftheseatsinasymphonyhall. If there isnotyetaGoldenAge inmusicalcomposition, thestage iscertainlysetforone.AsRoyHarrishassaid,thelong-playingrecord“istomusicwhattheprintingpresswastoliterature.”Thereisnothinglikeityetinthefieldofart.Thosewho are engaged in reproducing pictures lack the power and zeal of the electronicsindustry.Mostcopiesofpaintingsarefragile,awkwardtohandle,andhardtostoreordisplay.Fidelity isgenerally low.Howlittleattention ispaid to this isevident in thefact that new issues of reproductions are not reviewed in the popular press as newrecordsare.What is needed, however, is not only better reproductions but a basic change in

attitude.Copiesofworksofartaresuspect.Nooneisbotheredbythefactthatagoodphonographisnotactuallyanorchestra,butmanypeoplearebotheredbythefactthatapictureisnotgenuine.Indeed,averygoodcopysuggestsaforgery.Anotherdifficultyisthatpeopleseldomchangethepicturesontheirwalls.Picturesareregardedaspartofthedecorationofa room,or theyarepermanentlydisplayedasvaluablepossessions.Theytheneitherceasetobenoticedorbecomeasobjectionableasbackgroundmusic.(Itmaybearguedthatoneneednotlookatapicturewhileonecannothelplisteningtomusic,butbackgroundmusicisnotalwaysheard.)Picturesshouldbeenjoyedasmusicisenjoyed,andtheyshouldbeaseasily“played.”

Dedication

Theproperreinforcementofartisticbehaviorcanhaveanotherimportanteffect.Lifeisgreaterthanart,andbothproducerandconsumerofartare,aswehaveseen,underthecontrolofrelativelyweakreinforcers.Butitissignificantthatthenaturalcompetitorsofartarealsoweaklyreinforced.Farmorepeoplevisitracetracksthanmuseums,buylotteryticketsthanpictures,andlookattelevisedfootballgamesthanpictures,butthecompetingreinforcersarenotactuallystronger.Thenetreinforcementattheracetrackorinalotteryisindeedalmostalwaysnegative(thegamblereventuallyloses),andthevictory of a favorite team could scarcely be less important in itself. Something hasmade thesereinforcerseffective,andthankstorecentresearchweknowwhatit is.Aweak reinforcer exerts a powerful control when effectively scheduled. All gamblingsystems and all games and sports “pay off” in a special unpredictable way. Thebehaviorofplacingabetorplayingagameisreinforcedonaso-calledvariable-ratioschedule, and the schedule generates a high level of activity. Under such schedules

350

pigeonsaswellasmenbecomepathologicalgamblers.Wecancreate“pathological”artistsandviewersofartwiththesamesystem.Apersonwho,aswesay,livesforart,forwhomartisthemostimportantthingin

the world, is not somuch one who finds art reinforcing as one who has enjoyed afavorable history of painting or looking at pictures. The technique with which adishonest gambler “hooks” his victim shows what needs to be done. The firstreinforcementsmustbequickandeasy,buttheaverageamountofbehaviordemandedfor each reinforcement must then slowly increase. Eventually the behavior ismaintained for longperiodsof time,possiblywithoutany reinforcementwhatsoever.Instruction in the enjoyment of art should begin with arrangements in whichreinforcement is generous, andperhaps “cheap.”Even themeretriciousmayhave itsplace.Moredifficult(“better”)materialsshouldbeintroducedwithcareasinstructionproceeds. It is often hard to arrange effective programs, but the teacher must notneglect the possibility of doing so. The dedicated artist, like the dedicated hunter,fisherman, explorer, or scientist, is the product of a probably accidental but happyprogramofsuccesses.Aneffectiveprogrammayarisenaturallyasanartiststrivesformoreandmoredifficulteffectsorasaviewerturnstoreinforcingfeaturesofapicturewhicharetobefoundlessandlessoften.

Philistinism?

Wehaveconsideredonlysomeofthemoreobviousreasonswhypeoplepaintandlookatpictures.Wehavescarcelytouchedonthemanydifferentkindsofthingstobefoundinpicturesorontheeffectstheyhaveonartistorviewer.(Andwehavenot,ofcourse,consideredotherformsofartthanpainting.)Wehavenottracedtheeffectsofpicturestopersonalhistories.Itiseasy,therefore,tosaythatthisaccountisoversimplified.Butthe central point deserves consideration. People paint and look at pictures for goodreasons—which can be investigated. Prominent among the reasons are certainreinforcingconsequences,andrecentadvancesintheexperimentalanalysisofbehaviorhaveshowntheirimportance.Thereisapracticalimplication:whenconsequencescanbemanipulated,behaviorcanbechanged.Technologicalapplicationsarealreadywelladvancedinotherfields.Whyshouldwenotusethisknowledgetoinducemorepeopletotakeaninterestinart?Perhaps it is a kindofPhilistinism to suppose thatwe can “produce” artists, as if

they were some sort of commodity. Certainly such a position is incompatible withmany traditional conceptions.But is it unrealistic?Art is producedby a culture or amixtureofcultures.Peoplecommonlyacttochangethecultureinwhichtheylive,andtheydosoinordertochangeitseffects.Wechangeourcultureinanefforttofurtherartwhenwesubsidizeartists,teachorencouragetheteachingofart,makeworksofartmoregenerallyavailable,andsoon,andsomeof thesemeasurescannowbegreatlyimproved.Evenso,theywillaffectonlyasmallpartoftheconditionsresponsibleforanyartist, let aloneanygreatartist.Thesearecircumstancesoverwhichwehavenocontrol, butwe can give them a better chance. The occasional chess genius ismostlikelytoappearinacultureinwhichmanypeopleplaychess,andagreatartistismostlikelytoarisewhentheproductionandconsumptionofartareimportantpartsofawayoflife.Ifwearewilling toaccept theassignmentofmakingourculturemoreeffective in

351

this way, then we should turn to a formulation of human behavior which points tothingstobedone.Traditionalexplanationshaveseldomledtoeffectiveaction;theyaresupportedprimarilybytheweightoftraditionandthefascinationoftheinexplicable.PerhapsnothinglessthanaresolutePhilistinismwillpermitustobuildthebackgroundfromwhich,forreasonsweadmitwedonotfullyunderstand,morecreativeartistswillemerge.

FromOntheFutureofArt.NewYork:VikingPress,1970.1Ross,DenmanW.(1907).NewYork:PeterSmith,1933.

352

ALectureon“Having”aPoem

What I amgoing to say has the curious property of illustrating itself. The quotationmarksinmytitleareintendedtosuggestthatthereisasenseinwhichhavingapoemislikehavingababy,andinthatsenseIaminlabor;Iamhavingalecture.InitIintendtoraisethequestionofwhetherIamresponsibleforwhatIamsaying,whetherIamactuallyoriginatinganything,andtowhatextentIdeservecreditorblame.Thatisoneissue inBeyond Freedom and Dignity,1 but since I am having a verbal baby, theargumentgoesbacktoanearlierbook.InhisreviewofBeyondFreedomandDignityinTheNewYorkTimes,2Christopher

Lehmann-Haupt beginswith two sentences dear to the hearts ofmy publishers, andtheyhavenotallowedthemtobecomehiddenunderabushel.Butlaterinthereview,unhappyaboutsomeof the implications,he tries tofaultme.“Well then,”hewrites,“whataboutthemostserious(andbestadvertised)attackthathasbeenleveledagainstbehaviorism in recent years—namely, Noam Chomsky’s attempts to demonstrateman’s innate linguistic powers, which began with Chomsky’s famous review ofSkinner’sbookVerbalBehavior.SkinnersaysnothingexplicitonthematterinBeyondFreedom and Dignity. Indeed, Chomsky’s name is never brought up (which seemsdisingenuousonSkinner’spart).Havewegothimthere?”Letme tell you about Chomsky. I publishedVerbalBehavior in 1957. In 1958 I

receiveda55-pagetypewrittenreviewbysomeoneIhadneverheardofnamedNoamChomsky.Ireadhalfadozenpages,sawthatitmissedthepointofmybook,andwentnofurther.In1959,IreceivedareprintfromthejournalLanguage.3ItwasthereviewIhadalreadyseen,nowreducedto32pagesintype,andagainIputitaside.Butthen,ofcourse,Chomsky’s star began to rise.Generative grammar became the thing—and avery big thing it seemed to be. Linguists have always managed to make theirdiscoveries earthshaking. In one decade everything seems to hinge on semantics, inanother decade on the analysis of the phoneme. In the sixties, it was grammar andsyntax,andChomsky’sreviewbegantobewidelycitedandreprintedandbecame,infact,muchbetterknownthanmybook.Eventuallythequestionwasasked,whyhadInotansweredChomsky?Myreasons,I

am afraid, show a lack of character. In the first place I should have had to read thereview,andI found its tonedistasteful. Itwasnot reallyareviewofmybookbutofwhatChomskytook,erroneously,tobemyposition.Ishouldalsohavehadtoboneupongenerativegrammar,whichwasnotmyfield,and todoagood jobIshouldhavehad to go into structuralism, a theory which Chomsky, like Claude Lévi-Strauss,acquired from Roman Jakobson. According to the structuralists we are to explainhumanbehaviorbydiscoveringitsorganizingprinciples,payinglittleornoattentiontothecircumstancesunderwhichitoccurs.Ifanythingbeyondstructureisneededbywayof explanation, it is to be found in a creativemind—Lévi-Strauss’s savagemind orChomsky’s innate rules of grammar. (Compare the recent analysis of Shakespeare’s

353

sonnet “Th’ expence of spirit” by Jakobson and Lawrence Jones4 with my earlieranalysis inVerbalBehavior. Where Jakobson and Jones confine themselves to thestructureorpatternofthepoemasitappearstothereader,Iusedthesamefeaturestoillustratethebehavioralprocessesofformalandthematicstrengtheningwhich,toputitroughly,madewords available to the poet as hewrote.) No doubt I was shirking aresponsibility in not replying to Chomsky, and I am glad an answer has now beensupplied byKennethMacCorquodale in theJournal of theExperimentalAnalysis ofBehavior.5AfewyearsagoNewsweekmagazinecarriedthedisagreementfurther,goingbeyond

linguisticsandstructuralismtothephilosophyoftheseventeenthcentury.Iwassaidtobe a modern disciple of John Locke, for whom themind began as a clean slate ortabularasaandwhothoughtthatknowledgewasacquiredonlyfromexperience,whileChomskywassaidtorepresentDescartes,therationalist,whowasnotsureheexisteduntilhethoughtaboutit.Newsweeksuggestedthatthebattlewasgoingmyway,andthereactionbythegenerativegrammarianswassoviolentthatthemagazinefounditnecessary to publish four pro-Chomsky letters. Each one repeated a commonmisunderstanding of my position. One implied that I was a stimulus-responsepsychologist (which I am not) and another that I think people are very much likepigeons(whichIdonot).Onehadatleastatouchofwit.Goingbacktooursupposedseventeenth-century progenitors, thewriter advisedNewsweek to “Locke up SkinnerandgiveChomskyDescartesblanche.”(ButChomskycannotuseacarteblanche,ofcourse;itistoomuchlikeatabularasa.)Ironically,ChomskywaslaterinvitedtogivetheJohnLockeLecturesatOxford.I

wasatCambridgeUniversityatthetime,andtheBBCthoughtitwouldbeinterestingifweweretodiscussourdifferencesontelevision.Idon’tknowwhatexcuseChomskygave, but I agreed to participate only if themoderator could guarantee equal time. Isuggested thatweusechessclocks.Myclockwouldbe runningwhenIwas talking,andChomsky’swhenhewastalking,andinthatwayIplannedtohavethelastfifteenortwentyminutestomyself.TheBBCthoughtthatmysuggestionwouldnotmakeforaveryinterestingprogram.VerbalBehaviorwascriticized inadifferentwaybyanold friend, I.A.Richards,

whoseinterestinthefieldgoesback,ofcourse,totheMeaningofMeaning.Fornearlyforty years Ivor Richards and I have respected each other while disagreeing ratherviolently. I have never been able to understand why he feels that the works ofColeridgemakeanimportantcontributiontoourunderstandingofhumanbehavior,andhehasneverbeenabletounderstandwhyIfeelthesamewayaboutpigeons.Hehasattimesbeendeeplydistressed.Heonceaskedme to lecture tohis freshmancourse inGeneralEducation.Iturnedupattheappointedhour,hemadeafewannouncements,andthenhesaid,“InowpresenttheDevil,”andsatdown.AndIhadnotyetpublishedVerbalBehavior, that outrageous invasionofRichards’ territorywhichmight indeedhavebornethesubtitle,TheMeaninglessnessofMeaning.Whenmybookappeared,and in turnChomsky’s review, IvorRichards sentmea

poem.Itwasprefacedbytwoquotations,onefrommybookandonefromthereview,anditproceededtodocumenttheextraordinaryextenttowhicheachofusbelievedthathewasabsolutelyright.Thepoembegan:

Confidencewithconfidenceoppose.

354

KnowledgeducksunderinbetweentwoNo’sSofirmlyuttered.Lookagain.You’llseeUncertaintybesideuncertainty.

Someunacknowledgeduncertaintieswerethencitedandanalyzed.AfewmonthslaterIreceivedasecondpoem.Itwascalled“VerbalBehaviour”and

beganasfollows:

Nosenseinfrettingtobeofftheground,There’sneverhurrywhitherwearebound,Whereall’sbehaviour—andtherestisnaught,Notevenrest,butvoidbeyondallthought.

Itwentontoarguethatbehaviorismwillmeanthedeathoftheindividual,theendofman’sdivineimageofhimself.Thebehavioristcontendsthat

TheAngelsareasketchTheymadelongsincetocomfortthesorewretchCastoutofParadiseheknewnotwhyTostarthislongclimbbackintothesky.

Buthewillneverreachparadiseagainbecausethebehavioristwilltearoffhiswings,crying,tosethimfree:

Thesegleamingsailsarebuttheflatteringmeans,(Theologicgear,Pythagoreanbeans!)Wherebygrubsflitandfeedandlaytheireggs,Bymetaphor,beyondthereachoflegs.Nopsychemore!Homunculus-theory,out!Verbalbehaviour’sallit’sallabout.

Itseemedtomethatthishadgonefarenough,andsoIreplied—inkind—asfollows:ForIvorRichards

Yes,“all’sbehavior—andtherestisnaught.”

AndthuscompressedInto“therestOfall,”AthoughtIssurelyneitherbadnorwrong.

Orrightorgood?No,no.DefineAndthusexpungeTheought,Theshould!Nothingisso(SeeHistory.)

355

LetnotthestrongBecozenedByIsandIsn’t,WasandWasn’t.Truth’stobesoughtInDoesandDoesn’t.

DeclineTobe.

AndcallHimneitherbestNorblessedWhowroughtThatsillyjest,TheFall.

(ItwasaPlunge.)

A few days later Ivor Richards phoned. Why not publish our poems? I had noobjection,andsohesentthemtoanAmericanmagazine.Theeditoragreedthat theywereinterestingbutthat,sincewewerebothatHarvard,itwasasortofin-housejokewhichmightnotappealtotheirWestCoastreaders.StephenSpender,however,hadnoWest Coast readers to worry about, and our poems were eventually published inEncounter.6ThatistheonlypoemIhavepublishedsincecollege,anditmustserveasmyonly

credential in discussing the present topic. I am unwilling to let it stand withoutcomment,andsoIofferthefollowingexegesis,asitmightbewrittenbysomefuturecandidateforaPh.D.inEnglishliterature.

Thepoetbeginswithaquotationfromhisfriend’spoem,pickingupaslightredundancy.If all’s behavior, then of course the rest is naught.And it is perhaps just aswell, since athoughtreducedtonothingcanscarcelybebadorwrong.Butwhataboutthepossibilitythatitmightberightorgood?No,logicalpositivismwilltakecareofthat.Bydefiningourvaluesweexpungethem.A new theme then appears, perhaps best stated in the immortal words of Henry Ford,

“History isbunk,”buthereextended to thepresentaswell as thepast, as thepoetattacksexistentialismaswellastheusesofhistory.Thethemeisbroachedinthecontemptuouslines

Nothingisso(SeeHistory.)

but developedmore explicitlywhenwe arewarnednot to bedeceivedby Is and Isn’t (somuchfortheExistentialists)orWasandWasn’t(somuchfortheHistorians).Thenfollowsthatstirringbehavioristicmanifesto:

Truth’stobesoughtInDoesandDoesn’t.

Atthispoint,almostasifexhausted,thepoetentersuponanewmood.Behaviorismhassqueezedthoughttodeathandwithitconsciousnessandmind.GeorgeKatebmadethepoint

356

later inhis reviewofBeyondFreedomandDignity in theAtlanticMonthly,7 insisting that“Skinner foresees and condones the atrophy of consciousness.” But since foreseeing andcondoningareconsciousacts,thebehavioristisengagedinakindofintellectualsuicide.Touseastrangelyineptexpression,mindistodiebyitsownhand.Thepositionisstatedwithstunningeconomy:

DeclineTobe.

(Wenote here a certain infelicity.Tobe is a verb, and as such canbe conjugatedbut notdeclined.Butwemustrememberthatanintentionalsuicideislikelytobedistraughtandanunintentional one at least careless.One thinksofOphelia.The semantic blemish thereforesimplyaddstothetoneofthepassage.)The themeof suicidebecomesclearwhen thepoet turns tohis friend’s reference to the

fallenAngelandwarnsusagainstacceptinguncritically“thatsillyjest,TheFall.”(NoteinpassingthatsillyiscognatewiththeGermanselig,meaningholyorsacred.)“Fall”iswrongbecauseitsuggestschance.(Chance,ofcourse,comesfromtheLatincaderemeaningto fall—thefallofadieorpenny—andisitentirelyirrelevantthat“jest”isetymologicallyrelatedto “cast,” as in casting dice?) Make no mistake; the Fall was not an accident. It was adeliberateplunge.

Thusmightsomebeknightedgraduatestudentofthefuturewriteinsearchofpartialfulfillment.Whetherornothewillthusestablishmycompetenceindiscussingpoetry,Icannot say. If not, I must fall back upon that stock reply of the critic when theplaywrightwhohasreceivedabadreviewpointsoutthatthecritichasneverwrittenaplay,andthecriticreplies,“NeitherhaveI laidanegg,butIamabetter judgeofanomeletthananyhen.”Itisastaleandmustyjoke,andIshouldnotallowittoinjurethetoneofmylectureifitdidnotservetheimportantfunctionofbringingmetomypoint.Iamtocomparehavingapoemwithhavingababy,anditwilldonoharmtostartwithalowerclassoflivingthings.SamuelButlersuggestedthecomparisonyearsagowhenhesaidthatapoetwritesapoemasahenlaysanegg,andbothfeelbetterafterwards.But there are other points of similarity, and on one of themButler built a whole

philosophyofpurposiveevolution.Thestatementwascurrentinearlypost-Darwinismdays that“ahen isonlyanegg’swayofmakinganotheregg.” It isnot,ofcourse,aquestionofwhichcomesfirst, thoughthat isnotentirelyirrelevant.Theissueiswhodoeswhat,whoactstoproducesomethingandthereforedeservescredit.Mustwegivethehencreditfortheeggortheeggforthehen?Possiblyitdoesnotmatter,sincenooneisseriouslyinterestedindefendingtherightsofhenoregg,butsomethingofthesame sort can be said about a poet, and then it does matter. Does the poet create,originate,initiatethethingcalledapoem,orishisbehaviormerelytheproductofhisgeneticandenvironmentalhistories?Iraisedthatquestionanumberofyearsagowithadistinguishedpoetataconference

at Columbia University. I was just finishing Verbal Behavior and could not resistsummarizingmyposition.I thought itwaspossible toaccountforverbalbehavior intermsofthehistoryofthespeaker,withoutreferencetoideas,meanings,propositions,andthelike.Thepoetstoppedmeatonce.Hecouldnotagree.“Thatleavesnoplaceformeasapoet,”hesaid,andhewouldnotdiscussthematterfurther.Itwasacasualremarkwhich,Iamsure,hehaslongsinceforgotten,andIshouldhesitatetoidentifyhimifhehadnotrecentlypublishedsomethingalongthesamelines.

357

When Jerome Weisner was recently inaugurated as President of MassachusettsInstituteofTechnology,ArchibaldMacLeishreadapoem.8HepraisedDr.Weisneras:

Agoodmaninatimewhenmenarescarce,whentheintelligentforegather,followeachotheraroundinthefoglikesheep,bleatintherain,complainbecauseGodotnevercomes;becausealllifeisatragicabsurdity—Sisyphussweatingawayathisrock,andtherockwon’t;becausefreedomanddignity…

Oh,weep,theysay,forfreedomanddignity!You’renotfree:it’syourgrandfather’sitchyou’rescratching.Youhavenodignity:you’renotaman,you’rearatinavatofrewardsandpunishments,youthinkyou’vechosentherewards,youhaven’t:therewardshavechosenyou.

Aye!Weep!

I am just paranoid enough to believe that he is alluding toBeyond Freedom andDignity.Inanycase,hesumsupthemainissuerathereffectively:“Youthinkyou’vechosen the rewards; you haven’t. The rewards have chosen you.” To put it morebroadly,apersondoesnotactupontheenvironment,perceivingitanddecidingwhattodoaboutit;theenvironmentactsuponhim,determiningthathewillperceiveitandactinspecialways.GeorgeEliotglimpsedtheissue:“Ourdeedsdetermineus,asmuchaswedetermineourdeeds,”thoughshedidnotunderstandhowwearedeterminedbyourdeeds. Something does seem to be taken away from the poet when his behavior istraced to his genetic and personal histories. Only a person who truly initiates hisbehavior can claim that he is free to do so and that he deserves credit for anyachievement. If the environment is the initiating force, he is not free, and theenvironmentmustgetthecredit.The issue will be clearer if we turn to a biological parallel—moving from the

oviparoushentotheviviparoushumanmother.Whenwesaythatawoman“bears”achild,wesuggestlittlebywayofcreativeachievement.Theverbreferstocarryingthefetustoterm.Theexpression“givesbirth”goeslittlefurther;abitofaplatonicidea,birth,iscapturedbythemotherandgiventothebaby,whichthenbecomesborn.Weusually say simply that a woman “has” a baby where “has”means little more thanpossess.Tohaveababyis tocomeintopossessionof it.Thewomanwhodoessoisthenamother,andthechildisherchild.Butwhatisthenatureofhercontribution?Sheisnot responsible for the skincolor, eyecolor, strength, size, intelligence, talents,oranyother featureofherbaby.Shegave it half itsgenes,but shegot those fromherparents.Shecould,ofcourse,havedamagedthebaby.Shecouldhaveabortedit.Shecouldhavecaught rubellaat thewrong timeor takendrugs,andasa result thebabywouldhavebeendefective.Butshemadenopositivecontribution.Abiologisthasnodifficulty indescribing theroleof themother.She isaplace,a

locusinwhichaveryimportantbiologicalprocesstakesplace.Shesuppliesprotection,warmth,andnourishment,butshedoesnotdesignthebabywhoprofitsfromthem.The

358

poet isalsoa locus,aplace inwhichcertaingeneticandenvironmentalcausescometogether tohaveacommoneffect.Unlikeamother, thepoethasaccess tohispoemduring gestation. Hemay tinker with it. A poem seldommakes its appearance in acompletedform.Bitsandpiecesoccurtothepoet,whorejectsorallowsthemtostand,andwhoputsthemtogethertocomposeapoem.Buttheycomefromhispasthistory,verbal and otherwise, and he has had to learn how to put them together. The act ofcompositionisnomoreanactofcreationthan“having”thebitsandpiecescomposed.Butcanthisinterpretationbecorrectifapoemisunquestionablynew?Certainlythe

playsofShakespearedidnotexistuntilhewrotethem.PossiblyalltheirpartscouldbetracedbyanomniscientscholartoShakespeare’sverbalandnonverbalhistories,buthemust have served some additional function. How otherwise are we to explain thecreationofsomethingnew?Theanswerisagaintobefoundinbiology.Alittlemorethanahundredyearsago

the act of creationwas debated for a very different reason.The living things on thesurfaceof theearth showa fantasticvariety—farbeyond thevariety in theworksofShakespeare—and theyhad longbeenattributed toacreativeMind.Theanatomyofthehand,forexample,wastakenasevidenceofapriordesign.Andjustaswearetoldtoday that a behavioral analysis cannot explain the “potentially infinite” number ofsentences composable by a speaker, so it was argued that no physical or biologicalprocesscouldexplainthepotentiallyinfinitenumberoflivingthingsonthesurfaceoftheearth.(Curiouslyenoughthecreativebehaviorinvokedbywayofexplanationwasverbal: “In the beginningwas theword…,” supplemented no doubt by agenerativegrammar.)ThekeyterminDarwin’stitleisOrigin.Noveltycouldbeexplainedwithoutappeal

topriordesignifrandomchangesinstructurewereselectedbytheirconsequences.Itwasthecontingenciesofsurvivalwhichcreatednewforms.Selectionisaspecialkindofcausality,muchlessconspicuousthanthepush-pullcausalityofnineteenth-centuryphysics, andDarwin’s discoverymayhave appeared so late in the history of humanthoughtforthatreason.Theselectiveactionoftheconsequencesofbehaviorwasalsooverlookedforalongtime.Itwasnotuntiltheseventeenthcenturythatanyimportantinitiatingactionbytheenvironmentwasrecognized.Peopleactedupontheworld,butthe world did not act upon them. The first evidence to the contrary was of theconspicuouspush-pullkind.Descartes’s(paceChomsky)theoreticalanticipationofthereflex and the reflex physiology of the nineteenth century gave rise to a stimulus-responsepsychology inwhichbehaviorwassaid tobe triggeredby theenvironment.Thereisnoroominsuchaformulationforamoreimportantfunction.Whenapersonacts,theconsequencesmaystrengthenhistendencytoactinthesamewayagain.TheLaw of Effect, formulated nearly three quarters of a century ago by Edward L.Thorndike,owedagreatdealtoDarwiniantheory,anditraisedverysimilarissues.Itisnotsomepriorpurpose,intention,oractofwillwhichaccountsfornovelbehavior;itisthe “contingencies of reinforcement.” (Among the behaviors thus explained aretechniques of self-management, once attributed to “highermental processes,” whichfigureinthegestationofnewtopographies.)Thepoetoftenknowsthatsomepartofhishistoryiscontributingtothepoemheis

writing.Hemay,forexample,rejectaphrasebecauseheseesthathehasborroweditfromsomethinghehasread.But it isquite impossibleforhimtobeawareofallhishistory,andit isinthissensethathedoesnotknowwherehisbehaviorcomesfrom.Having a poem, like having a baby, is in large part a matter of exploration and

359

discovery, andboth poet andmother are often surprised bywhat they produce.Andbecausethepoetisnotawareoftheoriginsofhisbehavior,heislikelytoattributeittoacreativemind,an“unconscious”mind,perhaps,oramindbelongingtosomeoneelse—toamuse,forexample,whomhehasinvokedtocomeandwritehispoemforhim.Apersonproducesapoemandawomanproducesababy,andwecallthepersona

poetandthewomanamother.Bothareessentialaslociinwhichvestigesofthepastcome together in certain combinations. The process is creative in the sense that theproductsarenew.Writingapoemisthesortofthingmenandwomendoasmenandwomen,havingababyisthesortofthingawomandoesasawoman,andlayinganeggisthesortofthingahendoesasahen.TodenyacreativecontributiondoesnotdestroymanquamanorwomanquawomananymorethanButler’sphrasedestroyshenquahen.Thereisnothreattotheessentialhumanityofman,themuliebrityofwoman,orthegallityofGallusgallus.What is threatened,ofcourse, is theautonomyof thepoet.Theautonomous is the

uncaused,andtheuncausedismiraculous,andthemiraculousisGod.Forthesecondtimeinalittlemorethanacenturyatheoryofselectionbyconsequencesisthreateningatraditionalbeliefinacreativemind.Andisitnotratherstrangethatalthoughwehaveabandonedthatbeliefwithrespecttothecreationoftheworld,wefightsodesperatelytopreserveitwithrespecttothecreationofapoem?Butisthereanythingwrongwithasupportivemyth?Whynotcontinuetobelievein

ourcreativepowersifthebeliefgivesussatisfaction?Theanswerliesinthefutureofpoetry.Toacceptawrongexplanationbecauseitflattersusistoruntheriskofmissingarightone—onewhichinthelongrunmayoffermorebywayof“satisfaction.”Poetsknow all too well how long a sheet of paper remains a carteblanche. To wait forgeniusoragenieistomakeavirtueofignorance.Ifpoetryisagoodthing,ifwewantmoreofitandbetter,andifwritingpoemsisarewardingexperience,thenweshouldlookafreshatitssources.Perhaps the future of poetry is not that important, but I have been using a poem

simplyasanexample. I couldhavedeveloped the same theme inart,music, fiction,scholarship,science,invention—inshort,whereverwespeakoforiginalbehavior.Wesay thatwe“have” ideasandagain in thesimplesenseofcoming intopossessionofthem.Anidea“occurstous”or“comestomind.”Andifforideaweread“thebehaviorsaidtoexpressanidea,”wecomenoclosertoanactofcreation.We“have”behavior,astheetymologyoftheworditselfmakesclear.It“occurstous”toactinaparticularway,anditisnotanypriorintention,purpose,orplanwhichdisposesustodoso.Byanalyzing the genetic and individual histories responsible for our behavior, wemaylearnhowtobemoreoriginal.Thetaskisnottothinkofnewformsofbehaviorbuttocreateanenvironmentinwhichtheyarelikelytooccur.Somethingofthesorthashappenedintheevolutionofcultures.Overthecenturies

menandwomenhavebuiltaworldinwhichtheybehavemuchmoreeffectivelythanin a natural environment, but they have not done so by deliberate design.A cultureevolves when new practices arise which make it more likely to survive. We havereachedastageinwhichourcultureinducessomeofitsmemberstobeconcernedforits survival.A kind of deliberate design is then possible, and a scientific analysis isobviouslyhelpful.Wecanbuildaworldinwhichmenandwomenwillbebetterpoets,betterartists,bettercomposers,betternovelists,betterscholars,betterscientists—inaword,betterpeople.Wecan,inshort,“have”abetterworld.AndthatiswhyIamnotmuchdisturbedbythequestionwithwhichGeorgeKateb

360

concludeshisreviewofBeyondFreedomandDignity.Heisattackingmyutopianism,andheasks,“DoesSkinnernotseethatonlysillygeeselaygoldeneggs?”Thequestionbringsusbacktotheoviparousagain,butitdoesnotmatter,fortheessentialissueisraisedbyalllivingthings.Itischaracteristicoftheevolutionofaspecies,asitisoftheacquisitionofbehaviorandoftheevolutionofaculture,thatineffectiveformsgiverisetoeffective.Perhapsagooseissillyif,becauseshelaysagoldenegg,shegetstheax;but, silly or not, she has laid a golden egg. Andwhat if that egg hatches a goldengoose?There,inaneggshell,isthegreatpromiseofevolutionarytheory.Asillygoose,likeButler’shen,issimplythewayinwhichaneggproducesabetteregg.Andnowmylaborisover.Ihavehadmylecture.Ihavenosenseoffatherhood.If

my genetic and personal histories had been different, I should have come intopossessionofadifferent lecture. If Ideserveanycreditatall, it issimplyforhavingserved as a place inwhich certain processes could take place. I shall interpret yourpoliteapplauseinthatlight.

ThislecturewasgivenatthePoetryCenterinNewYorkCityonOctober13,1971.1BeyondFreedomandDignity.NewYork:Knopf,1971.2TheNewYorkTimes,September22,1971.3Language,1959,35,26–58.4Shakespeare’sVerbalArtin“Th’ExpenceofSpirit.”TheHague:Mouton,1970.5JournaloftheExperimentalAnalysisofBehavior,1970,13,83–99.6Encounter,Nov.1962.7AtlanticMonthly,Oct.1971.8BostonGlobe,October9,1971.

361

PARTVII

LiteraryandVerbalBehavior

HasGertrudeSteinaSecret?TheOperationalAnalysisofPsychologicalTermsTheAlliterationinShakespeare’sSonnets:AStudyinLiteraryBehaviorAQuantitativeEstimateofCertainTypesofSound-PatterninginPoetryTheProcessesInvolvedintheRepeatedGuessingofAlternatives

362

HasGertrudeSteinaSecret?

IfirstheardabouttheAutobiographyofAliceB.Toklas fromMaryLouiseWhite(Aswell).IthadreachedherdeskintheeditorialofficesoftheAtlanticMonthlyasabonafideautobiography,butthelastparagraphhadcomeasnosurprise:“AboutsixweeksagoGertrudeSteinsaid,itdoesnotlooktomeasifyouwereevergoingtowritethatautobiography.YouknowwhatIamgoingtodo.Iamgoingtowriteitforyou.IamgoingtowriteitassimplyasDefoedidtheautobiographyofRobinsonCrusoe.Andshehasandthisisit.”MissWhitewastobetheonlyAtlanticreadertoenjoyGertrudeStein’slittlejokefor

whenpartsof thebookwerepublishedin thatmagazineduring thesummerof1933,the author’s name appeared on the title page. I read the reference to Miss Stein’spsychological experiments while vacationing on Monhegan Island, Maine. Uponreturning toCambridge Imadeatonce for the libraryand thateveningreportedmy“discovery” toMissWhite,who suggested the present article. I called it “GertrudeSteinandAutomaticWriting.”TheeditoroftheAtlantic,EllerySedgwick,hadalesspedanticsuggestion.The article had the accidental distinction of being the first published by a Junior

PrizeFellowintheSocietyofFellowsatHarvard.TheSocietyhadbeenconceivedbyLawrenceJ.Hendersonandestablishedwithan“anonymous”giftfromA.LawrenceLowell,thenjustretiringfromthepresidencyofHarvard.Itsobjectwastobringyoungscholars in various fields into contact with each other and with some of thedistinguishedmembersof theHarvard faculty. Inaddition toHendersonandLowell,the first Senior Fellows included Alfred North Whitehead, John Livingston Lowes,JamesB.Conant,KennethB.Murdock,andCharlesP.Curtis.Weweretomeetweeklyforanexcellentdinner(ProfessorHendersonhadstockedthecellaroftheSocietywithanhonestBurgundyfromthevineyardsofafriendattheUniversityofDijon)andforthecross-fertilizationofideas.ThemountainlaboredandbroughtforthanarticleonGertrudeStein.Itprovedto

be an offspring not easily concealed. Conrad Aiken, writing in the New Republic,reviewedtheprevailingattitudetowardMissStein’sexperiment—“Likethesplittingoftheatom,orthetheoryofrelativity,MissStein’sdestructionofmeaningwasinevitablygoing tochange, ifnot theworld,atanyrate theword”—andexpressed theopinionthatthepresentarticlemadeofthewholethingaverycrueljoke.“Whatbecomesofallthispreciseanddetachedandscientificexperimentationwithrhythmandmeaning, if,after all, it has been nothing on earth but automaticwriting? Is itmerely onemoreinstance of the emperor’s new clothes? Have we been duped, and has Miss Steinherself,perhaps,beenduped?Itlooksverylikeit—thoughofcourseitisnotimpossiblethatMissSteinhasbeenpullingourlegs.”AmongthosewhocametothedefensewasSherwoodAnderson,whoquiteproperlyrejoinedthat“allgoodwritingis,inasense,automatic.”ButpartisansofGertrudeSteinhavebeenafraidofthisanswerandhavegenerallydenied thatherpublishedworkwaswrittenautomatically.Fortunately,we

363

haveGertrudeStein’sownversionofthefacts,expressedinalettertoEllerySedgwick:“Noitisnotsoautomaticashethinks.Ifthereisanythingsecretitistheotherway

too. I thinkIachievebyxtraconsciousness,xcess,but thenwhat is theuseof tellinghim that,hebeingapsychologistandIhavingbeenone.Besideswhenhe isnot tooseriousheisaprettygoodone.”In theAutobiographyofAliceB.ToklasGertrudeStein tells in the followingwayofsomepsychologicalexperimentsmadebyheratHarvard:

Shewas one of a group of Harvardmen and Radcliffe women and they all lived verycloselyandveryinterestinglytogether.Oneofthem,ayoungphilosopherandmathematicianwhowas doing researchwork in psychology, left a definitemark on her life. She and hetogether worked out a series of experiments in automatic writing under the direction ofMünsterberg. The results of her own experiments, which Gertrude Stein wrote down andwhichwasprintedintheHarvardPsychologicalReview,wasthefirstwritingofhersevertobe printed. It is very interesting to read because the method of writing to be afterwardsdevelopedinThreeLivesandTheMakingofAmericansalreadyshowsitself.

ThereisagreatdealmoreinthisearlypaperthanMissSteinpointsout.Itis,asshesays,ananticipationoftheprosestyleofThreeLivesandisunmistakablytheworkofGertrudeStein in spiteof the conventional subjectmatterwithwhich it deals.Manyturns of speech, often commonplace, which she has since then in some subtle waymadeherownarealreadytobefound.Butthereismuchmorethanthis.ThepaperisconcernedwithanearlyinterestofMissStein’swhichmusthavebeenveryimportantin her later development, and the work which it describes cannot reasonably beoverlookedbyanyonetryingtounderstandthisremarkableperson.Sincethepaperishardtoobtain,Ishallsummarizeitbriefly.Itwaspublishedinthe

Psychological Review for September, 1896, under the title, “Normal MotorAutomatism,”byLeonM.SolomonsandGertrudeStein,anditattemptedtoshowtowhatextenttheelementsofa“secondpersonality”(ofthesorttobeobservedincertaincasesofhysteria)weretobefoundinanormalbeing.Intheirexperimentstheauthorsinvestigated the limits of their own normal motor automatism; that is to say, theyundertooktoseehowfartheycould“split”theirownpersonalitiesinadeliberateandpurelyartificialway.Theyweresuccessfultotheextentofbeingabletoperformmanyacts(suchaswritingorreadingaloud)inanautomaticmanner,whilecarryingonatthesametimesomeotheractivitysuchasreadinganinterestingstory.

II

In the experiments with automatic writing a planchette of the ouija board type wasoriginallyused,but as soonas the authorshad satisfied themselves that spontaneouswriting movements do occur while the attention is directed elsewhere, an ordinarypencil and paperwere used instead. The subject usually began bymaking voluntaryrandomwritingmovements or bywriting the letterm repeatedly. In one experimentthiswasdonewhile thesubjectreadaninterestingstoryat thesametime,anditwasfoundthatsomeofthewordsreadinthestorywouldbewrittendowninanautomaticway. At first there was a strong tendency to notice this as soon as it had begun tohappenandtostopit,buteventuallythewordscouldbewrittendownunconsciouslyas

364

wellasinvoluntarily.(IshalluseMissStein’spsychologicalterminologythroughout.)“Sometimes thewritingof thewordwascompletelyunconscious,butmoreoften thesubjectknewwhatwasgoingon.Hisknowledge,however,wasobtainedbysensationsfromthearm.Hewasconsciousthathejusthadwrittenaword,notthathewasabouttodoso.”Inotherexperimentsthesubjectreadaninterestingstoryasbefore,andsinglewords

were dictated to him to be written down at the same time. These were difficultexperiments, but after considerable practice they were successful. The subject waseventuallyable towritedown“fiveorsix”wordsspokenbyanotherperson,withoutbeingconsciousofeithertheheardsoundsorthemovementofthearm.Ifhisattentionwerenotsufficientlywelldistracted,hemightbecomeawarethathishandwaswritingsomething. The information came from the arm, not from the sound of the dictatedword. “It is never the sound that recalls us. This, of course, may be an individualpeculiaritytoacertainextent….Yet,MissSteinhasastrongauditoryconsciousness,andsoundsusuallydeterminethedirectionofherattention.”In a third group of experiments the subject read aloud, preferably from an

uninteresting story, while being read to from an interesting one. “If he does not goinsaneduringthefirstfewtrials,hewillquicklylearntoconcentratehisattentionfullyonwhat isbeingread tohim,yetgoonreading just the same.The readingbecomescompletelyunconsciousforperiodsofasmuchasapage.”Automaticreadingofthissortisprobablypartoftheexperienceofeveryone.The fourth and last group brings out the relevance of the experiments to the later

workofGertrudeStein.IshallletMissSteindescribetheresult.

Spontaneousautomaticwriting.—This became quite easy after a little practice.We hadnow gained so much control over our habits of attention that distraction by reading wasalmostunnecessary.MissSteinfounditsufficientdistractionoftentosimplyreadwhatherarmwrote,butfollowingthreeorfourwordsbehindherpencil.Aphrasewouldseemtogetintotheheadandkeeprepeatingitselfateveryopportunity,

andhangoverfromdaytodayeven.Thestuffwrittenwasgrammatical,andthewordsandphrases fitted together all right, but there was not much connected thought. Theunconsciousnesswasbrokenintoeverysixorsevenwordsbyflashesofconsciousness,sothatonecannotbesurebutwhattheslightelementofconnectedthoughtwhichoccasionallyappearedwasduetotheseflashesofconsciousness.Buttheabilitytowritestuffthatsoundsallright,withoutconsciousness,wasfairlywelldemonstratedbytheexperiments.Hereareafewspecimens:“Hence there is no possible way of avoiding what I have spoken of, and if this is not

believedbythepeopleofwhomyouhavespoken,thenitisnotpossibletopreventthepeopleofwhomyouhavespokensoglibly….”Hereisabitmorepoeticalthanintelligible:“Whenhecouldnotbethelongestandthustobe,andthustobe,thestrongest.”Andhereisonethatisneither:“Thislongtimewhenhedidthisbesttime,andhecouldthushavebeenbound,andinthis

longtime,whenhecouldbethistofirstuseofthislongtime….”

III

365

Hereisobviouslyanimportantdocument.NoonewhohasreadTenderButtonsor thelaterworkinthesameveincanfailtorecognizeafamiliarnoteintheseexamplesofautomaticwriting. They are quite genuinely in themanner which has so commonlybeen taken as characteristic of Gertrude Stein. Miss Stein’s description of herexperimental result is exactly that of the average reader confronted with TenderButtons for the first time: “The stuff is grammatical, and the words and phrases fittogether all right, but there is notmuch connected thought.” In short, the case is sogood,simplyonthegroundsofstyle,thatwearebroughttotheswiftconclusionthatthe two products have a commonorigin, and that thework ofGertrudeStein in theTenderButtonsmanner iswrittenautomaticallyandunconsciously insomesuchwayasthatdescribedinthisearlypaper.This conclusion growsmore plausible aswe consider the case. It is necessary, of

course,todistinguishbetweentheGertrudeSteinofThreeLivesandtheAutobiographyand theGertrudeStein ofTenderButtons, a distinctionwhich is fairly easilymade,eventhough,asweshallseeinamoment,thereissomeofthefirstGertrudeSteininthe latter work. If we confine ourselves for the present to the second of these twopersons,itisclearthatthehypotheticalauthorwhomightbeinferredfromthewritingitselfpossessesjustthosecharacteristicswhichweshouldexpecttofindifatheoryofautomaticwritingwere the right answer.Thus there is very little intellectual contentdiscoverable.Thereader—theordinaryreader,atleast—cannotinferfromthewritingthatitsauthorpossessesanyconsistentpointofview.Thereisseldomanyintelligibleexpressionofopinion,andthereareenoughcapriciousreversalstodestroytheeffectofwhatevertheremaybe.Thereareevenfeweremotionalprejudices.Thewritingiscold.Strongphrasesarealmostwhollylacking,anditissodifficulttofindawell-roundedemotionalcomplexthatifoneisfounditmayaseasilybeattributedtotheingenuityoftheseeker.Similarly,ourhypotheticalauthorshowsnosignofapersonalhistoryorofa cultural background; TenderButtons is the stream of consciousness of a womanwithoutapast.Thewritingspringsfromnoliterarysources.Incontrastwiththeworkof Joyce, towhom a superficial resemblancemay be found, the borrowed phrase ispracticallylacking.Whenmemorizedpassagesoccur,theyarehumdrum—oldsawsorsimpledoggerel

recovered from childhood and often very loosely paraphrased: “If at first you don’tsucceedtrytryagain,”or“Pleasepalehot,pleasecoverrose,pleaseacreinthered….”If there is any character in the writing whatsoever, it is due to this savor of theschoolroom,andtheoneinferenceabouttheauthorwhichdoesseemplausibleisthatshehasbeentogrammarschool.Hersentencesareoftencastasdefinitions(“Whatisaspectacleaspectacleistheresemblance…”or“Asignisthespecimenspoken”)orascopy-book aphorisms (“An excuse is not dreariness, a single plate is not butter,” or“There is coagulation in cold and there is none in prudence”) or as grammaticalparadigms (“Ibeginyoubeginwebegin theybeganwebeganyoubegan Ibegan”).ThisheavydoseofgrammarschoolisespeciallystronglyfeltinAnElucidation,MissStein’sfirstattempttoexplainherself,andapieceofwritinginwhichtherearemanyevidencesofastruggleonthepartoftheconsciousGertrudeSteintoaccepttheoriginoftheTenderButtonsmanner.MissSteinwantedthevolumeLucyChurchAmiablytobeboundlikeaschoolbook,butIshallleaveittoamoreimaginativemindtoelaboratethismetaphorfurther.Thisisapparentlyasmuchofthewritingaswillhelptoilluminatethecharacterof

thewriter. For the rest, it iswhatMissStein describes as sounding all rightwithout

366

makingsense.Thereisnoparadoxaboutthis,thereisnosecretabouthowitisdone;butitgivesusverylittleinformationabouttheauthor.Grammariseverpresent—thatisthemainthing.Wearepresentedwithsentences(“sentencesandalwayssentences”),but we often recognize them as such only because they show an accepted order ofarticle,substantive,verb,splitinfinitive,article,substantive,connective,andsoon.Theframeworkofasentenceisthere,butthewordstackeduponitareanoddcompany.Inthe simplest type of case we have a nearly intelligible sentence modified by thesubstitutionforasinglewordofonesoundingmuchthesame.ThissortofsubstitutionwasreportedbyMissStein inconnectionwithherexperiments inautomaticreading:“Absurdmistakesareoccasionallymadeinthereadingofwords—substitutionssimilarinsoundbututterlydifferentinsense.”Thereaderwillrecognizeitasthesortofslipwhichismadewhenoneisverytired.Inmorecomplexcasesitcannot,ofcourse,beshownthattheunintelligibilityisduetosubstitution;ifmostofthewordsarereplaced,wehavenothingtoshowthatawordisaslip.Wemustbecontenttocharacterizeit,asMissSteinherselfhasdone:“Wehavemadeexcessreturntorambling.”

IV

Fromthisbriefanalysisitisapparentthat,althoughitisquiteplausiblethattheworkisduetoasecondpersonalitysuccessfullysplitofffromMissStein’sconsciousself,itisa very flimsy sort of personality indeed. It is intellectually unopinionated, isemotionallycold,andhasnopast.Itisunreadandunlearnedbeyondgrammarschool.It is as easily influenced as a child; a heard word may force itself into whateversentencemaybeunder construction at themoment, or itmaybreak the sentenceupaltogetherandirremediably.Itsliterarymaterialsarethesensorythingsnearestathand—objects,sounds,tastes,smells,andsoon.Thereadermaycompare,forthesakeofthestrongcontrast,thematerialsof“Melanctha”inThreeLives,apieceofwritingofquiteanothersort.InherexperimentalworkitwasMissStein’sintentiontoavoidtheproduction of a true secondpersonality, and she considered herself to be successful.The automatism she was able to demonstrate possessed the “elements” of a secondpersonality,itwasabletodoanythingwhichasecondpersonalitycoulddo,butitneverbecame the organized alter ego of the hysteric. The superficial character of theinferential author of Tender Buttons consequently adds credibility to the theory ofautomaticauthorship.The Gertrude Stein enthusiast may feel that I am being cruelly unjust in this

estimate.IadmitthattherearepassagesinTenderButtonswhicheludetheforegoinganalysis.ButitmustbemadeclearthatthetwoGertrudeSteinsweareconsideringarenotkeptapartbythecoversofbooks.ThereisagooddealoftheGertrudeSteinoftheAutobiographyinTenderButtons,intheformofrelativelyintelligiblecomment,oftenparentheticalinspirit.ThusattheendofthesectiononMutton(whichbegins“Aletterwhichcanwither,alearningwhichcansufferandanoutragewhichissimultaneousisprincipal”)comesthissentence:“Amealinmuttonmuttonwhyislambcheaper,itischeaperbecausesolittleismore,”whichiseasilyrecognizedasafavoriteprejudiceoftheGertrudeSteinoftheAutobiography.Similarlysuchaphraseas“thesadprocessionof the un-killed bull,” inAnElucidation, is plainly a reference to another of MissStein’sinterests.But,farfromdamagingourtheory,thisoccasionalappearanceofMissSteinherself ispreciselywhat the theorydemands.Inherpaper in thePsychological

367

Review shedealsat lengthwith the inevitablealternationofconsciousandautomaticselves,and in thequotationwehavegiven itwillbe remembered thatshecommentsuponthese“flashesofconsciousness.”EventhoughthegreaterpartofTenderButtonsis automatic, we should expect an “element of connected thought,” and our onlyproblemisthatwhichMissSteinherselfhasconsidered—namely,arewetoattributetoconsciousflashesalltheconnectedthoughtwhichispresent?There isacertain logicaldifficultyhere. Itmaybeargued that, sincewedispense

withalltheintelligiblesentencesbycallingthemconsciousflashes,weshouldnotbesurprisedtofindthatwhatisleftisthinandmeaningless.Wemustthereforerestateourtheory, in a way which will avoid this criticism. We first divide the writings ofGertrude Stein into two parts on the basis of their ordinary intelligibility. I do notcontendthatthisisahardandfastline,butitisasufficientlyrealoneformostpersons.Itdoesnot,itistobeunderstood,followtheoutlinesofherworks.WethenshowthattheunintelligibleparthasthecharacteristicsoftheautomaticwritingproducedbyMissStein in her early psychological experiments, and from this and many otherconsiderationsweconcludethatourdivisionoftheworkintotwopartsisrealandvalidandthatonepartisautomaticinnature.I cannot findanything in theAutobiographyor theotherworks Ihave readwhich

willstandagainstthisinterpretation.Onthecontrary,therearemanybitsofevidence,noneofwhichwouldbeveryconvincing in itself,whichsupport it.Thus (1)TenderButtonswaswrittenonscrapsofpaper,andnoscrapwaseverthrownaway;(2)MissStein likes towrite in thepresenceofdistractingnoises; (3)herhandwriting isoftenmorelegibletoMissToklasthantoherself(thatis,herwritingis“cold”assoonasitisproduced); and (4) she is “fond ofwriting the letterm,”withwhich, the readerwillrecall, the automatic procedure often began. In An Elucidation, her “first effort torealizeclearlyjustwhatherwritingmeantandwhyitwasasitwas,”therearemanyfitfulallusionstotheexperimentaldays:“Doyouallunderstandextraneousmemory,”“In this way my researches are easily read,” a suddenly interpolated “I stopped Istoppedmyself,”whichrecallsthemajordifficultyinherexperiments,andsoon.

V

ItisnecessarytoassumethatwhenGertrudeSteinreturnedtothepracticeofautomaticwriting(about1912?)shehadforgottenorwasshortlytoforgetitsorigins.Iacceptasmade inperfectlygoodfaith thestatement in theAutobiographythat“GertrudeSteinnever had subconscious reactions, nor was she a successful subject for automaticwriting,” even though the evidence to the contrary in her early paper isincontrovertible. She has forgotten it, just as she forgot her first novel almostimmediately after it was completed and did not remember it again for twenty-fiveyears. It isquitepossible,moreover, that themanner inwhichshewrites theTenderButtons sort of thing is not unusual enough to remind her of its origins or to beremarkedbyothers.Oneofthemostinterestingstatementsintheexcerptquotedfromher earlypaper is thatGertrudeStein found it sufficientdistraction simply to followwhatshewaswritingsomefewwordsbehindherpencil. If in thecourseof timeshewasabletobringherattentionnearerandnearertothepencil,shemusteventuallyhavereachedapointatwhichthereremainedonlythefinestdistinctionbetween“knowingwhatoneisgoingtowriteandknowingthatonehaswrittenit.”Thisisatransitional

368

state to which Miss Stein devotes considerable space in her paper. It is thereforereasonableforustoassumethattheartificialcharacteroftheexperimentalprocedurehascompletelywornoff,andthatthereremainsonlyanot-far-from-normalstateintowhichMissSteinmaypassunsuspectingly enoughand inwhich theTenderButtonsstyleisforthcoming.Havingbeguntoproducestuffofthissortagain,however,MissSteincouldnothave

failed to notice its peculiarities.Wehaveher ownopinion that the sentences quotedfromherautomaticwritingdonotshowmuchconnectedthought,andIbelievewearefully justified in our characterization of the greater part of Tender Buttons as“ordinarilyunintelligible.”Iknowthatitwouldbequitepossibleforanindustriousandingenious person to find any number ofmeanings in it, just as it is possible to findmeanings in any chance arrangement ofwords.But the conclusion towhichwe arenowledis that theworkwithwhichwearedealingisveryprobablyunintelligible inanyordinarysense,notonlytootherreaders,buttoMissSteinherself.Why,then,didshepublish?It is important forour theory thatbetween1896and1912MissSteinhadcometo

knowPicassoandMatisseandwasalreadylonginthepracticeofdefendingtheirworkagainstthequestion,“Whatdoesitmean?”Withsuchanexperiencebehindone,itisnot difficult to accept as art what one has hitherto dismissed as the interesting andrathersurprisingresultofanexperiment.Itwas,Ibelieve,onlybecauseGertrudeSteinhadalreadyprepared thedefenseas itapplied toPicasso that shecouldput forthherownunintelligibleproductasaseriousartisticexperiment.Forapersonof thesoundintelligenceofMissStein there isagreatnatural resistanceagainst theproductionofnonsense.Itwasthemajorprobleminherexperimentaltechnique:“IstoppedIstoppedmyself.”Butthewritingsucceededinthiscase,becausetheresistancehadbeenbrokendown, first by the procedure of the experiments, which permitted the sustainedproductionofmeaninglesssentences,andlaterbythechampioningofPicasso,whichpermitted their publication. This was a fortunate combination of circumstances. “Icould explain,” she says in An Elucidation, “how it happened accidentally thatfortunatelynoexplanationwasnecessary.”MissSteinhasnot,however,freedherselffromtheproblemofthemeaningofthe

things she writes. She is not above being bothered by criticism on the score ofunintelligibility.Sheoftencharacterizesherworkinthisveinasexperimental,butthatisinnosenseanexplanation.Beyondthisheranswerseemstobethatthewritingisitsownjustification.It was not a question, she told herOxford audience, ofwhether shewas right in

doing the kind ofwriting she did. “She had been doing as she did for about twentyyearsandnowtheywantedtohearherlecture.”AndshehadpreviouslydealtwiththematterinAnElucidation:

IfitisaneventjustbyitselfisthereaquestionTulipsisthereaquestionPetsisthereaquestionFursisthereaquestionFoldsisthereaquestionIsthereanythinginquestion.

Ithinkwemustacceptthisanswertotheethicalquestionofwhethersheisdoingright

369

by Oxford and the King’s English. The final test of whether it is right is whetheranyone likes it. But a literary composition is not “an event just by itself,” and theanswer toMiss Stein’s query is that there certainlyare questions, of a critical sort,whichmaylegitimatelyberaised.Meaningisoneofthem.Onekindofmeaningwhichmightbefoundifourtheoryisvalidispsychological.In

notingthepresenceofverbalslips(“substitutionssimilarinsoundbututterlydifferentinsense”)welayourselvesopentothecriticismoftheFreudian,whowouldarguethatthere areno true slips.According to thisview, there is always some reasonwhy thesubstitutionismade,andthesubstitutedwordwillhaveadeepersignificanceifwecanfindit.Butweareherenotprimarilyconcernedwithsuchpsychologicalsignificances.Ofliterarysignificancesitmaybeurgedthatfortheinitiatedorsympatheticreader

there is an intellectual content in this part of Miss Stein’s work which we haveoverlooked. Now, either this will be of such a sort that it could also be expressednormally,oritwillbeaspecialkindofcontentwhichrequirestheformgiventoitbyMiss Stein. A partisan could so easily prove the first case by translating arepresentativepassagethatwemayassumeitnottobetrue.Thesecondcaserequiresaverydifficulttheoryofknowledgeinitsdefense,andweshallnotneedtoinquireintoitanymoreclosely.ItisquitetruethatsomethinghappenstotheconscientiousreaderofTenderButtons.Partoftheeffectiscertainlydueeithertorepetitionortosurprise.These are recognized literary devices, and itmaybe argued that still a third kindofmeaning, which we may designate as emotional, is therefore to be found. But inordinarypracticethesedevicesaresupplementarytoexpressionsofanothersort.Themere generation of the effects of repetition and surprise is not in itself a literaryachievement.

VI

Wehaveallowedforthepresenceofanyorallofthesekindsofmeaningbyspeakingonlyofordinaryintelligibility.Idonotthinkthatacasecanbemadeoutforanyoneofthemwhich is not obviously the invention of the analyzer. In any event the presentargumentissimplythattheevidencehereofferedinsupportofatheoryofautomaticwritingmakes itmoreprobable thatmeanings arenot present, and thatweneednotbothertolookforthem.Atheoryofautomaticwritingdoesnot,ofcourse,necessarilyexcludemeanings. It ispossible tosetupasecondpersonalitywhichwillpossessalltheattributesofaconsciousselfandwhosewritingswillbeequallymeaningful.Butinthe present case it is clear that, as Miss Stein originally intended, a true secondpersonality does not exist. This part of her work is, as she has characterized herexperimentalresult,littlemorethan“whatherarmwrote.”Anditisanarmwhichhasvery little to say. This is, I believe, the main importance of the present theory forliterarycriticism.ItenablesonetoassignanorigintotheunintelligiblepartofGertrudeSteinwhichputsoneateaseaboutitsmeanings.Therearecertainaspectsofprosewriting,suchasrhythm,whicharenotparticularly

dependentuponintelligibility.Itispossibletoexperimentwiththemwithmeaninglesswords, and it may be argued that this is what is happening in the present case.ConsideringthefreedomwhichMissSteinhasgivenherself,Idonotthinktheresultisvery striking, although this is clearly a debatable point. It is a fairer interpretation,however,tosuppose,inaccordancewithourtheory,thatthereisnoexperimentationat

370

thetimethewritingisproduced.Theremaybegoodreasonforpublishingthematerialafterward as an experiment. For example, I recognize the possibility of a salutary,thoughaccidental,effectuponGertrudeStein’sconsciousproseoruponEnglishproseingeneral.InCompositionasExplanation,forexample,thereisanintimatefusionofthetwostyles,andtheconsciouspassagesareimitativeoftheautomaticstyle.Thisisalso probably true of parts of theAutobiography. It is perhaps impossible to tell atpresentwhether the effect uponher conscious prose is anythingmore than a loss ofdiscipline.Thecompensatinggainisoftenverygreat.We have no reason, of course, to estimate the literary value of this part ofMiss

Stein’swork.Itmightbeconsiderable,evenifourtheoryiscorrect.ItisapparentthatMissSteinbelievesittobeimportantandhasaccordinglypublishedit.Ifsheisright,ifthispartofherworkistobecomehistoricallyassignificantasshehascontended,thentheimportanceofthedocumentwithwhichwebeganisenormous.Forthefirst timewe should then have an account by the author herself of how a literary secondpersonalityhasbeensetup.I do not believe this importance exists, however, because I do not believe in the

importance of the part ofMiss Stein’s writing which does not make sense. On thecontrary,Iregrettheunfortunateeffectithashadinobscuringthefinerworkofaveryfinemind. Iwelcome thepresent theorybecause itgivesone the freedom todismissonepartofGertrudeStein’swritingasaprobablyill-advisedexperimentandtoenjoytheotherandverygreatpartwithoutpuzzlement.

FromAtlanticMonthly,January1984.

371

TheOperationalAnalysisofPsychologicalTerms

The symposium on operationism conducted in the September, 1945, issue of thePsychological Review was suggested by Edwin G. Boring, who proposed a set ofelevenquestionstowhichparticipantsmightaddressthemselves.Inthepresentversionofmycontributionseveralreferencestothesequestionsbynumberhavebeenomittedand others replaced by brief paraphrases. Part of the material appears in slightlymodified form in both Verbal Behavior and Science and Human Behavior, but itsbearingontheissueofoperationismmaymakeitworthincludinghere.ItisreprintedwiththepermissionofthePsychologicalReview.

“Is operationismmore than a renewed and refined emphasis upon the experimentalmethod (as understood already by Galileo, if not even by Archimedes)—that is, aformulationofmodernscientificempiricismandpragmatism(especiallyofthePeirce-Deweyvariety),mainlyofcriteriaof factualmeaningfulnessandempiricalvalidity?”An answer to this question will define the position to be taken in what follows.Operationismisnotregardedasanewtheoryormodeofdefinition.Theliteraturehasemphasized certain critical or hitherto neglected instances, but no new kind ofoperationhasbeendiscoveredandnoneshouldbesingledout.There isnoreason torestrict operational analysis to high-order constructs; the principle applies to alldefinitions.Thismeansthatwemustexplicateanoperationaldefinitionforeverytermunlesswearewillingtoadoptthevagueusageofthevernacular.Operationismmaybedefinedasthepracticeoftalkingabout(1)one’sobservations,

(2) the manipulative and calculational procedures involved in making them, (3) thelogical andmathematical stepswhich intervene between earlier and later statements,and(4)nothingelse.Sofar,themajorcontributionhascomefromthefourthprovisionand, like it, is negative. We have learned how to avoid troublesome references byshowing that theyareartifactswhichmaybevariously traced tohistory,philosophy,linguistics, and so on. No very important positive advances have been made inconnectionwiththefirstthreeprovisionsbecauseoperationismhasnogooddefinitionofadefinition,operationalorotherwise.Ithasnotdevelopedasatisfactoryformulationoftheeffectiveverbalbehaviorofthescientist.The operationist, like most contemporary writers in the field of linguistic and

semantic analysis, is on the fence between logical “correspondence” theories ofreferenceandempiricalformulationsoflanguageinuse.Hehasnotimproveduponthemixtureoflogicalandpopulartermsusuallyencounteredincasualorevensupposedlytechnicaldiscussionsof scientificmethodor the theoryofknowledge (e.g.,BertrandRussell’srecentAninquiryintomeaningandtruth).Definitionisakeytermbutisnotrigorously defined. Bridgman’s original contention that the “concept is synonymouswith the corresponding set of operations” cannot be taken literally, and no similarlyexplicitbutsatisfactorystatementoftherelationisavailable.Instead,afewroundaboutexpressionsrecurwithrathertiresomeregularitywheneverthisrelationismentioned.

372

We are told that a concept is to be defined “in terms of” certain operations, thatpropositionsare tobe“basedupon” operations, that a term denotes something onlywhen thereare“concretecriteria for itsapplicability,” that operationism consists in“referringanyconcept for itsdefinition to…concrete operations…,” and so on.Wemayacceptexpressionsofthissortasoutliningaprogram,but theydonotprovideageneralschemeofdefinition,much lessanexplicit statementof the relationbetweenconceptandoperation.The weakness of current theories of language may be traced to the fact that an

objectiveconceptionofhumanbehaviorisstillincomplete.Thedoctrinethatwordsareused to express or conveymeaningsmerely substitutes “meaning” for “idea” (in thehope thatmeanings can then somehow be got outside the skin) and is incompatiblewithmodernpsychologicalconceptionsoftheorganism.Attemptstoderiveasymbolicfunctionfromtheprincipleofconditioning(orassociation)havebeencharacterizedbyaverysuperficialanalysis.Itissimplynottruethatanorganismreactstoasign“asitwouldtotheobjectwhichthesignsupplants.”1Onlyinaverylimitedarea(mainlyinthecaseofautonomicresponses)isitpossibletoregardthesignasasimplesubstitutestimulusinthePavloviansense.Modernlogic,asaformalizationof“real”languages,retainsandextendsthisdualistictheoryofmeaningandcanscarcelybeappealedtobythepsychologistwhorecognizeshisownresponsibilityingivinganaccountofverbalbehavior.Itisnotmyintentiontoattemptamoreadequateformulationhere.Thefundamental

revision is toosweeping tobemadehastily. I should like,however, to try tomakeasmallbutpositivecontribution to thissymposiumbyconsideringa fewpointswhichariseinconnectionwiththeoperationaldefinitionofpsychologicalterms.Muchofthematerial which follows is adapted from amuch longerwork now in preparation, inwhichthenecessarygroundworkismorecarefullyprepared.Theoperationalattitude,inspiteofitsshortcomings,isagoodthinginanyscience

but especially in psychology because of the presence there of a vast vocabulary ofancientandnon-scientificorigin.Itisnotsurprisingthatthebroadempiricalmovementin the philosophy of science, which Stevens has shown to be the background ofoperationism, should have had a vigorous and early representation in the field ofpsychology—namely,behaviorism.InspiteofthedifferenceswhichStevensclaimstofind, behaviorism has been (at least to most behaviorists) nothing more than athoroughgoing operational analysis of traditional mentalistic concepts. We maydisagreewith someof theanswers (suchasWatson’sdispositionof images),but thequestionsaskedbybehaviorismwerestrictlyoperationalinspirit.Ialsocannotagreewith Stevens that American behaviorism was “primitive.” The early papers on theproblem of consciousness by Watson, Weiss, Tolman, Hunter, Lashley, and manyothers, were not only highly sophisticated examples of operational inquiry, theyshowed a willingness to deal with a wider range of phenomena than do currentstreamlinedtreatments,particularlythoseofferedbylogicians(e.g.,Carnap)interestedin a unified scientific vocabulary. But behaviorism, too, stopped short of a decisivepositive contribution—and for the same reason: it never finished an acceptableformulation of the “verbal report.” The conception of behavior which it developedcouldnotconvincinglyembracethe“useofsubjectiveterms.”Aconsiderable advantage is gained fromdealingwith terms, concepts, constructs,

and so on, quite frankly in the form inwhich they are observed—namely, as verbal

373

responses.There is thennodangerof including in the concept that aspectorpart ofnaturewhichitsinglesout.(Severalofthepresentquestionsseemtomixconceptandreferent;atleasttheyseemtobecometrivialwhen,inordertomakethemixturelesslikely,termissubstitutedforconceptorconstruct.)Meanings,contents,andreferencesare to be found among the determiners, not among the properties, of response. Thequestion “What is length?”would appear tobe satisfactorily answeredby listing thecircumstancesunderwhichtheresponse“length”isemitted(or,better,bygivingsomegeneraldescriptionofsuchcircumstances).Iftwoquiteseparatesetsofcircumstancesare revealed, then there are two responses having the form “length,” since a verbalresponse-class is not defined by phonetic form alone but by its functional relations.This is true even though the two sets are found tobe intimately connected.The tworesponsesarenotcontrolledbythesamestimuli,nomatterhowclearlyitisshownthatthedifferentstimuliarisefromthesame“thing.”Whatwewanttoknowinthecaseofmanytraditionalpsychologicaltermsis,first,

the specific stimulating conditionsunderwhich they are emitted (this corresponds to“finding the referents”) and, second (and this is a much more important systematicquestion),whyeachresponseiscontrolledbyitscorrespondingcondition.Thelatterisnotnecessarilyageneticquestion.Theindividualacquireslanguagefromsociety,butthereinforcingactionoftheverbalcommunitycontinuestoplayanimportantroleinmaintainingthespecificrelationsbetweenresponsesandstimuliwhichareessentialtotheproperfunctioningofverbalbehavior.Howlanguageisacquiredis,therefore,onlypartofamuchbroaderproblem.Wemaygeneralize theconditions responsible for the standard“semantic” relation

betweenaverbalresponseandaparticularstimuluswithoutgoingintoreinforcementtheory in detail. There are three important terms: a stimulus, a response, and areinforcement supplied by the verbal community. (All of these need more carefuldefinitionsthanareimpliedbycurrentusage,butthefollowingargumentmaybemadewithoutdigressingforthatpurpose.)Thesignificantinterrelationsbetweenthesetermsmaybeexpressedbysayingthatthecommunityreinforcestheresponseonlywhenitisemitted in thepresenceof thestimulus.Thereinforcementof the response“red,” forexample,iscontingentuponthepresenceofaredobject.(Thecontingencyneednotbeinvariable.)Aredobjectthenbecomesadiscriminativestimulus,an“occasion,”forthesuccessfulemissionoftheresponse“red.”This scheme presupposes that the stimulus act upon both the speaker and the

reinforcingcommunity;otherwisethepropercontingencycannotbemaintainedbythecommunity.Butthisprovisionislackinginthecaseofmany“subjective”terms,whichappear to be responses toprivate stimuli. The problem of subjective terms does notcoincideexactlywiththatofprivatestimuli,butthereisacloseconnection.Wemustknowthecharacteristicsofverbalresponsestoprivatestimuliinordertoapproachtheoperationalanalysisofthesubjectiveterm.The response “My tooth aches” is partly under the control of a state of affairs to

which the speakeralone isable to react, sincenooneelsecanestablish the requiredconnection with the tooth in question. There is nothing mysterious or metaphysicalaboutthis;thesimplefactisthateachspeakerpossessesasmallbutimportantprivateworldofstimuli.Sofarasweknow,hisreactionstothesearequitelikehisreactionstoexternalevents.Neverthelesstheprivacygivesrisetotwoproblems.Thefirstdifficultyis thatwecannot,as inthecaseofpublicstimuli,accountfor theverbalresponsebypointingtoacontrollingstimulus.Ourpracticeistoinfertheprivateevent,butthisis

374

opposedtothedirectionofinquiryinascienceofbehaviorinwhichwearetopredictresponsethrough,amongotherthings,anindependentknowledgeofthestimulus.Itisoften supposed that a solution is to be found in improved physiological techniques.Wheneveritbecomespossibletosaywhatconditionswithintheorganismcontroltheresponse “I am depressed,” for example, and to produce these conditions at will, adegreeofcontrolandpredictioncharacteristicofresponsestoexternalstimuliwillbemadepossible.Meanwhile,wemustbecontentwithreasonableevidenceforthebeliefthatresponsestopublicandprivatestimuliareequallylawfulandalikeinkind.But theproblemofprivacycannotbewholly solvedby instrumental invasion.No

matter how clearly these internal eventsmay be exposed in the laboratory, the factremains that in thenormalverbalepisode theyarequiteprivate.Wehavenotsolvedthe second problem of how the community achieves the necessary contingency ofreinforcement. How is the response “toothache” appropriately reinforced if thereinforcing agent has no contactwith the tooth? There is, of course, no question ofwhether responses toprivate stimuli arepossible.Theyoccurcommonlyenoughandmust be accounted for. But why do they occur, what is their relation to controllingstimuli,andwhat,ifany,aretheirdistinguishingcharacteristics?Thereareatleastfourwaysinwhichaverbalcommunitywhichhasnoaccesstoa

privatestimulusmaygenerateverbalbehaviorinresponsetoit:(1)Itisnotstrictlytruethatthestimuliwhichcontroltheresponsemustbeavailable

to thecommunity.Anyreasonablyregularaccompanimentwillsuffice.Consider, forexample,ablindmanwholearnsthenamesofatrayfulofobjectsfromateacherwhoidentifiestheobjectsbysight.Thereinforcementsaresuppliedorwithheldaccordingtothecontingencybetweentheblindman’sresponsesandtheteacher’svisualstimuli,buttheresponsesarecontrolledwhollybytactualstimuli.Asatisfactoryverbalsystemresultsfromthefactthatthevisualandtactualstimuliremaincloselyconnected.Similarly,inthecaseofprivatestimuli,onemayteachachildtosay“Thathurts”in

agreementwith theusageof thecommunitybymaking the reinforcementcontingentuponpublicaccompanimentsofpainfulstimuli (asmartblow, tissuedamage,andsoon). The connection between public and private stimuli need not be invariable; aresponsemaybeconditionedwithmerelyperiodicreinforcementandeveninspiteofanoccasional conflicting contingency.Thepossibilityof suchbehavior is limitedbythe degree of association of public and private stimuli which will supply a netreinforcementsufficienttoestablishandmaintainaresponse.(2) A commoner basis for the verbal reinforcement of a response to a private

stimulus isprovidedbycollateral responses to the samestimulus.Althoughadentistmayoccasionallybeable to identify thestimulus fora toothachefromcertainpublicaccompaniments as in (1), the response “toothache” is generally transmitted on thebasisofresponseswhichareelicitedbythesamestimulusbutwhichdonotneedtobesetupby an environmental contingency.The community infers theprivate stimulus,not from accompanying public stimuli, but from collateral, generally unconditionedandatleastnon-verbal,responses(handtojaw,facialexpressions,groans,andsoon).Theinferenceisnotalwayscorrect,andtheaccuracyofthereferenceisagainlimitedbythedegreeofassociation.(3)Someveryimportantresponsestoprivatestimuliaredescriptiveofthespeaker’s

ownbehavior.Whenthisisovert,thecommunitybasesitsinstructionalreinforcementupontheconspicuousmanifestations,butthespeakerpresumablyacquirestheresponseinconnectionwithawealthofadditionalproprioceptivestimuli.Thelattermayassume

375

practicallycompletecontrol,as indescribingone’sownbehavior in thedark.This isveryclose to the exampleof theblindman; the speaker and the community react todifferent,thoughcloselyassociated,stimuli.Suppose, now, that a given response recedes to the level of covert or merely

incipientbehavior.Howshallweexplainthevocabularywhichdealswiththisprivateworld?(Theinstrumentaldetectionofcovertbehavior isagainnotananswer,forweareinterestedinhowresponsestoprivatestimuliarenormally,andnon-instrumentally,set up.)There are two important possibilities.The survivingcovert responsemayberegarded as an accompaniment of the overt (perhaps part of it), in which case theresponsetotheprivatestimulusisimpartedonthebasisofthepublicstimulussuppliedbytheovertresponse,asin(1).Ontheotherhand,thecovertresponsemaybesimilarto, thoughprobably less intense than, theovert andhence supply the same stimulus,albeit in a weakened form. We have, then, a third possibility: a response may beemitted in the presence of a private stimulus,which has no public accompaniments,provided it isoccasionally reinforced in thepresenceof the samestimulusoccurringwithpublicmanifestations.Terms fallingwithin this class are apparently descriptive only of behavior, rather

thanofotherinternalstatesorevents,sincethepossibilitythatthesamestimulusmaybebothpublicandprivate(or,better,mayhaveorlackpublicaccompaniments)seemstoarisefromtheuniquefactthatbehaviormaybebothcovertandovert.(4)Theprincipleof transferorstimulus inductionsuppliesa fourthexplanationof

how a response to private stimuli may be maintained by public reinforcement. Aresponsewhich isacquiredandmaintained inconnectionwithpublicstimulimaybeemitted, through induction, in response to private events. The transfer is not due toidentical stimuli, as in (3), but to coinciding properties. Thus, we describe internalstatesas“agitated,”“depressed,”“ebullient,”andsoon,inalonglist.Responsesinthisclassareallmetaphors(includingspecialfigureslikemetonomy).Thetermmetaphorisnotusedpejorativelybutmerelytoindicatethatthedifferentialreinforcementcannotbe accorded actual responses to the private case. As the etymology suggests, theresponseis“carriedover”fromthepublicinstance.Insummary,averbalresponsetoaprivatestimulusmaybemaintainedinstrength

through appropriate reinforcement based upon public accompaniments orconsequences, as in (1) and (2), or through appropriate reinforcement accorded theresponse when it is made to public stimuli, the private case occurring by inductionwhenthestimuliareonlypartlysimilar.Ifthesearetheonlypossibilities(andthelistishere offered as exhaustive), thenwemay understandwhy terms referring to privateevents have never formed a stable and acceptable vocabulary of reasonably uniformusage.Thishistorical fact ispuzzling toadherentsof the“correspondenceschool”ofmeaning.Why is it not possible to assign names to the diverse elements of privateexperienceandthentoproceedwithconsistentandeffectivediscourse?Theanswerliesin the process bywhich “terms are assigned to private events,” a processwhichwehavejustanalyzedinaroughwayintermsofthereinforcementofverbalresponses.Noneoftheconditionswhichwehaveexaminedpermitsthesharpeningofreference

which is achieved, in the case of public stimuli, by a precise contingency ofreinforcement.In(1)and(2)theassociationofpublicandprivateeventsmaybefaulty;thestimuliembracedby(3)areoflimitedscope;andthemetaphoricalnatureofthosein (4) implies a lack of precision. It is, therefore, impossible to establish a rigorousscientificvocabularyforpublicuse,norcanthespeakerclearly“knowhimself”inthe

376

senseinwhichknowingisidentifiedwithbehavingdiscriminatively.Intheabsenceofthe “crisis” provided by differential reinforcement (much of which is necessarilyverbal),private stimuli cannotbeanalyzed. (Thishas littleornothing todowith theavailabilityorcapacityofreceptors.)Thecontingencieswehavereviewedalsofailtoprovideanadequatecheckagainst

fictional distortion of the relation of reference (e.g., as in rationalizing). Statementsabout private events may be under control of the drives associated with theirconsequencesratherthanantecedentstimuli.Thecommunityisskepticalofstatementsofthissort,andanyattemptbythespeakertotalktohimselfabouthisprivateworld(asinpsychologicalsystemmaking)isfraughtwithself-deception.Much of the ambiguity of psychological terms arises from the possibility of

alternativeormultiplemodesofreinforcement.Consider,forexample,theresponse“Iamhungry.”Thecommunitymayreinforcethisonthebasisofthehistoryofingestion,asin(1),orcollateralbehaviorassociatedwithhunger,asin(2),orasadescriptionofbehaviorwithrespecttofood,orstimulipreviouslycorrelatedwithfood,asin(3).Inadditionthespeakerhas(insomeinstances)thepowerfulstimulationofhungerpangs,which is private since the community has no suitable connectionwith the speaker’sstomach.“Iamhungry”maythereforebevariouslytranslatedas“Ihavenoteatenforalong time” (1), or “That foodmakesmymouth water” (2), or “I am ravenous” (3)(comparetheexpression“IwashungrierthanIthought”whichdescribestheingestionofanunexpectedlylargeamountoffood),or“Ihavehungerpangs.”Whileallofthesemaybe regardedas synonymouswith“Iamhungry,” theyarenot synonymouswitheach other. It is easy for conflicting psychological systematists to cite supportinginstancesortotrainspeakerstoemittheresponse“Iamhungry”inconformitywithasystem. With the balloon technique one might condition the verbal responseexclusively to stimulation from stomach contractions. Thiswould be an example ofeither(1)or(2)above.Oraspeakermightbetrainedtomakeniceobservationsofthestrengthofhisingestivebehavior,whichmightrecedetothecovertlevelasin(3).Theresponse “I am hungry” would then describe a tendency to eat, with little or noreference to stomachcontractions.Everydayusage reflectsamixed reinforcement.Asimilar analysis could bemade of all terms descriptive ofmotivation, emotion, andactioningeneral,including(ofspecialinteresthere)theactsofseeing,hearing,andsoon.Whenpublicmanifestationssurvive, theextent towhich theprivatestimulus takes

overisnevercertain.Inthecaseofatoothache,theprivateeventisnodoubtdominant,but this is due to its relative intensity, not to any condition of differentialreinforcement.Inadescriptionofone’sownbehavior,theprivatecomponentmaybemuch less important. A very strict external contingency may emphasize the publiccomponent, especially if the association with private events is faulty. In a rigorousscientificvocabularyprivateeffectsarepracticallyeliminated.Theconversedoesnothold.Thereisapparentlynowayofbasingaresponseentirelyupontheprivatepartofacomplexofstimuli.Adifferentialreinforcementcannotbemadecontingentuponthepropertyofprivacy.This fact isofextraordinary importance inevaluating traditionalpsychologicalterms.Theresponse“red”isimpartedandmaintained(eithercasuallyorprofessionally)by

reinforcementswhicharecontingentuponacertainpropertyofstimuli.Bothspeakerandcommunity(orpsychologist)haveaccesstothestimulus,andthecontingencymaybe made quite precise. There is nothing about the resulting response which should

377

puzzle anyone. The greater part of psychophysics rests upon this solid footing. Theolderpsychologicalview,however,wasthatthespeakerwasreporting,notapropertyof the stimulus, but a certain kind of private event, the sensation of red. This wasregardedasalaterstageinaseriesbeginningwiththeredstimulus.Theexperimenterwassupposedtomanipulatetheprivateeventbymanipulatingthestimulus.Thisseemslikeagratuitousdistinction,butinthecaseofsomesubjectsasimilarlaterstagecouldapparently be generated in other ways (by arousing an “image”), and hence theautonomyofaprivateeventcapableofevokingtheresponse“red”intheabsenceofacontrollableredstimulusseemedtobeproved.Anadequateproof,ofcourse,requiresthe elimination of other possibilities (e.g., that the response is generated by theprocedureswhichareintendedtogeneratetheimage).Verbal behavior which is “descriptive of images” must be accounted for in any

adequate science of behavior. The difficulties are the same for both behaviorist andsubjectivist.Iftheprivateeventsarefree,ascientificdescriptionisimpossibleineithercase.Iflawscanbediscovered,thenalawfuldescriptionoftheverbalbehaviorcanbeachieved,withorwithoutreferencestoimages.Somuchfor“findingthereferents”;theremainingproblemofhowsuchresponsesaremaintainedinrelationtotheirreferentsisalsosoluble.Thedescriptionofanimageappearstobeanexampleofaresponsetoaprivatestimulusofclass(1)above.Thatistosay,relevanttermsareestablishedwhentheprivateeventaccompaniesacontrollableexternalstimulus,butresponsesoccuratother times,perhaps in relation to thesameprivateevent.Thedeficienciesofsuchavocabularyhavebeenpointedout.Wecanaccount for the response“red” (at least aswell as for the“experience”of

red) by appeal to past conditions of reinforcement. But what about expandedexpressionslike“Iseered”or“Iamconsciousofred”?Here“red”maybearesponsetoeitherapublicoraprivatestimuluswithoutprejudicetotherestoftheexpression,but“see”and“conscious”seemtorefertoeventswhicharebynatureorbydefinitionprivate. This violates the principle that a reinforcement cannot be made contingentupontheprivacyofastimulus.Areferencecannotbenarroweddowntoaspecificallyprivateeventbyanyknownmethodofdifferentialreinforcement.The original behavioristic hypothesis was, of course, that terms of this sort were

descriptions of one’s own (generally covert) behavior. The hypothesis explains theestablishmentandmaintenanceofthetermsbysupplyingnaturalpubliccounterpartsinsimilarovertbehavior.The termsare ingeneralofclass(3).Oneconsequenceof thehypothesisisthateachtermmaybegivenabehavioraldefinition.Wemust,however,modifytheargumentslightly.Tosay“Iseered”istoreact,nottored(thisisatrivialmeaningof“see”),buttoone’sreactiontored.“See”isatermacquiredwithrespecttoone’s own behavior in the case of overt responses available to the community. Butaccording to the present analysis it may be evoked at other times by any privateaccompaniment of overt seeing. Here is a point at which a non-behavioral privateseeing may be slipped in. Although the commonest private accompaniment wouldappeartobethestimulationwhichsurvivesinasimilarcovertact,asin(3),itmightbesomesortofstateorconditionwhichgainscontroloftheresponseasin(1)or(2).The superiority of the behavioral hypothesis is not merely methodological. That

aspectofseeingwhichcanbedefinedbehaviorallyisbasictothetermasestablishedbytheverbalcommunityandhencemosteffectiveinpublicdiscourse.Acomparisonofcases(1)and(3)willalsoshowthattermswhichrecedetotheprivatelevelasovertbehaviorbecomescoverthaveanoptimalaccuracyofreference,asresponsestoprivate

378

stimuligo.Theadditionalhypothesisfollowsquitenaturallythatbeingconscious,asaformof

reacting to one’s own behavior, is a social product. Verbal behavior may bedistinguished, and conveniently defined, by the fact that the contingencies ofreinforcementareprovidedbyotherorganismsratherthanbyamechanicalactionuponthe environment. The hypothesis is equivalent to saying that it is only because thebehavioroftheindividualisimportanttosocietythatsocietyinturnmakesitimportanttotheindividual.Theindividualbecomesawareofwhatheisdoingonlyaftersocietyhas reinforced verbal responses with respect to his behavior as the source ofdiscriminativestimuli.Thebehaviortobedescribed(thebehaviorofwhichoneistobeaware)maylaterrecedetothecovertlevel,and(toaddacrowningdifficulty)somaythe verbal response. It is an ironic twist, considering the history of the behavioristicrevolution,thataswedevelopamoreeffectivevocabularyfortheanalysisofbehaviorwealsoenlargethepossibilitiesofawareness,sodefined.Thepsychologyoftheotheroneis,afterall,adirectapproachto“knowingthyself.”Themainpurposeofthisdiscussionhasbeentodefineadefinitionbyconsideringan

example.Tobeconsistentthepsychologistmustdealwithhisownverbalpracticesbydevelopinganempiricalscienceofverbalbehavior.Hecannot,unfortunately,jointhelogicianindefiningadefinition,forexample,asa“rulefortheuseofaterm”(Feigl);he must turn instead to the contingencies of reinforcement which account for thefunctionalrelationbetweenaterm,asaverbalresponse,andagivenstimulus.Thisisthe“operationalbasis”forhisuseofterms;anditisnotlogicbutscience.Thephilosopherwillcallthiscircular.Hewillarguethatwemustadopttherulesof

logicinordertomakeandinterprettheexperimentsrequiredinanempiricalscienceofverbal behavior. But talking about talking is no more circular than thinking aboutthinkingor knowing about knowing.Whether or notwe are lifting ourselves byourownbootstraps,thesimplefactisthatwecanmakeprogressinascientificanalysisofverbalbehavior.Eventuallyweshallbeabletoinclude,andperhapstounderstand,ourownverbalbehaviorasscientists.Ifitturnsoutthatourfinalviewofverbalbehaviorinvalidatesourscientificstructurefromthepointofviewoflogicandtruth-value,thensomuchtheworseforlogic,whichwillalsohavebeenembracedbyouranalysis.

The participants in the symposium were asked to comment upon all the paperssubmitted. Their comments were included in the same issue of the PsychologicalReview.Mycontributionfollows.

In the summer of 1930, two years after the publication of Bridgman’s Logic ofModernPhysics,Iwroteapapercalled“Theconceptofthereflexinthedescriptionofbehavior”[seepage475].Itwaslaterofferedasthefirsthalfofadoctoralthesisandwas published in 1931. Although the general method, particularly the historicalapproach,wasderivedfromMach’sScienceofMechanics,mydebt toBridgmanwasacknowledged in the second paragraph. This was, I think, the first psychologicalpublicationtocontainareferencetotheLogicofModernPhysics,2anditwasthefirstexplicitlyoperationalanalysisofapsychologicalconcept.Shortly after the paper was finished, I found myself contemplating a doctoral

examinationbeforeacommitteeofwhosesympathiesIwasnonetoosure.Notwishingtowaituntilanunconditionalsurrendermightbenecessary, Iputoutapeace feeler.

379

Unmindfulor ignorantof theethicsof theacademy, I suggested toamemberof theHarvarddepartmentthatifIcouldbeexcusedfromanythingbutthemostperfunctoryexamination,thetimewhichIwouldotherwisespendinpreparationwouldbedevotedtoanoperationalanalysisofhalf-a-dozenkeytermsfromsubjectivepsychology.Thesuggestionwasreceivedwithsuchbreathlessamazementthatmypeacefeelerwentnofurther.ThepointIwanttomakeisthatatthattime—1930—Icouldregardanoperational

analysisofsubjectivetermsasamereexerciseinscientificmethod.Itwasjustabitofhackwork,badlyneededbytraditionalpsychology,whichIwaswillingtoengageinasapublicserviceorinreturnfortheremissionofsins.Itneveroccurredtomethattheanalysiscouldtakeanybutasinglecourseorhaveanyrelationtomyownprejudices.Theresultseemedaspredeterminedasthatofamathematicalcalculation.Inspiteofthepresentsymposium,Iamofthisopinionstill.Ibelievethatthedataof

ascienceofpsychologycanbedefinedordenotedunequivocally,andthatsomeonesetofconceptscanbeshowntobethemostexpedientaccordingtotheusualstandardsinscientificpractice.Nevertheless,thesethingshavenotbeendoneinthefieldwhichwasdominatedbysubjectivepsychology,andthequestionis:Whynot?Psychology, alone among the biological and social sciences, passed through a

revolutioncomparableinmanyrespectswiththatwhichwastakingplaceatthesametimeinphysics.Thiswas,ofcourse,behaviorism.Thefirststep, likethat inphysics,wasare-examinationoftheobservationalbasesofcertainimportantconcepts.Butbythe timeBridgman’s bookwas published,most of the early behaviorists, aswell asthoseofusjustcomingalongwhoclaimedsomesystematiccontinuity,hadbeguntoseethatpsychologyactuallydidnotrequiretheredefinitionofsubjectiveconcepts.Thereinterpretationofanestablishedsetofexplanatoryfictionswasnotthewaytosecurethe tools thenneeded for a scientificdescriptionofbehavior.Historical prestigewasbeside the point. There was no more reason to make a permanent place for“consciousness,”“will,”“feeling,”andsoon,thanfor“phlogiston”or“visanima.”Onthe contrary, redefined concepts proved to be awkward and inappropriate, andWatsonianismwas,infact,practicallywreckedintheattempttomakethemwork.Thus it came about that while the behaviorists might have applied Bridgman’s

principle to representative terms from a mentalistic psychology (and were mostcompetent todoso), theyhadlostall interest in thematter.Theymightaswellhavespenttheirtimeinshowingwhataneighteenth-centurychemistwastalkingaboutwhenhe said that the Metallic Substances consisted of a vitrifiable earth united withphlogiston.Therewasnodoubtthatsuchastatementcouldbeanalyzedoperationallyortranslatedintomodernterms,orthatsubjectivetermscouldbeoperationallydefined.Butsuchmatterswereofhistoricalinterestonly.Whatwaswantedwasafreshsetofconcepts derived from a direct analysis of the newly emphasized data, and thiswasenoughtoabsorballtheavailableenergiesofthebehaviorists.Besides,themotivationoftheenfantterriblehadwornitselfout.IthinktheHarvarddepartmentwouldhavebeenhappierifmyofferhadbeentaken

up.WhathappenedinsteadwastheoperationismofBoringandStevens.Thishasbeendescribed as an attempt to climb onto the behavioristic bandwagon unobserved. Icannot agree. It is an attempt to acknowledge some of themore powerful claims ofbehaviorism(whichcouldnolongerbedenied)butatthesametimetopreservetheoldexplanatoryfictions.Itisagreedthatthedataofpsychologymustbebehavioralratherthanmental ifpsychologyis tobeamemberof theUnitedSciences,but theposition

380

taken ismerely thatof“methodological”behaviorism.According to thisdoctrine theworld isdivided intopublicandprivateevents;andpsychology, inorder tomeet therequirements of a science, must confine itself to the former. This was never goodbehaviorism,butitwasaneasypositiontoexpoundanddefendandwasoftenresortedtobythebehavioriststhemselves.Itisleastobjectionabletothesubjectivistbecauseitpermitshimtoretain“experience”forpurposesof“non-physicalistic”self-knowledge.The position is not genuinely operational because it shows an unwillingness to

abandon fictions. It is like saying that while the physicist must admittedly confinehimselftoEinsteiniantime,itisstilltruethatNewtonianabsolutetimeflows“equablywithoutrelationtoanythingexternal.”ItisasortofEpursimuoveinreverse.Whatislacking is the bold and exciting behavioristic hypothesis thatwhat one observes andtalksabout isalways the“real”or“physical”world(orat least the“one”world)andthat“experience” isaderivedconstruct tobeunderstoodonly throughananalysisofverbal(not,ofcourse,merelyvocal)processes.The difficulties which arise from the public-private distinction have a prominent

placeinthepresentsymposium,anditmaybeworthwhiletoconsiderfourofthem.(1)Therelationbetweenthetwosetsoftermswhicharerequiredhasprovedtobe

confusing. The pair most frequently discussed are “discrimination” (public) and“sensation”(private).Isonethesameastheother,orreducibletotheother,andsoon?Asatisfactoryresolutionwouldseemtobethatthetermsbelongtoconceptualsystemswhich are not necessarily related in a point-to-point correspondence. There is noquestionofequating themor their referents,or reducingone to theother,butonlyaquestionof translation—anda single term inone setmay require aparagraph in theother.(2) The public-private distinction emphasizes the arid philosophy of “truth by

agreement.”Thepublic,infact,turnsouttobesimplythatwhichcanbeagreeduponbecause it is common to two or more agreers. This is not an essential part ofoperationism; on the contrary operationism permits us to dispense with this mostunsatisfyingsolutionoftheproblemoftruth.Disagreementscanoftenbeclearedupbyasking for definitions, and operational definitions are especially helpful, butoperationismisnotprimarilyconcernedwithcommunicationordisputation.Itisoneofthemosthopefulofprinciplespreciselybecauseit isnot.Thesolitaryinhabitantofadesert isle could arrive at operational definitions (provided he had previously beenequippedwithanadequateverbalrepertoire).Theultimatecriterionforthegoodnessofa concept is not whether two people are brought into agreement but whether thescientist who uses the concept can operate successfully upon his material—all byhimselfifneedbe.WhatmatterstoRobinsonCrusoeisnotwhetherheisagreeingwithhimselfbutwhetherheisgettinganywherewithhiscontrolovernature.Onecanseewhy thesubjectivepsychologistmakessomuchofagreement. Itwas

onceafavoritesporttoquizhimaboutinter-subjectivecorrespondences.“HowdoyouknowthatO’ssensationofgreenisthesameasE’s?”Andsoon.Butagreementalonemeans very little.Various epochs in the history of philosophy and psychology haveseenwhole-heartedagreementonthedefinitionofpsychologicalterms.Thismakesforcontentment but not for progress. The agreement is likely to be shattered whensomeonediscovers that a setof termswill not reallywork,perhaps in somehithertoneglected field, but this does not make agreement the key to workability. On thecontrary,itistheotherwayround.(3) The distinction between public and private is by no means the same as that

381

betweenphysicalandmental.Thatiswhymethodologicalbehaviorism(whichadoptsthefirst)isverydifferentfromradicalbehaviorism(whichlopsoffthelatterterminthesecond). The result is thatwhile the radical behavioristmay in some cases considerprivate events (inferentially, perhaps, but none the less meaningfully), themethodologicaloperationisthasmaneuveredhimself intoapositionwherehecannot.“Science does not consider private data,” says Boring. (Just where this leaves mycontributiontothepresentsymposium,Idonotliketoreflect.)ButIcontendthatmytoothacheisjustasphysicalasmytypewriter,thoughnotpublic,andIseenoreasonwhyanobjectiveandoperationalsciencecannotconsidertheprocessesthroughwhichavocabularydescriptiveofatoothacheisacquiredandmaintained.Theironyofit isthat,whileBoringmustconfinehimself toanaccountofmyexternalbehavior, IamstillinterestedinwhatmightbecalledBoring-from-within.(4) The public-private distinction apparently leads to a logical, as distinct from a

psychological,analysisoftheverbalbehaviorofthescientist,althoughIseenoreasonwhyitshould.Perhapsitisbecausethesubjectivistisstillnotinterestedintermsbutinwhatthetermsusedtostandfor.Theonlyproblemwhichascienceofbehaviormustsolveinconnectionwithsubjectivismisintheverbalfield.Howcanweaccountforthebehavior of talking aboutmental events?The solutionmust be psychological, ratherthan logical, and I have tried to suggest one approach in my present paper. Thecompletelackofinterestinthisproblemamongcurrentpsychologicaloperationistsisnicelydemonstratedbythefactthattheonlyothermembersofthepresentpanelwhoseem to be interested in a causal analysis of verbal behavior are the two non-psychologists(oneofthemalogician!).Myreactiontothissymposium,then,istwofold.Theconfusionwhichseemstohave

arisenfromaprinciplewhichissupposedtoeliminateconfusionisdiscouraging.Butupon second thought it appears that the possibility of a genuine operationism inpsychology has not yet been fully explored.With a little effort I can recapture myenthusiasmoffifteenyearsago.(Thisis,ofcourse,aprivateevent.)

FromPsychologicalReview,1945,52,270–277.1Stevens,S.S.Psychologyandthescienceofscience.Psychol.Bull.,1939,36,221–263.

2LyleH.LanierhascalledmyattentiontothefactthatHarryM.JohnsonsummarizedBridgman’sargumentandappliedtheoperationalcriteriontotheconceptofintensityofsensationalmostayearbeforemyarticleappeared(Psychol.Rev.,1930,37,113–123).

382

TheAlliterationinShakespeare’sSonnets:AStudyinLiteraryBehavior

I have omitted articles from this collection when their substance has alreadyappeared in book form—for example, the series of experimental papers broughttogether in The Behavior of Organisms. It has not been easy to dispose of severalpapers on verbal and literary behavior according to this principle. The paper onGertrude Stein was only briefly mentioned in Verbal Behavior and is thereforereprinted.Muchof“Theoperationaldefinitionofpsychological terms”wasincludedinVerbalBehaviorandinScienceandHumanBehavior,buttheissueofoperationismseemed to justify reprinting thewholearticle.Apaper called“Theverbal summatorandamethodforthestudyoflatentspeech”(JournalofPsychology,1936,2,71-107)hasbeenomittedbecausepartofitwascoveredinVerbalBehaviorandtheremainingpart, concerning a relation between rank order and word frequency, did not seemworth salvaging. This was true of two other studies of the same relation—“Thedistributionofassociatedwords”(PsychologicalRecord,1937,1,71-76)and“Somefactorsinfluencingthedistributionofassociatedwords”writtenwithStuartW.Cook(PsychologicalRecord,1939,3,178-184).ThreeotherpaperswerenotfullycoveredinVerbalBehaviorbecauseofadecisionnottoreviewexperimentalorstatisticalstudies.One of these appeared in the Psychological Record (1939, 3, 186-192), and isreproducedherebypermissionoftheeditor.

Alliterationisoneofthemostfamiliarformsofsound-patterninginpoetryandprose.Itissaidtoexistwhentwoormoresyllablesbeginningwiththesameconsonantoccurnear each other in a given passage. Examples of alliteration are frequently cited ascontributingtotheeffectofaliterarywork,anditisusuallyimpliedthattheyrepresentdeliberateactsofarrangementonthepartofthewriter.Ifthisistrue,alliterationshouldthrow some light on the dynamics of verbal behavior and especially upon a processwhichmaybecalled“formalperseveration”or,better,“formalstrengthening.”Studiesofword-association,latentspeech,andsoon,haveindicatedthattheappearanceofasound in speech raises the probability of occurrence of that sound for some timethereafter. Stated in a different way: the emission of a verbal response temporarilyraises the strength of all responses of similar form. The principal characteristics ofpoetry(alliteration,assonance,rhyme,andrhythm)seemtobeexaggeratedcasesofthetendencytowardformalstrengthening,andtheyshouldsupplyusefulinformationwithregardtoit.Inordertodeterminetheexistenceortheimportanceofanyprocessresponsiblefor

acharacteristicpatterninasampleofspeech,itisnecessarytoallowfortheamountofpatterning to be expected from chance.We cannot assert, for example, that any oneinstanceofalliteration isdue toaspecialprocess in thebehaviorof thewriter ratherthan to an accidental proximity ofwords beginningwith the same sound.Proof thatthereisaprocessresponsibleforalliterativepatterningcanbeobtainedonlythrougha

383

statistical analysisof all the arrangementsof initial consonants in a reasonably largesample. In the case of alliterationwhatwewant to know is the extent towhich theinitial consonants are not distributed at random. If the distribution turns out to berandom, thennoprocessbyvirtueofwhichwords come tobe arrangedona formalbasiscanbeattributedtothebehaviorofthewriter,eventhoughselectedinstancesstillshowthegroupingcommonlycalledalliteration.Ifthereisanyprocessinthebehaviorofthewriterbyvirtueofwhichtheoccurrence

ofaninitialconsonantraisestheprobabilityofoccurrenceofthatsoundforashorttimethereafter,thentheinitialconsonantsinasampleofwritingwillbegrouped.Methodsare,ofcourse,available fordetectinga tendency towardgrouping,but in thecaseofpoetryamoreappropriatetechniquecanbebasedupontheuseofthelineasanaturalunit.Inanylargesampleofpoetrycertainlineswillcontainnooccurrencesofagiveninitialconsonant,andotherswillcontainone,two,three,andsoon,occurrences.Fromtherelativefrequencyoftheconsonantwemaycalculatethesenumbersifweassumethat theprobabilityofoccurrence remainsunchangedand that eachoccurrence is anindependent event. A process of alliteration, if it existed, would violate theseassumptionsandyieldagreaternumberoflinescontainingmorethanoneoccurrenceandalsoagreaternumberofemptylines.Thispaperpresentssomefactsconcerningthealliterativepatternsinablockofone

hundredShakespearesonnets.Thematerialisdrawnfromamoreextensiveresearchonanumberofdifferentkindsofsound-patterns,tobereportedinfulllater.Thesonnetswere first scanned according to a set of arbitrary rules, designed to preventunintentionalselectionandatthesametimetosingleoutthemostimportantsyllablesineachofthe1,400lines.Theaveragenumberofsyllablesperlinethusdesignatedwas5.036,whichagreeswellwiththepentametricformofthepoems.Therange,however,wasfromthreetoeight.Atabulationofinitialconsonantsbylinewasthenmade.1Theresultswereexpressedforeachconsonantseparatelyintheformof(1)thenumberoflines containing no occurrences, (2) the number containing one occurrence, (3) thenumbercontainingtwooccurrences,andsoon.The formula for the number of lines containing 0, 1, 2,… occurrences of a given

initialconsonantinvolvesthebinomialexpansionN(q+p)n,whereNisthenumberoflinesexamined,n thenumberofsyllablesper line,p theprobabilityofoccurrenceoftheconsonantunderconsideration(obtainedfromitsfrequencyinthewholesample),andqtheprobabilityofoccurrenceofanyothersound,or1—p.Thesuccessivetermsin the expansion give the numbers required.Agood approximation could have beenobtainedbylettingn=5,whichisclosetotheaveragenumberofimportantsyllables,but a more accurate estimate was obtained by calculating separately for lines ofdifferentlengthaccordingtothelengthsinthesample.Calculationsweremadefor277lines of four syllables (including a few in the original samplewhich contained onlythree), 830 lines of five syllables, 252 of six, and 41 of seven (including a feworiginally of eight).2 By adding the occurrences obtained from these separatecalculations,thetotalchanceexpectancyforthatconsonantwasobtained.Before the observed and calculated frequenciesmay legitimately be compared for

ourpresentpurposes,aspuriousalliterativeeffectintheobservedvaluesmustbetakenintoaccount.Shakespeare,perhapsmorethanmostotherEnglishpoets,tendstorepeata word (or to use an inflected form) within the space of a line. There are tworepetitions,forexample,intheline:

384

Sunsoftheworldmaystainwhenheaven’ssunstaineth

Intabulatinginitialconsonants, this linemustbecountedascontainingfours’s. It isclear,however,thatthelasttwomustbeattributednotonlytoformalstrengtheningbuttosomethematicsource.Thelineasheardisstronglysibilant,but twoofthes’sareduetosomethingbeyondasimplealliterativeprocess.Inalinecontainingarepeatedworditisatpresentimpossibletodeterminehowthe

responsibilityforthesimilarityofsoundistobedividedbetweenformalandthematicfactors.Toomitallrepeatedwordsfromthepresenttabulationwouldobviouslynotbejustified;atthesametimewecannotacceptatfullvaluetheinstancesofalliterationforwhich they are responsible. In the following summary the raw data (obtained bycounting all initial sounds regardless of repetition of thewholeword) are presentedfirst.The revisedvaluesobtainedbysubtracting the instancesarising fromrepetitionarethengiven.the successive terms giving the number of lines containing 0, 1, 2, 3, 4, and 5occurrencesoftheconsonantforwhichpandqwerecalculated.

TABLE1

InTable1 the initial consonants are arranged in order according to the frequencywithwhichtheyoccurintheblockof100sonnets.Thenumbersoflinescontaining0,1, 2, 3, 4, or 5 occurrences per line observed for Shakespeare and calculated asdescribed above are shown in their respective columns. The last column gives thesignificance of the difference between observed and calculated values expressed intermsoftheprobabilitythatthedifferenceisduetosampling.Fromthetableitwillbeseenthattheleastfrequentsound(qu)occursonly23times

and nevermore than once per line, as we should expect from chance. At the otherextremethesounds(whichoccurs938times)failstooccurin702lines(theexpected

385

numberofemptylinesbeing685)butoccursoncein501lines(expected:523),twicein161lines(expected:162), threetimesin29lines(expected:26),andfourtimesinsevenlines(expected:two).Ifweomitthecaseswhicharisefromrepeatedwords,weobtain the figures in parentheses in the table, which show a better agreement. Thecorrectionsrequiredinthetwo-per-linecolumnhavemerelybeenestimatedandarenotshown in the table. An examination of every tenth line showing two occurrencesindicates that about 19%of such lines are due to repetition.When this correction ismade for s, it appears that Shakespeare falls about 30 lines short of the expectednumberoflinescontainingtwos’s.Otherconsonantsinthetableshowvaryingdegreesofagreement.Theestimatesof

significance(whichareineverycasebasedupontherawdata)indicateapossible“useof alliteration” with n, k, h, t, f, b, and l; but these all involve repetition, and thecorrectedvaluesgiveverylittlesupporttothepopularnotion.Thereispossiblyatrendinthedirectiontobeexpectedfromaprocessofalliteration,buttheabsoluteexcessof“heavy”linesisveryslight.Someindicationofthisexcessmaybeobtainedfromthefollowingstatementsregardingthetableasawhole:Linescontainingfourlikeinitialconsonants.(Ex.:Borneonthebierwithwhiteand

bristlybeard.)Of these lines thereareonlyeightmore thanwouldbeexpected fromchance,and

fouroftheseareduetotherepetitionofthesamewordorwords.Notmorethanoncein twenty-five sonnets (350 lines) does Shakespeare lengthen a series of three likeconsonantsintofour,exceptwhenherepeatsaword.Linescontaining three like initial consonants. (Ex.:Save thatmysoul’s imaginary

sight.)Of these lines there are thirty-three toomany, but twenty-nine of these are due to

repetitionofthesameword.Onlyfourare,therefore,“pure”alliteration.Exceptwhenherepeatedawholeword,Shakespearechangedalineoftwolikeconsonantsintooneofthreenotoftenerthanonceintwenty-fivesonnets.Linescontainingtwolikeinitialconsonants.Thereareninety-twoexcesslinesofthissort,butthecorrectionforrepetitiongivesa

shortage of approximately forty lines. Allowing for eight lines extended to containthree or four occurrences, we may say that once in about every three sonnetsShakespearediscardedawordbecauseitsinitialconsonanthadalreadybeenused.Thesecorrectionsprobablygo too far, sincea repetitionof the samewordmay in

partexemplifyanalliterativeprocess.Moreover,wheninstanceshavebeenthrownoutbecausetheybelongtorepeatedwords,thewholetableshouldberecalculatedonthebasisofareducedtotalfrequency.Thisrecalculationwouldaffectchieflythevaluesforthe empty lines and for the lines containing one occurrence. As the table indicates,Shakespeare shows in general an excess of empty lines, butmost, if not all, of thisdifference would disappear under recalculation with a smaller total frequency.Similarly, Shakespeare’s shortage of one-occurrence lines would be reduced. Thesechanges cannot be made without an arbitrary estimate of the share contributed byalliterationwhenawordisrepeated,butbytakingtherawdataastheupperlimitandthefullycorrecteddataasthelower,themainquestionproposedinthisstudymaybeanswered.Inspiteoftheseemingrichnessofalliterationinthesonnets,thereisnosignificant

evidenceofaprocessofalliteration in thebehaviorof thepoet towhichanyseriousattention should be given. So far as this aspect of poetry is concerned, Shakespeare

386

might as well have drawn his words out of a hat. The thematic or semantic forceswhichareresponsiblefortheemissionofspeechapparentlyfunctionindependentlyofthisparticularformalproperty.It is scarcely convincing to argue that Shakespeare may have arranged certain

alliterativepatternsanddiscardedanequalnumberduetochance,sinceit isunlikelythattheexpectedfrequencieswouldbesocloselyapproximated.Itissimplertobelievethat we have been misled by the selection of instances and that no process ofalliteration should ever have been attributed to the poet. If “formal strengthening”provestobearealcharacteristicofnormalspeech,weshallhavetolookforthekeytoShakespeare’sgeniusinhisabilitytoresistit,therebyreversingtheusualconceptionofthiskindofpoeticactivity.Shakespeare’s“philosophyofcomposition”mightwellbeexpressedinthewordsof

theDuchess,whosaidtoAlice,“Andthemoralofthatis,‘Takecareofthesense,andthesoundswilltakecareofthemselves.’”

1ThetabulationwasmadebyMissMarianKruseandMissJanetteJones,FederalAidStudentsattheUniversityofMinnesota.

2Theformulaforthefive-syllablelines,forexample,is830(q5+5q4p+10q3p2+10q2p3+5qp4+p5),

387

AQuantitativeEstimateofCertainTypesofSound-PatterninginPoetry

ThetechniqueusedinestimatingcertainverbalprocessesinthecompositionoftheShakespeare sonnetswas extended tootheraspectsof sound-patterning inanalyzingthe work of a poet known to have favored formal devices. The following paper isreprintedwithpermissionoftheeditorofTheAmericanJournalofPsychology(1941,54,64-79).

Thephonetic elementswhich compose a representative sampleofnormal speech arenot distributed at randombut are to some extent grouped. This characteristic cannotalwaysbeaccountedforbytherepetitionofmeaningfulmaterial;anditmustthereforeberegardedasanexampleof“formalperseveration”or“formalstrengthening.”Itmaybedescribedby saying thatwhen a speech sound is once emitted, the probability ofemission of that sound is temporarily raised. In addition to normal speech, formalstrengthening isobserved inwordassociations, in speechobtained invacuowith theverbal summator, and in the verbal behavior characteristic of certain psychopathicdisorders. In literature it appears in an exaggerated form as rhyme, assonance,alliteration, and (if we include stress-pattern as a formal phonetic characteristic)rhythm.Theseextremecasesareinthemselvesworthyofinvestigation,andtheymayalsobeexpectedtothrowsomelightonthegeneralprocess.A satisfactory demonstration of formal strengthening in poetry or prose cannot be

madebypointingtoinstances,althoughthisiscommonpracticeinthefieldofliterarycriticism.Sevenoreightdifferentinitialconsonants,forexample,willusuallyaccountformorethanhalftheinstances,andmanyaccidental“alliterations”aretobeexpected.Theassertion thatanysampleofspeechdemonstratesalliterationasaprocess in thebehaviorofthewritermustrestuponastatisticalproofthattheexistingpatternsarenottobeexpectedfromchance.Onemethodofanalyzingapoemistoexaminethenumberoflinescontainingtwo

ormoreoccurrencesofagiveninitialconsonantandtodeterminewhetherthisexceedsthe expectation based upon the general frequency of the consonant in the wholesample.Inastudyof100Shakespeareansonnetsitwasfoundthattheslightexcessoflinescontainingmorethanoneoccurrencewaslargelyaccountedforbytherepetitionof whole words, where the strengthening or perseveration seemed to be principallyderivedfromthematicratherthanformalfactors[seepage431].Littleornothingcouldbelearnedofthenatureofalliterationinsuchacase,anditseemedadvisabletoturntoapoetwhoalmostcertainlyexemplifiestheprocessinanextremedegree.Thepresentreportisananalysisofthefirst500iambicpentameterlinesofSwinburne’s“AtalantainCalydon.”Bothalliterationandassonanceareconsidered.Notall thesounds inapoemcontributeequally to itseffect,nordo theyrepresent

equalopportunitiesofselectionandmanipulationonthepartofthewriter.Theformalanalysis of a poem is facilitated by a preliminary choice of the principal sounds,

388

especially the syllables receiving emphasis when the line is scanned, as well as anumber of other obviously important words. The omittedmaterial may be regardedeitherasnotcontributinginanyeffectivewaytothesoundpatternorasnotpermittinganyvariationinthebehaviorofthewriter.Inordertoavoidunconsciousprejudicewithregard to patterning, the material must be selected by rule, even though the bestinterpretivereadingisperhapsnotalwaysobtained.Inthepresentcaseeachlinewasfirst scanned strictly as iambic pentameter.When the accent fell on aweak syllable(e.g.,preposition,auxiliary,possessivepronoun,article,copula,orsuchanendingas-ness,-ing,-ance,or-ment),itwasshiftedforwardorbackwardwheneverpossibletoanadjacent strong syllable not included in the scanning.Otherwise itwas omitted. Forexample,intheline“Bitetothebloodandburnintothebone,”astrictscansiongave“Bite to theblood andburn into thebone.” The accent on to was shifted backward,accordingtorule,tothesyllablebite.However,theaccentontoinintocouldnotbesoshifted because there was no adjacent strong syllable, and it was hence eliminated,leavinga lineof four accents.This first stagegenerallyyielded fouror five selectedsyllablesper line.Eachlinewas thenexaminedforstrongsyllablesnotyet receivingstress,andallaccentedpartsofnouns,verbs,adjectives,andadverbswereaddedTheresultwasascannedlinewhichgaveanimportantplacetopracticallyeverysignificantsyllable.Afewspecificexceptionsweremade.Forexample, theprepositionsacross,because,above,about,against andupon were allowed to retain the beat in the firstscanning.Theadverbswhichwerelateraddeddidnot includenot,no,then,thus,so,but,only,all,nordidtheadjectivesincludeno,some,all,none,andsuch.Interrogativeandpersonalpronounswereaddedwhentheywerethesubjectsorobjectsofverbs.Theprocedureyielded27linesoffoursyllables,211linesoffivesyllables,193ofsix,54ofseven,and15ofeight,oratotalof2,819syllables.Alliteration by line. In determining Swinburne’s alliteration by line, the lines

containingone,two,three,four,orfiveoccurrencesofthesameinitialconsonantwerecounted,andthefrequencieswerecomparedwiththeexpectedmeanfrequencies.ThelatterwereobtainedfromthebinomialexpansionN(q+p)n,whereNisthenumberoflinesexamined,n thenumberofsyllablesper line,p theprobabilityofoccurrenceoftheconsonantunderconsideration(calculatedfromitsfrequencyinthesample),andqtheprobabilityofoccurrenceofanyothersound,or1—p.Separatecalculationsweremadefor the linesofvarious lengthsgivenabove,except that the linesofsevenandeightsyllablesweregroupedandtreatedasiftheycontainedsevensyllables.Thetotalexpectationwasobtainedbyaddingtheresultingfrequencies.Thedataforthetenmostfrequentconsonants(p>0.03)aregiveninTable1.

TABLE1LINESINASAMPLEOFSWINBURNECONTAININGVARIOUSNUMBERSOF

CERTAINCONSONANTSCOMPAREDWITHFREQUENCIESEXPECTEDINRANDOMSAMPLING

Numberofoccurrencesperline

389

Practicallywithoutexception,Swinburnehastoomanylinescontainingtwo, three,andfourinstancesofthesameconsonant.Forexample,accordingtoTable1,thereare27linescontainingpairsofb’sinplaceoftheexpected19,fourcontainingthreeb’sinplaceoftheexpectedtwo,andtwocontainingfourb’swherenoneisexpected.Thereisaconsequentshortageoflinescontainingonlyoneb(97inplaceof127)andanexcessof empty lines. The alliterative effect is especially strong for b, f, h, l, andw. Theconsonantsshowsapreponderanceoflinescontainingthreeandfourinstancesbutanactualshortageoflinescontainingtwo.Theeffectforth(voicingignored)isslight,andthatform,g,anddnegligible.Forthe10consonantscombined,thereisanexcessof65lines containing two instances of the same consonant, 23 containing three, and 11containingfour.AsinthecaseofShakespeare,itisdifficulttointerpretrepetitionsofwholewords.

Theseinvolveformalperseveration,butpresumablythemeaningof thepassageisofconsiderableimportanceindeterminingthesecondemissionineachcase.Thenumbersin parentheses in the table are the corrected frequencies obtained by subtracting allcasesoftherepetitionofawholeword.Thecorrectionoperatestoreducetheevidenceforalliteration,perhapsunduly,butsignificantdifferencesarestillobtained.Although the initial consonants in this sampleof500 lines areunquestionablynot

distributedatrandom,anactualestimateoftheamountofarrangementonthepartofthepoetisnoteasilymade.Adefinitemeaningcanbegiventothestatementthatthesamplecontainsso-and-somanyextralinescontainingtwo,three,orfourinstancesofthe same consonant, but in the face of an evident grouping the expected meanfrequencies usedhere are called in question. If some instances of the repetitionof aconsonantareduetoformalstrengthening,thetotalfrequencyusedinthecalculationofpisnotrepresentative,andthevalueofpistoohigh.Thisdefectinthecalculationwill be roughly proportional to the amount of grouping andwill operate to obscuresome of the alliteration actually present. Thus, if we conclude that there are eightexcess lines containing two instances of the sound b and two excess lines eachcontaining three and four instances, and that these lines would otherwise havecontainedonlyoneb(thisissuggestedbythefrequencyoftheone-occurrencelines),

390

then there are at least 18 instances which are due to formal strengthening, and thefrequency used in calculating the value ofpwhich applies to the sample in generalshould have been, not 171, but at least as low as 153. Ignoring the effect of thiseliminationon the size of the total sample,weobtain a value ofp equal to0.05428rather than0.06066, andwe thenexpectonly16 lines containing two instances, andonly one line containing three instances, instead of 19 and two, respectively. Theexcessexhibitedbythetableisthuslessthantherealexcessinthepoem.Whenlittlealliterationisindicated,nocorrectioniscalledforinthepresentdegree

ofapproximation.This isgenerally trueof theShakespearedata.Whereanexcessof“heavy” lines is demonstrated, it has probably been underestimated. For practicalpurposesthecalculationbaseduponthetotalfrequencyofeachconsonantprovidesforasatisfactorymeasureofalliteration.Alliterative span. The analysis of alliteration by line has a certain value in the

criticism of poetry, but it is not an exhaustive survey of alliterative strengthening.Theremaybeadetectablegroupinginsomeshorterunitwithinthelineor ina largeunitof,say,twoorthreelineswhichwouldnotnecessarilyappearwiththeprecedingmethod.Moreover,wherealliterationhasclearlybeendemonstrated,itisdesirabletodeterminetheactualalliterativerangeorspanbytaking thesizeof theanalyzedunitinto account. The perseverative strengthening which follows the emission of aconsonant must eventually die out, and the course of the effect should provide anindicationofitsnature.Thiscoursemaybetracedbyinquiringhowoftentherepetitionof a consonant follows immediately in the next syllable, howoften a single syllableintervenes,howoftentwosyllablesintervene,andsoon.Incollectingthenecessarydata,theinitialconsonants(orinitialvowels,wherethese

occurred) of all selected syllables were placed in order on a long tape. The breakscausedbytheomissionoftwochoruseswereignored.Thetapewasthenpassedundercards inwhichwindowswere cut to display consonantswith anydesired number ofintervening syllables, beginning with adjacent pairs. The numbers of pairs of eachconsonantatdistancesuptothatofeightinterveningsyllablesweretalliedbymovingthetapeundereachcardonespaceatatimethroughoutitslength.Thefiguresforthe10commonestconsonantsaregiveninTable2,togetherwiththeexpectedvalues.Thelatterwereapproximatedbysquaringthefrequencyforeachconsonantanddividingbythenumberofsyllablesinthetotalsample.Thus,wheneverybonthetapeiscomparedwith some other letter in a constant relative position,we expect to obtain, in roundnumbers,10pairs; for thereare171b’s,and theprobabilityof findingasecondb ineachcaseisapproximately171/2819or0.06066.When the second consonant immediately follows the first (with no syllable

intervening between the syllables being compared), 20 pairs of b’s are actuallyobservedinsteadoftheexpected10.Whenthepositioncomparedis thenextbutone(one syllable intervening), 21 pairs are observed. It is only when four syllablesintervenethattheobservedvaluedropstothechancelevel.(Thestandarderroroftheexpected mean frequency is in each case approximately the square root of thefrequency,andthismaybeusedinestimatingthesignificanceoftheobservedfigures.)Theseparatevaluesforeachconsonantarebasedupontoosmallasampletobeveryilluminating, but the total frequencies for all 10 consonants show a very clear trendfromamaximalinfluenceupontheimmediatelyfollowingsyllabletoapproximatelyachanceeffectaftertheinterventionoffoursyllables.ThesetotalshavebeenplottedasCurveAinFigure1,inwhichthechancelevelisalsoindicatedwithbrokenline.The

391

solidcirclesrepresentcolumnsinTable2whichdiffersignificantlyfromthecolumnofexpectedfrequencies;theobservedfrequenciesbeinglikelytooccurlessthanoncein100trialsunderrandomsampling(χ2test).

TABLE2NUMBEROFREPETITIONSOFACONSONANTATVARIOUSDISTANCESFROMTHE

FIRSTOCCURRENCECOMPAREDWITHTHEEXPECTATIONFROMCHANCE

The curve clearly demonstrates that the strengthening effect of the emission of aconsonantisgreatestinitsimmediatevicinityanddropstozeroatadistanceofaboutfour strong syllables. There are, however, certain qualifications to be applied to thisstatement. In the sample here examined only certain consonants show the effect.Moreover,therepetitionofwholewords,presumablydueinparttoconsiderationsofmeaning,mustbeallowedfor.Inthissamplethenumberofpairsofwholewords1withno intervening syllables is 14;with one intervening syllable, 33;with two, 27;withthree,21;andwithfour,12.CurveBinFigure1showsthespanobtainedafteralltheseinstances have been subtracted. Without defining the share contributed by purelyformalstrengthening to the repetitionofawholeword, thesecurvesgiveat least theupperandlowerlimitsoftheeffect.Inthecaseofthelowercurvethelineindicatingthesumoftheexpectedmeanfrequenciesissomewhattoohigh,sinceifweeliminatesomeof theobservedcases,wemustrecalculatewithsmallerfrequencies.Whenonesyllable intervenes, the indicated sum of the expected mean frequencies may be asmuchasfivepointstoohigh.TheeffectofsubtractingpairsofwholewordsfromthecurveforalliterativespanisthusnotquitesodrasticasFigure1maysuggest.

392

FIG.1.Swinburne’salliterativespan.CurveA:Thesumofthenumberofpairsofthe10commonestinitialconsonantsisplottedagainstthenumberofsyllablesinterveningbetweenthesyllablesexamined.Thedottedlineindicatesthesumofthemeanfrequenciesexpectedunderrandomsampling.Thesolidcirclesareforfrequenciesdifferingsignificantlyfromtheexpectedfrequencies.CurveB:Allinstancesoftherepetitionofwholewords(includingallformswithacommonrootwhichdeterminestheinitialconsonant)havebeensubtractedtoshowtheleast

possibleeffectofformalstrengthening.Thedottedlineisslightlyhighforcomparisonwiththiscurve.

Three adjacent instances of the same consonant. Further evidence of formalstrengtheningisprovidedbythepresenceofgroupsofthreeadjacentinstancesofthesameconsonant.Thevaluesofp leadustoexpectveryfewcases,butaconsiderablenumber should result from formal strengthening because bothmembers of a pair of

393

consonants presumably contribute to the heightened probability of occurrence in thethird position. In Table3 the observed and expected frequencies of groups of threeinstancesofthesameconsonantarecomparedforthe10commonestconsonants.Theexcessof28“triples”over theexpected14 isof sufficientmagnitude toconfirm thecumulative action of formal strengthening. A similar result has been obtained bytabulatinggroupsofthreeconsonantswhichcontainonesyllableofadifferentsort,asinthesequencesb-b-r-borb-r-b-b.

TABLE3GROUPSOFTHREEADJACENTINSTANCESOFTHESAMEINITIALCONSONANT

Shakespeare’salliterativespan.Acomparisonof instancesof thesameconsonantswithoutregardtolinearpositionmaybeusedtoexaminethe100Shakespearesonnetswhichpreviouslyyieldedonlyveryslightevidenceofalliterationwhentestedbyline.Itwas impossible toconstructasingle long tape in thiscase, since theendofone

sonnetandthebeginningofanothercouldnotbesupposedtohaveanyrelation,evenifwehadknowntheorderofcomposition.Henceatapeforeachsonnetwasconstructedand tabulated separately. All pairs up to and including those with four interveningsyllableswereexamined.Theprincipaleffectofbreaking thesample intoshortpartswastoreducethenumberofpairswithoneormoreinterveningsyllables;attheendofeachsonnetthereweresyllableswithwhichnosucceedingsyllablecouldbecompared,and there were more of these the greater the number of intervening syllables.Fullallowance for thiswasmadebycalculatinga separatemean frequency incomparingconsonants separated by different numbers of intervening syllables. The uncorrecteddata (no allowance beingmade for the repetition ofwholewords) are shown in theuppercurve inFigure2.Thesumsof theexpectedmean frequenciesarealso shownwithadashedline.Thesumsdeclinebecauseofthereductioninthenumberofpairsavailableforexamination,asjustnoted.

394

FIG.2.Shakespeare’salliterativespan.UpperCurve:Thesumofthenumbersofpairsofthe10commonestinitialconsonantsisplottedagainstthenumberofsyllablesinterveningbetweenthesyllablesexamined.Thedashedlineindicatesthesumsoftheexpectedmeanfrequencies;the

slopeisduetothefactthatwitheachadditionalinterveningsyllablefewerpairsareavailableforcomparison.Thesolidcirclerepresentsasetoffrequenciessignificantlydifferentfromthe

correspondingexpectedfrequencies.LowerCurve:Allinstancesoftherepetitionofwholewordshavebeensubtracted.Thedottedlineissomewhattoohighforcomparisonwiththiscurve.

At first sight the upper curve in Figure 2 may seem to indicate a considerablealliterative effect. Only the first point, however, is significantly different from theexpectedfrequency.Theobservednumbersofpairswhennosyllablesintervenewouldbeexpectedaboutoncein100trialsunderrandomsampling.Theotherpointsshowaplausible trend but are not statistically significant. The lower curve is for the datacorrected by subtracting instances of the repetition of whole words. As withSwinburne,the“expected”lineissomewhattoohighforcomparisonwiththecorrectedcurvebecauseof theneedof recalculationwith reducedfrequencies,but thecurve issafelywithintherangeofvaluescharacteristicofrandomsampling.In order to provide a rough comparison of Shakespeare and Swinburne, the data

obtained with this method were converted to percentages of the expected meanfrequencies.ThealliterativespansofthetwopoetsarerepresentedtogetherinFigure3.

395

It is obvious that with Swinburne alliteration is not only much more common, butextendsoveraconsiderablywiderspan,thanintheShakespearesonnets.

FIG.3.ComparisonofthealliterativespansofShakespeareandSwinburne.ThesumsplottedinFig.1andFig.2havebeenconvertedtopercentagesofthesumsoftheexpectedmean

frequencies.Thesolidcirclesindicatesignificantvalues.Nocorrectionhasbeenmadefortherepetitionofwholewords;therequiredcorrectionwouldbeslightlymoreextensivefor

Shakespeare.

Assonantspan.Aperseverativestrengtheningofvowels(or“assonance”)isalso tobeexpected from the casualobservationofmanypoems, and thepossibilitymaybeinvestigatedwiththepresentmethod.AtapewaspreparedforthevowelsintheselectedsyllablesoftheSwinburnepoem.

Inorder toavoid thenecessityofmakingrelativelyfinedistinctionsinpronunciationand to obtain larger frequencies, the phonemic boundarieswere somewhat enlarged.

AllvowelsmarkedinWebster’sInternationalDictionaryasŭ,ŭeẽr, and ,werecounted as a single“vowel” for present purposes.The samewas true of the sounds

markedôandä,ouandoi,andūand .Theothervowelstabulatedwereā,ă,ē,ĕ,

396

ī,ĭ,andō.Theresultingnumbersofpairsoftheninemostfrequentvowelsatvariousdistances, comparedwith the expectednumbers, arepresented inTable4 and in thelowercurveofFigure4.The result differs considerably from that for consonants. A (barely significant)

excess of pairs appears onlywhen three syllables intervene. The frequencies in thiscolumn(Table4)wouldbeexpectedfewerthanfivetimesbutmorethantwicein100trialsunderrandomsampling.Auniformtrendtowardthispointisevidentbeginning(wherethevowelsareadjacent)slightlybelowthechanceexpectation.Thefifthpointofthecurve,however,isbelowchanceandisthelowestpointinthecurveasawhole.Sucharapidchangefromapointsignificantlyabove, tooneprobablybelow,chancewouldbepuzzlingwereitnotforthefactthatitoccursinthatpartofthecurvewhichrepresentssyllablesoccurringinapproximatelythesamepositioninsuccessivelines.Itwouldappearthatthereisatendencynottousethesamevowelincomparablepositioninsucceedinglines,andthatthisconflictswithatendencytorepeatavowelaftertheinterventionofthreeorfoursyllables.

TABLE4REPETITIONOFAVOWELATVARIOUSDISTANCESFROMTHEFIRSTOCCURRENCE,COMPAREDWITHEXPECTATIONFROMCHANCE

397

FIG.4.Swinburne’sassonantspan.LowerCurve:Thesumofthenumbersofpairsoftheninecommonestvowelsisplottedagainstthenumberofsyllablesinterveningbetweenthesyllablesexamined.Thedottedlineindicatesthesumoftheexpectedmeanfrequencies.Thesolidcircle

representsasetoffrequenciessignificantlydifferentfromtheexpectedfrequencies.UpperCurve:Distributionoflinesofdifferentlength,placedonthehorizontalaxisinapositiontoindicatethenumbersoflineslikelytobringtwosyllablesintoacomparablepositioninsuccessivelineswhentheindicatednumberofsyllablesintervene.Itissuggestedthattheeffectofthisdistributionis,to

someextentatleast,tofillinthesharpbreakinthelowercurve.

A test of this hypothesiswasmade by comparing syllables on the basis, not of a

398

givennumberofinterveningsyllables,butofpositionwithintheline.Thefirstandlastsyllablesinsuccessivelineswereoptimalforthispurpose,becausethequestionofthedefinitionof“comparableposition” in thecaseof lineswhichdidnothave thesamenumberofsyllableswasavoided.Inthe499pairsofeitherinitialorfinalsyllables,weshouldexpect53pairsofthe

samevowel,butthenumberobservedintheinitialpositionisonly43andinthefinal,only34.Thegreatersensitivityofthefinalpositionisperhapsrelatedtotheavoidanceofperfectorimperfectrhymes.Ifthistendencyofsyllablesincomparablepositionnotto contain the same vowel is typical of the other syllables in the line, much of thesuddenchangeinthecurveforassonantspanisaccountedfor.Itmaybenoted thatacorrection for the repetitionofwholewordsbrings the first

two points of this curve to a position which is probably significantly below theexpectedmean,andthatthepeakofthecurveatthreeinterveningsyllablesisbroughtdownwellintotherangeofvaluestobeexpectedfromrandomsampling.Although the result is by no means as clear-cut as in the case of alliteration, a

tentativeconclusionmaybe stated.There isapparentlynoprocessacting to increasethenumberofpairsofvowelswhichfallwithinafewsyllablesofeachother;onthecontrarythereisapparentlysomesuppressionofsimilarwordsatthosedistances.Someformalstrengtheningisperhapsindicatedafterthreeorfourinterveningsyllables.Anobservedtendencytoavoidthesamevowelincomparablepositioninsuccessivelinesinterfereswith theevidenceforassonantstrengthening.Thereseems tobenosimilareffectofcomparablelinearpositionupontheagreementofconsonants.Non-alliterativegrouping.Alliterationdoesnot,ofcourse,exhaustthepossibilityof

consonant-patterning. A more subtle arrangement may well exist without revealingitselfinaconsiderationofsinglesounds.Apreliminarycheckonthispossibilitywasmadebyexaminingallsuccessivepairsofconsonantsregardlessofwhethertheywerethesameornot.Acompletetabulationwasmadefromwhichitcouldbedeterminedhowmany times anygiven consonantwas immediately followedby anyothergivenconsonant in the sample of 500 lines. Expected frequencies were calculated bymultiplyingthefrequencyofthefirstconsonantbythepof thesecond.Thestandarderror used in estimating the significance was computed as the square root ofNpq,whereN is taken as the frequency of the first consonant, andp andq are the usualprobabilities of occurrence and non-occurrence, respectively, of the second. Nodifference between observed and expected frequencies was found as great as threetimesthestandarderror.Anumberofcases,however,yieldeddifferencesgreaterthantwice the standarderror, and itmaybewell to record themhere for the sakeof anyfuturecomparison.AllcombinationsofwhichatleastfiveoccurrenceswereexpectedaregiveninTable

5.Thus,thereissometendencyforbtobefollowedbylandnottobefollowedbypands,andsoon.Itshouldbenotedthatatendencyforaconsonantnottobefollowedbyanotherconsonantmaybeinparttheresultoftheobservedtendencyofconsonantsto be followed by themselves. (In other words, we face the same difficulty inestimatingp in thepresenceof grouping.)For example, ifb is not followed by s asoftenasweshouldexpect,itispartlybecausebistoooftenfollowedbybandpartlybecause many of the s’s are tied up with other s’s. A positive tendency towardassociation,however,isallthemoresignificant.

TABLE5

399

ALLNON-ALLITERATIVECOMBINATIONSOFCONSONANTSOFWHICHATLEASTFIVEOCCURRENCESAREEXPECTEDANDINWHICHTHEDIFFERENCEBETWEENTHEOBSERVEDANDEXPECTEDFREQUENCIESISGREATERTHANTWICETHE

STANDARDERROR

Table 5 does not indicate any consistent sound-patterning over and abovealliteration.Thereseemstobenotendencyforonetypeofsound(dental,labial,andsoon)tobefollowedbyanotherornottobesofollowed.Takingsimilarityofsoundveryloosely,wemaynotethatbmaytendtoavoidpandftoavoidth,but,contrariwise,sapparentlyaccumulatesf’s.Non-assonantgrouping. A similar attempt wasmade to discover combinations of

vowels, in addition to repetitions, which showed exceptional frequencies. The onlysequencewhichexceededtheexpectedfrequencybymorethanthreetimesitsstandarderrorwas that inwhichāwas followedbyĕ. In twoother sequences theexcesswasgreater than twice the standarderror:ā tended tobe followedbyă andăbyō.Twosequences showed a deficiency which was greater than twice the standard error: ĭ

tended not to be followed byā and the groupū, not to be followed by ĭ.Theevidence is barely reliable, only a few sounds are involved, and no consistent trend(e.g.,fromlowtohigh,orfrombacktofront)isexhibited.A“coefficientofalliteration.”Anumerical“coefficientofalliteration”wouldenable

us to make a practical comparison of poems in the extent to which they indicatealliterative selection or arrangement in the behavior of the poet. A convenient andmeaningfulformwouldrepresentthenumberofsoundsinsomeunitnumberoflineswhichproduceanexcessof linescontaining two, three, four,or five instancesof thesameconsonant.Thevalueassumedbysuchacoefficientwoulddependuponthemeanlength of line and the range of lengths, as well as upon the arbitrary rules used inselectingsyllables.Wherethesearecomparable,asuitablecoefficientmaybesetupbyallowing one credit for every excess line containing two consonants, two credits foreveryexcesslinecontainingthree,andsoon,andbyconvertingthetotalexcessofthe10commonestconsonantsintoa(probablyfractional)excessperline.Theprocedureis

400

complicatedbythepresenceofrepetitionsofwholewords,butapracticalcompromisemaybemadebyattributingtopurelyformalperseverationone-thirdoftherepetitionsof wholewords. The precisemagnitude of the excess, in any event, is obscured byfactorsalreadyconsideredconcerningthevalueofp.IncalculatingthecoefficientofalliterationforSwinburne,wefindthatthenumber

of linescontaining two instancesof the sameconsonant, correctedby subtracting allrepetitionsofwholewords,is267.Ifweaddtothisone-thirdofthelinesaccountedforbytherepetitionofwholewords,wegetanobservedfrequencyof285,whichgivesusanexcessof36linesovertheexpected249,oracreditof36.Similarlythecorrectedexcessof21linescontainingthreeinstancesofthesameconsonantmaybeincreasedbyone line (approximatelyone-thirdof the two linescontainingrepetitionsofwholewords)toyieldanexcessof22linesor44credits.Thereisasimilarexcessof11linescontaining four instances of the same consonant, or 33 credits. The total number ofcreditsforSwinburneistherefore113.Dividingthisbythenumberoflines(500),weobtain 0.226 as Swinburne’s coefficient of alliteration. A similar calculation forShakespeareyieldsavalueofonly0.007.If therewerenoalliterationwhatsoeveronthepartofapoet,themeancoefficientwouldbezero,anditisclearthatShakespeareisveryclose to this.Theupper limitof thecoefficientwouldbeobtainedwhenall theconsonants were grouped in solid blocks, but this is a case which need not beconsidered.ItisprobablethatSwinburne’scoefficientof0.226isveryneartheupperlimit to be found in poetrywhich is not deliberately constructed (say, for humorouseffect) upon a principle of alliteration or where alliteration is not the chief poeticdevice,asinAnglo-Saxonpoetry.Toargue fromthestructureofapoemto thebehaviorof thepoet isdifficult.The

patternofapoem(insofarasitispatterned)ispossiblyduetosomethingmorethanaprocess of formal strengthening. Contemporary standards or personal verbalizationsconcerningthefunctionorstructureofpoetrywillhavetheireffects.Whenalliterationisinfashionasanornament,thepoetmaydeliberatelyseekitout,presumablythroughakindofcontrolledassociationpracticedatvariouspointsintheactofcomposition,orthroughtheuseofsuchanartificialdeviceasawordbook.Ontheotherhand,wherecurrenttasteisopposedtoalliteration,instanceswhichnaturallyarisefromchance(aswellasfromformalstrengthening)mayberejected.Nostatisticalanalysisofapoemwill supplydirect information concerning these activities, but itmayneverthelessberegardedasaprerequisite tosayingverymuchabout them.Wecannot trust thepoethimself regarding his practices. He may be unaware of his own behavior inencouraging or eliminating alliterative words. Many poems written “automatically”exhibit various kinds of formal patterning. In any event the poet’s conception of arandomorderofsoundsisprobablyinaccurate,andwemayassumehimincapableofestimatingtheextentofeitherprocessevenifheisawareofit.Thus,althoughwemaybe supplied with a statement of the poet’s philosophy of composition or anautobiographicalaccountofhisbehaviorasapoet,weare stillnotprovidedwithanaccurate statement of the structure of his poems. The selection of instances inillustrationoftheprincipleswhichheexpressesiswhollyunreliable.Wehavenoreasontoassumethattheamountorrangeofformalstrengtheninghere

exhibited is characteristic of the normal speech of eitherShakespeare or Swinburne.Thevalueof the research isnot in supplyinga rigorousmeasurementof thegeneralprocessofformalstrengtheningbutratherinestimatingthelengthtowhichwemaygoin citing these poetic devices as examples of certain processes.The fact that rhyme,

401

rhythm,alliteration,andassonance(thefourprincipalelementsofsound-patterninginpoetry)canbereducedtothesingleprincipleofformalstrengtheningissurelynotanaccident,norisitaninsignificantfact,butsomesuchanalysisasthepresentisrequiredinordertorelatethesecharacteristicsofpoetrytonormalverbalbehavior.It is tobehoped that thisstatementwillallay the fearsof thosewhoreact tosuch

researchasanintrusionintothefieldofcriticism.2Itisdifficulttoseehowananalysisof this sort can have any bearing upon questions of taste, “sensitivity of ear,” or, inshort,uponanyevaluativeestimateofpoemorpoet.Thequestionatissueissimplytheobjectivestructureofaliteraryworkandthevalidityofcertaininferencesconcerningliterarybehavior.

1Throughout,“wholewords”istobeunderstoodasincludinginflectedforms,nouns,andadjectivesfromthesameroot,oranycasewhereasimilarityofformcanberelatedtoasinglesubjectmatter.

2Stoll,E.E.Poeticalliteration.ModernLanguageNotes,1940,55,388–390.

402

TheProcessesInvolvedintheRepeatedGuessingofAlternatives

FromSeptember,1937,toJanuary,1938,theZenithFoundation(withthesupportoftheZenithRadioCorporation)conductedanexperimentin“telepathy”overanationalnetwork.Agroupofsendersinthestudioconcentratedononeofapairofcharactersorsubjectsselectedatrandom.Fromfive tosevencharactersweresentduringeachbroadcast. Each member of the radio audience was asked to write down theimpressionshereceivedandmailthemtothestudio.Thefirstbroadcastwasreportedtohaverevealedstartlingevidencefortelepathy:theradioaudiencescoredfarabovechance.Severallaterbroadcastsproducedasimilarresult.Butseveralshowedequallygoodevidence fornegative telepathy: the radioaudiencewas farbelowchance.TheresultswereanalyzedbyLouisD.Goodfellow (Journal of Experimental Psychology,1938,23,601–632),whodemonstratedthat“neithercoincidenceortelepathy,butthenatural response of an audience to secondary cues, caused the ‘highly successful’results of the Zenith Radio experiment in telepathy.” Goodfellow pointed to manyconditionsoftheexperimentwhichcouldconceivablyhaveinfluencedtheaudienceinchoosing between, say, a cross or a star. He also pointed out that when asked topredict a series of random events, such as tosses of coin, people show certainpreconceived notions of a chance sequence. When the actual selecting mechanismyielded a sequence similar to a preferred pattern, the responses of the audienceappeared to show telepathy. When the selecting mechanisms chose an unlikelysequence, negative telepathy seemed to be shown. The present paper is an effort toaccountfortheobservedpatternsofguessingintermsoftendenciestoalternatecalls.It is reprinted with the permission of the editors of the Journal of ExperimentalPsychology,inwhichitfirstappeared(1942,30,495–503).

In reporting the data from the Zenith experiments on telepathy, Goodfellow1 haspointedoutthatthesixteenpatternswhichresultwhensubjectsareaskedtomakefiveguesses between two alternatives occur with frequencies which are related to theirsymmetry.Goodfellowdefinessymmetryinaspecialwayandmakesnoclaimforitasa psychological process.Nevertheless, the term has recently appeared in a paper byYacorzynski2 as if it referred to a property of perceptual configuration, and patternshavebeenobtainedfrompsychoticsinordertofollowthedisintegrationof“perceptualprocesses.”Tospeakofaseriesoffiveguessesasasingleorganizedactisperhapsinlinewithonetrendinmodernpsychology,butapossiblealternativeview,inwhichaunit of behavior is taken at a lower level of analysis, needs to be stated. It is alsoimportanttocomparethetwolevelswithrespecttodescriptivepower.Guessing is a special kind of (usually verbal) behavior in which two or more

responses are about equally likely to be emitted. In guessing “heads or tails,” forexample, theconditionsof theexperimentstrengthen these twoverbal responses,butnocircumstancestrengthensonetotheexclusionoftheother.Theresponsesmayhave

403

slightlydifferent restingstrengths, so thatonewillbe“preferred” in the longrun,orsomesuchcircumstanceasabiasedinstructionmaymomentarilyalterthebalance;butthereisnocleardeterminerofeitherresponsesuchasexists,say,inthecaseofreadingacoinalreadytossed.Guessingrepaysstudybecauseitthrowsintoreliefcertainminorvariables in the determination of verbal behavior which are normally obscured byvariablesofgreatermoment.The first guess in a series of five, as in the Zenith experiments, is apparently

controlled by an abiding preference, by biased preliminary conditions, or by trivialcircumstanceswhich cancel out in the long run and are spoken of as “chance.”Thesecondguessraisesadifferentproblem,foritisundertheadditionalcontrolofthefirst.Likewise,thethirdguessisunderthecontrolofthesecondandperhapsthefirst,andsoon.Goodfellownotesthefactthatsubsequentguessesdependupontheprecedingbutdismissestherelationapparentlyastoocomplexforanalysis.However,ifwearenottofall back upon thewhole series as a configurational unit,wemust account for eachguess in turn, and this demands a statement regarding the effect of one guess uponanother. The additional variables noted in the preceding paragraph are presumablyoperativeuponallfiveguesses,butwemayeliminatethembyfollowingGoodfellow’spracticeofexpressingallsequencesintermsofthefirstcall.Thus,both“heads-tails-heads”and“tails-heads-tails”are recordedas“121.”Our problem is to describe thepercentagesofthefinalpatternsexpressedinthiswayintermsoftheeffectsofeachguessuponsubsequentguesses.In Table 1 the whole guessing process in the Zenith experiment has been

reconstructed.Thebold-facednumbersnotinparenthesesintheright-handcolumnarefromGoodfellow’s tablesandrepresent ineachcasethefrequencyof thecall-patternshownjustbelowinitalics.Thus,4.34percentofthepersonsreplyingtothebroadcastsreportedthesequence12121and5.70percent thesequence12122.Thesumof thesetwofigures(10.04)givesthepercentagewhoguessed1212onthefirstfourcalls,andthisfigure(alsoinbold-faceandwithoutparentheses)isenteredinthesecondcolumnfrom the right. Its corresponding four-placepattern is indicated just below.All four-placepercentageshavebeenenteredinthismanner.Fromthemweareabletocalculatepercentagesforthethree-placecallsandinturnforthetwo-place.Thus,21.94percentguessed121on thefirst threecallsand52.93percentguessed12on the first two. InTable1eachofthesefigureshasalsobeenconvertedintoapercentageofthefigureatthe immediately preceding stage. The results are given to one decimal place inparentheses(andalsoinbold-face).Thus,the21.94percentwhocalled121 represent41.5percentofthosewhocalled12,andthisfigureisshowninparenthesesjustbefore21.94 in the table. The corresponding percentage for those who called 122 is notneededinwhatfollows.Afurtherstepistodescribeeachcallintermsofwhetheritinvolvesalternationfrom

the preceding. The sequence 12 shows alternation and is accordingly marked A inTable1;thesequence11doesnotandismarkedO.ThedesignationAOAAmeansthatthe sequence12212 shows alternation followed by a failure to alternate followed bytwoalternations.Wearenowabletocollecttogetherthevariouspercentagesofalternationaccording

to thehistoryofalternation. In thecolumnheaded“Zen.” inTable2 thepercentageshavebeengroupedaccordingtotheconditionswhichprecedethelastA.Theyrevealasignificantuniformity.There isanoriginal tendencywhichproducesalternation52.9percentofthetimeonthesecondguess.(Thistendencyisactuallyveryslight,since50

404

percentwouldindicatenoneatall.)Whenacallfollowsanalternationalreadymade,thepercentageofalternationdropstothelowforties(Lines2,3,and4,Table2),butitrises to theseventieswhentheprecedingcallhasnotshownalternation(Lines5–11,Table2).(Thetwoexceptionsinthesixtiesareforfifth-placecalls.)

TABLE1RECONSTRUCTIONOFTHEGUESSINGPROCESSINTHEZENITHEXPERIMENTSThe

italicsindicatethepatternsofguesses;theprecedingA’sandO’sdescribeeachintermsofthealternationsinvolved.TheZenith-Goodfellowdataareinbold-face;the

numbersinparenthesesindicatethepercentageofalternationfromtheprecedingstage.Thenumbersnotinbold-faceareconstructedfromthefourselectedpercentagesin

Table2,Lines16-19.

405

406

TABLE2PERCENTAGESOFALTERNATIONATVARIOUSSTAGESINTHEGUESSING

PROCESS

WecangiveafairlygoodaccountofthetablebyroundingoutthethreeconstantsinLines16,17,and18,whichstatethepercentagealternationswhichobtainaccordingtowhether there isanimmediatelyprecedingcallandwhether it involvesalternationornot. It would be surprising if it were not necessary to go farther back than theimmediately preceding call, but the only cases which call for special treatment arethoseinwhichaprecedingAisitselfprecededbyanO(Lines12–14).ThepresenceoftheOapparentlyprevents the reduction to43percentwhichAotherwise implies.AtleastthisistrueforOAAandAOAA,andinordertoavoidmultiplyingconstantswemay let the same value of 54 percent apply to this group.OAAA is an outstandingexception,tobediscussedlater.TheadequacyoftheselectedpercentagesatthebottomofTable2maybeestimated

byreconstructingTable1exclusivelythroughtheiruse.Bybeginningwith100percentandtakingthepercentagesdictatedbytheconditionsofalternation,weobtainsixteen

407

final percentages. These, together with various percentages enroute, are shown inTable1 directly below the correspondingZenith figures,withwhich they should becompared.Withtwoexceptionsthecorrespondenceissatisfactory.Thefailuretodescribeonepairofpercentagesisacuriousone.Callsbeginningwith

1121aresatisfactorilyaccountedforatOAA,but thesubjectswhohavereachedthispoint tend to alternate about 74 percent of the time instead of 43 or 53 percent asexpected.TheprecedingOAAhashadthesameeffectasO(althoughthesimilarityinmagnitudemaybeacoincidence).There seems tobenospecial characteristicof thesequence11212toexplainthisanomaly.Itcannotbeinterpretedasanavoidanceoftheconspicuous symmetry of 11211 (not to be confused with Goodfellow’s symmetry,whichshowsadifferenceofonlyonepointbetween11212and11211), since there isonly negligible evidence of such avoidance elsewhere.3 Similarly, no appeal can bemadetothefactthat11212workstowardeveningthescorebetween1’sand2’s, sincethe table shows no tendency to exceed the calculated figures in the direction ofequalizingcalls.Ifwewishtousestatementsconcerningalternationtoaccountforthemajorpart of the table,wecannot appeal to these additionalprocesses to explain anatypical case since they are incompatible. (This does not exclude the possibility ofaccounting for thewhole tablewith someother set of statements, possibly involvingsymmetryorthebalancingof1and2.)IfweaccepttheZenith-Goodfellowfiguresasavalidindicationofgeneraltendencies(asthesizeofthesampleseemstowarrant),ouronlycourseistosetupaseparatestatementforthecaseofOAAA.Theeffectofanalternationmustbedistinguishedfromthatofasingleguess.It is

notaltogethertheprecedingresponsewhichdeterminestheresponsetobemadebutinadditionwhether this responsehas involvedalternation.This isadistinctionofsomeimportance. Studies of formal patterning in speech have frequently indicated asubstantialtendencytorepeat a responsealreadymade (see, forexample,page385),and various lines of evidence suggest that this is a primary characteristic of verbalbehavior. Nevertheless, a tendency is set up in the growing child, through readilyobservableprocessesofconditioning,whichopposesrepetition.4Inauniverselimitedto two responses,not-repeating is,ofcourse, alternating. In thepresentcase the firstguessmaybesupposedtostrengthenthesameresponseforasecondemission.Buttheeffect of reinforcement opposing repetition is to counteract this tendency so thatalternations occur 53 percent of the time.When this figure is compared, notwith achance 50 percent, but with a probable tendency to repeat the same form, greaterimportancemay be attached to the process which produces alternation. The presentdatademonstratetheadditionalfactthatalternationismorelikelytooccurafterfailuretoalternateandlesslikelyafterpreviousalternation.Thismayplausiblybeattributedtotheconditionsofreinforcementwhichestablishthetendencytoalternate,sincetheverbal community which provides the reinforcement opposing repetition willpresumablyreactmorevigorouslytorepeatedfailuretoalternateandlessvigorouslytofailurewhichfollowsduealternation.ThestatementsextractedfromTable2require littleorno interpretation,since they

have a very direct reference to behavior and describe processes which are quiteplausibly explained. For this reason the correspondence between the Zenith and thecalculatedpercentagesismoresignificantthanacorrespondencewithsymmetries.Itisalso probably amore complete correspondence.Although symmetrymay be said togivethecorrectrankorderofthepercentages,thereareasmanyasfivecaseswiththe

408

samenumberofsymmetries,andtherelativeorderoftheircorrespondingfrequenciesisnotpredicted.Nordoessymmetrygiveanyclueastomagnitude.TheadditionaldatasuppliedbyYacorzynskicannotbeasadequatelyaccountedfor

with a few numerical statements. The percentages of alternation exhibited by hiscontrol group, under conditions which were intended to approximate those of theZenith experiments, are given in Table 2 under “Con.” The original percentagealternation is lower (49.8 percent) and, in fact, shows no tendency to alternatewhatsoever.Thedropfollowingalternation(to37.4percent)iscomparablewiththatintheZenithcase,andtherisefollowingfailuretoalternateisofthesameorevenslightlygreatermagnitude.But inspiteof thisroughagreement, thecontrolfiguresaremuchlessconsistent,anditisimpossibletoselectroundedvalueswhichgiveagoodaccountofthefinalpercentages.Thedifficultyseemstobeinthesmallnessofsampleandisone which especially besets the fifth-place calls where the percentages are of verysmall groups. The anomalous result at OAAA in the Zenith-Goodfellow data alsoappears,butlessstrikingly,inYacorzynski’scontrolgroup.Yacorzynskihas shown that the sequencesgivenbypsychoticsdiffer significantly

fromhisowncontrolsandfromtheZenith-Goodfellowfigures.Doestheinterpretationjust advanced for the normal case throwany light on the difference betweennormaland psychotic behavior? The percentage alternations under “Schiz.” and “M.-D.” inTable2providewhateveranswerisavailablefromthesesmallsamples.The schizophrenics begin with a normal tendency to alternate but this is not

decreased when one or more alternations have been made. On the contrary thepercentagesrise,reaching83.2inthecaseofAAAA(ifthefigureisreliable).Suchanincrease would result in part from a mixed population, in which there were somememberswhoinvariablyalternated.Inanyevent,thenormaltendencynottoalternatefollowingalternationislacking.Theotherconspicuousfeatureofthenormalcase—theincrease in percentage-alternation after failure to alternate—is also lacking. It is truethatthevalueforOAis64.6,anincreaseover52.6,butitislessthanthevalueforAA,andthevalueforOOAisconsiderablylower.Ingeneral,theremaybesomeincreaseinalternation as the guessing proceeds, but this holds regardless of the nature ofpreceding calls. Essentially the same may be said of the manic-depressive group,exceptthattheoriginaltendencytoalternateishigh(61.3).Thereductionproducedbyearlieralternations(to59.2and58.7)isprobablynotsignificantand,asinthecaseoftheschizophrenics,averyhighvalue(thoughofdoubtfulsignificance) isreachedforAAAA.Any increase in tendency to alternate following failure to alternate (see thegroupendingOA)isalsoclearlylacking;indeed,adecreasemightbeclaimed.Sofaras thesesmallsampleswarrantgeneralization,wemaysummarize thepercentagesofTable 2 by saying that the psychotics show no effect of past alternation uponalternationinthecalltobemade.This is equivalent to saying that the percentages can be described with a single

statement,givingapermanenttendencytoalternate,ratherthanwiththethreeorfourdemanded in the caseof thenormal subject.From the combinedpercentages for thepresent groups (last column in Table 2) this tendency will be seen to be in theneighborhood of 60 percent, although there are large deviations. The adequacy of asingle statement may be tested by constructing final percentages on the basis of aconstant 60 percent alternation. The resulting sixteen percentages, together with theobserved,aregiveninTable3.Abouthalfthecasesshowacloseapproximation.Theothersdifferchieflyon the fifthcall,where the sample is smallest.Forexample, the

409

largestdifferencesappear in thefirst twopairsofpercentages,yet theirsumsshowavery good correspondence (22.0 and 21.6). All calls beginning with 1212 areadequatelydescribed,althoughthefinalcallisnot.Thediscrepancyisofsuchasortastobeproducedbyasmallgroupwhoinvariablyalternate,althoughthecorrespondenceat1212placesthisinquestion.Theonlyotherseriousdiscrepancyisfor11111,whichmayrepresentasmallperseverativegroupwhodonotalternateatall.

TABLE3PERCENTAGESFORTHESIXTEENPATTERNSOBTAINEDFROMPSYCHOTICSANDPERCENTAGESCALCULATEDONTHEBASISOFACONSTANTTENDENCYTO

ALTERNATE

One might conceivably describe this sort of psychotic behavior as showing adisintegration of perceptual processes, but by analyzing the pattern of a series ofguesses(withitsdubiousstatusasanaspectofbehavior)intothediscreteeventswhichmakeitup,asimplerandmoreusefulstatementmaybereached.Inmakingaseriesoffive calls between two alternatives the psychotic differs from the normal in beinguninfluencedbyhisownpastbehavior.Thestatement isplausibleenoughinviewofwhatisknownaboutpsychoticbehavioringeneral.

ANoteonMethod

Inamatterofthissortitisveryeasytodropintothejargonofpostulationalmethod.We might say that we assume certain tendencies to alternate and then validate ourassumptionsbypredictingthefinalpercentagescorrectly.Butthese“assumptions”areactuallynothingbutdescriptivestatements.Weexaminepercentagesofalternationatvarious stages andnote that they are related to certain anterior conditions.Thus,weassert thata subject tends to alternate 75 percent of the time if he has just failed toalternate,butatno timedoes theassertionappearasanassumption. It is, rather, theapproximate statement of a fact.Our net result is, not that certain assumptions havebeentestedandshowntobeadequate,butthatwehavefounditpossibletotalkaboutsixteenpatternsofbehaviorwithafewcomprehensivestatements.Itispossiblethatanyexampleofpostulationalmethodintheempiricalsciencesmay

beinterpretedinthesamewayandthat“predicting”afactfromasetofassumptionsisneverintendedtomeanmorethandescribingit.Butifthisisthecase,thepretenseofdeduction shouldbe abandoned, at least by thosewhoare interested in adescriptivescience of thought andwhowish, therefore, to see the number of thought processesreducedtoaminimum.

Summary

410

ParticipantsintheZenithexperimentsontelepathywereaskedtomakeaseriesoffiveguesses between two alternatives. The percentages of the sixteen resulting patternshavebeenreportedbyGoodfellowandarehere interpreted in termsofa tendencytoalternatecalls.Quantitativeestimatesofsuchatendencyaremadeforeachofseveralconditionsofalternation,and theseare testedbyconstructingpercentages forsixteenpatterns to be compared with the Goodfellow percentages. Yacorzynski’s data frompsychoticsdifferfromthenormalmerelyinshowingnoeffectofpastalternations.Theseinterpretationsareofferedasmoresatisfactorythanreferencesto“symmetry”

or,inthecaseofthepsychotics,tothedisintegrationofperceptualprinciples.

1Inthereferencegivenabove.2Yacorzynski,G.K.Perceptualprinciplesinvolvedinthedisintegrationofaconfigurationformedinpredictingtheoccurrenceofpatternsselectedbychance.J.Exp.Psychol.,1941,29,401–406.

3Ifitwereoperative,122oughttobefavoredover121,12122over12121,and12222over12221,butalthoughthepercentagesforthesenon-symmetricalsequencesexceedthecalculatedpercentages,theaverageexcessisonly0.3percentabovetheslightaverageexcessofthesymmetricalsequence.Thiswillnotgoveryfartowardaccountingforanexcessof4.33percentatOAAA.

4Reinforcementsappliedtospeechtoopposepriortendenciesarefairlycommonand,indeed,giverisetosomeofthemostimportantpropertiesofverbalbehavior.Thereisaspecialprobleminvolvedinseparatingopposedeffectsformeasurement,butitisnotinsoluble.

411

PARTVIII

TheoreticalConsiderations

WhyAretheBehavioralSciencesNotMoreEffective?TheConceptoftheReflexintheDescriptionofBehaviorTheGenericNatureoftheConceptsofStimulusandResponseTwoTypesofConditionedReflexandaPseudo-typeTwoTypesofConditionedReflex:AReplytoKonorskiandMillerAReviewofHull’sPrinciplesofBehaviorAReviewofBushandMosteller’sStochasticModelsforLearning

412

WhyAretheBehavioralSciencesNotMoreEffective?

Itisoftensaidthatsciencehasbecomethereligionofthetwentiethcentury—withitsBook,itsprophetsandpriests,anditscommunicants,nottomentionitsapostates.Likemostreligions,italsohasitsapocalypticvision:ourwayoflife,andpossiblymankinditself, may not long survive if we continue on our present course. An expandingpopulationwillexhaustourresourcesandpollutetheenvironmentandsoonerorlater(soonerifwesufferanuclearholocaust)putanendtothekindofworldinwhichthespecies can live.Thevision isnotdivine revelationbut an inference from facts, andalthoughthespeedwithwhichwearemovingtowarddestructionmaybedebated,fewscientistsquestionthedirection.Salvationmaycomespontaneouslyfromsomekindofinbuiltcorrectiveprocess,butitismorelikelythatweshallsaveourselvesonlyifwesolve our problems in a quite deliberate fashion. To do sowe need amuch clearerunderstanding of why people behave as they do. We need, in short, a science andtechnologyofbehaviorwhichwillpermitustodealwiththebehavioralaspectsofourproblems as effectively as other technologies dealwith their physical and biologicalaspects.Nothingofthesortis,ofcourse,athand.Whatwehavearetheso-calledsocialor

behavioral sciences—among them psychology, sociology, anthropology, politicalscience, and economics.They are clearly not adequate to the useswe should like tomake of them. Why not? Philip Handler, president of the National Academy ofSciences, has said that they are “too young,” but they are actuallymuch older thanmanybranchesofphysics,chemistry,andbiologywhichhaveprovedveryhelpful.Acommon explanation is that there is something about human behaviorwhich puts itbeyondthereachofscience.Thatwouldbedisastrousiftrue,buttheargumentpointsin a useful direction. There is something about human behavior which has made ascientificanalysisdifficultanddelayedthedevelopmentofatechnology.With any other species we should explain behavior by pointing to a genetic

endowment—to the anatomy and physiology of the species—and to events in thehistoryof itsmembers.But there is somethingspecial about thehumancase:wearemembersofthespecieswearestudying.Thatfactshouldnotbetroublesome.Onthecontrary,itshouldmeanthatweareinaparticularlyfavorablepositioninaccumulatingfactsaboutbehaviorandabout the relationofbehavior togeneticandenvironmentalhistories.Moreover,totheextentthatwecanobserveourownbodies,itshouldmeanthatwehaveaspecialkindofinformation—insideinformation,asitwere—abouttheeffectsofthosehistoriesonanorganism.Alongtimeago,however,thatinformationwas woven into a very different kind of explanation: the kind we give when weattribute what a person does to his feelings, thoughts, states of mind, purposes,expectancies,andsoon.Ishallnotgointothestandardobjectionstothatkindofthing,butIhopetoshowthatthebehavioristicpositionhastakenonanewsignificanceinitsbearing on current problems. I shall argue, in short, that the social sciences are notmoreeffectivepreciselybecausetheyarenotfullybehavioral,andforthatreasonnot

413

really scientific, and for that reason not commensurate with the problems they areaskedtosolve.Consider the fact that increasingly largenumbersofyoungpeoplearebehaving in

wayswhich deeply disturb their elders. They stay aloof from their families or leavehomealtogether.Theyneglectthesocialamenities:theyarepoorlydressed,dirty,andrude.Theytake timeofffromschoolor theuniversity,ordropoutcompletely.Theywork,ifatall,irregularlyandindifferently,andmanyarecontenttobegforthethingstheyneed.Theystealandcondonestealing,callthepolice“pigs”andattackthemwhenthey enforce the law, desecrate flags, burn draft-cards, refuse to serve in the armedservices, andsometimesdefect toothergovernments.Suchbehavior is shockingandincomprehensible toalmosteveryoneandamatterofconcernevenforyoungpeoplethemselveswhentheyhavesampledthewayoflifetowhichitleads.And what have the behavioral sciences to say about it? In general, this: there is

somethingwrongwith peoplewhobehave thatway.Theyoungperson is said to besufferingfromadisturbedpersonality.Heisalienatedandrootless.Helackspurpose.He is suffering from “misperception” of the world around him. He has no sense ofachievement.Heisfrustratedanddiscouraged.Hehasnosenseofidentity.Hecanfindnothing that interests him (technically speaking,he is suffering fromanhedonia).Helacks rules to liveby (technically speaking,he is suffering fromanomie).Hisvaluesarewrong.According to such an explanation, our task is to correct disturbed personalities,

changetroubledstatesofmind,makepeoplefeelwanted,givethempurposeorasenseof pride in their work, allay their frustration, and teach them the value of order,security, and affluence. But we have no direct access to states of mind, feelings,purposes,attitudes,opinions,orvalues.Whatwedoistrytochangethebehaviorfromwhichweinferthingsofthatsort,andwechangeitonlybychangingtheenvironment,verbal andnonverbal, inwhichyoungpeople live.By responding in certainways towhatapersondoes,wemaychangethebehaviorfromwhichweinferthathedoesnotfeelwanted,anditisquitepossiblethathemaythenfeelwanted.Byteachingapersonto produce things of which he can be proud, we may make him productive andindirectlycreate“asenseofcraftsmanship.”Byarrangingbetterincentiveconditions,wemayinducehimtobehavemoreindustriously,andwhathefeelswillprobablyalsochange.Bymakingsurethathewillbegenerouslyreinforcedforworkingforacause,wemaycreatethekindofbehaviorwecallloyal,and,incidentally,createaconditionwhich may be felt as a “sense of loyalty.” By reinforcing nonverbal and verbalbehaviorinparticularways,wechangewhatapersonsaysordoes,butwhathesaysordoesisnotduetohisopinionsorattitudesbuttothecontingenciesofreinforcementwehavearranged.The traditional view misrepresents our task. It suggests that by changing an

environment we first change feelings or states of mind, and that these, in turn,determinewhatapersondoes.Thefeelingorstateofmindseemstobeanecessarylinkinacausalchain,butthefactisthatwechangebehaviorbychangingtheenvironment,and,indoingso,changewhatisfelt.Feelingsandstatesofmindarenotcauses,theyareby-products.Byturningdirectlytotheenvironmentalhistory,ratherthantoitsperceivedorfelt

effects,wemaytakeadvantageofcertainrecentadvancesintheexperimentalanalysisofbehavior.Inhundredsof laboratoriesthroughouttheworld,complexenvironmentsarearrangedandtheireffectsstudied.Theevidencegrowsmoreandmoreconvincing

414

thatapersonbehavesashedoesbecauseof(1)whathashappenedinthedistantpastashisspeciesevolved,and(2)whathashappenedtohiminhislifetimeasanindividual.Practicalsolutionsofsomeimportantproblemsthenfollow.One fact has become particularly clear: people do things because of the

consequences.Ithaslongbeenrecognizedthatsomeeffectsofaperson’sbehavioraresatisfying or rewarding, but a special significance is emphasizedwhenwe call theseeffects “reinforcing”: they strengthen the behavior they are contingent upon, in thesense of making it more likely to occur again. (Other consequences are said to bepunishing,andtheireffectsarerathermorecomplex.)Thebehavioristisoftensaidtotreatbehaviorsimplyasresponse tostimuli,but thatviewhas longbeenoutofdate.Three thingsmustbe taken into account: the situation inwhichbehavioroccurs, thebehavior itself, and its consequences. These three things are interrelated in veryintricatewaysinwhatarecalledthe“contingenciesofreinforcement.”Someextremelycomplexcontingencieshavebeenanalyzed,andtheresultshelpininterpretingsomeofthecontingencieswhichprevailindailylife.Howreinforcingisayoungperson’shomesimplyasaphysicalplace?Howdoesit

lookorsoundorsmell?Howoftendoothermembersofhisfamilyreinforcehimwithattention, approval, or affection—and for what behavior? How often do theydisapprove of him and punish him? What competing contingencies await himelsewhere?Do theapprovalanddisapprovalof thepeoplehemeets fromday todayshape and maintain the kind of behavior which, as we say, observes the socialamenities,ordotheymakehimrudeandselfish?Contingenciesofthissort,ratherthanany feeling of commitment or alienation, create the problem we are examining.Alienation is not a state of mind; it is a state of behavior attributable to defectivecontingenciesofreinforcement.Whatisfeltisaby-product.Howreinforcingisaschoolsimplyasaphysicalplace?Howdoesitlook,sound,or

smell? What happens when a student behaves well toward his teachers and otherstudents,andwhathappenswhenhebehavesbadly?Doeshestudymainlytoavoidtheconsequences of not studying or are there positive effects—such as conspicuousprogress toward the mastery of a subject or skill? How long will it be before thebehavior acquired in school has reinforcing consequences elsewhere which willcontributetoitsstrength?Whatcontingenciesintheoutsideworldcompetewiththosein school? The answers bear directly on our problem. If a young person often staysawayfromschoolordropsout,itisnotbecauseheisshiftless,orlackscuriosity,orisdull;itisbecausethecontingenciesofreinforcementdonotkeephimatschool.How reinforcing is the place in which a young person works? How do his

supervisorsand fellowworkers treathim? Iswhathedoesdangerousorexhausting?Doesitlackvariety?Isthereanyreinforcingconnectionwithafinalproduct?Iswhathe is paid sensitively contingent upon how hard or how carefully he works? Howreinforcing to him is money? How serious is the loss when he takes a day off orchangesjobsorquitsworkaltogether?Theanswersarecrucial.Ifayoungpersondoesnotworkproductively,itisnotbecausehedoesnotlikehisjoborislazy;itisbecausethecontingenciesaredefective.Hisfeelingabouthisjob,andthetraitsofcharacterhedisplaysinit,areby-productsofthesamecontingencies.And so with governmental sanctions. People obey the law primarily to avoid

punishment,andtheymayavoiditorescapeinotherways,suchasbydefecting.Theextenttowhichapersonusesthelawalsodependsupontheconsequences.Whetherheturnstothegovernmentortakesthelawintohisownhandsdepends,inpart,uponthe

415

results;aretheyquickandaretheyreinforcing?Theconsequencesofparticipating ingovernment are also important. What is the effect of casting a vote? Is a personreinforcedbywhatagovernmentdoeswithhishelpwhenhejoinsthearmedservicesor accepts a draft call? Behavior traditionally attributed to loyalty or disloyalty,affection or disaffection, commitment or anomie, is the product of specifiablecontingencies.Howapersonfeelsabouthisgovernmentisaby-product.Somuchforaroughinterpretation.Letusturntoapracticalsolution.Disaffection

andrevoltcouldscarcelybebetterillustratedthanbythebehaviorofyoungoffenderslivinginaschoolforjuveniledelinquents.Fewhavehadfamiliestheyhavelivedwithclosely,almostallhavedroppedoutofschoolwithlittleornoeducation,fewhaveeverheldajobforanylengthoftime,andallhavebrokenthelawsooftenorsoviolentlythat it has been necessary to lock them up. It is hard to imagine a group of youngpeoplemorecompletelyoutofcontrolofthecultureoftheircountry.Buttheyarenotout of control of their own culture, and theymay be brought under the control of abetterone.A group of about forty juvenile offenders—teenage armed robbers, rapists, and

murderers—who were at the time “students” in the National Training School inWashington,D.C., since relocated at theRobert F.KennedyCenter inMorgantown,WestVirginia,participatedinanexperimentdirectedbyHaroldCohenoftheInstituteforBehavioralResearch in Silver Spring,Maryland.A new social environmentwasconstructedinwhichnoboywasrequiredtodoanything.Hecouldsleeponapadinadormitory,eatnutritiousifnotverypalatablefood,andsitonabenchallday.Buthecouldgreatlyimprovehislotbyearningpointsexchangeableformoredeliciousfoodatmealtimes,admissiontogamesrooms,therentalofaprivateroomortelevisionset,oreven a short vacation away from the school. He could earn points by doing simplechores, but muchmore easily by learning things. Correct responses to programmedinstructionalmaterialsandcorrectanswersinexaminationsafterstudyingotherkindsofmaterialmeantpoints.Theresultsweredramatic.Boyswhohadbeenconvincedbytheschoolsystemthat

theywereunteachablediscoveredthattheywerenot.Theylearnedreadingandwritingand arithmetic, and acquired other verbal and manual skills. They did so withoutcompulsion, and the hostile behavior characteristic of such institutions quicklydisappeared.Theydiscovered that itwaspossible toearn rather thansteal the thingstheywanted,andtheybegantoparticipateintheorganizationandmanagementoftheschool.Asaresult,theywerepreparedtoleadmoresociallyacceptablelivesaftertheyleft. Inthenormalcourseofevents,85percentof themwouldhavebeenbackintheschoolbytheendofoneyearbecauseoffurtherviolations,butthefigurewasonly25percent.Attheendofanotheryearithadrisento45,butthenormalfigurewouldhavebeenpractically100percent.Itwouldbepleasanttoreportthatthechangewaspermanent,butafterthreeorfour

yearstherewaslittleevidenceofanyfurthereffect.Theboyshadbeenexposedtothisexceptionalenvironmentforonlyafewmonthsoratmostayear,whichwasapparentlynotenoughtooffsetdeficienciesintheenvironmentstowhichtheyreturned.Butwhiletheywereinschool,andforsometimethereafter,thedesignedculturehaditspredictedeffects.Inabetterworld,theylivedbetterlives.Theworldgavethembetterreasonsforbehaving well, for working to produce some of the goods they needed, and foracquiringbehaviorwhichmadethemsuccessfulinotherways.Atrainingschoolisacultureinminiature.Itcanbesuccessfullydesigned.Andso

416

can the culture in the world at large. Hundreds of experiments are now in progressexploring the range of what has come to be called “behavior modification.” Newpractices in child care, in the management of institutionalized retardates andpsychotics, in individual psychotherapy, in classroommanagement, in the design ofincentivesystems in industryandelsewherearebeing tested.Weare, I think,onourwaytothetechnologyweneedtosolvemanyofourproblems.Butprogressisdishearteninglyslow.Socialscientistshavenotyetfullyunderstood

thesignificanceof thebehavioristicposition.Mostof themstill lookforsolutions totheir problems inside the people they study. In psychotherapy, the medical analogypersists: the problem ismental illness, and it is the patientwhomust be cured. Thetherapisttriestoreachhispatientbymakinganinterpersonalcontact,notbychanginganenvironment.Educatorsplace theirbetson thenaturalcuriosityof thestudent,hislove of learning, his creative spirit. The teacher is at best amidwifewho helps thestudentgivebirthtoknowledgeandwisdom.Schoolsanduniversitiesstillselectgoodstudents;theyhavenotyetbeguntocreatethem.Andindustrystillselectsworkerswhoareindustrious,skilled,andcareful;ithasnotgivenseriousattentiontothedesignofcontingencies underwhich everyoneworks hard and carefully and enjoys hiswork.Governmentsstillholdtheindividualresponsibleandaresaidtobebestiftheygovernleast, because a person is then free to behavewell because of inner virtues.All thiscontinues todivert attention from the taskofbuildinga social environment inwhichpeoplebehavewellwithrespecttoeachother,acquireeffectiverepertoires,producethegoodstheyneed,andenjoylife.Itobscuresthefactthattheproblemistodesignbettercultures—notbetterpeople.Thebehavioralscienceshavebeenslowto take thisstepformanyreasons,among

them a characteristic response to the possibility of a technology of behaviorcomparable in power with the technologies of physics and biology. Perhaps humanbehaviorcanbecontrolledviatheenvironment,butwhowillexertthecontrol?Thosewhoaskthatquestiondonotexpecttobegivenapropername.Itisnotthepersontheyareworriedaboutbuthis intentions,hispurposes,hisgoodwillorbenevolence.Butthisisonlyanotherexampleofwhatiswrong.Theyarestilllookingforassurancetothestatesofmindorthefeelingsofpotentialcontrollers.Whattheyshouldbeaskingis: “What kinds of cultural contingencies induce people to engage in the control ofother people? Under what contingencies do people act like tyrants? Under whatcontingenciesdotheyactlike‘menandwomenofgoodwill’?”Wemusthopethataculturewillemerge inwhich thosewhohavepowerwilluse it for thegeneralgood.Suchaculturewouldprobablybemostlikelytosurvive,andthatisanimportantpoint.Geneticistsarebeginningtospeakofchangingthecourseofhumanevolution,butwehave long been able to change the evolution of cultures.Aswe begin to understandwhatacultureis,wemaybegintomovetowardbetterdesigns.Atechnologyofbehaviormaygiveaidandcomforttocultureswhichbreeddespots,

butitwillnotgivethemanadvantageunlessthecultureswhichbreedmenofgoodwillrefusetoaccepttheaidandcomfortwhicharealsoofferedthem.Ascienceofbehaviorhasemergedinthekindofcultureweliketocallfree,butitisquitepossiblethatitwillnot be put to use there.A historical parallelmay be significant. In the fifth centuryB.C.,Chinawasasadvancedinphysicaltechnologyasanyotherpartoftheworld,andit retained its positionuntil about 1400A.D.Among its great contributionswere thecompass, gunpowder, andmovable type. But for the next three hundred years, very

417

little use wasmade of them.Military power remained ceremonial, astrological, andgeomantic,longvoyageswereforbidden,andanideographicsystemofnotationgainedlittle frommovable type.A neo-Confucianist system of thought emphasized passiveknowing.TheWestseizedupontheseChineseinventions,however,andmadefantasticprogress.Thecompassenabledmentoexploretheglobe,andgunpowdertoconquerit.Movable type led to a flood of books which brought about a revival of learning.Meanwhile, China remained a medieval society. The difference was not in theavailability of technologicalmeans but in the cultural contingencies governing theiruse. Something of the same sort could happen again. It is possible that our currentaggrandizementoftheindividualwillobscurethepossibilityofbuildingabetterwayoflife.Theevolutionofourdemocraticculturewillthenhavetakenadisastrousturn.

FromTheListener,September30,1971.

418

TheConceptoftheReflexintheDescriptionofBehavior

This paper, written in the summer of 1930, still seems to me important for threereasons.Inthefirstplace,itwasanearlyexampleoftheoperationalanalysisoftermsdescribingbehavior.Ibelieve theclue to thedefinitionof reflexcame fromBertrandRussell.Somewhere,possiblyinaseriesofarticlesintheDialinthelate20’s,Russellpointed out that the concept of the reflex in physiology had the same status as theconceptofforceinphysics.AddthattoBridgman’streatmentofforceinTheLogicofModern Physics and you have the present point. I supported the argument with aMachian analysis of the history of the reflex to explain the traditional definition asunconscious, involuntary, and unlearned behavior. The operational analysis ofSherrington’s synapse and the more generalized statement in Chapter 12 of TheBehaviorofOrganisms,inwhichIsuggestedthatC.N.S.mightbetakentostandfortheConceptualNervousSystem,havebeeninterpretedasshowingananti-physiologicaloranti-neurological bias. I was, however,merely protesting the use of inferences frombehavior to explain behavior, while arguing for the solid status of behavioral factsapartfromimaginedphysiologicalcounterparts.Animportantpointisthatoperationaldefinitionsare suggestednotonly for reflex,but fordrive, emotion,andother termsappropriatetotheintactorganism.Secondly,thearticleinsistsupontheappropriatenessoftheconceptofthereflexin

describing behavior.One consequence is a statement of behavioral facts in a formwhichmost readily makes contact with physiological concepts andmethods. In thissensethepaperis,Ibelieve,apositivecontributiontophysiology.True,thetermreflexproved too rigid. It implieda complete specificationofpropertiesof behaviorwhichwas not in fact supported by the data either from whole or surgically subdividedorganisms (a pointwhichwasmade subsequently in the paper on page504),and itinsisted upon a demonstrable and presumably controllable stimulus as the principalindependentvariable(apointeventuallyrejectedinthepaperonpage535andinTheBehaviorofOrganisms).Buta similar sharpreference tobehavior inphysical termswasmaintainedby theconceptof the“operant” introduced toremedy these defects,andafurtherelaborationofthisconceptininterpreting,forexample,verbalbehaviorpresents the behavioral facts to be accounted for by the physiologist in the simplestpossibleway.Psychologicalfactsremainontheplaneofthephysicalandbiological.Thirdly,thepaperoffersaprogramandageneralformulationofascientificanalysis

of behavior. As to program, the last paragraph on page 501 describes withconsiderableaccuracya subsequent investigationof“conditioning,”“emotion,”and“drive,” some results of which are sketched in the paper on page 132. As toformulation,theobservablefactsunderlyingtheconceptofthereflexpermitustowritetheequation:

R=f(S)whereR=responseandS=stimulus.Changesinthisfunctionprovideanothersortof

419

datum.Forexample,ifwerepeatedlyelicitaresponse,thereflexundergoes“fatigue.”Wemayrewritetheequation:

R=f(S,A).InmypapertheletterA isnoncommittally referred toasa “thirdvariable,”apointwhichisrelevanttocurrentcontroversiesaboutinterveningvariables.ThevariableAismerelyanothervariabletobetakenintoaccount;itdoesnotinterveneeventhoughthetraditionalpracticeistoassignthechangeinthefunctionwhichitaccountsfortosome inner state or condition. Some experimental studies on rate of eating (offered,togetherwiththepresentpaper,asmydoctoralthesis)exemplifiedthisdistinction.Theconcept of “hunger” as an inner statemay be useful for certain purposes, but in astrictly operational definition we must confine ourselves to changes in “reflexstrength”asafunctionofdeprivation,satiation,andsimilaroperations.EdwardC.Tolmanmissed thispointwhen, fiveyears later,heattempted tosetup

hisownversionofan“operationalbehaviorism.”1Tolman’sequation:

B=f(SPHTA)is patterned after the equations above (with B for “behavior” in place of R for“response,”andS for “environmental stimulus condition” rather than “stimulus”).But what about P, H, T, and A? Tolman is sufficiently operational in saying thatgroups of these variables are all he finds in the way of mental processes (thusdisposing of the mentalistic nature of such processes), but he is still looking forsubstitutes and hence (and for no other reason) calls the additional variables“intervening.”Thepresent paperwas published in the Journal ofGeneral Psychology, (1931,5,

427-458)andisreprintedherebypermissionoftheeditor.

IntroductoryNote

Theextensionoftheconceptof thereflextothedescriptionofthebehaviorof intactorganismsisacommonpracticeinmoderntheorizing.Nevertheless,weowemostofourknowledgeofthereflextoinvestigatorswhohavedealtonlywith“preparations,”andwhohaveneverheld themselves tobeconcernedwithanythingbuta subsidiaryfunctionof thecentralnervoussystem.Doubtless, there isample justificationfor theuseofrelativelysimplesystemsinanearly investigation.But it is true,nevertheless,thattheconceptofthereflexhasnotemergedunmarkedbysuchacircumstanceofitsdevelopment. In its extension to the behavior of intact organisms, that is to say, thehistorical definition finds itself encumberedwithwhatnowappear tobe superfluousinterpretations.Thepresentpaper examines the conceptof the reflex andattempts to evaluate the

historicaldefinition.Itundertakeseventuallytoframeanalternativedefinition,whichisnotwhollyindespiteofthehistoricalusage.Thereaderwillrecognizeamethodofcriticism first formulated with respect to scientific concepts by ErnstMach [inTheScienceofMechanics] andperhapsbetter statedbyHenriPoincaré.To theworksof

420

these men and to Bridgman’s excellent application of the method [in The Logic ofModernPhysics] the reader is referredforanydiscussionof themethodquamethod.Probablythechiefadvantage,firstexploitedinthisrespectbyMach,liesintheuseofahistorical approach. But the reader should understand that in the present case noattempt is made to give an exhaustive account of the history of the reflex. Certainhistorical facts are considered for two reasons: to discover the nature of theobservationsuponwhichtheconcepthasbeenbased,andtoindicatethesourceoftheincidentalinterpretationswithwhichweareconcerned.

I

ItwasDescartes2whofirstproposedamechanismbywhichthecharacteristicsofthelivingorganismcouldplausiblybeproduced.Hecameveryneardescribingitsactionasthetruemodeofoperationoftheanimalbody,butthecriticismofhiscontemporary,NicolasStensen,probablyexpressedhisintentioncorrectly.“Descartes,”saidStensen,“wastoocleverinexposingtheerrorsofcurrenttreatisesonmantoundertakethetaskof expounding the true structureofman.Therefore inhis essayonManhedoesnotattemptsuchadelineation,butiscontenttodescribeamachinecapableofperformingallthefunctionsofwhichmaniscapable.”ThisinterpretationisborneoutbythetextoftheTraité,where,althoughpartsoftheanatomyareagainandagainpointedoutassuitable for the functions of themechanism, and themachine and body are, indeed,almost identified, the reader is, nevertheless, invitedonly to suppose the truthof thedetails.Descartes’interestlayprimarilyinfurtheringhisphilosophicalnotions,andtheinventionwhichisusuallytakenastheearliestexpressionofthereflexwaslittlemorethananinstrumentofpersuasion.Indesigningaconvincingmodelof the livingorganismDescartesfacedapeculiar

difficulty. Movement in itself was easily enough obtained, for there were manymechanisms available as sources of energy. There was, for example, the currentexplanation of muscular contraction upon a hydraulic analogy, which, in fact,Descartes adopted. But if the energy itself was conveniently accounted for, thedirection and order of its release were, on the contrary, critical. In meeting thisdifficulty Descartes introduced a novel device—the mechanism of the stimulus, bymeansofwhichexternalforcesreleasedthemovementsofthemachine.Thestimulusdistinguished the model of la běte machine from that of a mere activated doll. Itenabledthemodeltosimulatetheappropriatenessandtheapparentspontaneityofthemovementsofthelivingorganism.SofarasDescartes’purposewasconcerned,itwassuccessful in supplanting certain metaphysical concepts as causal agents leading tomovement.Heretofore,thesuppositionhadbeenthatananimalmovedbecauseoftheactionof,letussay,a“soul.”Descartesproposedthatthebodyberegardedasasystemof stored energy, and he pointed tominute, hitherto unobserved forces which actedupontheorganisminsuchawayastoserveasreleasingmechanisms.The principle of the stimulus was, of course, little more than a guess. With an

enthusiasm for the new physics, Descartes contended that the movements of anorganismwerefunctionsoftheforcesactinguponit,buthecouldinpracticepointtoonly the roughest demonstration of this relationship. Subsequent investigation of thereflex has revealed the extraordinary difficulty of identifying the stimulating forces

421

correlated with particular movements. The information available in Descartes’ timewas so scant, and the principle so far-reaching that one is tempted to regard thediscovery of the stimulus simply as another example of the insight with whichDescartesanticipatedlaterthought.Butthiswouldbetooverlooktheinfluenceofanunusualanalogy.Descartes sought amechanicalmodelof the livingorganism for the support of an

argument. For other reasons, namely, for the sake of the entertainment which theyafforded,suitablemodels(utilizingtheactionofastimulusasasourceofspontaneityand appropriateness) had already been constructed by the engineers of the royalfountainsinFrance.Descartesdescribestwoofthesefountainfiguresandtheactionoftheirreleasingmechanisms,whichareoperatedunwittinglybytheobservers.

Forinenteringtheynecessarilytreadoncertaintilesorplates,whicharesodisposedthatiftheyapproachabathingDiana,theycausehertohideintherose-bushes,andiftheytrytofollowher, theycauseaNeptune tocomeforward tomeet themthreatening themwithhistrident.Oriftheypassinanotherdirectiontheyoccasionthespringingforwardofamarinemonsterwhospoutswaterintotheirfacesorthingsofalikesortaccordingtothecapriceoftheengineerswhoconstructedthem.

Acontemporaryengineer,SalomondeCaus,3publishedanaccountoftheoperationofsimilarfigures,althoughhedidnotdescribethetwogroupsreferredtobyDescartes.Themechanical principles are few in number, but among them can be found all ofthoseusedbyDescartesinhis“reflexarc.”InDescartes’proposedmodeltheorganofsense is set inmotion “even ever so little” by the external object and pulls upon athread,whichinturn,actinglikeabellrope,opensavalveatacentralreservoir,lettingthecontainedfluidflowoutwardalongapipelineintothemuscles,whichitactivates.With a plate or lever substituted for the organ of sense and awaterwheel or similardevice for the muscles, the description applies as well to the fountain figures. Soslightly does Descartes depart from the details of the fountain mechanism that itspositionastheprototypeofhismodelseemsunquestionable.ItwastheaccidentofaconvenientanalogywhichledDescartestothediscoveryof

an important principle, and so great a mutation was it in the evolution of humanthought that itprovedlethal. Inspiteoffrequentassertions to thecontrary,Descartesseems to have exerted no influence upon the development of the reflex. Instead, thediscoveryofthestimuluswasmadeagain,withgreatdifficulty,astheculminationofacentury of experimentation, and another century and a half had elapsed before theprinciplehadagainbeencomparably extended to thebehaviorof the total organism.Thislackofhistoricalinfluencemaybevariouslyexplained.Descarteswas,asFosterhas said, a “retrograde” physiologist, who accepted the more convenient theory, asagainst the more accurate, for the sake of a broader consistency. His interest wasultimatelyphilosophical,eveninhisphysiologicalexplanations,andhedidnotattempttodiscoverthetrueactionofthenervoussystem.Descartesisimportanttoanunderstandingofthereflex,notbecauseofanorganic

connectionwithsubsequenthistory,butasasymbol.Thestimulusisanessentialpartof a mechanistic theory of behavior, whether the notion is arrived at throughobservation, as it was with Marshall Hall, for example, or argued from physicalnecessityormechanicalanalogy,asitwaswithDescartes.Furthermore,theanalysisofbehaviorwhichisaccomplishedinthemeredescriptivephrase,“withdrawingthefoot

422

from fire,” became a critical part of later method. But a further characteristic ofDescartes’ position must be noted: although he substituted the stimulus for ametaphysical concept in his description of the animal,Descartes could not eliminatemetaphysicalconceptsfromhisdescriptionofman.Hereheregardedthemechanicalprinciples as at work, but under the control of the soul, which might suspend thephysical necessitiesmuch as the engineermightmodify the activity of the fountainfigures.Descartesreservedafieldofactionfortheconceptofsoul,notbecausethephysical

factswere anymore lacking in the case ofman than elsewhere, but because of thepressureofcertainmetaphysicalnotions.Fragmentsofsimilarreservationsstillprevail.But the history of the reflex can almost be told by describing the progressiveencroachmentofthestimulusuponthem.ThelinewhichDescartesdrewbetweenthefieldsofactionofhisphysicalandmetaphysicalconceptswasatemporaloneonly.Amovementmightfollowatonetimetheactionofastimulusandatanothertheactionofsoul.ThelaterdistinctionwhichwasfirstdefinitelyestablishedbyMarshallHallsetinpartananatomicalboundary.Butbothlinesweredrawnforthesamepurpose,namely,to resolve, by compromise, the conflict between an observed necessity andpreconceptionsoffreedom in thebehavioroforganisms. Inoneformoranother, thiscompromiseaccountsalmostwholly for theaspectsof thehistoricaldefinitionof thereflexwhichweareattemptingtoreconsider.

II

The concept of the reflex arose again from investigations which had already begunduring Descartes’ lifetime. They were concerned with animal movement andrepresentedasuddenturninthehistoryoftheconceptsdealingwiththatphenomenon.Itisageneralizationsufficientlyaccurateforourpurposestosaythatthemovementsofan organism had generally been taken as coexistent with its life and as necessarilycorrelated with the action of some such entity as soul. The necessary relationshipbetweentheactionofsoulandthecontractionofamuscle,forexample,wasexplicit.As a consequence, itwas disturbing to find, experimentally, that amuscle could bemadetocontractafterithadbeenseveredfromalivingorganismorevenafterdeath.This,however,wasthecontentionofthenewphysiology.Thedemonstrationthatthevolumeofamuscledoesnotappreciablyincreaseduring

contractionwasmadebyFrancisGlisson4inthemiddleoftheseventeenthcentury.Hewas, as Fulton has shown, probably anticipated here by the Dutch naturalist,Swammerdam, whose influence was less immediately felt. Swammerdam’sexperimentsonexcisednerveandmuscleweremoreclear-cutthanthoseofGlissononthe intact limb,buteitherprocedurewasconvincingandwassomethingmore thanamere disproof of the “animal spirits” hypothesis of muscular contraction. Theexperiment pointed to the existence of a “property of contractility” resident in themuscular tissue and independentof any remote sourceof energy.Contemporary andsubsequent experimentation was of the same import. Glisson himself experimentedupon intestinal and skeletal muscle after death, when metaphysical concepts hadsupposedly ceased to act. Swammerdam’s experiments indicated a characteristicactivity in excised nerve and muscle, and in 1700 Giorgio Baglivi, the Italian

423

physician,reportingthecontractionsof isolatedmuscular tissue,emphasizedthat thiswas “without the soul’s having any share in it, or even being sensible of it.” Theexperiments published by von Haller in 1739 and 1742 permitted him to make thefollowingclaims:

Bymy experiments I separated this irritable nature on the one hand from amere deadforce,andontheotherhandfromthenervousforceandfromthepowerofthesoul.Ishewedthatthemovementoftheheartandtheirritablenatureoftheintestinesdependedonitalone.I confined it entirely to themuscular fibre…. I also shewed that that forcewas somethingperpetually living, and that it often broke out intomovement though no external stimulussuchascouldberecognizedbyuswasacting.Byastimulus,however,itcouldatanytimebecalledbackfromrestintoaction.InamovementproducedthroughitIdistinguishedbetweenthe stimulus which might be very slight, and the movement called forth by the stimuluswhichmightbeverypowerful.

The doctrine of irritability was the theoretical accompaniment of thisexperimentation.Asapropertyassignedtolivingtissue,irritabilitywasfromthefirstclearly defined in terms of the experimental operations which revealed it. In hisfundamentalexperiment,Glissonnotedthatthegallbladderandthebiliaryductbringaboutagreatersecretionwhentheyareirritated.“Heargues,”asFosterhasnoted,“thattheycannotbeirritatedunlesstheypossessthepowerofbeingirritated.Thispowerofbeing irritated he proposes to denote by the term irritability.”Although the conceptwasnot immediately freedofnon-physicalcounterparts, itwasessentiallyaphysicalhypothesis,whichultimatelyledtothescienceofthegeneralphysiologyofnerveandmuscle.Movement,farfrombeingtheobjectivemanifestationoftheactivityofsoul,hadbecomeanorganicprocesssubjecttoexperimentalinvestigation.AsVerwornhassaid,stimulationandirritabilitycannotbeseparated.Irritability,by

itsdefinition, implies the actionof a stimulus.Thedoctrineof irritability,moreover,assigns an autonomy of function to the parts of an organism. These were theprerequisites for a formulation of the concept of the reflex. The first expression inharmonywiththeexperimentalmaterialwasmadebyRobertWhytt.5Thegenesisoftheideaisapparentinasinglesentencefromthatwork.Theobservationismadethatmuscles will contract not only upon direct stimulation but “whenever a stimulus isappliedtothecoatsormembranescoveringthem,tothenerveswhicharesenttothem,or to someneighboringor evendistant part.6 Step by step, the point of stimulationrecedes from the locus of the phenomenon with which it is identified. When thestimulus has been spatially distinguished from the response, the inference of aconductingmediumisnecessary,andafurtherexperimentbyWyatt,suggestedtohimby Stephen Hales, showed that the spinal marrow was a necessary part of thisconductingpath,whichcouldnotfunctionifthemarrowweredestroyed.Weshallneedto return later to discuss the significance of these experiments and certain points ofmethodwhichtheyexemplify.Whytt,itistrue,regardedtheconductingforceasa“sentientprinciple,”which,ifit

wasnon-rational,wasalsonon-physical.HethusfollowedtheexampleofGlissoninhis compromisewitholder concepts.Nevertheless, theobservations themselveswereindependentofWhytt’sinterpretation.Hispsychicalqualifications,moreover,wereinthis sense useful: that they permitted him to generalize his principle to various

424

functions of the intact organism, free of the resistance which would have beenencounterediftheprinciplehadbeenwhollyphysical.Theapplicationtothevegetativefunctionswaseasilymade,forithadalreadybeenstronglyforeshadowed.Heextendedhisdoctrinealsotothefieldofactionofsomeofthedepartmentsofthesoul.

III

WhyttlefttheconceptofreflexactionverymuchasGlissonhadleftirritability:intheposition partly of a description of observed fact and partly of a superfluousinterpretation.WhatvonHallerhaddoneinestablishingirritabilityasaphysiologicaldatumindependentofanyaspectofsoul,MarshallHall7nowrepeatedonbehalfofthereflex.Not only, he suggested, couldmuscular tissue contract solely by virtue of itspropertyofcontractility,butagivenmuscleinsitu couldbebrought intoactionbyatrainofnervouseventswhichwereinthemselvesactingonlybyvirtueofanintrinsicproperty. “Thisprinciple,”he said, “is that termedvisnervosabyHaller,motorischeKraft or vis motoria by Professor Müller, and excitabilité by M. Flourens.”8 Todescribe the series of eventswhich a single instance of this activity comprisedHallaccepted the word reflex, and he spoke of the principle in general as “the reflexfunction.”Thehypothesis that thephenomenaof the reflexarcareonlyaspectsof irritability

seemstohavebeenoriginalwithMarshallHall.Ofthecontributionswhichheclaimedtohavemadeitresistedmostsuccessfullythechargeofplagiarismleveledagainsthimby his contemporaries and was most often cited in his defense. The hypothesisinvolved,notadenialof theoperationofnon-physicalconcepts,butanexclusionoftheir operation from the reflex field. In Hall’s own estimation, as Sherrington9haspointedout,“hischiefadvancelayinthedoctrineofseparatenessinthecentralnervoussystemof thegreat sub-system forunconscious reflex action, and anothergreat sub-system for sensation and volition.” This estimate of himself has been more or lessconfirmed historically. He is, it seems fair to say, the acknowledged author of thealmostimmutabledistinctionbetweenvoluntaryandreflexactionandoftheresultingnegativedefinitionofthereflexasaformofmovementunconscious,involuntary, andunlearned.Theemphasiswhichthisbriefaccounthasthusfarplacedupontherelationof thereflextonon-physicalconceptsmaynowappearmorereasonable.Hall’sbasichypothesis was simply a restatement of the relationship and must be understoodaccordingly.Hall distinguished between fourmodes ofmuscular action.Volition, he said, acts

throughthecerebrum,isspontaneous,andaffectsthemusclesinadirectlinethroughthe spinal marrow and motor nerves. Respiration acts in the same way, is alsospontaneous(sic),butitsseatisinthemedulla.Involuntaryactionistheresponseofamuscle to direct stimulation (the phenomenon of irritability). The fourth mode (thereflexfunction)involvesthespinalmarrowanddiffersfromvolitionandrespirationinthatitisneitherspontaneousnordirectinitscourse.

It is, on the contrary, excited by the application of appropriate stimuli, which are not,however, applied immediately to the muscular or nervo-muscular fibre, but to certainmembranousparts,whencetheimpressioniscarriedtothemedulla(spinalis),reflected, andreconductedtothepartimpressed,orconductedtoapartremotefromit,inwhichmuscular

425

contractioniseffected.

Wemayneglectthesecondclass,whichseemstohavebeenincludedbecauseofthecurrentknowledgeoftherespiratorycenterbutwasexclusiveofneithervoluntarynorreflex action, and the third (unfortunately called involuntary), which was only astatementofirritability.Forthedistinctionbetweenthetworemainingclasses(volitionandreflexaction),wemayconsiderHall’savailableevidence.Hall defined volition as a formofmovementwhichwas (a) spontaneous, and (b)

dependent upon the integrity of the cerebrum. The second characteristic was not initself sufficiently distinguishing. Hall was familiar with the segmental nature of thespinalfunctions,anditwasafairsuppositionthatthedifferentialactivityfoundinsouniformastructuremightbeaccentuatedinthehighersegments.Ifitwasamatterofobservationthatcharacteristicmovementsof theanimaldisappeareduponablationofthe cerebrum, thiswas also true for any given part of the cord. Spontaneity, on theotherhand,wasthemorecriticalfactorinthedelimitationofvolition.Butspontaneity,as Hall used it, described in effect only thosemovements for which no appropriatestimuli could be observed, and theword seems to have had no othermeaning. Thedistinctionbetweenreflexandvoluntaryactionrested,then,uponthepossibilityortheimpossibilityoftheexperimentaldemonstrationofstimulatingforces.TothesupportofhisdistinctionHallmusteredavarietyofotherfacts.Henotedthat

drugs(suchasopiumorstrychnine)discriminateintheiractionbetweenvoluntaryandreflexactivity,thatthebrainsleeps,whilethespinalmarrowneversleeps,andsoon.Hewas,ofcourse,describinganobservablephenomenon,whichwerecognizeasthedifferentialfunctioningofseparatereflexsystems.10Theobservationsdonot,however,indicateanessentialdifferencebetweenbrainandcord,especiallysincethebrain-corddistinction is seldom strictly respected physiologically. Hall’s appeal to physiology,likethat toanatomy, lackedcogency.Hisdistinctionrestedprimarilyuponthesingleitemofspontaneity.In defining volition as the hypothetical antecedent of movement for which no

corresponding stimulus could be observed, Hall left the concept open to extensivemodification,foritwasimplicitinthenatureofthereflexthatitshould,inthecourseof its growth,disfranchisevolition.So far as it concernsbehavior, thehistoryof thereflexhasbeen, infact,essentially theaccountof thediscoveryofstimuliandof theconcurrent passage of the corresponding behavior from the field of volition into thefieldofreflexaction.Furthermore,inopposingvolitionandreflexactionasmutuallyexclusive terms,Hall identified the reflexwith scientificnecessity, andvolitionwithunpredictability.Thiswasthepatternforfuturecontroversy,ofwhichwemaynotetwoinstances.In1853,Pflüger,11asiswellknown,questionedthereflexnatureofthemovements

ofthespinalfrogonthebasisofunpredictability.Inseparateinstancesoftheflexionreflex,hepointedout,themovementofthelegvarieswidely,althoughthestimulusisheld constant. On the basis of the observed variability, Pflüger postulated a spinalmind,hisfamousRückenmarkseele.Note that theexperimental justification formind(asforHall’svolition)wastheabsenceofdemonstrablenecessity,thatthefunctionofthenon-physical conceptwas, asheretofore, to account forvariability.RefutationofPflüger’scriticismneededonlyademonstrationthattheobservedvariabilitywasitselfafunctionofcollateralstimulation.Thenecessaryobservationswerefirstsuppliedby

426

vonUexküll,andtheprinciple(Shaltung)hasbeenelaboratedbyMagnus.Briefly,theyhaveshownthatagivenresponsemaybemodifiedthroughproprioceptivestimulationarisingfromthepostureoftheanimal.TheeffectswhichPflügerobservedinthespinalfrogareconsequentlysubjecttoadequateprediction,and,inthisparticularinstanceatleast,thevariabilityhasdisappeared.WithitdisappearedalsoitscorrespondingSeele.ShortlybeforethebeginningofthecenturyPavlovwasengagedintheinvestigation

of the activity of the digestive glands. For much of this activity it was possible toidentify the necessary antecedent events (themechanical or chemical changes actingdirectlyorreflexlyupontheglands).Thegreaterpartofthenormalsecretion,however,wasunfortunatelynotunderthecontroloftheexperimenter.Thereadershouldnotbesurprised that thiswascalled“psychic”secretion.Pavlovundertook the investigationof this activity. His findings are too well known to call for more than the briefestcommenthere.Essentially,itwasthediscoveryoftheoperationof“substitute”stimuli.Thenatureoftheprocessofsubstitution(conditioning)andtheuseoftheprincipleasamethodarenotimportantatthispoint.Wemayemphasize,however,theaspectsofthediscoverywhichexemplifytheusualcourseofreflexinvestigation.Givenaparticularpartofthebehaviorofanorganismhithertoregardedasunpredictable(andprobably,as a consequence, assigned to non-physical factors), the investigator seeks out theantecedentchangeswithwhichtheactivityiscorrelatedandestablishestheconditionsofthecorrelation.Hethusestablishes,aswesay,thereflexnatureofthebehavior.Intraditional practice, upon the demonstration of such a correlation, non-physicalconceptsdealingwiththesamesubjectmatterarediscarded.

IV

ThesubsequentdevelopmentofHall’s formulationcouldby itsverynatureapproachonly one end, namely, the hypothesis that the total behavior of the intact organismmightbedescribed in termsof the reflex.Thisextensionwaspossibleonlyupon thedemonstrationofnecessarycorrelations ina largebodyof residualbehavior,mostofwhich was mediated by the distance receptors of the head segments. The work ofPavlovmaythereforebetakenashistoricallyfundamental.Hisevidencewasdecisive,if necessarily incomplete. It led to two achievements. The principle of conditioningsuppliedtheextendedrangeofstimulationneededtoaccountforthecomplexbehaviorof the total organism, and the demonstration of the reflex activity of the cortex laidsiege to the last stronghold of the old anatomical distinctions. The extension of theconcepthasbeen further facilitatedby theworkofMagnusupon reflexes concernedwiththemaintenanceofposture.Magnus,likePavlov,broadenedthefieldofoperationof the stimulus by discovering a large number of specific stimulus-responsecorrelations, and, again like Pavlov, he attacked the anatomical distinctions bydemonstrating the reflex nature of the activity of the higher, although chieflysubcortical,centers.Thereflexasaconceptinthedescriptionofbehaviorhasreceiveditsmostextendedsystematicsupportfrombehaviorism.Theadequacyorinadequacyofthereflexinthedescriptionoftotalbehaviorseems

tobebeyondimmediateexperimentaldemonstrationand, inanyevent, isbeyondthescopeofthispaper.Someofitsimplicationsalsoneednotgreatlyconcernus,as,forexample, the fate of the non-physical concepts which are deprived of their field ofoperation.Wemayregardtheultimatevalidityoftheconceptofvolition(as,indeed,of

427

that of the reflex) as beyond any immediate estimation.We are concernedwith thereflex as a working concept. What is its nature and how shall it be defined? Inparticular,wehavesetourselvestoresolvecertaindifficultiesofdefinitionimposedbytheextension to totalbehavior,wherevolition (or thepracticewhich it represents) isimportantforitseffects.Butperhapswehavereviewedenoughofitshistoryandmayturndirectlytoastatementoftheargument.Inthehistoryofthereflexonepositivecharacteristichasalwaysbeengivenbythe

facts—theobservedcorrelationoftheactivityofaneffector(i.e.,aresponse)withtheobservedforcesaffectingareceptor(i.e.,astimulus).Thenegativecharacteristics,onthe other hand,which describe the reflex as involuntary, unlearned, unconscious, orrestricted to special neural paths, have proceeded from unscientific presuppositionsconcerning the behavior of organisms.WhenMarshall Hall decapitated his famousnewt,hepointedquitecorrectlytothereflexactivityofthepartsoftheheadlessbody,totheobservedfactthatmovementfollowed,inevitably,theadministrationofspecificstimuli.Buthisassumptionthathehadimprisonedintheheadofthenewtthesourceofanotherkindofmovementwasirrelevantandunsupported.Thefactbeforehimwasademonstrablenecessity in themovementof theheadlessbody;his failure toobservesimilarnecessitiesinthemovementoftheintactorganismwastheaccidentofhistimeandofhiscapabilities.Tentatively,then,wemaydefineareflexasanobservedcorrelationofstimulusand

response. When we say, for example, that Robert Whytt discovered the pupillaryreflex,12we do notmean that he discovered either the contraction of the iris or theimpingement of light upon the retina, but rather that he first stated the necessaryrelationshipbetween these twoevents.So farasbehavior isconcerned, thepupillaryreflexisnothingmorethanthisrelationship.Oncegivenaspecificstimulus-responsecorrelation,wemay,ofcourse,investigatethephysiologicalfactsofitsmediation.Theinformation there revealed will supplement our definition, but it will not affect thestatusofthereflexasacorrelation.Thesearematters,however,whichwillbearamoredetailedtreatment,fortheypresentmanyproblems.

V

The notions of both stimulus and response were, as we have seen, essential to theprincipleof irritability,so that thecorrelationwhichweareemphasizingwasalreadypresent(initsmosteasilyobservedform)intheolderconcept.Thereflexemergedasaseparate principle when a correlated stimulus and response could be spatiallydistinguished, and we have already commented upon Whytt’s insistence upon thepossibility of a spatial differentiation. The observation of a correlation between twospatiallydiscreteactivitiesledatoncetotheinferenceofaseriesofinterveningevents,to the inference, that is, of conduction.Subsequently, the investigation of the eventsinterveningbetweenastimulusanditscorrelatedresponsebecametheparticularfieldofreflexphysiology.Weshallneedtoreviewcertaincharacteristicsofitsmethod,andwemayturnfirst to theprocedurebywhichtheanatomicalstructuresunderlyingthemediationofareflexareidentified.Even for so early an investigator as Robert Whytt convenient material from the

physiology of the nervous system was available in interpreting reflex phenomena.

428

Rough descriptions of the activity of end-organ and effector and of the conductingactionofnervewereathand.Theinvestigatorsofthereflexappropriatedthiscurrentknowledgebutbegan immediately to refine the references toanatomy.Whytt, aswehavealreadyseen,firstdemonstratedthatthenecessaryrelationshipbetweenstimulusandresponse(or,asheexpressedit,the“sympathybetweendifferentmusclesorotherpartsofthebody”)waslackingafterdestructionofthecord;“fromwhenceitseemstofollow,”headded,“thatthenerves….havenocommunicationbutattheirterminationinthebrainorspinalmarrow.”Hisconclusion,ifitwasnotstrictlylogical,wasmadeextremelyprobablebyBell’s subsequent differentiationbetween the functionsof theanterior and posterior spinal roots.AlthoughBell did not expressly subscribe to thereflex doctrine, his experiments are more to the point in its support than in hisgeneralizationto“sensationandthepowerofmotion.”The“diastalticarc”ofMarshallHallappliedBell’sdiscoverymoreexplicitlytotheprincipleofthereflex.The diastaltic arc (eventually spoken of as the reflex arc) embraced an “esodic”

(afferent)nerve,aspinalcenter,andan“exodic”(efferent)nerve.Itwasananatomicalterm.Theexperimentalevidenceforitsclosecorrespondencetothereflexwasofthesortwehavenoted:theimpairmentofreflexfunctionafteranatomicalinjury(Whytt),the fractional functioning of surgically isolated parts of the arc (Bell), and so on.Subsequently (especially in the neuron theory), the argument was extended tomicroscopic levels. By similar procedures the gross location of the central part of agivenarcisdetermined.Thepracticeisessentialtothedoctrineofthesegmentalactionofthecordandwasutilizedearlyinthehistoryofnervephysiology(independentlyofreflextheory)inthelocalizationofhighercenters,forexample,therespiratorycenter.Typical examples in reflex investigation may be found in the work of Magnus onvariousmid-brain preparations.The basic assumption is that,when the ablation of aparticularpartofthenervoussystemimpairsorabolishesareflexfunction,theablatedstructureisessentialtothereflexandincludespartofthearc.Thatinference,ofcourse,isnotunavoidable.Moreover,intheinterpretationofsuchexperimentstheprobabilityof operative artifactmust unfortunately be regarded as proportional to the degree ofspecificity desired. For the broader inferences, however, such as that of theparticipationofthespinalcord,theprobabilityofseriousartifactisnegligible.The notion of the reflex arc as the anatomical counterpart of a reflex has been

generallyaccepted.Anend-organ(oranervetrunkactinginthatcapacity),anafferentnerve,aninterconnectionbetweennervesinthecordorbrain,anefferentnerve,andaneffectorareusuallyregardedasessentialtothemediationofareflex.Weshallhavenooccasion togobeyond these rathergeneralassumptions.Themethodwhichwehavereferred to is not peculiar to the reflex, and we shall not need to estimate it moreclosely.Wemaynote,however,thatthedescriptionofareflexinfunctionalterms(asacorrelationofstimulusandresponse)isalwayspriortothedescriptionofitsarc.Inanyavailable procedure the anatomical inference must always be drawn from anexperimentinwhichtheintegrityofafunctioniscritical.In its simplest form, the concept of the reflex arc satisfies the need for continuity

between stimulus and response, but the arc must serve as a locus not only for acommunicationbetweenend-organandeffectorbutformodificationsintheformofthecommunication. The statement of any reflex (for example: “the flexion of a limbfollowing electrical stimulation of the skin of the foot”) implies the possibility of aquantitative description of both stimulus and response.The statement thus expressestheobservedcorrelationoftwoevents,butbydescribingtheseeventsitdescribesalso

429

thespecialconditionsof thecorrelation.Agivenstimulusanditsresponsediffer, forexample,intimeofinception,induration,andintheformandamountofenergy;andthesemodificationsandconversionsmustbeaccountedforbytheinterveningevents.Byproceduresofthesortwehavealreadydescribed,thecharacteristicsofareflex

havebeenassignedtoparticularpartsofthearc.Thegrossconversionsofenergyhave,ofcourse,beenreferredtoend-organandeffector,partoftheelapsedtimetoafferentandefferentnerve,andsoon.Byaprocessoflogicalandsurgicalisolation,however,acertain group of the conditions of a reflex correlation have been shown to beindependent of the activity of end-organ, effector, and nerve-trunk. These are thespecial characteristics of reflex conduction. They have been classically described bySherringtoninTheIntegrativeActionoftheNervousSystem.InSherrington’slisteachcharacteristic is expressed as a difference between nerve-trunk and reflex-arcconduction, which means simply that certain of the inferences noted above havealready been made. In reproducing the list here, we shall reword it in order toemphasizethenatureofeachitemasanobservedconditionofacorrelationofstimulusandresponse.(1)Thereisalatentperiodbetweenapplicationofthestimulusandappearanceofthe

endeffect;(2)thedurationoftheresponseisgreaterthanthedurationofthestimulus;(3) if stimuli areapplied rhythmically (betweencertain limiting rates), the rhythmofthe response does not closely correspond with the rhythm of the stimulus; (4) theintensityoftheresponsedoesnotvaryrectilinearlywiththeintensityofthestimulus;(5) a single brief stimulus is often not effective, but succeeding stimuli followingcloselyuponitare;(6)afferentandefferentpathscannotfunctioninterchangeably;(7)repetitionofastimulus(withcertaintimespecifications)evokesprogressivelyweakerresponses;(8)thestrengthofstimulusjustsufficingtoelicitaresponseisvariable;(9)(a) a second stimulus is ineffective for a short interval after a first and subnormallyeffective for a succeeding short interval, (b) two stimuli at separate points ofstimulationmayfacilitateeachother,(c)astimulusmayacttoproducetheabsenceofaresponse, (d) injury to the nervous system may temporarily destroy or weaken theeffectivenessofastimuluswhichissubsequentlyfoundtobenormallyeffective;(10)theeffectivenessofastimulusdependsupontheintegrityofthebloodsupply;(11)theeffectivenessofastimulusispartiallyorwhollyabolishedbyanaesthetics.Althoughall thecharacteristicsexcept(6)andpartsof(9)arerepresentedtosome

extent innerveconduction, thedegree towhichtheyarepresent inreflexconductioncannotbeexplainedbyreferencetonerve-trunkalone.Ithasbeenassumed,therefore,thatthecharacteristicsrepresentthefunctioningofaspecialstructure,whichhasbeencalled thesynapse, andwhich has been hypothetically located at the interconnectionbetweenneurons.Sincetheseinterconnectionsaregroupedtogetherinthegraymatterofthecordandbrain, theexperimentalpracticeassociatedwiththeotherpartsof thearcmaybeusedintestingthehypothesis.Itispossible,forexample,toshowthatthesynaptic characteristics are present in conduction through the cordwhen an afferentnerve is stimulated close to the cord and the response taken electrically from theefferentroot.Again,thecharacteristicscanbeshowntovarywiththetemperatureofthecord,buttheyarealmostwhollyindependentofthetemperatureofeitherafferentorefferentnerve.Examplescouldbemultipliedindefinitely.Moreover,asSherringtonhasshown,thelocationofthesynapseattheinterconnectionbetweenneuronsreceivesconsiderablesupport fromthehistologyof thenervoussystem, fromstudies innervedegeneration,fromthephysicalchemistryofsurfacesofseparation,andfromvarious

430

othersources.Reflexphysiologyseeksaphysico-chemicaldescriptionoftheeventspeculiartothe

mediation of a stimulus-response correlation. It regards the synapse, therefore, as aphysico-chemicalsystem.Theoriesof thedetailsof that systemhavebeenofvarioussorts.KeithLucas13soughtanexplanationofsynapticphenomenaintermscommontonerve-trunkconduction.Inhistheorythesynapsewasregardedasaregionofimpairedconduction, and its description in physico-chemical terms waited only upon thedescriptionoftheconductionofthenervousimpulse.Sherringtonhasrecentlyregardedthe synapseas the locus for thedepositingofexcitatoryand inhibitory substancesorstates.14 For Lapicque the synapse is the boundary between neural structures ofindependentchronaxies.Weshallhavenoneedofevaluatingtheoriesofthissort,norshallwefinditnecessary,inthelightofourbriefexaminationofmethod,tojustifyordiscredit thehypothesisof thesynapseitself.Weareinterested,not inthevalidityofthatconcept,butinitsnature.Hereweareledtooneconclusion.Our present information concerning the synapse is derivedwholly from observed

instancesofreflexconduction.Thereisnothinginourdescriptionofthesynapsewhichhasnotalreadyservedtodescribeexperimentaldata,butwetranslateourdescriptionsofdataintothelawsofthesynapseforconvenienceofexpression.Thesynapse,thatisto say, described in terms of its characteristics, is a construct. It is the conceptualexpression for the conditions of correlation of a stimulus and response, where theincidentalconditionsimposedbyaparticularstimulusandaparticularresponsehavebeeneliminated.Thereisnothinginthephysiologyofthereflexwhichcallsinquestionthenatureof

the reflexas a correlation,because there is nothing tobe found therewhichhas anysignificancebeyondadescriptionoftheconditionsofacorrelation.Itmaybeobjectedthat,shouldreflexphysiologysucceedindescribingthesynapseasaphysico-chemicalsystem,thesynapsewouldbenolongerconceptual.Actually,thedescriptionwouldbetranslated into concepts of another order, which would possess the tremendousadvantageofbeingcommontoallthephysicalsciences.Butwearehereveryclosetocertain fundamental questions of scientific method which we shall not attempt toanswer.The physiological study of the reflex supplements and restricts our definition. It

begins by identifying and describing certain of the events which intervene typicallybetween stimulus and response, and it then arbitrarily restricts the use of the wordreflex to correlations which employ that kind of event. Physiologically, the wordimplies the participation of at least two neurons with a synaptic junction. The bestpractical criterion is irreversibility of conduction, which is not in any degree acharacteristic of nerve-trunk conduction and is therefore the clearest evidence of theoperationofasynapse.Thisrestrictionin theuseof thewordreflex excludes (1) themovement of the organism solely under the influence of mechanical forces (forexample,themovementofthepawofadogwhenitis“shaken”);(2)theactivityofaneffector in response to direct stimulation, a distinction which is more often ofimportanceindealingwiththeinternaleconomyoftheorganism;(3)thoseresponsesmediatedbyothertypesofnervoussystemthanthesynaptic;and(4)thosecorrelationsbetween discrete activities which are mediated by non-nervous mechanisms, as, forexample,byhormones.Theadvantageofthesequalifications(evenofthethird,whichissometimesfeltto

431

beanotabledifficulty)isthattheyinsureauniformityofmaterialintheinvestigationof the reflex. It would be impossible to state with any degree of specificity a lawdescribing the course of reflex fatigue if it were to apply to the exhaustion of suchdiversemechanismsasaneuralstructureandaconcentrationofhormonicsubstance.Therestricteddefinitionlimitstheapplicationoftheprincipleofthereflex,butwithinboundarieswhichmaybejustifieduponindependentgrounds.Wehave tried toemphasizeanessential continuitybetween reflexphysiologyand

the special science of the description of behavior. We must not, however, fail torecognizeawell-groundeddistinctionbetweenthetwofields,whichisbasedprimarilyuponadifference in immediatepurpose.Theoneseeksadescriptionof the reflex intermsofphysico-chemicalevents, theotheradescriptionofbehavior in termsof thereflex. It is assumed that theword reflex refers to the same thing in both instances.Historically,however, the investigationof thephysiologyof the reflexproceeded,aswehaveseen,almost independentlyof thedescriptionofbehaviorandwasadvancedalmosttoitspresentstatusatatimewhenthepossibilityofaquantitativedescriptionofbehaviorwasconsideredtooremoteforscientificconsideration.Theconceptsofreflexphysiology,consequently,cannotbetransposedtothedescriptionofbehaviorwithoutmodification,whichwillinmostcasesbefound,however,toproceednaturallyfromananalysisofthephysiologicalmethod,as,forexample,inthediscussionofthesynapsegivenabove.Moreover,manyassumptionswhichareunimportantandthereforetacitinreflex physiology are critical in the description of behavior and must be given anexplicit and independent statement. This is true of the procedure bywhich a singlereflexisisolated,asweshallseelater.Again,anumberofproblemswhichariseinthedescription of behavior have no parallel at all in reflex physiology. Some of thesematterswehavealreadytouchedupon;otherswillenteratonetimeoranotherintothefollowingdiscussion.

VI

We may summarize this much of the argument in the following way. A reflex isdefinedasanobservedcorrelationoftwoevents,astimulusandaresponse.Asurveyofthehistorydisclosesnoothercharacteristicuponwhichadefinitioncanlegitimatelybebased.Thephysiologicalinvestigationdoesnotquestionthecorrelativenatureofthereflex,foritsdataanditsconceptsdealessentiallywiththeconditionsofacorrelation;but heterogeneous instances of correlations which would be embraced by thedefinition,readliterally,areexcludedbythephysiologicalrefinementsofusage.Itnowremainsforustodealmorespecificallywiththereflexinthedescriptionofbehavior.Whatisthedescriptionofbehavior,andhowdoesthereflex,asacorrelation,enterintoit?Here(thereadermayagainbewarned)weshallbeconcernednotsomuchwiththevalidityortheadequacyoftheconceptaswithitsnatureandthemethodpeculiartoit.Lackingsomearbitrarydistinction,thetermbehaviormustincludethetotalactivity

of theorganism—the functioningof all its parts.Obviously, its proper application ismuchlessgeneral,butitisdifficulttoreachanycleardistinction.Thedefinitionofthesubjectmatter of any science, however, is determined largely by the interest of thescientist, and this will be our safest rule here. We are interested primarily in themovementofanorganisminsomeframeofreference.Weareinterestedinanyinternalchangewhichhasanobservableandsignificanteffectuponthismovement.Inspecial

432

caseswearedirectly interested inglandularactivity,but thiswillusuallyconcernusonly secondarily in its effect uponmovement. The unity and internal consistency ofthissubjectmatterishistorical:weareinterested,thatistosay,inwhattheorganismdoes.Butthedescriptionofbehavior,ifit istobeeitherscientificorsatisfying,mustgo

further.Asascientificdiscipline,itmustdescribetheeventnotonlyforitselfbutinitsrelation to other events; and, in point of satisfaction, it must explain. These areessentiallyidenticalactivities.Inthebriefsurveyatthebeginningofthispaperitwasoccasionally necessary to regard the stimulus as a newly discovered cause ofmovementforwhichvariousconceptualcauseshadpreviouslybeendesigned.Inthiswaywerepresentedarealaspectofthehistoryofthereflex.Butwemaynowtakethatmore humble view of explanation and causation which seems to have been firstsuggestedbyMachandisnowacommoncharacteristicofscientificthought,wherein,inaword,explanationisreducedtodescriptionandthenotionoffunctionsubstitutedforthatofcausation.Thefulldescriptionofaneventistakentoincludeadescriptionofitsfunctionalrelationshipwithantecedentevents.Inthedescriptionofbehaviorweareinterested in the relationshipswithina regressiveseriesofeventsextendingfromthebehavioritselftothoseenergychangesattheperipherywhichwedesignateasstimuli.We stop here in the regression only because further steps are beyond the field ofbehavior. The two end events, the behavior and the stimulus, have, moreover, aparticularimportance,becausetheyalonearedirectlyobservableinanintactorganism,and because they limit the series.With the relationship of these two end terms thedescriptionofbehaviorischieflyconcerned.The reflex is important in thedescriptionofbehaviorbecause it isbydefinition a

statementof thenecessity of this relationship. The demonstration of the necessity isultimatelyamatterofobservation:agivenresponseisobservedinvariablytofollowagiven stimulus, or exceptions to this rule may be independently described. In itsextension to total behavior the principle generalizes the statement of the necessityobserved in a particular reflex, the form of the expression remaining essentially thesame.That is tosay, thehypothesis that“thebehaviorofanorganism isanexact, ifinvolved, functionof the forcesactingupon theorganism”states thecorrelationofastimulus and a response, both ofwhich remainwholly undifferentiated. It is, in thissense,thebroadestpossiblestatementofareflex,butitisnotanobservedcorrelationandisthereforeahypothesisonly.It is,nevertheless, solely the faultofourmethod thatwecannotdealdirectlywith

thissinglecorrelationbetweenbehaviorasawholeandalltheforcesactingupontheorganism stated in the hypothesis. Quantitative statements of both stimulus andresponseandastatisticaldemonstrationofthecorrelationaretheoreticallypossiblebutwould be wholly unmanageable. We are led, for lack of a better approach, toinvestigate thecorrelationofpartsof thestimuluswithpartsof theresponse.For thesakeofagreaterfacility(andinthiscasetheverypossibilty)ofdescription,weturntoanalysis.Originally, the use of analysis was quite accidental and unrecognized, but it has,

nevertheless,alwaysbeennecessary.Theearlyobservationswerepossibleonlyafterithad been achieved in some form or other. This is not difficult to understand if werememberthatthecorrelationwhichwecallareflexrestsultimatelyuponobservation.In an intact newt, to return to Hall’s experiment, it would have been very nearlyimpossible to observe a correlation between the movement of the tail and the

433

applicationofaprobingneedle,becausethemovementofthetailwasalsocorrelatedwithotherstimuliandtheactionof theprobingneedlewithothermovements.In theisolatedtail,however,onekindofmovementfollowedagivenstimulusandwasabsentintheabsenceofthestimulus.Thecorrelationwasobviousandthereforeobserved.MarshallHall and his few predecessors divided the behavior of an organism into

partsbytheexpedientmethodofdividingtheorganism.Thisbecame,ingeneral, themethodofreflexphysiology,although,forobviousreasons,thedivisionofthenervoussystem supplanted the division of the whole organism. The best-known group ofreflexes to be studied in surgical isolation are those surviving in the body of theorganismaftersectionofthecordjustbelowthebulb.Thisisthe“spinal”preparation,whichhasbeenthebasisforthegreaterpartofphysiologicalinvestigation,notablythatofSherrington.Othercommonreflexsystemsarethedecerebrate,inwhichthemedullaandthecerebellumremainintact,andthevariousmid-brainandthalamicpreparations,as, for example, those of Magnus. A further extension of the method involves thesurgicalorphysiologicalexclusionofend-organs,asbyextirpationoranaesthetization(forexample,ofthelabyrinth),orbysectionofafferentnerves.Thecommonobjectoftheseproceduresistopermittheinvestigationofaparticularresponseinrelationtoacontrolledvariable,independentofothervariablesalsorelatedtothatresponse.But the same resultmay be obtained in another way. The experimentmay be so

designed that the undesired variables do not vary. The distinction between the twomethods will appear in the following example from the work of Magnus. Certainpostural effects in a mid-brain animal are correlated partly with the position of thelabyrinthsrelativetotheearthandpartlywiththeconditionofflexionorextensionofthemusclesoftheneck.Thecorrelationbetweenthepostureandthestateoftheneckmusclescanbestudiedaloneifthelabyrinthsarecocainizedorextirpated.ButMagnuswasalsoable toobtain the isolationbydesigninghisexperiments insuchawaythatthe position of the labyrinths relative to the earth did not change. Perhaps the bestexamples of thismethod, however, are to be found in thework of Pavlov.Here theorganismisintactandtheveryactivereceptorsoftheheadsegmentsfullyfunctional.Bycontrollinglight,sound,odor,andotherconditionsintheexperimentalchamber,itis possible to observe in isolation the correlation between a given response and aselected stimulus.Placing an animal in a dark room, that is to say, is equivalent forpurposesofisolationtoblindingit, tosectioningtheoptictracts,ortodestroyingthevisual projection areas in the cortex, andhas thegreatadvantageover these surgicalmethodsofbeingrelativelyfreefromunknownartifacts.Thepracticalmeritsofboththesemethodsareobvious;butweareconcernedwitha

broader aspect of analysis. For the physiologist, the isolation of a reflex is apreliminary matter of method and is relatively insignificant. In the description ofbehavioritisoffirstimportance.Howlegitimate,then,istheprocessofanalysis,andwhatisthenatureofitsproduct?Let us deal entirely with the flexion reflex in the spinal dog, as a familiar and

convenientexample.Wehavealreadyanalyzed,ofcourse,whenwehaveoncenamed,sothatwemustgobackforamomenttothebehaviorasawhole.Withoutregardtoitscorrelationwith stimulating forces, behavior, aswe have seen, is simply part of thetotalfunctioningoftheorganism.Theproblemofanalysisat this leveliscommontophysiology and anatomy. We shall not need to solve it, but shall assume that forpurposes of description the body of an organismmay be divided into parts (thatwemayspeak,forexample,ofaleg),andthatthefunctioningofaparticularpartmaybe

434

described in isolation (that we may speak, for example, of the flexion of a leg).Moreover,weshallassumethattheforcesactingupontheorganismmaybeanalyzedanddescribedinthemannercommontothephysicalsciences.Ourownproblemliesbeyondtheseassumptions.In the flexion reflex our first experimental datum is the nearly simultaneous

occurrenceoftheflexionofalegand,letussay,theelectricalstimulationoftheskinofthefoot.Ifwemeasurebotheventsverycarefullyandrepeatthestimulation,weobtainasecondflexionwhichcloselyresemblesthefirst,andwefindthatwemaycorroboratetheobservation,within limits,asoftenaswelike.Wecall theobservedcorrelationareflexand,forconvenienceofreference,giveitaspecialname,theflexionreflex.Thequestionthenarises:whatistheflexionreflex?Ifwetrytoanswerbydescribing

indetailastimulusandaresponse,wemeetembarrassingdifficulties.Wefindthattheexact degree and direction of flexion may vary with many factors. We find, forexample,thatitwasveryimportantforouroriginalmeasurementsthatthetorsooftheanimalhadaparticularposition,thatthecontralaterallegwas,say,unsupported,andsoon. But we cannot specify these incidental conditions in our description withoutdestroying its generality. Thereupon we shall probably resort to surgical methods.Theoretically,atleast,wemayparedownthestructuresunderlyingtheflexionreflexuntilthecollateralvariablesarenolongereffective.Butwecanneverbesurethatthereflexwhichwehavethuscarvedoutofthebehavioroftheorganismwouldnothavebeengrosslyotherwiseifouroperativeprocedurehadbeendifferent.Wearenotsure,that is to say, thatwhat turnsup at the endofourprocessof isolation is the flexionreflex. There is another method open to us. In the flexion reflex we are dealingessentiallywithagroupofcorrelationsshowingmanycharacteristicsincommon.Theyinvolvethesameeffectorsactingroughlyinthesamewayandstimuliwhichresembleeachotheratleastintheirgrossanatomicalreference.Wemay,therefore,ifwewish,construct a flexion reflex by a statistical treatment of many of these separatecorrelations.Wemay, in otherwords, determine and state a correlation between thecharacteristicscommontoallourobservedresponsesandthecharacteristicscommontoallourobservedstimuli,andwemaynamethisconstructtheflexionreflex.Buttheresultingdescriptionof thisstatisticalentitywill likewisedependuponourchoiceofobservationsanduponourmethodofanalysis.Wehavebeenproceeding,ofcourse,uponanunnecessaryassumption,namely,that

thereisaflexionreflexwhichexistsindependentlyofourobservationsandwhichourobservations approximate. Such an assumption is wholly gratuitous, but it isremarkablyinsistent.Itarisesinpartfromthenatureofthereflex.Ifweremainatthelevel of our observations, we must recognize a reflex as a correlation. But theimmediateuncriticalreactiontoadefinitiononthatbasisisthatacorrelation,inpointof satisfaction, is not enough. There is an urge toward solidification, clearly evidentthroughout the history. We turn instantly to the reflex arc for material support.Although our knowledge of the critical part of the arc is, as we have seen, derivedwholly from the observation of a correlation, we much prefer to regard thecharacteristicsofthecorrelationaspropertiesofthesynapseratherthantoretainthemas characteristics of a correlation. Under the same pressure, then, but with lessjustification, we are led to assume that there are isolated reflexes concealed in thebehaviorofanorganism,whichbyproperinvestigatorymethodswemaydiscover,andin the description of behavior to state the corollary of this proposition, namely, thatbehavioristhesumortheintegrationoftheseunits.

435

Herewearetouchinguponthesubjectofawidespreadcurrentcontroversy,butwemay, by virtue of what we have already said, dispose of the matter briefly. Let usphrase two typical questions. Is a reflex a unitarymechanism? Is behavior a sumofsuchmechanisms?Then,ifbyreflexwemeanahypotheticalentitywhichexistsapartfrom our observations but which our observations are assumed to approach, thequestionsareacademicandneednotdetainus;if,ontheotherhand,wedefineareflexasagivenobservedcorrelationorasastatisticaltreatmentofobservedcorrelations,thequestions are meaningless, for they ignore the process of analysis implied in thedefinition.Areflex,thatistosay,hasnoscientificmeaningapartfromitsdefinitionintermsofsuchexperimentaloperationsaswehaveexamined,and,sodefined,itcannotbethesubjectofquestionsofthissort.There is a certain practical advantage, it is true, in regarding a reflex as a unitary

mechanism—anadvantage,asMachmighthavesaid,whichmayhavegivenrisetothepractice.Itisonlywhenwemisconstrueapurelypracticaldeviceandtakeittobeanintegralpartofourdefinitionthatthepossibilityoftheoreticalmisunderstandingarises.Oursamplequestionsdealnecessarilywiththereflexdefinedintermswhichwehaveseentobewellbeyondanyobservationaljustification.AsPoincaréhassaidofasimilarissue, “ces questions ne sont pas seulement insolubles, elles sont illusoires etdepourvuesde sens.”15A commonmistake in the present case has been to supposethat,becauseananswerislacking,theprincipleofthereflexissomehowimpeached.Aswehave repeatedly noted, the validity of the reflex as a scientific concept is nothereinquestion.Thereflexremains,asithasalwaysbeen,anobservedcorrelationofstimulusandresponse.

VII

Itremainsforustoconsiderhowareflexasacorrelationisdealtwithexperimentally.Thefirststep,aswehaveseen,istheisolationofaresponseandtheidentificationofitscorrelatedstimulus.Inpractice,thedemonstrationofthecorrelationisusuallyleftatanelementarylevel.Itisbasedupontheappearanceofthetwoeventstogetherandtheirfailure to appear separately.As an experimental datum of this sort, a reflexmay begiventheexpression

R=f(S)[1]whereRisaresponseandSastimulus.Theoretically,theexactnatureofthefunctionisdeterminable,althoughforanypresentpurposecorrespondingvaluesofSandR areobtainable by observation only.Choosing convenientmeasures of both stimulus andresponse,wemayvary the strengthofS andobservevariations in the strengthofR.Thisiscommonpractice,althoughverylittlehasbeendonetowarddetermininghowagivenR varieswith its correspondingS.One characteristic of the relationship is thethreshold:forvaluesbelowagivenvalueofS,R=O.Therearealsotemporalaspectsofthefunction,whichhavebeeninvestigatedundertheheadingsoflatencyandafter-discharge.Threshold, latency, after-discharge, and the order of variation ofS andR are thus

descriptionsofthecorrelationwecallareflex.Theymaybeinvestigatedwithonlyoneelicitationofthereflexor,atmost,withasinglesetofcorrespondingvaluesofSandR.Thereisasecondfieldofinvestigation,however,whichisconcernedwithvariationsin

436

any aspect of a correlation, as they may appear in the comparison of successiveelicitations. If, for example, we select a value of S and repeat the elicitation of thereflexat agiven rate,weshallobserveaprogressivedecrease in thevalueofR.Or,again, if the interval between two successive elicitations bemade brief enough, thesecondRmaybeofgreatlyreducedmagnitudeorwhollylacking.HerearesignificantvariationsinthevalueofthetermsinEquation[1].Theydonotchallengethenecessityoftherelationshipexpressedtherein(astheymightwelldoiftheywerelessorderly),but they do require that, in the description of a reflex, account be taken of thirdvariables.Wemayindicatetherequiredchangebyrewritingourequationas

R=f(S,A)[2]whereAisavariabledesignedtoaccountforanygivenobservedchangeinthevalueofR.As it appears in such an experiment,A is properly either time or the number of

elicitationsatagivenrate.Theinferenceiscommonlymadethatitrepresentsafactorofanothersort,whichvarieswithtimeorthenumberofelicitationsinthesameway.Inthefirstexamplenotedabovethephenomenonhasbeencalledreflexfatigue,whichisregarded as a synaptic change—as the exhaustion of a substance or state, or as anincrease in resistance, according to one’s preference in synaptic theory. But in thedescriptionofbehavior,whereweareonlysecondarilyinterestedinthesephysiologicalinferences, reflex fatigue is nothingmore than an orderly change in somemeasuredaspect of a given correlation.A lawdescribing the course of that change,where theindependentvariable is timeor thenumberofelicitationsorsomeotherconditionoftheexperiment,ispeculiarlyalawofbehavior.Itmaybecomealawofthesynapse,byvirtue of certain physiological inferences, but it has by that time passed beyond thescopeofthedescriptionofbehavior.Nevertheless, ifweare to followcurrentusage,adefinitionof reflex fatigueasan

observed variation in one aspect of a correlation is too narrow, for we know fromobservation that, when such a change has taken place, the other aspects of thecorrelationhavealsochanged.Ifwehaveobserved,forexample,achangeintheratioofaparticularRandS,wemayexpecttofindallotherratios,aswellasthethreshold,latency, and after-discharge of the reflex, likewise changed. It is usual, therefore, toregard the particular change which we chance to observe as a sample of a greaterprocess.Occasionally,whereachangeinoneaspectofacorrelationisaloneimportant(as in summation, which is chiefly a matter of threshold), the characteristic maypossibly be defined in terms of a single change. But such a characteristic as reflexfatigue,ortherefractoryphase,orfacilitation,isbyintentionadescriptionofagroupofconcurrentchanges.Ifwearetospeakintermsofthesegroupchanges,itisalmostnecessarytohavea

term describing the state of a correlation at any given time with respect to all itsaspects.Thephysiologist, of course,mayuse the synapse for thispurpose.Whenhehas once described reflex fatigue as the exhaustion of a synaptic substance, forexample,hemayattributeachange inany aspectofacorrelation to thatexhaustion.Althoughhemayobserveandmeasureatonetimeachangeinafter-dischargeandatanother a change in themagnitude ofR, he may reasonably consider himself to bedealingwith thesameprocess inbothcases.Fortunately, there isalsoa termservingthesamepurposeatthelevelofbehavior.If,inagivenreflex,thethresholdislow,thelatency short, the after-discharge prolonged, and the ratio R/S large, the reflex is

437

ordinarily said to be strong. If, on the other hand, the threshold is high, the latencylong,theafter-dischargeshort,andtheratioR/Ssmall,thereflexissaidtobeweak.Anattributeofstrengthisimputedtothereflex.Thestrengthoftheresponse,ofcourse,isnotmeant;aweakresponsemayindicateastrongreflexifitbeelicitablewithaveryweakstimulus.“Reflexstrength”expressesinaverygeneralwaythestateofagivencorrelationata

giventimewithrespecttomanyofitscharacteristics.Itisausefulterm,foritpermitsus to deal with reflex fatigue, for example, as a change in reflex strength, withoutstopping to specify the particular changes which compose it. Nevertheless, itsusefulness does not extendbeyond this qualitative level.The concept is subject to amajorobjection,whichholdsaswellfor theparalleluseofthesynapticstate.Wedonot know, since it has never been determined, whether the changeswhich composesuchacharacteristicasreflexfatigueallproceedatthesamerate.Ifthethreshold,letus say, and themagnitude ofR do not vary in precisely the same way, we are notjustifiedintakingeitherasameasureofasupposedcommonvariable,nor,indeed,incontinuingtoregardreflexfatigueasaunitaryprocess.Thestudyofthereflex,then,leadstotheformulationoftwokindsoflaw.Thefirst

arelawsdescribingcorrelationsofstimulusandresponse.Areflex,aswehavedefinedit, is itself a law, and of this sort. It has a considerable generality in spite of thespecificityofitsterms,butitmustbesupplementedbyotherlawsdescribingtheexactconditionsofacorrelation.Secondly,therearelawsdescribingchangesinanyaspectoftheseprimaryrelationshipsasfunctionsofthirdvariables,wherethethirdvariableinanygivencase is a conditionof the experiment.These secondary lawsmaybedealtwith in groups, according as they involve the same experimental third variable, andtheymaybespokenof,forconvenience,asdescribingchangesinreflexstrength.Inthebehavior of intact organisms the apparent variability of specific stimulus-responserelationships emphasizes the importance of laws of the second sort. Conditioning,“emotion,”and“drive,”sofarastheyconcernbehavior,areessentiallytoberegardedaschanges in reflex strength,and theirquantitative investigationmaybeexpected toleadtothedeterminationoflawsdescribingthecourseofsuchchanges,thatis,tolawsofthesecondsort.16Itisdifficulttodiscoveranyaspectofthebehavioroforganismswhichmaynotbe

describedwith a law of one or the other of these forms. From the point of view ofscientificmethod,at least, thedescriptionofbehavior isadequatelyembracedby theprincipleofthereflex.

Summary

Thepresentanalysisof thereflexasaconcept in thedescriptionofbehaviorfollowsthemethodfirstformulatedwithrespecttoscientificconceptsbyMachandPoincaré.Itexaminesthesourceofthehistoricaldefinitionandpointsouttheincidentalnatureofmostofitscriteria.Eventually,itoffersanalternativedefinitionandconsidersindetailsomeofthequestionswhicharisefromthenatureoftheconceptsodefined.I. Descartes “discovered the stimulus” and designed a mechanism which could

account for animal movement upon the basis of the appropriate release of storedenergy.Buthewasinterestedlessindescribingtheactionofthenervoussystemthan

438

insupportingmetaphysicalcontentionsoftheautomaticityofanimals.Headvancedthestimulusasasubstituteforsoul,butonlywithinafieldwhichomittedthegreaterpartoftheactivityofman.II. The notion of the reflex developed, independently of Descartes, from the

investigationof“irritability.”Theactionofastimuluswas implicit in theconceptofirritability,whichalsoassignedanautonomyoffunctionto thepartsofanorganism.Theconceptofthereflexarosequitenaturallywhenastimulusanditsrelatedresponsewere to be spatially distinguished.RobertWhyttmade the first historically effectiveobservations.III.ItremainedforMarshallHalltocleartheconceptofpsychicalcounterparts.This

hedidbysettingupadistinctionbetweenreflexandvoluntaryaction,whichresultedeventuallyintheunfortunatehistoricaldefinitionofthereflexasaformofmovementunconscious,involuntary,andunlearned.Volition,inHall’ssense,wasessentiallythehypothetical antecedent ofmovement forwhich no corresponding stimulus could beobserved,adefinitionwhichservedtoidentifythereflexwithscientificnecessityandvolitionwithunpredictability.IV.Thehistoryofthereflexhasknownonlyonepositivecharacteristicbywhichthe

concept may be defined: the observed correlation of two events, a stimulus and aresponse.Thenegativecharacteristics,ontheotherhand,whichdescribethereflexasinvoluntary, unconscious, and unlearned, have proceeded from unscientificpresuppositionsconcerningthebehavioroforganisms.Thereflexistentativelydefinedhereinasanobservedcorrelationofstimulusandresponse.V.Reflexphysiologyundertakes todescribe theeventswhich intervenebetweena

stimulusandaresponse.Thephysiologicalusagedoesnotquestionthedefinitionofareflex as a correlation, for the synapse is only a conceptual expression for the“reduced”characteristicsofagivencorrelation.VI. The essence of the description of behavior is held to be the determination of

functional laws describing the relationship between the forces acting upon, and themovementof, agivensystem.The reflex is,bydefinition, theprecise instrument forthisdescription.Itsanalyticalnatureisdiscussed,andexistingmethodsofanalysisareexamined.Currentobjectionstoanalysisareheldtohavenoscientificmeaning.VII.Theexperimental studyof the reflexmaybedivided into twoparts.There is,

first,theinvestigationofthecharacteristicsofacorrelation—latency,threshold,after-discharge,andtheorderofvariationofSandR.Secondly,thereistheinvestigationofvariations in thesecharacteristicsas functionsof thirdvariables.Thenotionof reflexstrengthisusefulindealingwiththissecondgroup.Thequestionofthirdvariablesisofextremeimportanceinthedescriptionofthebehaviorofintactorganisms.From the point of view of scientific method, any law describing the behavior of

organismsmustbereducibletooneoftheformshereindiscussed.Thedescriptionofbehavior,thatistosay,isadequatelyembracedbytheprincipleofthereflex.

1Tolman,E.C.Operationalbehaviorismandcurrenttrendsinpsychology.InProc.25thAnniv.Celebr.Inaug.Grad.Stud.LosAngeles:Univ.South.Calif.Press,1936.ReprintedinMarx,MelvinH.Psychologicaltheory:ContemporaryReadings.NewYork:Macmillan,1951.

2Descartes’accountistobefoundintheTraitédel’homme,whichdiffersinmany

439

respectsfromtheearlierPassionsdel’âmeinitsrepresentationoftheactionofthenervoussystem.

AconvenientaccountofDescartesasphysiologistisgivenbySirMichaelFosterinhisLecturesontheHistoryofPhysiology(London,1901).ThequotationsfromNicolasStensenandthetranslationfromDescartesaretakenfromthiswork.

3deCaus,Lesraisonsdesforcesmouvantesavecdiversemachines…degrottesetdefontaines.Thedate(1624)isconsiderablyearlierthanthatoftheTraité,andtheexamplestreatedbydeCausare,ingeneral,simplerthanthosedescribedbyDescartes.

4Foranaccountoftheeventsantecedenttotheexplicitformulationoftheconceptofthereflexsee:(a)Foster,op.cit.;(b)thehistoricalintroductioninJ.F.Fulton’sMuscularContractionandtheReflexControlofMovement.Baltimore,Md.:Williams&Wilkins,1926;(c)Ch.1inVerworn’sIrritability:APhysiologicalAnalysisoftheGeneralEffectofStimuliinLivingSubstances.NewHaven,Conn.:YaleUniv.Press,1913;and,formanyquotationsfromthesources,(d)F.Fearing’sReflexAction:AStudyintheHistoryofPhysiologicalPsychology.Baltimore,Md.:Williams&Wilkins,1930.

5“Anessayonthevitalandotherinvoluntarymotionsofanimals.”Edinburgh,1751.6Thequotationisfromthesecondedition,publishedin1763,page267.Thefirsteditionomits“tothenerveswhicharesenttothem.”

7Hall,M.Onthereflexfunctionofthemedullaoblongataandmedullaspinalis.London,1833,Phil.Trans.Roy.Soc.,readJune20,1833.

8FromHall’sMemoirsontheNervousSystem.MemoirII.OntheTrueSpinalMarrowandtheExcito-motorySystemofNerves,London,1837.Whytthaddistinguishedbetweentheactionofthewillandtheactionofastimulus,butbythelatterheprobablyintendedonly“mechanicalaction”uponamuscle.Whyttalsousedthewordspontaneoustoaidinthedistinction,butinthereversesense!Reflexmovementswereforhimspontaneous,sincetheyareperformedbytheseveralorgans“asitwereoftheirownaccord.”

9QuotedbyStirling,W.SomeApostlesofPhysiology.London,1902.10AnexhaustiveaccountofthediscriminatingactionofdrugsuponseparatereflexsystemsisavailableinR.Magnus’sKörperstellung.Berlin:Springer,1924.

11Pflüger,E.DiesensorischeFunctionendesRückenmarksderWirbelthierenebsteinenneuenLehreüberdieLeitungsgesetzederReflexionen.1853.Thiscontroversy,ofcourse,extendedwellbeyondPflüger’spersonalparticipation.Itwas,infact,thecontinuationofaphilosophicalreactionagainstHall’sconceptandwasonlyaccentuatedbyPflüger’sinterpretationofhisexperimentalfindings.

12DisregardingthesupposeddiscoverybyGalenandDescartes.13Lucas,K.TheConductionoftheNervousImpulse.London:Longmans,Green,1917.

14Sherrington,C.S.Remarksonsomeaspectsofreflexinhibition.Proc.Roy.Soc.,1925,97B,519–544.

15Poincaré,H.Lascienceetl’hypothése.Paris,1903.

440

16Thesecondhalfofmythesis,ofwhichthispaperwasthefirsthalf,describesexperimentson“hungerdrive”fromthispointofview.

441

TheGenericNatureoftheConceptsofStimulusandResponse

Inextendingtheconceptofthereflextothebehavioroftheorganismasawhole,itwasnecessarytoabandonthesurgicalprocedureswhichhadtraditionallybeenusedto generate “preparations” yielding orderly data. The specification of stimulus andresponse became a crucial issue. This paper, which appeared in The Journal ofGeneralPsychology,1935,12,40-65,wasaneffort to resolve itbyappealing to thelawfulness of changes in reflex strength as a criterion for the specification ofproperties.

I

Inthedescriptionofbehavioritisusuallyassumedthatbothbehaviorandenvironmentmaybebrokenintoparts,whichmaybereferredtobyname,andthatthesepartswillretain their identity from experiment to experiment. If this assumption were not insome sense justified, a science of behavior would be impossible; but it is notimmediatelyclear towhatextent it is supportedbyourobservations.Theanalysisofbehavior isnotanactofarbitrarysubdividing,andwecannotdefine theconceptsofstimulusandresponsequiteassimplyas“partsofbehaviorandenvironment”withouttakingaccountof thenatural linesoffracturealongwhichbehaviorandenvironmentactuallybreak.Ifwecouldconfineourselves to theelicitationofa reflexupona singleoccasion,

this difficulty would not arise. The complete description of such an event wouldpresent a technical problem only; and, if no limit were placed upon apparatus, anadequateaccountofwhatmightbetermedthestimulusandtheresponsecouldinmostcasesbegiven.Theadvantagewouldbethatweshouldbefreeofthequestionofwhatweweredescribing.Butwhenwe insistupona reproducibleunit,aswecannothelpdoing ifweare tohavea scienceofbehavior, the accountof a single elicitation,nomatterhowperfect,isinadequate.Foritisverydifficulttofindastimulusandresponsewhich maintain precisely the same properties upon two successive occasions. Thepossible (andvery rare) exceptions to this rule concernonlyvery simple stimulatingforces acting upon simple (and usually simplified) preparations. In the intact andunhamperedorganism(towhichourlawsmust,eventuallyatleast,apply)moststimuliare subject to the momentary orientation of receptors or to similar factors; andespeciallywherethestimulusisselectedthroughtheactionofprepotency(whichisthecaseinthegreaterpartofnormalbehavior),it isextremelydifficulttogiveanyclearaccountofhowthestimulatingenergiesaregoingtoact.Thereasonsarenotquitethesame on the side of the response, since the stimulus-response relationship is notsymmetrical, but the rule is equally well obeyed. Even in such a relatively simpleexampleastheflexionreflex,twosuccessiveresponseswillbefoundtodifferwidelyif

442

thecharacterofthemovementiscloselyenoughexamined.Weareaccustomedtodealwiththisproblembymainforce.Weconfineourstudyto

a reflex in which the response is originally of a very simple sort or may be easilysimplified (flexion, for example, or salivation) and in which the stimulus is of aconvenientform,maybelocalizedsharply,andisapplied,ratherthanselectedthroughprepotency. It is easier to restrict the stimulus than the response, since the stimuluspresents itself as the independent variable, but we are able by technical means tocontrol some of the properties of the response also. In thiswaywe devise a sort ofreproducibility;thatistosay,wearefrequentlyabletodescribearestrictedpreparationinwhichastimulus iscorrelatedwitharesponseandallpropertiesofboth termsarecapableofspecificationwithinasatisfactorilynarrowrange.Formany purposes a preparation of this kindmay be an adequate solution of the

problemofreproducibility.Asweshallseelater,somedegreeofrestrictionisprobablyalways required before successful experimentation can be carried on. But severerestrictionmust be rejected as a general solution. It necessarily implies an arbitraryunit, which does not fully correspond to the material originally under investigationbecauseitsexactcharacterdependsinpartupontheselectionofproperties.Likewise,itisnotasolutionwhichcanbeextendedtoaverylargenumberofreflexes.Aboveall,itsuppresses, by virtue of the very act of restriction, an important characteristic of thetypicalreflex.Itiswiththislastobjectionthatweshallbeespeciallyconcerned.

II

One way to show the inadequacy of the restricted preparation is to determine howmuchofeither the stimulusor the response isessentialor relevant to thecorrelationbetweenthem.Inapreparationoftheflexionreflexweareable,byreasonofcertaintechnical restrictions, to state a correlation between two terms fairly completelydescribed. But on the side of the stimulus we must admit that, so far as a merecorrelationisconcerned,theexactlocationwhichwehavegivenisunimportant—thatthecorrelationcouldbeshowneventhoughthestimuluswereappliedelsewherewithinaratherwiderange.Similarly,weneednotspecifytheformoftheenergy(whetheritisheat,forexample,orpressure,orelectriccurrent)orthedurationofitsadministrationor its amount within rather wide limits. A reduction to terms of afferent nervousimpulseswouldeliminatepartoftheproblemoftheformofenergybutnotthatoftheirrelevanceoftheotherproperties.Onthesideoftheresponse,likewise,weneednotspecifytherateordegreeofflexion;andifwehavenotsimplified,wecannotspecifythe exact direction or, having simplified, we cannot justify the selection of onedirectionasagainstothers.Mostofthepropertiesofthetwoeventsinthecorrelationare, so far as the mere elicitation of the reflex is concerned, irrelevant. The onlyrelevantpropertiesareflexion(thereductionoftheanglemadebyadjacentsegmentsofa limbatagiven joint) andagiven (“noxious”)kindof stimulationappliedwithinaratherlargearea.If we turn, then, from the exact reproducibility of stimulus and response to the

criterion of simple elicitability, we arrive at nothingmore than a correlation of twodefiningproperties. Inordinarypractice thesepropertiesalonemaintain their identifyfromexperiment to experiment.But itwouldbe inconvenient to regard a reflex as acorrelation of properties. We cannot produce one defining property at a given

443

elicitation without giving incidental values to the non-defining properties whichcomposetherestoftheevent.Astimulusoraresponseisanevent,thatistosay,notaproperty; and we must turn, therefore, to a definition on the principle of classes.Accordingly,ifwearetocontinuetoregardtheflexionreflexasasingleentity,boththestimulusandtheresponsemustbetaken(tentatively,atleast)asclassterms,eachofwhichembracesanindefinitelylargenumberofparticularstimuliorresponsesbutissufficientlywelldefinedbythespecificationofoneortwoproperties.Thealternativetoacknowledgingthisgenericnatureistoarguethateverypossible

restricted correlation is an independent unit in itself. On this hypothesis there arepracticallyaninfinitenumberofflexionreflexes,correspondingto theproductof thenumber of ways in which an effective stimulus can be applied into the number ofparticular responseswhich can be obtained through differentmethods of restriction.We may contrast these two views by saying that either a reflex is a broad termexpressing the correlation of a class of stimuliwith a class of responses (where thereproducibilityofnon-definingpropertiesisunimportant)oritappliestoanyoneofagroupofparticularcorrelations(wherethetermshavebeenseverelyrestrictedtoobtainthereproducibilityofallproperties).Inthesecondcasewemaystillgroupourspecificcorrelations together on the basis of a defining property without implying thefunctionallygenericnatureofeitherstimulusorresponse:eveniftherearepracticallyaninfinitenumberofflexionreflexes,forexample,theyallhavesomethingincommonnotsharedbyanyother,inthattheirresponsesareexamplesofflexion.Ifwewishtoassignthetermreflexforthemomenttoagroupofthissort,ratherthantoaparticularexample,ourproblemmaybestatedinthefollowingform:isareflexacorrelationofclassesoraclassofcorrelations?There is a statement of the subject which differs only slightly from the present

(althoughitismuchlessflexible),inwhichwhatwehavecalledtheirrelevanceofthenon-definingpropertiesofastimulusisexpressedbyspeakingofagroupofstimuli,allofwhichareequivalentintheelicitationofaresponse.Thekindofproofusuallygivenforthisviewisbaseduponthefactthatintheprocessofconditioning(Pavlov’stype)anewreflexiscreated.It is thenpossibletoprovetheirrelevanceofcertainproperties(or the equivalence of stimuli) in the following way. Let a conditioned reflex beestablishedtoalight,forexample,whichissoplacedthatonlyalimitedregionintheretinaofoneeye is illuminated.Then itmaybeshown that,after theconditioning iscomplete,abeamof lightstrikingotherpartsofeither retinawillelicit the response.Theeffectivenessofthenewlyconditionedstimulusisindependentofthepropertyoflocation,andsofarasthesimplecorrelationofstimulusandresponseisconcerned,weneednotspecifyitslocationinourdescription,atleastwithinwidelimits.Wemayalsofindthatthepropertiesofbrilliance,hue,shape,andsizemaynotbesignificantoverconsiderableranges,andthattheonlyimportantpropertiesare,indeed,thosedenotedroughlyby“spot”and“light.”Here, then,wehaveaclassofstimuli,definedbytwoproperties,themembersofwhichareequivalentsofarastheelicitationofaresponseisconcerned.Theadvantageinusingaconditionedreflexliesinbeingabletoshowthatmembers

of the group differing from the particular stimulus used for conditioning cannot beelicitingresponses“ontheirownaccount”sincetheywerenotabletodosobeforetheconditioningwassetup.Butunfortunatelythisproofisoflimitedscope.Itisnoteasilyapplied to the case of the response and is of no value for unconditioned reflexes orconditionedreflexesthehistoryofwhichisnotknown.Aninfinitelylargenumberof

444

stimulimay,throughtheuseofconditioning,bemadetoevokethesameresponse(aspotoflightandatone,forexample,maybothelicitsalivation),butthereneedbenocommonpropertyamongthemexceptthatofbeingastimulus,whichisnotinitselfapropertywhichguaranteestheeffectivenessofanuntriedstimulusknowntopossessit.The “equivalence” of a spot of light and a tone is the product of an experimentalprocedureandisclearlynottheequivalencefoundinthecaseoftwospotsoflight;buttheabilitytoelicitacommonresponsedoesnotdistinguishbetweenthetwosorts,andnodistinction is, aswehave said, possiblewhenwedonot know thehistoryof theorganism.Abetterproof,whichisapplicabletoallcases,makesuseofthesecondarylawsof

the reflex [see page502]. It is often true in the investigation of these laws that thenumberofelicitationsofareflexisimportant,as,forexample,whenwearemeasuringarate.Itisthenpossibletotesttheirrelevanceofanon-definingpropertybyshowingthat two responses, one of which possesses the property, the other not, contributeequallywell toatotalnumber.Supposethatwearestudyingthebehaviorofsuchanorganismasaratinpressingalever.Thenumberofdistinguishableactsonthepartoftheratwhichwillgivetherequiredmovementoftheleverisindefiniteandverylarge.Exceptforcertainrarecasestheyconstituteaclass,whichissufficientlywell-definedby the phrase “pressing the lever.” Now it may be shown that under variouscircumstancestherateofrespondingissignificant—thatistosay,itmaintainsitselforchangesinlawfulways.1Buttheresponseswhichcontributetothistotalnumber-per-unit-timearenotidentical.Theyareselectedatrandomfromthewholeclass—thatis,by circumstanceswhich are independent of the conditions determining the rate.Notonly, therefore, are themembersof the class all equally elicitableby the stimulationarisingfromthelever,theyarequantitativelymutuallyreplaceable.Theuniformityofthechangeinrateexcludesanysuppositionthatwearedealingwithagroupofseparatereflexesandforcestheconclusionthat“pressingthelever”behavesexperimentallyasaunitarything.Analmostparallelargumentcouldbemade fromthesamedataon thesideof the

stimulus, yielding a stimulus-class sufficientlywell denoted for ourpresent purposesby the term lever.The proof by appeal to secondary laws ismuch stronger than theargument for equivalence of stimuli based upon the behavior of newly conditionedreflexes.Itisofgeneralvalidityandgoesbeyondtheuseofmere“abilitytoelicit”toaquantitativemeasure.Thusinourtestcasewecoulddistinguishbetweentheseparatecorrelations of a single response with a tone and a spot of light by showing, forexample,thattheextinctionofoneofthemdoesnotmodifythestateoftheother.An exception may be taken to this last example on the ground that there will

probably be some influence between the two, and this brings us abruptly to animportantpoint.Theargumentonthebasisofsecondarylawswouldbeunanswerableifitwereasclear-cutaswehavegivenit,anditwoulddecidethequestionclearlyonthe side of the reflex as a correlation of two generic terms rather than as a class ofdistinctcorrelationsoranyonememberofsuchaclass.Butunfortunatelytheargumentmustbequalified,andinsuchawayastostrengthentheoppositeview.Foritistruethatthenon-definingpropertiesareoftennotwhollynegligibleandthatthemembersofour classes are consequently not exactly mutually replaceable. On the side of theresponse, of which we have less control, our data will not show this inmost casesbecause of the present lack of precision. But it is certain that there are outlying

445

membersofaclasswhichhavenota full substitutivepower; that is tosay, thereare“flexions”and“pressings”whicharesounusualbecauseofotherpropertiesthattheydonot fullycountas such. It ought to be supposed that lesser differenceswould besignificantinamoresensitivetest.Ifweshouldexaminealargenumberofresponsesleadingtothemovementofthelever,mostofthesewouldberelativelyquitesimilar,buttherewouldbesmallergroupssetoffbydistinguishingpropertiesandafewquiteanomalousresponses.Itisbecauseofthehighfrequencyofoccurrenceofthefirstthatthey are typical of the response “pressing the lever,” but it is also because of thisfrequency that any lack of effectiveness of atypical responses is not at presentsufficientlystronglyfelttobenoted.On the side of the stimulus, on the other hand, small differences may be

demonstrated.Sincewemayherecontrolthevaluesofournon-definingproperties,wemaymasstheeffectofagivenexample.Thuswecanshowthat in theflexionreflexfatiguefromonelocusofstimulationdoesnotresultincompletefatigueofthereflexfromanotherlocus.Herewehavesegregatedparticularstimuliintotwogroupsonthebasisofthepropertyoflocation,andhaveshowntherelevanceofthepropertytothecourse of a secondary change. A similar and very important example of the use ofsegregation arises in the behavior of the intact organism in the process ofdiscrimination.Supposewehaveestablishedaconditionedresponsetoalever,asintheaboveexample.Uponanyoneoccasionthestimulusis,aswehaveseen,anymemberofanindefinitelylargeclassofstimuliarisingfromtheleverandthesurroundingpartsoftheapparatus.Itispossibletocontrolsomeofthepropertiesofthesemembers.Forexample,thelevermaybemadetostimulateeitherinthelightorinthedark,sothatallpropertieswhicharise asvisible radiation canbe introducedor removedatwill.Werequiretoshowthattheyarenotwhollyirrelevant.Thismaybedonebysettingupadiscrimination,sothatthestrengthoftheresponsetothelever-plus-lightremainsatagiven(say,nearlymaximal)value,whilethestrengthoftheresponsetotheleveralonedeclinestoanothervalue(say,nearlyzero).Althoughadiscriminationofthissortisinpart thedevelopmentof adistinctionwhichdidnotoriginally exist, it canbe shownthatsomesignificanceoriginallyattachedtothedifferentiatingproperty.2Ineitherofthesecasesifwehadallowedthestimulustovaryatrandomwithrespect

to thenon-definingproperty,we shouldhaveobtained reasonably smooth curves forthesecondaryprocess,accordingtoourpresentstandardsofsmoothness.Itisonlybyseparatingthestimuliintogroupsthatwecanshowtheirlackofcompleteequivalence.Butoncehavingshownthis,wecannolongerdisregardtheimportanceoftheproperty,even in the absence of grouping.A similar argumentwould apply, of course, if ourcriterionweresimplyabilitytoelicit.Heretherelevanceofnon-definingproperties(orlackofequivalence)canperhapsbeshownonlyatnear-thresholdstatesofthereflex,sincethemeasureisall-or-noneandthereforecrude,butwecannotassumethatatotherstates a similar relevance would not be detected with a more sensitive measure. Inneither case havewe a clear indication that the argument for a generic definition iswhollyvalid.Inregardingeverydiscretecorrelationasaseparateentity,bothoftheaboveproofs

maybeexplainedawaybyappealto“induction”—aprocessthroughwhichachangeinthe stateofone reflex is said to inducea similarchange in the stateofanother.Theapparent mutual replaceability of a number of flexion reflexes in the course of asecondary change is explained by holding that something done to one of them (in

446

fatigue, for example) is done to others also through induction. The principle isobviously designed to dealwith the effectswe have just appealed to, and it has theadvantage that where the argument for equivalence or a generic term falls short ofcompleteexperimentalsupport,theargumentforinductionisstrengthened:foritmightbeexpectedthatamutualinfluenceofthissortwouldbeonlypartial,asitprovestobe,andwould,moreover,dependuponthedegreeofcommunityofproperties,asitcanbeshown to depend. On the other hand induction is under the present circumstancesclearlyanadhoc device, and its use should lead us to suspect the view that everyparticularcorrelationisadiscreteandautonomousentity.We have, in short, no clear basis for choosing either of these two views, and the

decision we are likely to make is free to follow our personal prejudices. If we areinterestedinthephysiologicaleventsmediatingareflexweshallveryprobablywanttodealwithseverelyrestrictedpreparationsandweshallbewillingtoexplainawaytheproofsforthegenericnatureofthereflexbybringinginthedeviceofinduction.If,onthe other hand, we are interested in the behavior of the intact organism, whererestrictionismuchmoredifficultandinmanycasesimpossibleunlessthematerialisseriously disturbed, we shall be anxious to prove the irrelevance of non-definingpropertiesandshallwanttodefineourunitwithoutrespecttothem.Butitoughttobeclear fromour failure to findavalidproof foreitherof theseextremeviews that thetruthliesbetweenthem.Thereisnoreasonwhyacleardefinitionofaunitofthissortisnotpossibleinourpresentstateofknowledge.Theproblemofdefinitionis,afterall,anexperimentalone,andtheentitieswhichwearetouseinthedescriptionofbehaviorareexperimentalentities.Wehaveplacedourselvesatagreatdisadvantageintryingtofind among our data evidence for a preconceived term, when our primary concernought to be simply with putting the data in order; and we may well suspend for amomentthequestionofthenatureofthesetermsandturndirectlytoanexaminationoftheavailableexperimentalmaterial.

III

1.One factwhich seems tobe sufficientlywell established is that there aredefiningproperties.Nothingwehaveconsideredof the importanceofnon-definingpropertiesmodifiesthisintheleast,norareweprejudgingthepresentissue,sinceapropertymaybetaken,aswehaveseen,todefineeitheronereflexoraclassofreflexes.Adefiningpropertyappearsonthesideoftheresponseinthefirststeptowardwhatiscalledthediscoveryofareflex.Someaspectofbehavior isobservedtooccurrepeatedlyundergeneralstimulation,andweassignanametoitwhichspecifies(perhapsnotexplicitly)adefiningproperty.Ourcontrolovertheresponseisalmostexclusivelyofthissort—specification.Wehavetherefusalofallresponsesnotfallingwithintheclasswehaveset up.Sincewe are completely free in this first choice, it is easy to select awrongdefiningproperty,but the followingstepscannot thenbe takensuccessfully.Whenadefiningpropertyhasbeendecidedupon,thestimuliwhichelicitresponsespossessingitarediscoveredbyexploration.Onestimulusmaybeenoughtodemonstratethesortof correlation sought for, but (either deliberately or through lack of control) thepropertiesareusuallyvaried in laterelicitationsandothermembersof theclass thusadded. Subsequently the defining property of the stimulus is inferred from the partcommontothedifferentstimuliwhicharethusfoundtobeeffective.

447

There must be defining properties on the sides of both stimulus and response;otherwiseourclasseswillhavenonecessary reference to realaspectsofbehavior. Ifthe flexion reflex is allowed to be defined simply as the class of all reflexes havingflexionasaresponse(orasareflexhavingforitsresponseaclassdefinedbyflexion),thereisnothingtopreventthedefinitionofaninfinitenumberofreflexesuponsimilarbases.Forexample,wecouldsaythatthereisareflexorclassofreflexesdefinedbythisproperty:thatintheelicitationthecenterofgravityoftheorganismmovestothenorth.Sucha class is experimentallyuseless, since it brings togetherquiteunrelatedactivities.Butwemustbereadytoshowthatallflexionsarerelatedinawayinwhichallmovementsof thecenterofgravityarenot, and todo thiswemustappeal to theobservedfactthatallflexionsareelicitablebystimuliofafewclasses.Assoonasthisrelationisapparentourtentativeresponse-classbeginstotakeonexperimentalrealityasacharacteristicofthebehavioroftheorganism.Itisdifficult,however,tosaypreciselywhatdefiningpropertiesare.Wefrequently

definethestimulusbytheverydoubtfulpropertyofitsabilitytoelicittheresponseinquestion rather than by any independent property of the stimulus itself. Thus, in thebehavioroftheunhamperedorganismwithrespecttosomeobjectinitsenvironment,weoftencannotdescribetheactualstimulatingenergies,butweassumethat,whenevera response is elicited, some member of the class of effective stimuli has acted.Similarlyintheflexionreflexthebasisforourdefinitionoftheproperty“noxious”isprobablyonlytheeffectivenessofacertainformofenergyinelicitingaresponse.Itisalwaysimplied,ofcourse,thataparalleldefinitionintermspeculiartothestimuluscanbegiven.Anexception is thecasealreadynotedof the typeofconditionedreflex inwhich we cannot define the stimulus except by ability-to-elicit or by appeal to thehistoryof theorganism.Fortunatelywedonotreverse thedirectionof thisargumentand define a response as any behavior elicited by a given stimulus.Behavior is lessunderexperimentalcontrolthanenvironment,anditwouldbemoredifficulttodetectasignificant correlation in that direction. But at this level of analysis the response isseldom clearly defined in any way. A rigorous definition without regard to non-defining properties is, in fact, probably impossible because, as we have seen, thedefiningpropertycanbemadetofailbytakingextremevaluesofotherproperties.Noraretheactualmembersofeitherclasseverexhaustivelyinvestigated;sothatitmaybesaid that these broad terms are defined neither by specification of properties nor byenumeration.2.Asidefromavoidingawrongdefiningproperty,whichwillnotyieldacorrelation

withasinglestimulus-class,wehaveacertainfreedominspecifyingtheresponse.Byincludingothernon-definingproperties inour specificationwemaysetupotherandless comprehensive classes, forwhich corresponding stimulus-classesmay be found.The latterwillbe lesscomprehensivealso since, aswehaveseen, the stimulus-classwhichwearriveatisalwayscloselyadjustedtotheresponse.Forexample,ifwebeginwith “flexion in a specific direction only,” we obtain a stimulus-class embracing asmallerstimulatingarea.Now,thereisnothingtopreventourtakingsucharestrictedunitatthestart,solongasforanysuchclassastimulus-classmaybefound,andifarestrictedunitistakenfirsttheverybroadesttermcanbearrivedatonlybyremovingrestrictions.Oursecondexperimentalfactisthatwithintheclassgivenbyadefiningpropertywe

may set up subclasses through the arbitrary restriction of other properties. Thisprocedure yields a series of responses, generated by progressive restriction, each

448

memberofwhichpossessesacorrespondingstimulusinamoreorlessparallelseries.Atoneendweapproachasa limit thecorrelationofacompletelyspecifiedresponseand a stimulus which is not necessarily strictly constant but may be held soexperimentally. If at this point both terms are in fact unit classes, one part of ourproblem vanishes, since with a perfectly restricted preparation there is no practicaldifferencebetweenaclassofcorrelationsandacorrelationofclasses.Butthisstateis,as we have argued, probably always impractical and in any event never fullyrepresentative.Ourinterestinithereisasanideallimit.Theotherendoftheseries,theunrestrictedclass,wehavealsoseentobeideal,sothatanyexperimentallyvalidunitmustbesoughtforamongthepartiallyrestrictedentitieslyingbetweentheseextremes.Inspeakingofaseriesgeneratedbyrestrictionweare,ofcourse,usingtoosimplea

term.Our technique of restrictionmust respect the defining property, but that is ouronlyimportantlimitation.Throughtheselectionofdifferentnon-definingpropertieswemaysetupdifferentrestrictedentitieswithinasingleclass;forexample,inrestrictingtheflexionreflexbyfixingthelocusofstimulation,wemayobtainseparateentitiesbyselectingdifferentloci.Thereisnouniquesetofnon-definingpropertiespeculiartoagiven defining property, and we have to deal, not with a single series, but with acomplex set of ramifications from a single virtual source, approaching as limits anindefinitenumberofdifferentcompletelyrestrictedentities.Part of the difficulty of definition which we encounter in dealing with a single

definingproperty(point1above)maydisappearinthepartiallyrestrictedpreparation.Usuallythefirstrestrictionsaredesignedtoprotectthedefiningpropertybyexcludingextremecases.Theyclarifythedefinitionandaddweighttotheexpressedcorrelationwithastimulus-class.Ingeneral,asweprogressivelyrestrict,ourdescriptioncomestoinclude more and more of the two events and is consequently so much the moresuccessful.At the same timeagreater andgreater restrictionof the stimulus-class isdemanded, so that the increase in the validity and completeness of the correlation ispaidforwithaddedexperimentaleffort.3.Ourthirdfactisinduction,whichitisnowpossibleforustodemonstratewithout

raisingthequestionofaunit.Wehaveseenthatitispossibletoobtainvariouskindsofentitieswithinasingleclassthroughtherestrictionofnon-definingpropertiesandthatmanyofthesemayexistatthesametime.Theyareexperimentallyrealandoperable,andtherecanbeshownbetweentwogivenexamplessomedegreeofmutualinfluenceofthesortwehavealreadyexamined.Achangetakingplaceinoneofthemisfoundtohavetakenplacealsointheother.Theonlyimportantruleofinductionthatweneedtonoteisthattheextentofthemutualinfluenceisafunctionofthedegreetowhichtheentities possess their non-defining properties in common.We shall not review otherinformationinanydetail.Theliteratureisverylarge,especiallyifweinclude(aswerightfullymay)allworkondiscrimination.Itisanimportantfieldofanalysis,althoughitsrelationtotheproblemofthedefinitionofareflexhasusuallynotbeenmadeclear.4. In turning to induction we have necessarily taken up new criteria. Classes or

subclasses may be demonstrated simply by showing correlations of stimuli andresponses andby listing thepropertiesof these events, but the influences exertedbyone restricted entity upon another are felt principally in the course of secondarychanges.Ourfourthpointisthat,inthemeasurementofthesemoreadvancedaspectsof a correlation,movement along a series in the direction of a completely restrictedentity isaccompaniedbyan increase in thesimplicityandconsistencyofourdata. Ifwearemeasuringfatigue,forexample,weshallnotobtaintoosmoothacurveifour

449

stimulusvariesinsuchawayastoproduceatonetimeonedirectionofflexionandatanothertimeanother;butaswerestrictourstimulustoobtainalessvariableresponse,thesmoothnessofthecurveincreases.This is not really a separate point but rather a special case of point 3. In such a

secondaryprocessasfatigueorextinctionweareexaminingtheeffectofoneelicitationuponanotherfollowingit.Butthisisonlyinduction,sincewearenotyetassuminganykind of identity from one occasion to another.We look for this effect to follow themainruleofinduction:itwillbeafunctionofthedegreeofcommunityofproperties.In a completely restricted preparationwe should therefore have complete induction,sincetwosuccessiveelicitationswouldbeidentical.Eachelicitationwouldhaveitsfulleffect upon a secondary change, and the curve for the secondary change would besmooth. But if we are using only a partially restricted entity, successive elicitationsneednothaveidenticallythesameproperties,andsecondaryprocessesmayormaynotbe advanced full steps through induction. From our third point, therefore, we couldhavededucedaformofthefourth,namely,thatanimprovementindatafollowsfromanychangewhichmakessuccessiveelicitationsmorelikelytoresembleeachother.5. If induction followed properties quite literally and without prejudice, its study

would not add anything to our knowledge of the relationship between two entitieswhichwecouldnotinferfromacomparisonofpropertiesalone.Butpropertiesarenotallequallyimportantsofarastheinductionbetweentwomembersisconcerned.Thestructurewehavesetuphassofarbeenbasedsolelyoncommunityofproperties.Anydistinction whatsover between responses has been allowed, so long as the test ofcorrelationwithastimulus-classwasforthcoming.Now,itmayhavebeennoticedthatanattempttodistinguishbetweentworesponse-classesonthebasisofsomepropertyhas failed. Itmaynothavebeenpossible to find twocorresponding stimulus-classeswhichelicitedthemseparately.Butifwehaveonestimulus-classcorrespondingwithtwo response-classes,we cannot be sure of confirming either correlation on a givenoccasiontotheexclusionoftheother.Wemustconclude,therefore,thatthepropertyupon which the two classes have been distinguished is not effective. This variableimportancecomesoutclearlyinthestudyofinduction,anditisimportantenoughtobestatedseparatelyasourfifthpoint.The most general form of the rule, in agreement with the present ordering of

experimental data, is as follows:practically complete inductionmayprevail betweentwoentitiesdifferingevenwidelywithrespecttosomenon-definingproperties.Aswehavejustseen,itmaybestatedinrelationtooursecondpointinthisway:somenon-definingpropertiesdonotestablishsubclasses.Amorelimitedexpression,whichtakesthe formofa qualification of point 4, is as follows: asweproceedwith the gradualrestrictionofapreparation,notingacorrespondingimprovementintheconsistencyofourdata,thepointatwhichanadequateconsistencyisreacheddoesnotcoincidewiththe final complete restriction of all properties of the preparation.The proofs for thisveryimportantrule(especiallytheproofbyappealtosecondarylaws)havebeengivenabove inarguingfor thegenericnatureof thereflex,andweshallnotneed torepeatthem.We are now, however, including somenon-defining properties in the terms towhich theyapply,andwe thereforeavoid theobjectionspreviously raised. In fact, itwillbeapparentthatwehavebasedourselectionofnon-definingpropertiesuponjustthecriterionappealedtointhoseobjections—namely,completenessofinduction.Thisisapracticalrule,whichdoesnotpretendtogobeyondthelimitsofourpresent

degreeofprecision.Butitsmainfeaturesaretoowellmarkedtobeseriouslydisturbed

450

under limiting conditions. A practical consistency may appear at such a relativelyunrestrictedlevel—and,asonemightsay,sosuddenly—thatextrapolationtocompleteconsistencyappearstofallfarshortofcompleterestriction.Itwouldbeidletoconsiderthepossibilityofdetailswhichhaveatpresentnoexperimentalrealityorimportance.Itmaybethatthelocationofthespotoflightortheidentityofthemuscle-fibreswhichcontract as the lever is pressed are somehow significant up to the point of completespecification;butwearehereinterestedonlyinthedegreeofconsistencywhichcanbeobtainedwhiletheyarestillbynomeanscompletelydetermined.Thisconsistencyissoremarkablethatitpromisesverylittleimprovementfromfurtherrestriction.As amatter of fact, whenwe have reached the point at which orderly secondary

changesappear,wecannotgobeyonditwithfurtherrestrictionwithoutdestroyingthisdesiredresult.Intheexampleofthelever,wemayobtainsmoothcurvesbyrestrictinguptoacertainpointonly; ifwefurther limit theresponsebyexcludingallexamplesexceptthoseofonegivenkind(pressingwithacertainmuscle-group,forexample),wedestroy our curves by eliminating many instances contributing to them. The set ofpropertieswhichgivesus“pressingthelever”isuniquelydetermined;specifyingeitherfewerormorewilldestroytheconsistencyoftheresultobtained.Thisfollowsnaturallyfromthenatureofourcontroloverresponsespecificationandrefusaltoaccept.

IV

These,then,aretheimportantaspectsoftheanalysisofbehaviorwhichbearuponthedefinitionofaunit.Wehavelistedthem,notinrelationtoadefinition,butintheorderinwhichtheyappearinactualexperimentation.Buttheproblemofdefinitionhasnowbeen practically solved. We have arrived at a structure of entities having anexperimentalfoundation,andwehaveonlytodecidetowhatpartofitwearetoassignthetermreflex.The twoextremeviewswithwhichwebeganmaybe related to thepresent result

without difficulty. The extreme generic view is that a stimulus or a response is thewholeclassgivenbyadefiningproperty.Butwehaveseenthatthisisprobablyneversharply defined without appeal to secondary properties, and its members are neverexhaustivelyinvestigated.Asastructureitmaybecomeprodigious:inthebehaviorofthe intact organism the number of subclasses which could be set up throughdiscrimination is often practically infinite. This kind of unit yields a sort ofreproducibility (that of its defining property), but it is not enough to insure uniformsecondaryprocesses.Itisnot,inshort,anexperimentalconcept,andalthoughitmightbewelltogiveitaname(“surreflex”forexample),weoughttoreservethetermreflexitself for an observable entity. For the same reasonwe cannot accept the definitionproposedbytheextremeparticularist;thefullydeterminedentityapproachedwiththetechniqueofrestrictionisalso,aswehaveseen,ideal.Thematerialwhichweactuallyobserve,andwhichexhibits significantuniformity, is thebehaviorof thepreparationrestrictedtothepointofgivingsimpleandconsistentdata.Here,ifanywhere,itwillbeconvenienttoapplyourterm.We may restrict a preparation for two quite different reasons, either to obtain a

greaterprecisionofreferenceforourterms(sothatourdescriptionofaresponse,forexample, will describe it more completely and accurately) or to obtain consistentcurves forsecondaryprocesses.The increase inprecisiongivesagreaterauthority to

451

our statementof a correlation,which isdesirable; but itwill nothelpus indecidinguponaunit.Itleadsultimatelytoacompletelyrestrictedentity,whichwehaveseentobeusuallyunreproducibleandotherwiseimpractical,andtoobtainaunitweshouldbeforcedtostopatsomearbitrarylevel—forexample,atacompromisebetweenprecisionof reference and the experimental effort of restriction. Our second criterion, theorderlinessofsecondaryprocesses,givesus,ontheotherhand,aunitwhichis innosense arbitrary. As we have seen, the appearance of smooth curves in secondaryprocessesmarksauniquepointintheprogressiverestrictionofapreparation,anditistothisuniquelydeterminedentitythatthetermreflexmaybeassigned.Areflex,then,is a correlationof a stimulusanda responseat a level of restrictionmarkedby theorderlinessofchangesinthecorrelation.Incertainrespectsthisisnotassimpleadefinitionasonemightwishfor.Itmeans

that sincemany equally consistent preparationsmay be set upwithin a single class,therewillbealargenumberofreflexespassingunderasinglename.Thismayseemtorob the principle of the reflex of much of its simplicity, but it is a necessaryconsequenceofthecomplexityofthematerial,whichcannotbechangedbytheoreticalconsiderations.Ifweshallnotbeabletoreferunequivocallytoasingleexperimentalentitywiththetermflexionreflex,atleastwemayknowthatthishasneverreallybeenpossible. A great deal of misunderstanding has arisen from the practice of namingreflexes,whichaninsistenceuponasupplementarylistofspecificationsmayavoid.Likewise,itisnotnecessarilytruethattheentitiesresultingfromthisdefinitionare

souniformthatalawbasedupononeexamplewillhavecompletegenerality.Acertainlatitude isallowedbyourpresentdegreeofprecision. It isnotalwayseasy toprovefromthedegreeoforderlinessofachangethatasignificantpropertyisnotvaryingatrandom,althoughthepresenceofsuchavariablewillprobablyaffecttheshapeofthecurve for the change. Aside from this matter of precision, it is also probable thatpreparationshavingdifferentcontrolledvaluesofagivenpropertywillyielddifferentcurves. In the case of restriction through the removal of properties (where this ispossible),wehavea seriesofpreparationsof increasingsimplicityandof increasingeaseofcontrolbutnotnecessarilyof increasingconstancy.Weshouldnotexpectanincreaseinsmoothnessalongsuchaseries,butitisprobablethatthenatureofacurvewillshowachange.Theseare,however,experimentalquestionsandouronlypresenttask(formulation)hasbeensufficientlywellcarriedout.Ourdefinitionisnot, inanyevent, dependent upon the generality of the laws obtained with a single example,althoughthegreatestpossiblegeneralityisobviouslydesirable.Indecidinguponthisdefinitionwechoosesimplicityorconsistencyofdataagainst

exactreproducibilityasourultimatecriterion,orratherwetempertheextenttowhichexact reproducibility is to be demanded and use the consistency of our data in ourdefense.Thiswouldbeonlygoodscientificmethodifwewerenotforcedtoitforotherreasons.Toinsistupontheconstancyofpropertieswhichcanbeshownnot toaffectthemeasurementsinhandistomakeafetishofexactitude.Itisobviouswhythishassooftenbeendone.What iswanted is the“necessaryandsufficient”correlationofastimulus and a response. The procedure recommended by the present analysis is todiscover the defining properties of a stimulus and a response and to express thecorrelationintermsofclasses.Theusualexpedienthasbeentoholdallpropertiesofagiven instance constant so far as this is possible. In a successful case all propertiesseemtoberelevantbecausetheyinvariablyoccuruponalloccasions.(Itisalmostasif,faced with the evident irrelevance of many properties, we had invented the highly

452

restricted preparation to make them relevant.) In giving a complete account of anarbitrarilyrestrictedpreparation,wedescribeatthesametimetoolittleandtoomuch.Weincludematerialirrelevanttoourprincipaldatum,sothatpartofourdescriptionissuperfluous,andwedeliberately ignore thebroadercharacterof thestimulusand theresponse.Thecompletedescriptionofoneactofpressingaleverwouldhaveverylittleusefulness,sincemostoftheinformationwouldbeirrelevanttothefactofelicitation,with which we are chiefly concerned, and would tell us nothing about the set ofpropertiesyieldingaconsistentresult.Wedonot,ofcourse,avoidorwishtoavoidrestriction.Itisanindispensabledevice,

forithasthemeritofholdingadefiningpropertyconstanteventhoughthepropertyhasnotbeen indentified.Untilwehavediscoveredadefiningproperty, it isnecessary toresort to restriction to guarantee ultimate validity.And since, aswe have seen, it isoftendifficulttodesignatedefiningpropertiesclearly,especiallywhereextremevaluesof other properties interfere, some measure of precautionary restriction is usuallynecessary.Itisnotoftenobviousthatitisbeingused.Wehavespokenofthenumberofwaysinwhichalevermaystimulatearatandthenumberofwaysinwhichtheratmayrespond.Weshouldfinditverydifficulttodefineeitheroftheseclasseswithoutconsiderable precautionary restriction of essentially non-defining properties—concerningthesizeoftheleverandsoon.Theuseofauniformleverfromexperimenttoexperimentisinitselfaconsiderableactofrestrictionandisapparentlynecessarytoassureaconsistentresult.Assigningthetermreflextotheentitiesinthispartofourstructuremeans,ofcourse,

that the reflex is a generic term. That is to say, the “stimulus” and the “response”entering into a given correlation are not to be identified with particular instancesappearinguponsomegivenoccasionbutwithclassesofsuchinstances. In thissensethe generic view has been borne out as against the autonomy of the completelyrestricted preparation. This is perhaps the most important characteristic of thedefinition. Freedom from the requirement of complete reproducibility broadens ourfieldofoperationimmeasurably.Wearenolongerlimitedtotheveryfewpreparationsinwhich some semblance of completeness is to be found, forwe are able to define“partsofbehaviorandenvironment”havingexperimental realityand reproducible intheirownfashion.Inparticular thebehaviorof the intactorganismismadeavailableforstudywithanexpectationofprecisioncomparablewiththatoftheclassicalspinalpreparation.(Indeed,ifsmoothnessofcurveistobetakenasanultimatecriterion,theintact organism often shows much greater consistency than the usual spinalpreparation,eventhoughthenumberofuncontrollednon-definingpropertiesismuchsmaller in the latter case. That is to say, the generic character is more marked inreflexespeculiartotheintactorganism.)

V

The generic nature of stimulus and response is in no sense a justification for thebroadertermsofthepopularvocabulary.Wemaylayitdownasageneralrulethatnoproperty isavaliddefiningpropertyofaclassuntil itsexperimental realityhasbeendemonstrated. This excludes a great many terms commonly brought into thedescriptionofbehavior.Forexample,supposethatitbecasuallyobservedthatachildhideswhenconfrontedwithadog.Thenitmaybesaid,inanuncriticalextensionofthe

453

terminologyofthereflex,thatthedogisastimulusandhidingaresponse.Itisobviousatoncethatthewordhidingdoesnotrefertoauniquesetofmovementsnordogtoaunique set of stimulating forces. In order tomake these terms validly descriptive ofbehavioritisnecessarytodefinetheclassestowhichtheyrefer.Itmustbeshownwhatproperties of a stimulus give it a place in the class “dog” and what property of aresponsemakesitaninstanceof“hiding.”(Itwillnotbeenoughtodignifythepopularvocabularybyappealingtoessentialpropertiesof“dogness”and“hidingness”andtosuppose them intuitively known.) The resulting classes will meanwhile have beenshown to be correlated experimentally, but it ought also to be shown that secondarychangesinthecorrelationarelawful.Itisnotatallcertainthatthepropertiesweshouldthusfindtobesignificantarethosenowsupposedlyreferredtobythewordsdog andhiding,evenafterallowingforthevaguenessinevitableinapopularterm.For reasons to be noted shortly, the existenceof a popular termdoes create some

presumptioninfavoroftheexistenceofacorrespondingexperimentallyrealconcept.Butthisdoesnotfreeusfromthenecessityofdefiningtheclassandofdemonstratingtherealityifthetermistobeusedforscientificpurposes.Ithasstilltobeshownthatmostofthetermsborrowedfromthepopularvocabularyarevalidlydescriptive—thatthey lead to a consistent and reproducible experimentation. We cannot legitimatelyassumethat“ridingabicycle,”“seeingone’sfriends,”or“heartbreak”areresponsesinanyscientificsense.Thisrestrictionupontheuseof thepopularvocabulary inbehaviorismisoftennot

felt because the partial legitimacy of the popular term frequently results in someexperimental consistency. The experimenter is more likely than not to hit uponexperimentally real terms, and hemay have some private set of properties resultingfromhisowntrainingwhichwillserve.Thusthewordhidingmayalwaysbeusedbyhim in connectionwith eventshaving certaindefiniteproperties, andhisown resultswillbeconsistentbyvirtueofthisdefinitionperaccidens.Butitisamistakeforhimtosuppose that thesepropertiesarecommunicated inhisuseof thepopular term. Ifnomoreaccuratesupplementaryspecificationisgiven,thedifficultywillbecomeapparentwheneverhisexperimentsarerepeatedbysomeonewithanothersetofprivatedefiningproperties andwill be the greater thewider the difference in backgroundof the twoexperimenters.Wearehereveryclosetoaprobleminepistemology,whichisinevitableinafieldof

this sort. For the relation of organism to environment with which we are primarilyconcernedmustbesupposed to include thespecialcaseof the relationofscientist tosubject-matter. If we contemplate an eventually successful general extension of ourmethods,wemustsupposeourselves tobedescribinganactivityofwhichdescribingitselfisonemanifestation.Itisnecessarytoraisethisepistemologicalpointinordertoexplain why it is that popular terms so often refer to what are later found to beexperimentallyrealentities.Thereasonisthatsuchtermsareinthemselvesresponsesofagenericsort:theyaretheresponsesofthepopulaceofwhichtheexperimenterisamember.Consequently,whentheorganismunderinvestigationfairlycloselyresemblesman (for example, when it is a dog), the popular term may be very close to theexperimentally real entity. We may hit immediately upon the right property of thestimulus,notbecausewehavemanipulateditexperimentallyinthemannerdescribedabove,butbecauseweourselvesreactinameasuresimilarlytothedog.Ontheotherhand, if theorganismis, letussay,anantoranamoeba, it ismuchmoredifficult todetectthe“real”stimuluswithoutexperimentation.Ifitwerenotforthisexplanation,

454

thepartiallegitimacyofthepopulartermwouldbeastrikingcoincidence,whichmightbe used (and indeed has been used) as an argument for the admission of a specialmethod(suchas“empathy”)intothestudyofbehavior.Ininsistingthatnoamountofrealityinthepopulartermsalreadyexaminedwillexcuseusfromdefininganewtermexperimentallyifitistobeusedatall,weareofcourserejectinganysuchprocess.Ourrule that the generic termmay be used onlywhen its experimental reality has beenverifiedwillnotadmitthepossibilityofanancillaryprinciple,availableinandpeculiarto the study of behavior, leading to the definition of concepts through some othermeansthanthesortofexperimentalprocedurehereoutlined.

VI

Throughoutthisdiscussionwehavekepttoourintentionofdealingwiththereflexanditsassociatedprocessessolelyatthelevelofbehavior.Wehavemadenoreferencetointermediating events in the central nervous system, and, here as elsewhere, this hasapparently not caused the slightest inconvenience. But the reader may feel that thepresentcasehasinvolvedaspecialdifficulty:thedefinitionwearriveatmayseemtobelogicalorstatisticalrather thanphysiological.Totakeasimpleexample,whatweshould call the defining property of a stimulus is actually the logical product of allobserved instances. This is easily said, so far as a consideration of the stimulus isconcerned; but when we come to deal with what this means in terms of a centralnervoussystem,itismuchmoredifficult.Theanswerofthestudentofbehavioroughttobethatthisisnothisproblem.Heis

interestedinasetofconceptsadaptedtothedescriptionofbehavior.Thenotionofaclassorofadefiningpropertyisjustifiedinadescriptionofthissortbecause,sofarasbehavior is concerned, all problems arising from its use are soluble. We havetechniques available for demonstrating defining properties, for showing the relativeimportance of non-defining properties, and formeasuring induction.The problemofdiscrimination,withitssubsidiaryproblemoftheestablishmentofnewclasses(or,inabroadsense,concepts),canbe formulatedequallywellwithout reference toacentralnervoussystem.Andifthesearerealaspectsofbehavior(ifnothinghasgonewronginouranalysis), theymustalsobeaspectsof theactivityof thecentralnervoussystem,which it is the business of the reflex physiologist to discover—through some othermeans, incidentally, than inference from behavior. This is a division of laborwhichoughttobeaspleasingtothephysiologistastothebehaviorist.Arigorousformulationof thepresentproblemat the levelofbehavior shouldbe themostdesirable startingpointforaphysiologicalstudyandisanecessaryconditionfortheeventualsynthesisofthetwofields.

VII.Summary

1. In breaking behavior and environment into parts for the sake of description, wecannottakeasingleinstanceoftheelicitationofaresponseasaunitbecauseitisnotafully reproducible entity. The usual solution of this problem through forcedsimplificationisinadequate.2.Inareflexpreparationtheobservedcorrelationisneverbetweenallpropertiesof

455

bothstimulusandresponse.Somepropertiesareirrelevant.The relevant properties are accordingly taken to define classes, and the reflex is

regardedasacorrelationofgenericterms.Thealternativeviewisthateverypossiblecorrelation of a particular stimulus and a particular response is a unit in itself. Oneargument against this alternative, frequently offered in support of the notion of“equivalence of stimuli,” is incomplete.A better proof of the generic view is baseduponthesecondarylawsofthereflex.Ifsmoothcurvescanbeobtainedinsecondarychanges while the stimulus and response vary in composition with respect to givenproperties, these properties may legitimately be regarded as non-defining. In theexamplesgiven,however,thenon-definingpropertiescanbeshowninseveralwaystobenotwhollyirrelevant.Thisfailureaddsweighttothealternativeview,butthedeviceusedtoexplainawaytheproofsforthegenericviewmakestheparticularunitequallydoubtfulasanautonomousentity.3. The observed facts are: (a) that there are defining properties (not rigorously

described)whichestablishgrossclassesofstimuliandresponses;(b)thatbyspecifyingotherpropertieswemaysetupotherandlesscomprehensiveclassesinaprogressiveseries or set of ramifications extending from the gross class to completely restrictedentities(thelatternotnecessarilyoperable);(c)thatbetweenanytwomembersofsuchafamilywemaydemonstrateinductionandshowthatitisafunctionofthedegreetowhich the entities possess their properties in common; (d) that in restricting apreparationwe obtain greater consistency of result because, from (c), we make twosuccessiveelicitationsmorelikelytoresembleeachother;but(e)thatsomepropertiesare largely irrelevant so far as induction is concerned, so that in the progressiverestriction of a preparation a pointmay be reached beyondwhich further restrictiondoesnotyieldanimprovementinconsistencyandmayyieldtheopposite.4. These phenomena, properly considered, lead to a definition of stimulus and

response.Bothextremesofaseriesofpreparationsarenonexperimental.Thereisonlyoneotherpointinsuchaseriesuniquelydetermined:thatatwhichsmoothcurvesforsecondaryprocessesareobtainable.Areflexisaccordinglydefinedasacorrelationofastimulusandaresponseatalevelofrestrictionmarkedbytheorderlinessofchangesinthecorrelation.Ifthisisnotanideallysimpledefinition,itisatleastinaccordwithourdata. It is based upon consistency of result rather than exact reproducibility; and itutilizesrestrictiononlyinmoderation.5.Thegenericnatureofstimulusandresponseisnotajustificationfortheuseofa

popular termuntil ithasbeendefinedexperimentally.Theobjection isnotoften feltbecause thepopular termmayhavesome legitimacy,due to the fact that the term isitself a generic response—of the populace. Its partial legitimacy is consequently nocoincidence,noranargumentfortheadmissionofaprinciplepeculiartothestudyofbehavior which will allow for the definition of concepts through other thanexperimentalmeans.

1TheoriginalpaperreferredheretoseveralexperimentalreportslaterincludedinTheBehaviorofOrganisms.

2Hereagain,referencewasmadetoexperimentslaterreportedinTheBehaviorofOrganisms.

456

TwoTypesofConditionedReflexandaPseudo-type

TheprincipleofconditioninghadbeenadvancedbyPavlov toexplainall learnedbehavior.ManyAmericanpsychologists,particularlythebehaviorists,hadcometousethe term in the same comprehensive sense. But an act acquired through whatThorndikecalled“LawofEffect” learningcouldnotbe interpretedasaconditionedreflexwithoutstrainingthePavloviannotionofsignalizationorsubstitutionofstimuli.The followingpaperpointedout thata reinforcementmayhave twokindsof effects.Bothof themmaybecalled learning,butonlyone follows thePavlovianpattern. Inlistingdifferencesbetweenthetwotypes,Iwasstilltryingtopreservethenotionofthereflex,with its eliciting stimulus. The typeswere distinguished, however, in termsoftemporal and other kinds of relations among observable events, and the distinctionsurvivedthelaterrevisiondiscussedonpage535.ThepaperappearedinTheJournalofGeneralPsychology(1935,12,66-77)andisreprintedbypermission.

Aconditionedreflex issaid tobeconditionedin thesenseofbeingdependentfor itsexistenceorstateupontheoccurrenceofacertainkindofevent,havingtodowiththepresentationofareinforcingstimulus.Adefinitionwhichincludesmuchmorethanthissimple notionwill probably not be applicable to all cases.At almost any significantlevel of analysis a distinction must be made between at least two major types ofconditioned reflex. Thesemay be represented, with examples, in the following way(whereS=stimulus,R=response,(S—R)=reflex,→=“isfollowedby,”and[]=“thestrengthof”theinclosedreflex):

Givensuchasequence,where[S1—R1]≠o,conditioningoccursasachangein[S0—R0]—anincreaseinstrength(positiveconditioning)in(a)andadecrease(negative

conditioning)in(b).1

GiventhesimultaneousorsuccessivepresentationofS′0andS′1,where[S′1—R′]≠

457

o,conditioningoccursasanincreasein[S′0—R′1].Differencesbetweenthetypesareasfollows:1.InTypeI,S0→R0→S1,whereR0necessarily intervenesbetween the stimuli; in

Type II, S′0→S′1, where R′0 is ignored.2 In I, R0 is important; it becomes the

conditionedresponse.InII,R′0isirrelevant(exceptwhenitisrelevantinanothersensebyconflictingwithR′1)andmayactuallydisappear.SinceconditioningofthesecondtypemaytakeplaceevenwhenS′1occursafterR′0,

ParadigmII,Example(C),maybewrittenforthiscaseasfollows:

whenitisidenticalwithI.Buttheresultisnottoreducethetwotypestoasingleform.Both kinds of conditioning proceed simultaneously but separately. If r is “turningtoward the light,” for example, and if the food iswithheld until turning takes place,[light—turning]will increaseaccordingtoIwhile[light—salivation]will increaseaccordingtoII.Thesameresult isobtainedwithnegativeconditioning.Example(D)gives,upondelayingS′1,

where [light — r] will decrease according to I, while [light — flexion] increasesaccordingtoII.In the special case inwhichR′0andR′1 are of the same form, the two kinds can

apparentlynotbeseparated.If,forexample,someunconditionedsalivationissupposedtobeelicitablebyalight,3wemaysubstitute“salivation”forr,toobtain

Both[light—salivationI]and[light—salivationII]willincrease,withapparentlynopossibledistinction.4Thisisaveryspecialcaseandisalsoinnosenseareductiontoasingletype.2.InI,(S0—R0)occursnormallyintheabsenceof(S1—R1),anditsstrengthmaybe

measured without interfering with the reinforcing action of S1. In II S′1 must bewithheld whenever a measurement of [S′0—R′1], the conditioned reflex, is taken,because S′1 also evokes R′1. Some amount of extinction necessarily ensues in thesecondcase.3.SinceS′1mustbewithheldinmeasuring[S′0—R′1],R′1mustbe independentof

any property of S′1 not possessed by S′0. In Example (C) salivation may becomeattached to the light as a conditioned responseofType II; but seizing, chewing, andswallowing, which are also responses to S′1, must not be included in the paradigmsincetheyrequirethepresenceofpartsofS′1whichcannotbesuppliedbyS′0.Aspecial restrictiononParadigmII is thereforenecessary.WhereS′0 isofavery

simplesort(atone,forexample),thepropertiespossessedinadditiontoS′0byS′1 arepractically equal to S′1, and we may express the restriction in terms of a generaldistinctionbetweentwokindsofresponses.Thefirstkindrequirenoexternalpointof

458

reference in theirelicitationordescription.Typicalexamplesare:glandularactivities(salivation),localmuscularresponses(flexion,wink,breathingmovements,productionofsounds),andfacilitationandinhibition.5Thesecondkindrequirepointsofreferencefortheirelicitationordescriptionwhich

arenotsuppliedbytheorganismitself,butbythestimulus.Examplesare:orientationtoward the source of a sound, approaching a light, and touching, seizing, andmanipulatingobjects(suchasaleverorfood).OurpresentruleisthatresponsesofthesecondkindcannotbesubstitutedforR′1 inParadigmII,unlessS′0 also supplies therequiredpointsofreference.4. In Type I, [S0—R0] ≠ o before conditioning takes place. The reflex-to-be-

conditionedmustbeelicitedatleastonceasanunconditioned“investigatory”reflex.InTypeII,[S′0—R′1]maybeginatzeroandusuallydoes.InTypeIthestateofthereflexis“conditioned”bytheoccurrenceofthereinforcingsequence,butitsexistenceisnot.A distinction between a conditioned and an unconditioned reflex is here lesssignificant,becauseallexamplesoftheformerhavenecessarilybeenexamplesofthelatter.Therearenoexclusivelyconditionedreflexesinthistype.Since[S′0—R′1]maybeginatzero,anewreflexmaybecreatedinconditioningof

thesecondtype.AndsincepracticallyanystimulusmaybeattachedtoR′1inParadigmII,averylargenumberofnewreflexescanthusbederived.ConditioningofTypeI,ontheotherhand,isnotadeviceforincreasingtherepertoryofreflexes;R0continuestobeelicitedbytheonestimuluswithwhichitbegan.TherearethreereflexesinParadigmII,butonlytwoinI.5. The significant change in Type I may be either an increase or a decrease in

strength;inTypeIIitisanincreaseonly,evenwhen[S′0—R′1]doesnotbeginatzero.In Type I stimuli may be divided into two classes, positively and negatively

conditioning,accordingtowhethertheyproduceanincreaseordecreasewhenusedasreinforcement.6 The distinction cannot be made in Type II, where a reflex may benegativeinanothersense(areflexof“avoidance,”forexample),butwhereitsstrengthonlyincreasesduringconditioning.6. InType I theconditioned reflex (S0—R0)maybeassociatedwithanydrive; in

TypeIIthereflex(S′0—R′1)isnecessarilyattachedtothedrivespecifiedbyR′1.Thispointmayrequiresomecomment.Inthepresentuseofthetermadriveisan

inferredvariableofwhichthestrengthofagroupofreflexesisafunction.7Hunger,forexample,isavariable(H)achangeinwhichisresponsibleforconcurrentchangesinthestrength(a)ofallunconditionedreflexesconcernedwiththeingestionoffood,(b)of all conditioned reflexes (of either type) in which the reinforcing stimulus isconcerned with the ingestion of food, and (c) to a much lesser extent of all“investigatory” reflexes. In Paradigm I, ExampleA (lever—pressing) is originally afunctionofHtosomeslightextentunder(c)above.AfterconditioningitvarieswithHaccordingto(b),overawiderangeprobablyequaltothatofanyunconditionedreflexunder(a).ConditioningofTypeIisreallythebecomingattachedtoagroupofreflexesvaryingasafunctionofsomedrive.Thisisamuchmorecomprehensivedescriptionoftheprocessthantodefineitasanincreaseinstrength,wherethedriveisassumedtoremain constant at a significant value. But the identity ofH in the present case isdeterminedonlybyourchoiceofareinforcingreflex.Given(S1—R1)ofanotherdrive,

459

say thirst, then (S0—R0) will become conditioned by attaching itself to the groupvaryingwiththirst,andwillnotvarywithHexcepttosomeslightextentunder(c).ThisisacharacteristicwhollylackinginTypeII.HereR′1 isoriginallypartof the

unconditionedreflexandthedrivetowhichitbelongsisdefinitelyfixed.7.Aminordifferenceisinthewayinwhichthestimulus-to-be-conditionedusually

acts.InTypeI,S0 isusuallypartofa larger field,andR0occursas the resultof theeventual prepotency of S0 over other stimuli. In Type II, S′0 is usually suddenlypresented to theorganism.Thesignificanceof thisdifference,which isnotabsolute,willappearlater.We shall nowconsider a third typeof relationwhich involves adiscrimination. It

maybebaseduponaconditionedreflexofeither type,butweshallbeginwithI.Toestablish a discrimination subdivide S0 into two classes on the basis of a selected

property or component member.8 For example, let the lever stimulate either in thepresence of a light (L), when the stimulus may be written as SAB..L.. (subscriptsindicatepropertiesorcomponents),orinthedark,whenthestimulusisSAB..Continueto reinforce the response to one of them, say SAB..L.., and extinguish or negativelyconditiontheresponsetotheotherbybreakingthesequenceatS1orbyintroducinganS1 of the negatively conditioning kind (Difference 5). When this has been done,[SAB..L..—R0]>[SAB..—R0].Andatanyvalueoftheunderlyingdrivesuchthat(SAB..L..—R0) is usually elicited but (SAB—R0) is not, there exists the following condition:givenanorganisminthepresenceofSAB..ordinarilyunresponsive,thepresentationofL will be followed by a response. For the sake of comparison we may set up aparadigminimitationofIIasfollows:

The relation between the light and the response to the lever might be called apseudo-conditioned reflex. It has some of the characteristics of Type II: the originalresponse to the light is irrelevant (Difference 1); the relationmay bewholly absentprior to the “conditioning” (Difference 4); it changes in a positive direction only(Difference5);andthe“stimulus”isusuallyofthepresentedkind(Difference7).Inallthese respects itdiffers fromType I, although theexample isbasedupona reflexofthattype.Inmanyotherrespectsitdiffersfrombothtypes.Areinforcingreflexisnotincludedintheparadigm,butmustbeaddedasathirdorfourthreflex.Theresponseisnotprincipallytothelight,buttothelever;thelightisonlyacomponentmemberofthewholestimulus,and“light—pressing”isnotlegitimatelytheexpressionofareflex.Thelevercannotbe removed toshowtheconditionedeffectivenessof the lightas inTypeII;instead,theresponsetotheleveralonemustbeextinguished—acharacteristicwehavenotmetbefore.Inspiteofthesedifferencesitisoftensaid(insimilarcases)thatthelightbecomes

the“conditionedstimulusfortheresponsetothelever”justasitbecomesthestimulusforsalivation.ThisisaconfusionwithTypeIIwhichobviouslyarisesfromaneglect

460

of the extinguished reflex. The relation of pressing the lever to the lever itself isignored and only the relation to the light taken into account.The lever comes to betreated, not as a source of stimulation, but as part of the apparatus, relevant to theresponse only for mechanical reasons. When the discrimination is based upon aresponsenotrequiringanexternalpointofreference(Difference3),thechanceofthisneglectincreasesenormously.Ifwesubstitute“flexionofaleg”for“pressingalever”(andcontinueforthemomentwithTypeI),S0inParadigmIisnotdirectlyobservable;wesimplywaituntilaflexionappears,thenreinforce.Havingestablished(S0—R0)asa conditioned reflex of some strength, we subdivide our inferred S0 as before,extinguish(SAB..—R0),andreinforce(SAB..L..—R0).Whenthediscriminationhasbeenset up, we have a condition in which the organism is ordinarily unresponsive butimmediatelyrespondswithflexionuponpresentationofthelight.OurinabilitytodemonstrateS0makesitdifficulttoshowthediscriminativenature

ofthisrelation;butitisbynomeansimpossibletofindothergrounds,aswemayseeby comparing it with a true reflex of Type II. Let the presentation of the light befollowedbyashocktothefootuntilthelightaloneelicitsflexion.Theresultingreflexissuperficiallysimilartotherelationoflightandflexionwhichwehavejustexamined,butfundamentallythetwocasesareunlike.Assumingthatnoimmediatedifferencecaninfactbedetected,9wemaystillshowdifferencesbyreferringforwardorbackwardtothehistoryoftheorganism.Thetworelationshavebeenestablishedindifferentwaysandtheircontinuedexistencedependsuponreinforcementfromdifferentstimuli.Thediscriminative relation also varies with an arbitrarily chosen drive, while theconditionedreflexisnecessarilyattachedtothedrivetowhichshock-flexionbelongs.These differences are chiefly due, however, to the use of a conditioned reflex of

Type I in setting up the discrimination. In a pseudo-conditioned reflex based uponType II thedistinction ismuch less sure.Herewe are invariably able to neglect theextinguished member because R′1 is of the kind not requiring an external point ofreference(Difference3),andwecanminimize its importanceinotherways.Givenaconditionedreflexofthiskind:

ifweestablishadiscriminationbetweenthetoneandthetone-plus-a-light(reinforcingthe response to the latter), we obtain the following condition: an organism in thepresence of the tone, ordinarily unresponsive,will respond upon presentation of thelight. The only difference between this relation and a true reflex of Type II is theextinctionoftheresponsetothetone,whichisevidencethatadiscriminationhastakenplace.Thereinforcementoftoneandlightshouldconditionresponsestobothofthesestimuli;butweobserve that theorganismisunresponsive in thepresenceof the tonealone.Now,thissurvivingdifferencemaybereducedatwillbyreducingthesignificance

ofS′0 in thebasic reflexof thepseudo-type. Ifwe lower the intensityof the toneorchooseanother stimulusofa less importantkind,wemayapproachascloselyaswe

461

pleasetoaconditionedreflexofTypeII.WecannotactuallyreachTypeIIinthisway,butwecaneasilyreachapointatwhichourpseudo-reflexisidenticalwithanyactualexperimentalexampleofthattype.Thisistruebecausesomeamountofdiscriminationis practically always involved in cases ofType II.Whenwe put a dog into a stand,presentalightandthenfood,thefoodreinforcesnotonlythelightbutthestimulationfrom the stand.Merely putting the dog into the stand again should elicit salivationaccording to Paradigm II. In practice this is a disturbing effect, which must beeliminatedthroughextinction.Solongasitoccurs,anyactualcaseofTypeIImustbeformulated as a pseudo-conditioned reflex. If SG is the stimulation affecting theorganisminadditiontoS0,thenS0inParadigmIIshouldreadSG+S0.TheeffectuponSGisextinguishedthroughlackofreinforcementintheabsenceofS0,andtheresultisadiscrimination:anorganisminthepresenceofSG,ordinarilyunresponsive,respondswhen S0 is added. The importance of this criticism will depend upon the relativemagnitudesofSG andS0. In the optimal experiment SGmay be reduced to a valueinsignificantincomparisonwithordinaryvaluesofS0.The partially discriminative nature of Type II is inevitable. It is not important in

TypeIbecauseofDifference1.ParadigmIcontainsanimplicitspecificationthatS0 isactiveorhasjustactedatthemomentofreinforcement,sinceitspecifiesthatS1 is tobewithhelduntilR0hasoccurred.ThereinforcedstimulusisreallyS0andnotSG+S0(it is the lever, in our example, not the whole stimulating field presented by theapparatus). Paradigm II contains no specification of the activity of S0; and thereinforcingactionofS′1mustbesupposedtoextendtoSGaswellastoS0.Inpracticean active state at the moment of reinforcement is usually insured by presenting S0suddenly.10ThismightbeincludedasanadditionalprovisioninParadigmII,buttheprovision really required is that S0, and no part of SG, be active at the moment ofreinforcement. This is not easily arranged.We cannot wholly avoid the generalizedactionofthereinforcementinTypeIIbecauseofthelackofdependenceofS′1uponR′0.Onecharacteristicofthepseudo-conditionedreflexisthevarietyoftheformsofits

“stimulus.”Wehaveassumedthatinourtwofundamentalparadigmsanystimulushadultimatelythedimensionsofenergy(althoughwehaveoftenusedtheshorthanddeviceofspeakingofthesourceoftheenergy—as,forexample,“lever”).Inthepseudotype,however,the“stimulus”canbeasingleproperty.Itcanbetheintensityofthestimulus,orsomesuchqualitativeaspectaspitchorhue.Itcanbeachangefromonevalueofapropertytoanother,ortheabsenceofaproperty,oraduration.Thereasonwhythisispossible is that the other properties of the stimulus can be relegated to SG forextinction.Ifthepitchofatoneistobeaconditioned“stimulus,”thetoneitselfmustfirstbecomeonealso,andtheresponsetoitsotherpropertiesmustbeextinguishedbyextinguishingtheresponsestotonesofotherpitches.Inatrueconditionedreflexthiscannotbedone.Althoughitiscommontospeakofpropertiesasstimuli[asPavlov,forexample, does], the presence of a property in the position of a stimulus is a certainindication that a pseudo-conditioned reflex is really in question. A property alonecannot be used in either true type because it implies extinction; most of the realstimulus must be relegated to SG, and the requirement that the value of SG be

462

negligiblecannotthereforebesatisfied.Thepositionofapseudo-conditionedreflexmaybesummarizedasfollows.When

thepseudo-reflexisbaseduponareflexofTypeIandwhenR0requiresexternalpointsof reference, there are important practical and theoretical reasons why a separateformulationisdemanded.WhenR0doesnotrequireexternalpointsofreference,therearefewerdifferences,butaseparateformulationisstillnecessary.Whenthepseudo-typeisbaseduponareflexofTypeII, thedistinctionisweakenedbutshouldstillbemade,exceptwhenSGcanbereduced toavery lowvaluerelative toS′0. In the lastcase a practical distinction is impossible, not because of an identity of types, butbecauseofthefailureofTypeIItoappearexperimentallyinapureform.It isa temptinghypothesis thatII isnotanauthentic typebutmaybereducedtoa

discriminationbasedonTypeI.Butthishasnotbeenshown;wehavenotreducedthepseudo-type to Type II or viceversa. Nor have we come very near it. The presentpseudo-reflex which resembles II most closely requires a reflex of that type for itsestablishment.ItisprobablymorethanacoincidencethatadiscriminationbaseduponTypeIhassomanyof thepropertiesofII,but thereductiontoasingle typeappearsfromourpresentevidencetobehighlyimprobable,desirablethoughitwouldbeasanimmensesimplification.Thedifferenceswehavenotedarenoteasilydisposedof.StillmoreimprobableisareductionofI toII,sincethefirststepsuppliedbythepseudo-typeisthenlacking.Tothedifferenceswehavelistedmightbeaddeddifferencesinthepartsplayedby

the two types in the economy of the organism. The essence of Type II is thesubstitution of one stimulus for another, or, as Pavlov has put it, signalization. Itprepares the organism by obtaining the elicitation of a response before the originalstimulushasbeguntoact,anditdoesthisbylettinganystimuluswhichhasincidentlyaccompaniedoranticipatedtheoriginalstimulusactinitsstead.InTypeIthereisnosubstitutionofstimuliandconsequentlynosignalization.TypeIacts inanotherway:theorganismselects froma large repertoryofunconditioned reflexes thoseofwhichthe repetition is important with respect to certain elementary functions and discardsthoseofwhichitisunimportant.TheconditionedresponseofTypeIdoesnotprepareforthereinforcingstimulus,itproducesit.Thestimulus-to-be-conditionedisneverinanysenseincidental.TypeIplaysthemoreimportantrole.Whenanorganismcomesaccidentally(thatis

tosay,astheresultofweakinvestigatoryreflexes)uponanewkindoffood,whichitseizes and eats, both kinds of conditioning presumably occur. When the visibleradiationfromthefoodnextstimulatestheorganism,salivationisevokedaccordingtoParadigmII.Thissecretionremainsuselessuntilthefoodisactuallyseizedandeaten.Butseizingandeatingwilldependupon thesameaccidental factorsasbeforeunlessconditioningofTypeIhasalsooccurred—thatis,unlessthestrengthofthereflex(food—seizing) has increased. Thus, while a reflex of Type II prepares the organism, areflex of Type I obtains the food for which the preparation ismade. And this is ingenerala faircharacterizationof the relative importanceof the two types.AsPavlovhassaid,conditionedstimuliareimportantinprovidingsalivabeforefoodisreceived,but “even greater is their importance when they evoke themotor component of thecomplex reflex of nutrition, i.e., when they act as stimuli to the reflex of seekingfood.”11Although“thereflexofseekingfood”isanunfortunateexpression, it refersclearlyenoughtobehaviorcharacteristicofTypeI.

463

1Later,inTheBehaviorofOrganisms,therealityofadecreasein(b)wasquestioned.2ForconvenienceweomitthecaseofsimultaneousstimuliinTypeII.3SeeDifference4belowforthisgeneralrequirementinTypeI.4InTheBehaviorofOrganismsthesuitabilityofanautonomicresponsesuchassalivationforwhatisherecalledTypeIconditioningwasquestioned.

5WhereconditionedfacilitationandinhibitionaredefinedbysubstitutingforR′inParadigmIItheexpressions“Incr.[S″—R″]”and“Decr.[S″—R″]”respectively.(Incr.=“increasein”;Decr.=“decreasein”).

6Seefootnote1onpage525.7Skinner,B.F.Driveandreflexstrength:I.J.Gen.Psychol.,1932,6,22–37.8ReferencewasoriginallymadeheretoapaperdescribingexperimentsondiscriminationlaterreportedinTheBehaviorofOrganismsandtothepaperonpage504.

9ThisisagenerousassumptionsincesomeevidenceforthepresenceofS0canusuallybefound.Adifferenceinthecharacteroftheresponsemightalsobeshown(inthecaseofthetruereflexitmaybeaccompaniedbychangesinbreathingrate,forexample,whichwouldbelackinginthepseudo-reflex).

10ThisisourexplanationofDifference7.Anotherexplanationmightbeadded.IfS′0isactiveforanylengthoftimepriortoS1itwillhaveanextinguishingeffect.ThiscannotbesaidofTypeI.

11Pavlov,I.P.ConditionedReflexes.Trans.&ed.byG.V.Anrep.London:OxfordUniv.Press.1927,p.13.ThisisadoublyinterestingstatementbecausePavlovhasconfinedhisowninvestigationspracticallyexclusivelytoconditionedreflexesofthesecondtype.Itoughttobesaidthatheusuallyregardsthistypeasadequateforthewholefield.Thushesaysthatthe“functionofthehemispheres”issignalization(p.17),althoughsignalizationis,aswehaveseen,acharacteristicofTypeIIonly.

464

TwoTypesofConditionedReflex:AReplytoKonorskiandMiller

A distinction between two types of conditioned reflex had been made as early as1928bytwoPolishphysiologists,JerryKonorskiandS.Miller,withwhoseworkIwasnot familiar when the preceding paper was written. In The Journal of GeneralPsychology (1937, 16, 264-272) they questioned the characterization of my Type I,citing several of their experiments. In one of these a dog placed in an experimentalstandand shocked lightly in one foot eventually lifted the foot before the shockwasadministered. In another the same movement in response to electric shock wasreinforcedwith food and eventually occurredwithout the shock. The principal issuewasthestatusofthestimulusoriginallyresponsiblefortheconditionedmovement.TwoyearsintervenedbetweenmypaperandtheKonorskiandMillercommentonit.

Myanswer(whichappearedinthesameissue,thankstothekindnessoftheauthorsinprovidingmewith an advance copy) referred to“awork now in preparation.”ThiswasTheBehaviorofOrganisms,inwritingwhichIhadalreadyabandonedthenotionofanelicitingstimulusintheanalysisofoperantbehavior.Myanswerwas,therefore,lessacommentonthepointsraisedbyKonorskiandMillerthanarevision-in-progressofmy earlier formulation. In it the operant-respondent distinctionwasmade for thefirst time. The paper is reprinted by permission from The Journal of GeneralPsychology (1937,16,272-279). A translation of the paper byMiller andKonorski(“OnaParticularFormofConditionedReflex”)maybefoundinTheJournaloftheExperimentalAnalysisofBehavior,1969,12,187–189.

Before considering the specificobjections raisedbyKonorski andMiller againstmyformulation of a second type of conditioned reflex, I should like to give a morefundamentalcharacterizationofbothtypesandofthediscriminationbaseduponthem.Let conditioning be defined as a kind of change in reflex strength where the

operationperformedupon theorganism to induce thechange is thepresentationofareinforcingstimulusinacertaintemporalrelationtobehavior.Allchangesinstrengthso induced come under the head of conditioning and are thus distinguished fromchangeshavingsimilardimensionsbutinducedinotherways(asindrive,emotion,andsoon).Differenttypesofconditionedreflexesarisebecauseareinforcingstimulusmaybepresentedindifferentkindsoftemporalrelations.Therearetwofundamentalcases:inonethereinforcingstimulusiscorrelatedtemporallywitharesponseandintheotherwithastimulus.For“correlatedwith”wemightwrite“contingentupon.”ThesearethetypesIhavenumberedIandIIrespectively.KonorskiandMillerrefertothesecondasType I and to a complex case involving the first (see below) as Type II. To avoidconfusionandtogainamnemonicadvantageIshallrefertoconditioningwhichresultsfromthecontingencyofareinforcingstimulusuponastimulusasofTypeSandtothatresultingfromcontingencyuponaresponseasofTypeR.Ifthestimulusisalreadycorrelatedwitharesponseortheresponsewithastimulus,

465

areinforcementcannotbemadecontingentupontheonetermwithoutbeingputintoasimilar relationwith the other. That is to say, if a reinforcing stimulus is correlatedtemporarilywiththeSinareflex,itisalsocorrelatedwiththeR,orifwiththeR,thenalsowiththeS. It isnotpossible toavoid thisdifficulty (whichseems todestroy thevalidityof the foregoingdefinition)byspecifyingakindof temporal relation. If, forexample,we should distinguish between the cases inwhich the reinforcing stimulusprecedesS(andhencealsoprecedesR)andthoseinwhichitfollowsR(andhencealsofollowsS), the resulting classeswould be close to those of TypesR and S but theywouldnotbeidenticalwiththem,andthebasisfor thedefinitionwouldnotpermitadeductionoftheothercharacteristicsofthetypes.Thecontingencyofthereinforcingstimulusuponaseparatetermisnecessary.It may be noted, therefore, that in both paradigms of conditioning as previously

given[page525]theconnectionbetweenthetermtobecorrelatedwiththereinforcingstimulusandanothertermisirrelevant.Noconnectionneedexistatthestart.InTypeSwemayuseastimulus(S0)elicitingnoobservableresponseandinTypeRaresponse(R0) elicitedbynoobservable stimulus (for example, the “spontaneous” flexionof aleg).Or,ifaconnectionoriginallyexists,itmaydisappearduringconditioning.InTypeS,ifS0elicitsadefiniteresponse[say,where(S0—R0)is(shock—flexion)],R0maydisappear(Eroféeva);andinTypeR,ifR0isapparentlyelicitedbyadefinitestimulus[say,where(S0—R0)isthesame],R0willeventuallyappearwithoutS0,asKonorskiandMillerhaveshown.Theparadigmsmaythereforeberewrittenasfollows:

wherethearrowsindicatethetemporalcorrelationresponsibleforconditioning,1andwhere the terms written in lower case either (a) cannot be identified, (b) may beomitted, or (c) may disappear. The correlation of the reinforcing stimulus with aseparatetermishereachievedandfromitthepropertiesoftwo(and,incidentally,onlytwo) typesofconditioned reflexmaybededuced.Thedifferencesbetween the typesgiveninmypaper,whichneednotberepeatedhere,arenolongeruseful indefiningthetypes,buttheyserveasconvienenthallmarks.Thissolutiondependsuponthestatementthatthereareresponsesuncorrelatedwith

observablestimuli—astatementwhichmustnotbemadelightlybutcannot,sofarasIcansee,beavoided.Itisanecessaryrecognitionofthefactthatintheunconditionedorganism two kinds of behavior may be distinguished. There is, first, the kind ofresponsewhichismadetospecificstimulation,wherethecorrelationbetweenresponseand stimulus is a reflex in the traditional sense. I shall refer to such a reflex as arespondentandusethetermalsoasanadjectiveinreferringtothebehaviorasawhole.Butthereisalsoakindofresponsewhichoccursspontaneouslyintheabsenceofanystimulationwithwhichitmaybespecificallycorrelated.Weneednothaveacomplete

466

absenceof stimulation inorder todemonstrate this. It doesnotmean thatwecannotfindastimuluswhichwillelicitsuchbehaviorbutthatnoneisoperativeatthetimethebehavior is observed. It is the nature of this kind of behavior that it should occurwithoutanelicitingstimulus,althoughdiscriminativestimuliarepracticallyinevitableafter conditioning. It is not necessary to assume specific identifiable units prior toconditioning,butthroughconditioningtheymaybesetup.Ishallcallsuchaunitanoperantandthebehavioringeneral,operantbehavior.Thedistinctionbetweenoperantandrespondentbehaviorandthespecialpropertiesoftheformerwillbedealtwithatlength in a work now in preparation. All conditioned reflexes of Type R are bydefinition operants and all of Type S, respondents; but the operant-respondentdistinctionisthemoregeneralsinceitextendstounconditionedbehavioraswell.Aformulationofthefundamentaltypesofdiscriminationmayalsobecarriedoutin

terms of the contingency of the reinforcing stimulus. Discrimination differs fromconditioningbecausetheexistingcorrelationcannotbeunequivocallyestablishedwithany one set of properties of the stimulus or response. The effect of a given act ofreinforcement is necessarilymore extensive than the actual contingency implies, andthe relation must be narrowed through extinction with respect to the properties notinvolvedinthecorrelation.Therearethreebasictypesofdiscrimination.1.DiscriminationoftheStimulusinTypeS.a.S1iscontingentuponlessthanalltheaspectsorpropertiesofS0presentuponany

givenoccasionofreinforcement.Forexample,letS1becontingentuponthepitchofatone.Beforethisrelation(andnotmerelyarelationbetweentheresponseandthetoneitself) can be established, responses to tones of other pitches which have beenconditionedthroughinductionmustbeextinguished.b.S1iscontingentuponagroupofstimulibutnotuponsubgroupsorsupergroups.

For example, let SA and SB be reinforced together but not separately. Before therelationwillbereflectedinbehavior,theresponsestoeitherstimulusalonewhicharestrengthenedthroughinductionmustbeextinguished.2.DiscriminationoftheStimulusinTypeR.S1iscontingentuponR0inthepresence

ofa stimulusSD.Forexample, let thepressingofa leverbe reinforcedonlywhenalightison.Beforethisrelationcanbeestablishedinthebehavior,theresponsesintheabsence of the light developed through induction from the reinforcement in thepresenceofthelightmustbeextinguished.3.DiscriminationoftheResponseinTypeR.2S1iscontingentuponanR0havinga

givenvalueofoneormoreofitsproperties.Forexample,letS1becontingentuponaresponse above a given level of intensity.Responses of lower intensity strengthenedthroughinductionmustbeextinguished.(ThereisnofourthcaseofadiscriminationoftheresponseinTypeS.)Bothdiscriminationsofthestimulus(butnotthatoftheresponse)yieldwhatIhave

calledpseudo-reflexes,inwhichstimuliarerelatedtoresponsesinwayswhichseemtoresemblereflexesbut requireseparate formulations ifconfusion is tobeavoided.3 InTypeS(Caseb,above),giventheorganisminthepresenceofSA, thepresentationofSBwillbe followedbya response.Thesuperficial relation (SB—R) isnota reflex,becausetherelevanceofSAisoverlooked.SimilarlyinTypeR,thesuperficialrelationbetween the light and pressing the lever is not a reflex and exhibits none of the

467

propertiesofonewhenthesearetreatedquantitatively.The distinction between an eliciting and a discriminative stimuluswas notwholly

respected in my earlier paper, for the reflex (lever—pressing) was pseudo. As adiscriminatedoperantthereflexshouldhavebeenwritten(s+lever—pressing).SinceI did not derive the two types from the possible contingencies of the reinforcingstimulus, it was not important that R0 in Type R be independent of an elicitingstimulus. But the treatment of the lever as eliciting an unconditioned response hasprovedinconvenientandimpracticableinotherways,andtheintroductionofthenotionof the operant clears up many difficulties besides those immediately in question. Iteliminates the implausible assumption that all reflexes ultimately conditionedaccordingtoTypeRmaybespokenofasexistingasidentifiableunitsinunconditionedbehavior and substitutes the simpler assumption that all operant responses aregenerated out of undifferentiated material. Certain difficulties in experiments uponoperants are also avoided. Operant behavior cannot be treated with the techniquedevisedforrespondents(SherringtonandPavlov)becauseintheabsenceofanelicitingstimulus many of the measures of reflex strength developed for respondents aremeaningless. In an operant there is properly no latency (except with respect todiscriminative stimuli), no after-discharge, andmost important of all no ratio of themagnitudesofRandS.Inspiteofrepeatedeffortstotreatitassuch,themagnitudeoftheresponseinanoperantisnotameasureofitsstrength.Someothermeasuremustbedevised, and from the definition of an operant it is easy to arrive at the rate ofoccurrenceoftheresponse.Thismeasurehasbeenshowntobesignificantinalargenumberofcharacteristicchangesinstrength.There is thus an important difference between the Konorski andMiller sequence

“shock—flexion→food”andthesequence“s+lever—pressing→food.”Thefirstcontainsarespondent,thesecondanoperant.Theimmediatedifferenceexperimentallyisthatinthesecondcasetheexperimentercannotproducetheresponseatwillbutmustwait for it to come out. A more important difference concerns the basis for thedistinction between two types. Since there is no eliciting stimulus in the secondsequence,thefoodiscorrelatedwiththeresponsebutnotwiththeleverasastimulus.Inthefirstsequencethefoodiscorrelatedasfullywiththeshockaswiththeflexion.The Konorski and Miller case does not fit the present formula for Type R, and adivergent result neednotweighagainst it.Thecasedoesnot, as amatterof fact, fiteither type so long as the double correlationwith both terms exists.ConditioningofTypeSwilloccur(theshock-salivationreflexofEroféeva),butthereisnoreasonwhyconditioning of Type R should occur so long as there is a correlation between thereinforcement and an eliciting stimulus.Nothing is to be gained in such a case; theoriginalsequenceoperatesasefficientlyaspossible.ThecasecomesunderTypeRonlywhenthecorrelationwithS0isbrokenup—that

is, when a response occurs which is not elicited by S0. The complex experimentdescribedbyKonorskiandMillermaybeformulatedasfollows.Intheunconditionedorganism there is operant behaviorwhich consists of flexing the leg. It isweak andappearsonlyoccasionally.Thereisalsothestrongrespondent(shock—flexion),whichhasmoreorlessthesameformofresponse.InKonorskiandMiller’sexperimentwemayassumethatanelicitationoftherespondent(S—R)bringsoutat thesametimetheoperant(s—R),whichsumswithit.Wehaveinrealitytwosequences:(A)shock—flexion,and(B)s— flexion→food.Here the respondent (A)neednot increase in

468

strengthbutmayactuallydecreaseduringconditioningofTypeS,whiletheoperantinBincreasesinstrengthtoapointatwhichitiscapableofappearingwithouttheaidofA.AsKonorskiandMillernote, “…the stimulusS0 [shock]playsonlya subsidiaryrole in the formationofaconditioned reflexof thenew type. It servesonly tobringabouttheresponseR0[bysummationwiththeoperant?],andoncetheconnectionSG— R0 is established [read ‘once the operant is reinforced’], it loses any furtherexperimentalsignificance.”TheexistenceofindependentcompositepartsmaybeinferredfromthefactsthatB

eventuallyappearswithoutAwhenithasbecomestrongenoughthroughconditioningandthatitmayevenbeconditionedwithouttheaidofA,althoughlessconveniently.KonorskiandMillerseemtoimplythataschemewhichappealstothespontaneous

occurrence of a response cannot be generally valid because many responses neverappearspontaneously.Butelaborateandpeculiarformsofresponsemaybegeneratedfrom undifferentiated operant behavior through successive approximation to a finalform.Thisissometimestrueoftheexampleofpressingthelever.Aratmaybefound(very infrequently)not topress the leverspontaneouslyduringaprolongedperiodofobservation. The response in its final form may be obtained by basing thereinforcementuponthefollowingstepsinsuccession:approachtothesiteofthelever,liftingthenoseintotheairtowardthelever,liftingthefore-partofthebodyintotheair,touchingtheleverwiththefeet,andpressingtheleverdownward.Whenonestephasbeenconditioned,thereinforcementiswithdrawnandmadecontingentuponthenext.Withasimilarmethodanyvalueofasinglepropertyoftheresponsemaybeobtained.Theratmaybeconditionedtopresstheleverwithaforceequaltothatexertedby,say,100grams(althoughspontaneouspressingsseldomgoabove20grams)ortoprolongtheresponseto,say,30seconds(althoughtheleverisseldomspontaneouslyhelddownfor more than two seconds). I know of no stimulus comparable with the shock ofKonorskiandMillerwhichwillelicit“pressingthelever”asanunconditionedresponseorelicititwithabnormalvaluesofitsproperties.ThereisnoS0availableforelicitingtheseresponsesinthewaydemandedbyKonorskiandMiller’sformulation.Where an eliciting stimulus is lacking, Konorski and Miller appeal to “putting

through.” A dog’s paw is raised and placed against a lever, and this “response” isreinforced with food. Eventually the dog makes the response spontaneously. But agreat dealmay happen herewhich is not easily observed. Ifwe assume that tensionfrompassiveflexionistosomeextentnegativelyreinforcing,anythingwhichthedogdoeswhichreduces the tensionwillbe reinforcedasanoperant.Suchaspontaneousresponseasmovingthefootinthedirectionofthepassiveflexionwillbereinforced.Wethushaveaseriesofsequencesofthisgeneralform:

s+SD(touchingandflexingofleg)—R(movementoflegincertaindirection)→S1(reliefoftension).

The effect of “putting through” is to provide step-by-step reinforcement for manycomponent parts of the complete response, each part being formulated according toTypeR.Thesubstitutionoffoodasanewreinforcementiseasilyaccountedfor.This interpretation of “putting through” is important becauseKonorski andMiller

base their formulation of the new type upon the fact that proprioceptive stimulationfrom the responsemay become a conditioned stimulus of Type S since it regularly

469

precedesS1.Oneof theconditions for this second type is that“themovementwhichconstitutesitseffectisaconditionedfoodstimulus.”ThatconditioningofthissortdoestakeplaceduringconditioningofTypeRwasnotedinmypaper,butitsrelevanceintheprocessofTypeRdoesnotfollow.Perhapsthestrongestpointagainstitisthefactthat conditioning of Type R may take place with one reinforcement, where aprerequisite conditioning of Type S could hardly have time to occur. AnyproprioceptivestimulationfromR0actsasanadditionalreinforcementintheformulaforTypeR.Whereit ispossibletoattachconditionedreinforcingvaluetoaresponsewithoutelicitingit,thereinforcementisaloneinitsaction,butthecasestillfallsunderType R. In verbal behavior, for example, we may give a sound reinforcing valuethroughconditioningofTypeS.Anysoundproducedbyachildwhichresemblesitisautomaticallyreinforced.Thegeneralformulaforcasesofthissortis

s—R0→stimulationfromR0actingasaconditionedreinforcement.

Iassumethatthisisnotaquestionofpriority.ThebehaviorcharacteristicofTypeRwasstudiedasearlyas1898(Thorndike).Thepointatissueistheestablishmentofthemost convenient formulation, and Imay list the following reasons for preferring thedefinitionof typesgivenherein. (1)Aminimalnumberof terms is specified.This isespecially important in Type R, which omits the troublesome S0 of Konorski andMiller’s formula. (2) Definition is solely in terms of the contingency of reinforcingstimuli—otherpropertiesofthetypesbeingdeducedfromthedefinition.(3)Noothertypesaretobeexpected.WhatKonorskiandMillergiveasvariantsorpredictasnewtypesarediscriminations.(4)Thedistinctionbetweenanelicitingandadiscriminativestimulus is maintained. Konorski andMiller’s variants of their Type II are pseudo-reflexes and cannot yield properties comparable with each other or with genuinereflexes.Two separate points may be answered briefly. (1) It is essential in this kind of

formulationthatonereflexbeconsideredatatimesinceourdatahavethedimensionsof changes in reflex strength. The development of an antagonistic response when areinforcementinTypeRisnegativerequiresaseparateparadigm,eitherofTypeRorTypeS.(2)ThatresponsesofsmoothmuscleorglandulartissuesmayormaynotenterintoTypeR,Iamnotpreparedtoassert.Iusedsalivationasaconvenienthypotheticalinstanceofsimultaneousfusedresponsesofbothtypes,butaskeletalresponsewouldhavedoneaswell.Thechildwhohasbeenconditionedtocry“realtears”becausetearshavebeenbeen followedbypositive reinforcement (e.g., candy) apparentlymakes aglandularconditionedresponseofTypeR,butthematterneedstobecheckedbecausean intermediatestepmaybe involved.Such is thecase in theHudgins’experiment,4where the verbal response “contract” is an operant but the reflex (“contract”—contractionofpupil)isaconditionedrespondent.Thequestionatissueiswhetherwemayproducecontractionofthepupilaccordingtos—contraction→reinforcement,where(forcaution’ssake)thereinforcingstimuluswillnotitselfelicitcontraction.Itisaquestionforexperiment.

1TheuncommoncaseinwhichS1followsS0isaminorexceptiontothedirectionofthearrow,whichmaybeaccountedforwiththenotionofthetrace.

470

2Laterreferredtoas“differentiation”ofR.3Notallpseudo-reflexesarediscriminative,ifweextendthetermtoincludeallsuperficialcorrelationsofstimulusandresponse.Forexample,letatetanizingshocktothetailofadogbediscontinuedassoonasthedogliftsitsleftforeleg.Thediscontinuanceofanegativereinforcementactsasapositivereinforcement;andwhenconditioninghastakenplace,ashocktothetailwillbeconsistentlyfollowedbyamovementoftheforeleg.Superficiallytherelationresemblesareflex,butthegreatestconfusionwouldarisefromtreatingitassuchandexpectingittohavetheusualproperties.

4Hudgins,C.V.Conditioningandthevoluntarycontrolofthepupillaryreflex.J.Gen.Psychol,1933,8,3–51.

471

AReviewofHull’sPrinciplesofBehavior

WhenClarkL.Hull first approachedbehavior theory inearnest,more thanadecadeago, thescienceofbehaviorwas inadifficultposition.Twentyyearsof the“naturalsciencemethod”heraldedbyBehaviorismhadfailedtoprovideaconsistentandusefulsystematicformulation.Thecommonestlaboratoryinstrumentswerestillthemazeandthediscriminationbox,andexperimentaldatareflectedmanyarbitrarypropertiesoftheapparatus. Acceptable conclusions of any degree of generality referred to aspects,characteristics, or limiting capacities.While many of these were valid enough, fewwere logically compelling, and individual preferences had led to many individual“sciences”ofbehavior.Hull saw theneed fora logicalanalysisof thevocabularyofbehavior, for an explicit definition of fundamental terms and an unambiguousstatement of principles. He adopted the rather extreme procedures of “postulatetheory,”andproposed toputorder into thescienceofbehaviorbycarefullydefiningcertainprimitivetermsandsettingupa(necessarilyratherlarge)numberofpostulates,fromwhich thousandsof theoremscouldbededucedandexperimentally tested. It isclearfromhisearlymemorandathathefirstregardedhispostulatesasingeneralnotdirectly testable, a condition which gives point to the postulate method. Deducedtheoremsweretobecheckedagainstthefacts,butthevalidityofapostulatewastobeestablishedbythesuccessofthetheoremstowhichitgaverise.The cogency of the postulate method may be said to vary inversely with the

accessibility of a subjectmatter.When a process cannot be directly investigated, itsproperties may often be inferred from a study of consequences. This conditionprevailedinthefieldofbehaviorintheearly30’s,andHull’sproposalwasthereforejustified.Thesituationwas,however,changing.ThegrowinginfluenceofPavlovandotherdevelopments in the fieldat this timerevealed thepossibilityofadirectattackuponfundamentalprocesseswhichwouldleadtoadifferentsortofanalysis.Hullwasone of the first to recognize this possibility and to encourage relevant research.Moreover, he was not insensitive to its bearing upon postulational procedures. In amemorandum of November 28, 1936, he confesses that “the ‘geometrical’ type ofdeductiondoes not permit the readyuse of the calculus and thus has limitations notcharacteristicofthehigherformsofscientifictheory.”Insteadofabandoningthepostulatemethodandturningtothehigherformsoftheory

appropriate to a functional analysis, Hull attempted to salvage his program bycombining the twomethods. In 1937hewrote that “the postulates of the theoreticalconstructsaretobethebasiclawsorprinciplesofhumanbehavior.Theselawsaretobe determined directly by experiment, so far as feasible, in advance of their use intheoretical constructs…. The experimental determinations will be quantitative,designedtorevealfunctionalcurvesofbasicrelationships.Mathematicalequationswillbefittedto…thesecurves;suchequationsconstitutethepostulatesofthesystem.”Thisprogram,whichhecalled“logicalempiricism,”provedtobeanunstableblendoftwowidelydifferentprinciplesofanalysis.Anydemonstratedfunctional relationbetween

472

behavior and its controlling variables is not a postulate but a law, and there is littlereason tocontinuewith the ritualofpostulates.The resulting split inHull’s thinkinghasleftitsmarkonhisPrinciplesofBehavior.Atonetimeheappearstobeworkingwithinaformalpostulatesystem;atanotherheisconsideringbehaviorasadependentvariableandrelatingittocontrollingvariablesintheenvironment.Theseactivitiesarenotalwayssupplementary;onthecontrarytheyareoftenmutuallyharmful.The transition from“principle” to “process,”or fromaspect to functional relation,

maybetracedinthechangingpostulatesystemswhichHullhassponsored.Hebegan,in the caseof rote learning,with a selected setof currenthypotheses, assembled forpurposesofclarificationandheldtogetherbyacommonrelationtoafieldofresearch.By1939aroughprecursorofthepresentsethadbeenconstructed.Itrevealedthenewinterestinfunctionalanalysisbyreferringtoagroupofbasicprocesses(e.g.,stimulusexcitation, conditioning, motivation) in presumably an exhaustive way (the currentPostulate16issaidtocomplete“thestatementofprimaryprinciples”).Someofthesepostulateswere ineffect inverteddefinitions;othersdescribedquantitativeprocesses.Themethodwasnowbeing used, notmerely for the logical rectification of existingprinciples,butfortheisolationofasystemofvariables.Thenewpostulatesprovedembarrassinglyun-hypothetical,however,andsometime

after1939Hullretreatedtoamorespeculativelevelofanalysis—theneurological.Thisisasurprisingchange,forwearetoldthatthePrincipleswillremainatthemolarlevel,asthattermisusedbyTolman.“Theobjectofthepresentwork,”Hullwrites(p.17),“istheelaborationofthebasicmolarbehaviorallawsunderlyingthesocialsciences.”Yet Postulate 1 begins, “When a stimulus energy impinges upon a suitable receptororgan,anafferentimpulseisgenerated…,”andtheotherpostulatesholdessentiallytothesamelevel.Hullisnotalwaysathomeinwritingofneurology(as,forexample,inthecarelessstatement,onpage54,thatinreciprocalinnervationonemusclereceivesaneural discharge which is inhibitory rather than excitatory in nature), and we canscarcely explain themaneuver except by assuming that he is determined to keephispostulates hypothetical at any cost. The exigencies of his method have led him toabandontheproductive(andatleastequallyvalid)formulationofbehavioratthemolarlevelandtoalignhimselfwiththesemi-neurologists.The same pressure has led to a misrepresentation of many of his scientific

procedures. His “deductions” are often concerned merely with showing thatcomplicatedinstancesofbehaviormaybeanalyzedintosimplerinstanceswhichhavebeen,orareat leastcapableofbeing,studiedexperimentally.This isanunavoidabletask in a science of behavior, but to regard the simple case as postulate and thecomplicatedastheorem(orviceversa)istoextendthepostulationalframeworkbeyondits sphere of usefulness. In the case of the postulateswhich aremerely quantitativepredictions of processes yet to be studied, “deduction” often turns out to refer to adirectexperimentaldetermination,but thisshouldnotbeconfusedwith thetestingofpostulates via deduced theorems. To force simple scientific inferences into thepostulative mold does not contribute to clarity, but rather to awkwardness andconfusion.Becauseoftheunsuccessfulattempttoembraceafunctionalanalysis,thebookwill

hardly stand as an example of postulate method. Since it is likely to be receivedprimarilyasanexampleofmethod,afewdefectsfromamorestrictlylogicalpointofview may be listed. (1) There are no formal definitions, although Hull originallyrecognized this responsibility. (2) Some postulates contain asmany as five separate

473

statements,sothatreferencestoapostulateduringthecourseofaproofareambiguous.(3)Thesymbolsnever“payoff”inconvenienceorprogress.Sixandone-halfpagesareneededtodefinethem,buttheyareusedonlytoparaphrasewhathasalreadybeensaidinwords.Thereisapparentlynoinstanceinthebookofaproductivemanipulationofsymbols. (4) Although the distinction between primary and derived principles isfrequentlyinvoked,thereisnoadequatediscussionofthecriteriaofprimacyorofthelevelofanalysisofthelanguageofthepostulates.(5)Thegroundsforadmissionofanew postulate are not stated. Postulates are generally brought in when facts cannototherwisebeaccountedfor,butthisisnotalwaysthecase.Thetendencyisnowtowardaminimalset,butthisisnotexplicitlydiscussed.(Atleasttwopostulates,whichreferto “inhibition,” could be dispensed with on logical grounds. Although they arecarefully evaluatedwith respect to the status of inhibition as an “unobservable,” noquestion is raised as to whether they are needed. Inhibition, as the obverse ofexcitation, requires no separate reference, and none of the facts in Hull’s chapterdemandstheterminanyothersense.)Twopostulatesmustbeobjectedtoonmorethangeneralgrounds.ThePostulateof

AfferentNeuralInteractionassertsthattheimpulsesgeneratedbyastimulusenergyarechanged by each other into “something different.” Since no provision is made fordeterminingwhat this something is, thepostulatemaybe adequatelyparaphrasedbysaying that it is impossible to predict behavior from thephysical stimulus. This is awell-known difficulty in behavior theory, but we have come to expect from Hullsomething more than an explanatory fiction. Until the properties of the resultingpsychologicalorbehavioralstimulusareatleastsuggested,thepostulateservesmerelytoaccountforfailuretoachievearigorousanalysisandmakesnopositivecontribution.ThesameobjectionappliestothePostulateofBehavioralOscillation,whichassertsthepresence of an oscillatory “inhibitory potentiality” (it might as well have been“excitatory”) which blurs “the concrete manifestation of empirical laws.” This isanother neural fiction, with the single negative function of accounting for failure topredict. In his introductory chapter Hull inveighs against certain traditionalpsychological ghosts, but it is doubtfulwhether anyof them is quite so ghostlike infunctionasAfferentNeuralInteractionorBehavioralOscillation.Predilection for a givenmethod is not in itself objectionable. In the present case,

however, it has unquestionably diverted the author from a frontal attack on crucialissues. The important task of formulating behavior as a system of variables isperformed only indirectly. The “pivotal theoretical construct” is Effective ReactionPotential,whichissaidtobemanifestedbyprobabilityofreactionevocation,latency,resistancetoextinction,andreactionamplitude.Thesemeasuresdonot,unfortunately,alwaysvarytogether,andinthefaceofthisdifficultyHullselects“probability”asthebest single indicator. This notion appears very late in the book, and almost as anafterthought.Itisnotincludedinanearlierlistofthemanifestationsofhabitstrength(where, incidentally, the lack of covariation of the other measures is dismissed asunimportant), yet these manifestations should be identical with those of effectivereactionpotential.Probabilitylacksthephysicaldimensionsoflatency,amplitude,andso on, and might better be described as the thing manifested. Effective reactionpotentialcould, in fact,beusefullydefinedas theprobabilityofevocation.AlthoughHull would doubtless wish to retain some reference to a physical substratum, theconcept does little more than assert that the business of a science of behavior is topredict response. This prediction is to be achieved by evaluating the strength of a

474

response (the probability that it will occur) and relating this to other variables,particularlyinthefieldsofreinforcementandmotivation(“emotion”doesnotappearinthe index).FollowingTolman,Hullprefers to fractionate thisprobability, identifyingonepartwith reinforcement (which is thencalled “habit strength”) andanotherwithmotivation(called“drive”).Asimilarlyglancingandineffectivetreatmentofothercurrentproblemsseemsdue

tothesamemethodologicaldifficulties.Weexpectsomethingnewandhelpful in theanalysis of conditioning, but are presented instead with three demonstrationalexperimentswhichare intendedtorevealessentialrelations in learning.Byusingthecomplicatedandunexploredmotiveofescape(whichislikelytoconfuseelicitingandreinforcingstimuli),HulltriestosteeramiddlecoursebetweenPavlovianandoperantconditioning. But the essential characteristics are admittedly difficult to find in thePavlovianexperiment,andthehighlyverbalresolutionisunsatisfactory.Similarly,thechapteronpatterningisheavilymethodologicalbuttonorealeffect,sincetheproblemis virtually disposed of in the original definition, which limits patterning to thecompoundingofstimuliinthePavlovianmanner.Although the book is not intended as a factual survey, the quantitative relations

whichtheauthorwouldliketoseeinascienceofbehaviorareextensivelyillustrated.Exceptforacertainautistictendencytocreateappropriatedata(one-thirdofthegraphsrepresent hypothetical cases), Hull shows his characteristic willingness to abide byexperimental facts. The present volume probably sets a record for the use ofexperimentalmaterial inaprimarily theoreticalwork.Theheavyuseofmathematicsdoes not imply, as one might suppose, a more rigorous insistence upon a factualcorrespondence,forBengtCarlson,whoisresponsibleforthe“complicatedequations”attheendofmanychapters,hasbeengiventoomuchfreedominhiscurvefitting.Thefactthatheisabletofind“simplegrowthfunctions”whichapproximateanumberofselectedsetsofdataofferslittleassuranceoftheultimateusefulnessofthatfunction,sincehehasbeenallowedtousethreeconstantstowhichfreshvaluesareassignedatwill. (In one instance—p. 276—Carlson describes three experimental pointswith anequation containing three constants!) The mathematics is also occasionally ratherwishful, as, for example, when detailed instructions are given for “calculating habitstrength,” although no techniques have been discovered for making the necessarymeasurements.Inspiteofanextensiveperiodofdevelopment,thePrinciplesrevealsaprogramstill

in transition.There iseveryevidencethat thepostulatemethodisbeingsloughedoff(compare theearlierMathematico-deductiveTheory), butHull has not yetmade fulluse of an outright functional analysis.The uncertain theoretical position of the bookwill not lessen its stimulating effect in the field of behavior theory.More importantperhaps, is the research which will certainly follow. The book is wide open toexperimentalattack,anditisonlyfairtoaddthattheauthorplanneditthatway.

FromTheAmericanJournalofPsychology,1944,57,276–281.ThebookreviewedisClarkL.Hull,PrinciplesofBehavior.NewYork:Appleton-Century-Crofts,1943.

475

AReviewofBushandMosteller’sStochasticModelsforLearning

As the preface of this interesting book points out, there are those who “feel thatanswerstotheimportantquestionsofpsychologyaretobefound…inthecollectionofmore and better data rather than inmathematical formulas.” The reviewer is one ofthese.Thebookwasnotwrittenfortheday-to-dayworkerinthefieldoflearning,buthereisthereactionofoneforwhatitmaybeworth.Thewordstochastic is a newcomer inmathematical statistics and is not yet fully

naturalized.Itappearsonlyafewtimesinthepresenttext(mainlyinconnectionwith“stochastic matrices”) and seems to have been added to the title almost as anafterthought. The authors speak of stochastic processes as identicalwith the field ofprobabilityandstatethattheyusethewordstochastictoemphasizethetemporalnatureoftheprobabilityproblemstheyconsider,butthereareatleastfourdifferentsensesinwhichthewordprobabilitymaybeappliedtotheirwork.Probability–1.IntheAmericanScientist forApril1955LinusPaulingexplainshis

use of stochastic to describe a particular scientific method. The Greek means “good at hitting a target or at guessing.” Scientific

hypotheseswhicharelittlemorethanguessesabouttheresultsoffutureinquiryaretobedistinguishedfromthosewhichmayneverbedirectlyconfirmedbutwhichmerelylead topossiblyconfirmable theorems. In thiscontextstochastic refers to statementsmade in advance of adequate information and means ‘currently probable’ asdistinguishedfrom‘eventuallycertain.’Althoughthismaynotbethesenseinwhichthepresentauthorswouldwishtouse

the term, it does in fact describe their method. They point out that they are notinterested in setting up a model of the organism. Such a model would be a set ofhypothesesleadingatbesttothedeductionoftheoremstobetested.Instead,theyareconstructingmodels for experiments. They prepare theway for the experimenter bysupplying himwith equationswhich hemay find useful. (Meanwhile they test theirluckonafewavailablecases.)Ifwearetohavemathematical theoriesof learning—andthefourfoundationswhichsupportedtheworkinthisbook,likemanyothersinapositiontodeterminethecourseofscientificresearch,appeartobedeterminedthatweshall—thenit is tobehopedthatmoreandmoretheoristswillpracticethestochasticmethodinthissense.Probability–2. The authors reject a position of strict determinism with respect to

behavior and its prediction and express their belief that behavior is intrinsicallyprobabilistic.Thatistosay,themacroscopicfactsofbehaviorseemtothemtosuggesttheoperationofrandom,chance,orstochasticfactors.Inthisuseofthetermtheyarestill close to etymology,but theyare talkingnot about the aimof themarksmanbutaboutthescatterofhitsonatargetfromwhichdirectionofaimmaybeinferred.Thustheauthorsset theproposition,“behavior isstatisticalbyitsverynature,”against thealternativeform,behavior“appearstobesobecauseofuncontrolledoruncontrollable

476

conditions.” In describing the application which the experimental psychologist maymake of their book, they characterize themselves as presenting “procedures forestimating parameters from data.” How the parameters so estimated depend uponexperimental variables can be determined only by parametric studies,which they donotundertake.Actually,notmuchofthebookisconcernedwithestimatingparametersfromdatasufferingfromstochasticorrandomdisturbances.Probability–3. If probability–1 may be said to refer to the behavior of the

experimenterandprobability–2tothebehavioroftheorganism,thenprobability–3,towhich most of the book is devoted, does not refer to either, although the authorscontend that they are considering processes in the organism. In order to deal withprobability–3itisnecessarytomakesomearbitraryandhighlyrestrictiveassumptions.First, it is necessary to define a trial. The authors suggest: “an opportunity for

choosingamongasetofmutuallyexclusiveandexhaustivealternativesorresponses.”Thesumof theprobabilitiesofall such responsesat anygivenmomentadd to1—anumerical elegance not commanded by the first two probabilities. Separateprobabilities are sometimes arrived at by an analysis of the responses possible in agivenapparatus,asoneanalyzesanydiscretesamplespace.Forexample,inaTmaze“weassumethatnoneoftheratshasaninitialpositionpreference,andsoweletp1=0.5.”Suchaprobabilitydescribesanopportunitypresentedtoaratbuthasnoreferencetotherat’sinclination.Perhapstheratwillnotgoeitherway;butinthatcasetherehasbeennotrial,andhenceprobability–3isnotapplicable.When separate possibilities depend upon the rat’s history in the apparatus, the

sample space loses cogency. Since the authors define the learning process as any“systematicchangeinbehavior”(thus,incidentally,missingtheattractivestablestateswhich prevail under particular conditions of reinforcement), they cannot evaluateprobabilitiesfromobservedfrequencies.Thereis,unfortunately,onlyone‘tenthtrial.’Aratmayturnrightonthistrial,butwhatistheprobabilitythatitwilldoso?Heretheauthorsfallbackuponaratherdiscrediteddevice:“Theobviouswaytotestamodelistocollectdataonalargenumberofnearlyidenticalrats,sayonehundred.”Identicalatbirth, or just before the first trial, or on the tenth trial?Andwhere can such rats bepurchased? Although the authors supply a chapter on Distributions of ResponseProbabilities,theyneversucceedinsupportingtheirassumptionthatonanygiventrialaratmaybeconceivedofasspinningadiskordrawingaballfromanurntodetermineitschoiceofrightorleft.Whenwesaythattheprobabilityofturningrightmovesfrom0.5to1.0asaratlearnsaTmazetocompletemastery,theseexplicitvaluesareonlyspuriously reassuring. They bear no useful relation to the magnitude of the changetakingplaceintherat.Conceiveofacoindroppedverticallyashortdistanceontoahorizontalsurface,and

defineatrialasanycaseinwhichthecoinfallsononeface.Makethecoin‘learn’inthefollowingway:Whenit fallswithagivenfaceup, tilt itslightly in thatdirectionbeforedroppingitthenexttime.Let‘speedoflearning’berepresentedbyhowmuchthecoinistiltedateachtrial.Eventually,ofcourse,thecoinwillalwaysfallwithoneface up. Arguing from frequencies before and after learning, we could say that theprobabilityhasmovedfrom0.5to1.0,butwhatdoesthattellus,quantitatively,aboutthefinaltiltofthecoin?Amathematicalanalysisofthepossibilitiesinasamplespaceis no substitute for, andmay be of little help in, the empirical study of a scientificsubjectmatter.

477

Probability–4.ProfessorsBushandMostellerexplainintheirintroductionthattheyareattempting“todescriberesponsetendenciesbysetsofprobabilityvariables.”Thereader is likely to suppose that they mean tendencies to respond, such as thedeterminingorexcitatorytendenciesofpsychologicaltheory.Butsuchisnotthecase.That kind of probability—number 4 in our series—is dealtwith only indirectly in averyshortchapter.Theauthors’“tendency”referstoprobability–3—toachoiceamongmutually exclusive responses. The question of the inclination tomake any responsewhatsoeverisexcludedbythedefinitionofatrial—themathematicsdonotapplyuntila response has occurred. An effort is made to account for bar-pressing (which theauthorsdescribeas“thesimplesttypeoflearning”!)withinthisframeworkbydefining“bar-pressing”and“not-bar-pressing”asmutuallyexclusiveandexhaustiveclassesofresponses.The difficulty heremight be expressed by saying that “not-bar-pressing” requires

infinite time for its completion. The notion of a trial is not applicable. The timeproblem inprobabilityestimates isdiscussedonly in reference to latencyand rateofrunningonarunway.Iftimeis“quantized”intointervalsofagivenlength,respondingandnot-respondingbecomemeaningfulalternatives.Theauthorsadmit,however,that“ourtaskis toconstructamodelfor therunwaybycreatingachoicesituationinourminds,eventhoughtheremaynotbeoneintherat’s.”The procedures described in this chapter contribute little toward the much more

difficult problem of rate of responding, which is dismissed in this single sentence:“Rateofrespondingcanalsoberelatedtoprobabilitiesofappropriateresponseclasses,ashasbeendemonstratedintheliterature.”ThereferencesaretoapaperbyEstesandanearlypaperbytheauthors.Itisunfortunatethatsomethingmorealongthislinewasnot included, since themodels actually presented are seldom applicable to behavioroutside the laboratory, where trials satisfying the authors’ definition are almostunknown.Theauthorsundertooktomakethisextensivestudyupondiscoveringthe“largebulk

of empirical information on learning”whichwas available. They have tried to helppsychologistsmakethemostofthisinformation.Itwouldbeunfortunateiftheirbookwere instead to obscure amore attractive possibility—namely, that of discarding allsuchearlyworkascrudeexplorationandproceedingtothecollectionoffreshdatawithnewinstrumentsandmorerigorousmethods.Oneofthegreatdisservicesofstatisticsto science is likely to be just this: in showing the scientist that something may besignificantly inferredfromasetofdata,statisticsencourageshimtoholdon to thesedata, and the methods responsible for them, long after they might better have beendiscarded in favor ofmore expedientmeasures. It is to be hoped that in adding theprestigeofmathematics toearlywork in learning, theauthorshavenot inadvertentlyretardedprogressinthefield.A day-to-day researcher can scarcely object to the activities ofmathematicians as

such—eventotheirrewritingthefirst lineofthegospelaccording toSt. John in thisform: “First there must be a mathematical theory or system” (p. 1). If science hasseldom,ifever,begunwithamodel,itdoesnotfollowthatmodelsarenotworthwhile,or that someone cannot profitably begin with one. Nevertheless those who areconcerned with progress in the field of learning should not worry if their practicesfollowadifferentpattern.Itiswelltounderscorethemanyfrankdisclaimersscatteredthroughsuchabookasthis.Wearetoldinanearlychapter,forexample,that“closeagreementdoesnotprovethatthemodeliscorrect,butsuggeststhatitmaybeuseful;

478

poor agreement indicates that the specific model, including the identification, isinappropriate.” This is only one example of the authors’ modest evaluation of theircontribution. Indeed, they would probably agree that for one who is primarilyconcernedwithfindingoutmoreaboutthelearningprocess,theBaconianadmonitionstillholds:“Studynature,notbooks.”

FromContemporaryPsychology,1956,1,101–103.ThebookreviewedisRobertR.BushandFrederickMosteller,StochasticModelsforLearning.NewYork:JohnWiley&Sons,1955.

479

PARTIX

AMiscellany

AParadoxicalColorEffectSomeQuantitativePropertiesofAnxiety(withW.K.Estes)“Superstition”inthePigeonASecondTypeof“Superstition”inthePigeon(withW.H.Morse)Two“SyntheticSocialRelations”ConcurrentActivityUnderFixed-IntervalReinforcement(withW.H.Morse)SustainedPerformanceDuringVeryLongExperimentalSessions(withW.H.Morse)JohnBroadusWatson,BehavioristHowtoTeachAnimalsBabyinaBoxAWordAboutBoxesThePsychologyofDesignPigeonsinaPelicanSomeResponsestotheStimulus“Pavlov”SquirrelintheYard

480

FIG.1

481

AParadoxicalColorEffect

Whentheaccompanyingfigureisheldabouttwofeetfromtheeyeandlookedatunderlowwhite-lightillumination,thediscsappearasifcolored.Asuitableilluminationcanbeconvenientlyobtainedbyviewing the figure in roomlight throughapinhole.Thecolorsareusuallyof lowsaturationbutarequitedefinite.Mostof thepersonsIhavetested have reported a rose or a purplish red, a yellow, and a green or blue-green.Others have reported only a blue and yellow.Each disc takes on a single color andretainsitaslongasthefixationisnotchanged.Adjacentdiscsmaybeandusuallyare,ofdifferentcolors.An explanation suggests itself in terms of a functional-element theory of color

vision.At the necessary low illumination the number of functional coneswithin theretinalimageofasingledisccannotbeverylarge.Ifthenumberissufficientlylow,aproportionatedistributionofelementarytypesisimprobableorevenimpossible.Butadisproportionate activity is, by hypothesis, the characteristic effect of stimulation bycoloredlight.Itshouldnotbedifficulttotestthisexplanation.Aquantitativeinvestigationofthe

optimalilluminationandtheoptimalvisualanglesubtendedbyadiscisbeingplanned.Thedistributionof the severalhuesandpossibledifferences in saturationarealsoofobvioussignificanceincolorvisiontheory.SofarasIamaware,nocomparableeffecthasbeenreported.

FromTheJournalofGeneralPsychology,1932,7,481–482.

482

SomeQuantitativePropertiesofAnxiety

(WITHW.K.ESTES)The“mechanizationoftheLatinSquare”referredtoonpage120wasdescribedin

detailinapaperpublishedjointlywithW.T.HeroninthePsychologicalRecord(1939,3,166-176).Steelwires leadingfrom24cumulativerecordersweredividedintofourgroups,eachofwhichenteredasystemofpulleysasshowninFig.1.ThesixwiresatA,A….,allofwhichmoveupwardastheorganismsrespond,contributeequallytothemovementofB,whichis1½timestheaveragedmovementofalltheA’s.Thismuchofthe system gives amean curve for a subgroup of six rats. By combining fourmeancurvesinvariouswaysothercurvesareobtained.LeadsfromGroupsIandII,passingoverasinglepulley,recordthemeanfortwelverats(againactuallyoneandone-halftimesthemean).FourmeancurvesareobtainedforgroupsIandII,IandIII,andIIIand IV, and II and IV. Leads from the first and third of these groups, passing overanotherpulley,givethemeancurveforalltwenty-fourrats.

483

FIG.1

Meancumulativecurves,evenwhenarrangedinsubgroups,concealmorefactsthanthey clarify. Moreover, the whole system suffers if any one of the twenty-fourapparatuses breaks down, or if any one of the twenty-four animals is carelesslyhandledorfed,orotherwiseincorrectlytreated.Itisnotsurprising,therefore,thatthesystem did not prove productive. The present paper is one of a few examples of anapparentlylegitimateandsuccessfuluse.The “conditioned suppression” of behavior (as it is now generally called) has

provedtobeausefulbaselineinstudyingmeasureswhich“relieveanxiety”inhumansubjects (see page 153). The paper appeared in the Journal of ExperimentalPsychology(1941,29,390-400)and is reprintedwith thepermissionof theeditorofthejournalandoftheco-author.

Anxietyhasatleasttwodefiningcharacteristics:(1)itisanemotionalstate,somewhatresemblingfear,and(2) thedisturbingstimuluswhichisprincipallyresponsibledoesnotprecedeoraccompanythestatebutis“anticipated”inthefuture.Bothcharacteristicsneedclarification,whether theyareapplied to thebehaviorof

484

manor,asinthepresentstudy,toalowerorganism.Onedifficultyliesinaccountingforbehaviorwhicharises in“anticipation”ofa futureevent.Sincea stimuluswhichhas not yet occurred cannot act as a cause,wemust look for acurrentvariable.Ananalogywiththetypicalconditioningexperiment,inwhichS1,havinginthepastbeenfollowedbyS2,nowleadstoan“anticipatory”responsetoS2,putsthematteringoodscientific order because it is a current stimulus S1, not the future occurrence of S2,whichproducesthereaction.PastinstancesofS2haveplayedtheirpartinbringingthisabout,butitisnotS2whichiscurrentlyresponsible.Although the temporal relationships of classical conditioning provide for an

acceptable definition of anticipation, the analogy with anxiety is not complete. Inanxiety,theresponsewhichisdevelopedtoS1neednotbeliketheoriginalresponsetoS2. In a broader sense, then, anticipationmust be defined as a reaction to a currentstimulusS1which arises from the fact thatS1 has in the past been followed by S2,where the reaction is not necessarily that which was originally made to S2. Themagnitude of the reaction to S1 at anymoment during its presentationmay dependupontheprevioustemporalrelationsofS1andS2.Theconceptof“emotionalstate”alsoneedsclarificationinviewoftheexperiments

tobedescribed.Ithasbeensuggested[intheBehaviorofOrganisms] that in treatingemotion purely as reaction (either of the autonomic effectors or of the skeletalmusculature), a very important influence upon operant behavior is overlooked. Inpracticewearemostofteninterestedintheeffectofastimulusinalteringthestrengthofbehaviorwhichisfrequentlyotherwiseunrelatedtotheemotion.Astimulusgivingriseto“fear,”forexample,mayleadtomuscularreactions(includingfacialexpression,startle, and so on) and a widespread autonomic reaction of the sort commonlyemphasizedinthestudyofemotion;butofgreaterimportanceincertainrespectsistheconsiderablechangeinthetendenciesoftheorganismtoreactinvariousotherways.Some responses in its current repertoire will be strengthened, others weakened, invaryingdegrees.Ourconcern ismostoftenwithanxietyobserved in thisway,asaneffect upon the normal behavior of the organism, rather than with a specificsupplementaryresponseinthestrictsenseoftheterm.Theexperimentstobedescribedfollowthisinterpretation.Anemotionalstateisset

up in “anticipation” of a disturbing stimulus, and the magnitude of the emotion ismeasuredby its effect upon the strengthof certainhunger-motivatedbehavior,morespecifically upon the rate with which a rat makes an arbitrary response which isperiodically reinforcedwith food.Sucha ratehasbeen shown tobe avery sensitiveindicatorofthestrengthofbehaviorunderavarietyofcircumstances,anditisadaptedheretothecaseofemotion.Mowrer’srecentsummaryoftechniquesformeasuringthe‘expectation’ofastimulusdoesnotincludeacomparableprocedure.1In these experiments the disturbing stimulus to be “anticipated” was an electric

shockdeliveredfromacondenser throughgrids in thefloorof theexperimentalbox.The stimulus which characteristically preceded the disturbing stimulus and whichthereforebecametheoccasionforanxietywasatone,producedbyphonesattachedtoa60cycleA.C.transformer.The apparatus, which provided for the simultaneous investigation of twenty-four

rats, has been described in detail elsewhere. Each rat was enclosed during the

485

experimental period in a light-proof and nearly soundproof box containing a leverwhichcouldbeeasilydepressed.Acurve(numberofresponsesvs.time)foreachratandmechanically averagedcurves for thegroupand for certain sub-groupsof sixortwelveratswererecorded.Undertheprocedureofperiodicreconditioning,thecontrolclockwassettoreinforcesingleresponsestothelevereveryfourminutes,interveningresponses going unreinforced. The rats came to respond at a relatively constant rateduringtheone-hourexperimentalperiod,andthesummatedresponsecurvestendedtoapproximate straight lines, except for local cyclic effects resulting from a temporaldiscriminationbaseduponthefour-minuteperiodofreinforcement.CurvesAandCinFigure5areforgroupsoftwenty-fourratsandrepresentthesortofbaselineavailablefortheobservationoftheeffectofanxiety.Thesubjectsweretwenty-fourmalealbinoratsundersixmonthsofage,takenfrom

anunselectedlaboratorystock.Recordsweretakenforonehourdailyduringtheentireexperiment.After preliminary conditioning of the pressing response, two sub-groupswere formed;onegroupof twelve ratswaskept at a relativelyhighdrive,while theothertwelvewereheldatadrivewhichproducedaverylowrateofresponding.Thesoundandshockwerefirstintroducedaftertwoweeksofperiodicreinforcement.

ConditioningofaStateofAnxiety

Theaveragedperiodiccurvefortwelveratsonahighdriveontheoccasionofthefirstpresentations of the tone (T) and shock (S) is shown in Figure 2. On this firstpresentationthetonewasallowedtosoundforthreeminutes.Eachratwasthengivenashockandthetonewasstopped.Itwillbeobservedthatneitherthetonenortheshock(attheintensityusedthroughouttheexperiment)producedanydisturbanceinthemeanperiodic rate at either presentation. This orderly baselinemade it possible to followwith ease the development of the “anticipation” of the shock during subsequentrepetitionsofthesituation.

486

FIG.2.Firstpresentationsoftoneandshock.Mechanicallyaveragedcurvefortwelveratsunderperiodicreinforcement.ThetonewasturnedonatT,andatStheshockwasadministeredandthetoneturnedoff.Thereisnonoticeableeffectofeithertoneorshockupontherateofrespondingat

thisstage.

The tone-shock combination was presented twice during each of six consecutivehourlyperiods.Then,inordertoclarifyanychangesinthebehavior,theperiodofthetonewaslengthenedtofiveminutesandthecombinationwasgivenonlyonceduringeachensuingexperimentalhour.Theprincipalresultofthispartoftheexperimentwastheconditioningofastateof

anxietytothetone,wheretheprimaryindexwasareductioninstrengthofthehunger-motivatedlever-pressingbehavior.Theratioofthenumberofresponsesmadeduringtheperiodofthetonetotheaveragenumbermadeduringthesamefractionofthehourincontrolexperimentswas1.2:1.02forthefirstexperimentalhours;ithaddroppedto0.3:1.0bytheeighth.The changes in behavior accompanying anticipation of the shock are shown in

Figure3,whichgives the averaged curves for the groupof six ratswith the highestperiodic rate during the first four days of the five-minute tone. A number ofcharacteristics of these records should be noted. The progressively more markedreductioninperiodicrateduringtheanticipatoryperiodisobvious.Theeffectupontherate is felt immediately after the presentation of the tone and remains at a constantvalueuntiltheshockisgiven.(Thisconstancymightnotbemaintainedifthesituationwererepeatedoftenenoughtoallowtherattoformatemporaldiscrimination.)EffectsalsoappearaftertheshockwhichwerenotpresentinFigure2astheresultoftheshockalone.EspeciallyinCurvesAandBofFigure3,theshockisseentobefollowedbyadepressionandirregularityofratewhichareatleastmuchgreaterthananyeffectinthecontrolrecords.Withcontinuedrepetitionoftheexperiment,thisdisturbancetendstoadaptout, althoughnotcompletely. InCurvesC andDofFigure3, the distortion ismuchlessmarked.CurveBofFigure5givesasimilarexampleatarelativelylatestageofconditioning.

487

FIG.3.Reductioninrateofrespondingduringsuccessiveperiodsofanxiety.Averagedcurvesforsixratsonfourconsecutivedays.Bythethirdorfourthdayrespondingpracticallyceasesduring

thepresentationofthetone.

Themodificationinbehaviorcorrelatedwiththeanticipationofadisturbingstimuluscannotbeattributedtoanegativereinforcementoftheresponsetothelever,sincetheshockwasalwaysgivenindependentlyoftherat’sbehaviorwithrespecttothelever.Onlyupon rareoccasions could the shockhave coincidedwith a response.Thiswasespecially true in the experiments upon the group at a lower drive, where a similareffectwasobtained.Figure4showsaveragedcurvesforagroupofsixratswhichhadbeensubjectedtotheprocedurejustdescribedexceptthattheirdrivewassolowduringconditioningthattherateofrespondingwasvirtuallyzero.ThelowercurveinFigure4is for the first dayonwhich the five-minute rather than the three-minuteperiodwasgiven.Uptoandincludingthisrecord,noeffectoftheanticipationoftheshockcouldbedetected,sincetheanimalswerenotrespondingatasignificantrate.Thedrivewasthenraised,andtheuppercurveofFigure4showstheperformanceofthesamegroupon the following day. By sighting along the curve, one may observe a markeddepression in the rate of respondingduring the period of the tone.ComparisonwithCurveB in Figure3 shows that, although the baseline at the higher drive is moreirregular, a depression of relatively the same magnitude is obtained. In this case,coincidentalpresentationsof shockand responsemaysafelybe ignored,yet the tonehasacquiredthesamedepressingeffectuponthebehavior.

488

Another characteristic which deserves attention is the compensatory increase inperiodic rate following the period of depression. This appears to some extent in allrecordsobtained;but itmaybe seenmostclearly inCurveBofFigure5, aperiodiccurve for all 24 rats after the emotional conditioningwas quite complete.The curvewas obtained about two weeks after the records in Figure 3. Curves A and C arecontrolstaken(ataslightlyhigherdrive)onadjacentdays.BysightingalongCurveB,onemayobserveaclear-cutincreaseinratesubsequenttotheshock,whichcontinuesuntiltheextrapolationofthecurveprecedingthebreakisreached.Evidentlytheeffectof the emotional state is a temporary depression of the strength of the behavior, thetotal amountof respondingduring the experimental period (the “reserve”) remainingthe same. Similar compensatory increases have been described under a number ofcircumstances,includingphysicalrestraintoftheresponse.

FIG.4.Reductioninrateduringanxietyfollowingexperimentsataverylowdrive.ThelowerrecordisacurveforsixratsataverylowdrivebutotherwisecomparablewiththecurvesinFigs.2and3.Theuppercurveisforthesamegroupatahigherdriveonthefollowingday.Thetone

hasanobviouseffect,althoughallpreviouspresentationshavebeenmadeatadrivesolowthatnoeffectwasobservable.

EffectsofAnxietyuponExtinction

When reinforcementwith food is withheld, the rat continues to respond, but with adecliningrate,anddescribesthetypicalextinctioncurve.Theeffectsofanxietyupon

489

thiscurvehavebeeninvestigated.Thefirsthourofa typicalextinctioncurve,duringwhichthecombinationoftoneandshockwaspresented,isshowninthegroupcurvesofFigure6 and the individual curvesofFigure7. By sighting along either curve inFigure6,onemayobserveadistinctdepression inrateduring theperiodof the tone,and(followingtheshock)anequallydistinctcompensatoryincrease,whichappearstobe maintained until an extrapolation to the first part of the curve is approximated.Figure7 contains sample records from four rats which showed different degrees ofdepressionduringthetone.

490

FIG.5.Subsequentcompensationforthereductioninrateduringanxiety.Thecurvesareaveragesfortwenty-fourratstakenonthreeconsecutivedays.AandCweretakenunderperiodic

reinforcement,whileBshowstheeffectofthetoneatalatestageintheexperiment.Thereductioninrateisfollowedbyacompensatoryincrease,bringingthecurvebacktotheextrapolationofthe

firstpart.

During extinction, then, a state of anxiety produces a decrease in the rate ofrespondingandtheterminatingstimulusisfollowedbysuchacompensatoryincreaseinratethatthefinalheightofthecurveisprobablynotmodified.

491

FIG.6.Effectsofanxietyuponextinction.Thelowercurveisanaverageforsixrats,theupperfortwelve.Thetone,whichhadpreviouslybeenfollowedbyshockduringperiodicreinforcement,

depressestheslopeoftheextinctioncurves,andacompensatoryincreasefollowstheadministrationoftheshock.

FIG.7.Effectsofanxietyuponextinction.IndividualrecordsfromtheexperimentrepresentedinFig.6.

TheExtinctionofaStateofAnxiety

Afurtherpropertyofanxietywasinvestigatedbypresentingthetoneforaprolongedperiodwithouttheterminalshock.Inoneexperiment,whiletheratswererespondingunderperiodicreinforcement,thetonewasturnedonaftertwenty-sevenminutesofthe

492

experimentalperiodhadelapsedandallowedtosoundfor theremainderof thehour.TheresultisshowninFigures8and9.Itwillbeobservedthattherecoveryofanormalperiodicrateisdelayedconsiderablybeyondtheaccustomedfive-minuteperiodofthetone.When the time is taken from theonsetof the tone to thepointatwhich the ratagain reaches his previous periodic rate (measurements being made on individualcurves), themeanperiodrequiredforrecoveryisfoundtobe8.6minutes.Thegroupcurve for twelve rats (the upper record in Figure8) shows a definite compensatoryincreaseinratelaterinthehour,althoughtheextrapolationofthefirstpartofthecurveisnotquitereachedbytheendoftheperiod.

FIG.8.Extinctionoftheeffectofatonewhentheterminatingshockfailstoappear.Theupperrecordistheaveragecurvefortwelveratsunderperiodicreinforcement.Thetonewasturnedon

atTandcontinuedtosoundduringtherestofthehour.Noshockwasgiven.Therateofrespondingreturnstonormal(andperhapsshowssomecompensatoryincrease)withinten

minutes.Thelowercurveshowsarepetitionoftheexperimenttendayslater.

Thesameexperimentwasrepeatedtendayslateratasomewhatlowerdrivewiththeresult shown in the lowercurve inFigure8.Themeandelay in recovery ishere9.1minutes, and recovery is less complete. Except for the effects of the difference inmotivation, the tworecordsappearquitesimilarandexemplify the reproducibilityofbehaviorofthissort.

493

494

FIG.9.IndividualcurvesfromtheexperimentdescribedinFig.8.

Because the period of depressed activity varies among rats, individual records areneeded in order to observe the course of the recovery of normal strength during theextinction of anxiety. Figure9 shows a number of individual records with differentperiodicrates, thedifferencesbeingattributablemainly todifferences inhunger.Thelaginrecoveryappearsinnearlyallrecords,andthecompensatoryincreaseinperiodicrateinthemajority.Insomecurves,notablyE,F,andG,anextrapolationofthefirstpartisreachedbeforetheendofthehour.Itisnotclearthatthiswouldhavebeenthecasewiththeotherratsiftheexperimentalperiodcouldhavebeenprolonged,butthecurvesingeneralappeartobepositivelyaccelerated.Spontaneousrecoveryfromtheextinctionoftheanxietyisfairlycomplete.Thedaily

recordwhichprecededtheupperfigureinFigure8showedaratioof0.6:1.0betweentheaverageperiodicrateduringtheperiodofthetoneandthenormalrateforsuchaninterval.Onthedayfollowing thefigure, theratiowas0.7 :1.0forasimilarperiod,indicatingthatlittleornoeffectofextinctionsurvived.

Summary

Anxiety is here defined as an emotional state arising in response to some currentstimuluswhichinthepasthasbeenfollowedbyadisturbingstimulus.Themagnitudeofthestateismeasuredbyitseffectuponthestrengthofhunger-motivatedbehavior,inthis case the ratewithwhich rats pressed a lever under periodic reinforcementwithfood.Repeatedpresentationsofatoneterminatedbyanelectricshockproducedastateofanxietyinresponsetothetone,theprimaryindexbeingareductioninstrengthofthehunger-motivated behavior during the period of the tone.When the shockwas thuspreceded by a period of anxiety, it produced amuchmore extensive disturbance inbehavior than an “unanticipated” shock. The depression of the rate of respondingduringanxietywascharacteristicallyfollowedbyacompensatoryincreaseinrate.During experimental extinction of the response to the lever the tone produced a

decrease in the rate of responding, and the terminating shock was followed by acompensatoryincreaseinratewhichprobablyrestoredtheoriginalprojectedheightoftheextinctioncurve.Theconditionedanxietystatewasextinguishedwhen the tonewaspresentedfora

prolonged period without the terminating shock. Spontaneous recovery from thisextinctionwasnearlycompleteonthefollowingday.

1Mowrer,O.H.Preparatoryset(expectancy)—somemethodsofmeasurement.Psychol.Monogr.,1940,52,43.

2Theratioisnotexpectedtobeexactly1:1sincethenumberofresponsesmadeduringaperiodoffiveminuteswilldependuponwheretheperiodbeginswithrespecttothefour-minuteintervalofreinforcement.

495

“Superstition”inthePigeon

ThefollowingarticleisreprintedwithpermissionfromTheJournalofExperimentalPsychology(1948,38,168–172).

Tosaythatareinforcementiscontingentuponaresponsemaymeannothingmorethanthatitfollowstheresponse.Itmayfollowbecauseofsomemechanicalconnectionorbecauseofthemediationofanotherorganism;butconditioningtakesplacepresumablybecauseofthetemporalrelationonly,expressedintermsoftheorderandproximityofresponseandreinforcement.Wheneverwepresentastateofaffairswhichisknowntobereinforcingatagivenlevelofdeprivation,wemustsupposethatconditioningtakesplaceeventhoughwehavepaidnoattentiontothebehavioroftheorganisminmakingthepresentation.Asimpleexperimentdemonstratesthistobethecase.A pigeon is reduced to 75 per cent of itsweightwhenwell fed. It is put into an

experimentalcageforafewminuteseachday.Afoodhopperattachedtothecagemaybeswungintoplacesothatthepigeoncaneatfromit.Asolenoidandatimingrelayholdthehopperinplaceforfivesec.ateachpresentation.If a clock is nowarranged to present the foodhopper at regular intervalswith no

referencewhatsoevertothebird’sbehavior,operantconditioningusuallytakesplace.In six out of eight cases the resulting responses were so clearly defined that twoobserverscouldagreeperfectlyincountinginstances.Onebirdwasconditionedtoturncounter-clockwiseaboutthecage,makingtwoorthreeturnsbetweenreinforcements.Another repeatedly thrust its head intooneof theupper cornersof the cage.A thirddeveloped a “tossing” response, as if placing its head beneath an invisible bar andliftingitrepeatedly.Twobirdsdevelopedapendulummotionoftheheadandbody,inwhich the head was extended forward and swung from right to left with a sharpmovement followed by a somewhat slower return. The body generally followed themovement and a few stepsmight be takenwhen itwas extensive.Another birdwasconditioned tomake incompletepeckingorbrushingmovementsdirected towardbutnot touching the floor.None of these responses appeared in any noticeable strengthduringadaptationtothecageoruntilthefoodhopperwasperiodicallypresented.Intheremainingtwocases,conditionedresponseswerenotclearlymarked.Theconditioningprocessisusuallyobvious.Thebirdhappenstobeexecutingsome

response as the hopper appears; as a result it tends to repeat this response. If theintervalbeforethenextpresentationisnotsogreatthatextinctiontakesplace,asecond“contingency” is probable.This strengthens the response still further and subsequentreinforcementbecomesmoreprobable.It is truethatsomeresponsesgounreinforcedandsomereinforcementsappearwhentheresponsehasnotjustbeenmade,butthenetresultisthedevelopmentofaconsiderablestateofstrength.Withtheexceptionofthecounter-clockwiseturn,eachresponsewasalmostalways

repeatedinthesamepartofthecage,anditgenerallyinvolvedanorientationtowardsomefeatureofthecage.Theeffectofthereinforcementwastoconditionthebirdto

496

respond to someaspectof the environment rather thanmerely to execute a seriesofmovements. All responses came to be repeated rapidly between reinforcements—typicallyfiveorsixtimesin15sec.Theeffectappearstodependupontherateofreinforcement.Ingeneral,weshould

expect that the shorter the intervening interval, the speedier and more marked theconditioning.Onereasonis that thepigeon’sbehaviorbecomesmorediverseas timepasses after reinforcement. A hundred photographs, each taken two sec. afterwithdrawalofthehopper,wouldshowfairlyuniformbehavior.Thebirdwouldbeinthesamepartofthecage,nearthehopper,andprobablyorientedtowardthewallwherethehopperhasdisappearedorturningtoonesideortheother.Ahundredphotographstakenafter10sec.,ontheotherhand,wouldfindthebirdinvariouspartsofthecageresponding to many different aspects of the environment. The sooner a secondreinforcement appears, therefore, the more likely it is that the second reinforcedresponsewill be similar to the first, and also that theywill both have one of a fewstandardforms.Inthelimitingcaseofaverybriefintervalthebehaviortobeexpectedwould be holding the head toward the opening through which the magazine hasdisappeared.Anotherreasonforthegreatereffectivenessofshortintervalsisthatthelongerthe

interval, the greater the number of intervening responses emitted withoutreinforcement. The resulting extinction cancels the effect of an occasionalreinforcement.According to this interpretation the effective intervalwill dependupon the rateof

conditioningandtherateofextinction,andwillthereforevarywiththedeprivationandalsopresumablybetweenspecies.Fifteensec.isaveryeffectiveintervalatthelevelofdeprivationindicatedabove.Onemin.ismuchlessso.Whenaresponsehasoncebeenset up, however, the interval can be lengthened. In one case itwas extended to twomin., and a high rate of responding was maintained with no sign of weakening. Inanother case,many hours of respondingwere observedwith an interval of onemin.betweenreinforcements.Inthelattercase,theresponseshowedanoticeabledriftintopography.Itbeganasa

sharp movement of the head from the middle position to the left. This movementbecamemoreenergetic,andeventuallythewholebodyofthebirdturnedinthesamedirection,andasteportwowouldbetaken.Aftermanyhours, thesteppingresponsebecamethepredominantfeature.Thebirdmadeawell-definedhoppingstepfromtherighttotheleftfoot,meanwhileturningitsheadandbodytotheleftasbefore.Whenthesteppingresponsebecamestrong, itwaspossible toobtainamechanical

recordbyputtingthebirdonalargetambourdirectlyconnectedwithasmalltambourwhichmadeadelicateelectriccontacteachtimesteppingtookplace.Bywatchingthebirdandlisteningtothesoundoftherecorderitwaspossibletoconfirmthefactthatafairlyauthenticrecordwasbeingmade.Itwaspossibleforthebirdtoheartherecorderat each step, but thiswas, of course, in noway correlatedwith feeding. The recordobtainedwhenthemagazinewaspresentedonceperminuteresemblesineveryrespectthecharacteristiccurveforthepigeonunderfixed-intervalreinforcementofastandardselectedresponse.Awell-markedtemporaldiscriminationdevelops.Thebirddoesnotrespondimmediatelyaftereating,butwhen10or15oreven20sec.haveelapsed, itbeginstorespondrapidlyandcontinuesuntilthereinforcementisreceived.Inthiscaseitwaspossibletorecordthe“extinction”oftheresponsewhentheclock

was turned off and the magazine was no longer presented at any time. The bird

497

continued to respond with its characteristic side to side hop. More than 10,000responseswererecordedbefore“extinction”hadreachedthepointatwhichfewifanyresponses were made during a 10 or 15 min. interval. When the clock was againstarted, the periodic presentation of the magazine (still without any connectionwhatsoever with the bird’s behavior) brought out a typical curve for reconditioningafterfixed-intervalreinforcement,showninFigure1.The recordhasbeenessentiallyhorizontalfor20min.priortothebeginningofthiscurve.Thefirstreinforcementhadsome slight effect and the second a greater effect. There is a smooth positiveaccelerationinrateasthebirdreturnstotherateofrespondingwhichprevailedwhenitwasreinforcedeveryminute.

498

FIG.1.“Reconditioning”ofasuperstitiousresponseafterextinction.Theresponseofhoppingfromrighttolefthadbeenthoroughlyextinguishedjustbeforetherecordwastaken.Thearrowsindicatetheautomaticpresentationoffoodatone-min.intervalswithoutreferencetothepigeon’s

behavior.

When the response was again extinguished and the periodic presentation of foodthen resumed, a different response was picked up. This consisted of a progressivewalking response inwhich the birdmoved about the cage.The response of hoppingfromside tosidenever reappearedandcouldnot,ofcourse,beobtaineddeliberatelywithoutmakingthereinforcementcontingentuponthebehavior.Theexperimentmightbesaidtodemonstrateasortofsuperstition.Thebirdbehaves

as if therewere a causal relation between its behavior and the presentation of food,although such a relation is lacking. There are many analogies in human behavior.Rituals for changing one’s luck at cards are good examples. A few accidental

499

connectionsbetweenaritualandfavorableconsequencessufficetosetupandmaintainthebehavior inspiteofmanyunreinforced instances.Thebowlerwhohasreleasedaballdownthealleybutcontinuestobehaveasifhewerecontrollingitbytwistingandturninghisarmandshoulderisanothercaseinpoint.Thesebehaviorshave,ofcourse,no real effect upon one’s luck or upon a ball halfway down an alley, just as in thepresent case the food would appear as often if the pigeon did nothing—or, strictlyspeaking,didsomethingelse.It is perhaps not quite correct to say that conditioned behavior has been set up

without any previously determined contingency whatsoever.We have appealed to auniformsequenceofresponsesinthebehaviorofthepigeontoobtainanover-allnetcontingency.Whenwearrangeaclocktopresentfoodevery15sec.,weareineffectbasing our reinforcement upon a limited set of responseswhich frequently occur 15sec.afterreinforcement.Whenaresponsehasbeenstrengthened(andthismayresultfrom one reinforcement), the setting of the clock implies an even more restrictedcontingency.Somethingofthesamesortistrueofthebowler.Itisnotquitecorrecttosaythatthereisnoconnectionbetweenhistwistingandturningandthecoursetakenbytheballatthefarendofthealley.Theconnectionwasestablishedbeforetheballleftthebowler’shand,butsinceboththepathoftheballandthebehaviorofthebowleraredetermined,somerelationsurvives.Thesubsequentbehaviorofthebowlermayhavenoeffectupontheball,butthebehavioroftheballhasaneffectuponthebowler.Thecontingency, thoughnotperfect, is enough tomaintain thebehavior in strength.Theparticular form of the behavior adopted by the bowler is due to induction fromresponses in which there is actual contact with the ball. It is clearly a movementappropriate to changing the ball’s direction. But this does not invalidate thecomparison,sincewearenotconcernedwithwhatresponseisselectedbutwithwhyitpersists in strength. In rituals for changing luck the inductive strengthening of aparticular form of behavior is generally absent. The behavior of the pigeon in thisexperiment is of the latter sort, as the variety of responses obtained from differentpigeons indicates. Whether there is any unconditioned behavior in the pigeonappropriatetoagiveneffectupontheenvironmentisunderinvestigation.The results throw some light on incidental behavior observed in experiments in

which a discriminative stimulus is frequently presented. Such a stimulus hasreinforcing value and can set up superstitious behavior.Apigeonwill often developsomeresponsesuchas turning, twisting,peckingnear the locusof thediscriminativestimulus, flapping itswings,andsoon. Inmuchof thework todate in this field theinterval between presentations of the discriminative stimulus has been onemin. andmanyofthesesuperstitiousresponsesareshort-lived.Theirappearanceastheresultofaccidentalcorrelationswiththepresentationofthestimulusisunmistakable.

500

ASecondTypeof“Superstition”inthePigeon

(WITHW.H.MORSE)

The following article is reprinted with permission from The American Journal ofPsychology(1957,70,308-311).

When food is given to a hungry organism, any behavior in progress at themomentmust be assumed to be reinforced by this event. When small amounts of food arerepeatedlygiven,a“superstitiousritual”maybesetup.Thisisduenotonlytothefactthat a reinforcing stimulus strengthens any behavior it may happen to follow, eventhough a contingency has not been explicitly arranged, but also to the fact that thechangeinbehaviorresultingfromoneaccidentalcontingencymakessimilaraccidentsmore probable. In an earlier experiment the automatic operation of a food-magazineevery15sec.wasfoundtoinducehungrypigeonstoengageinsuchritualisticbehaviorasbowing,scraping,turning,anddancing[seepage570].Insomecasesthebehaviorwas stable, in others the topography slowly changed; but in all cases superstitiouseffects survived indefinitely. Similar effects from the adventitious reinforcementarising from thepresentationofdiscriminative stimuli have recentlybeenobserved.1Such effects must always be allowed for in designing experiments on complexbehavior.Accidental,butneverthelesseffective,relationshipsmayariseinthesensory control

of operant behavior. For example, a stimulus present when a response is reinforcedmay acquire discriminative control over the response even though its presence atreinforcementisadventitious.Suppose,forexample,thatanorganismisrespondingatamoderaterateonavariable-intervalscheduleofreinforcement,andletanincidentalstimulus (A) occasionally appear for a brief period.Even though there is no explicittemporal relation between the appearance ofA and the program of reinforcement, aresponsewill occasionallybe reinforced in thepresenceofA.For a brief period thefrequencyofsuchreinforcementmaybeappreciablygreaterthanintheabsenceofA.Anorganismwhichissensitivetoslightdifferencesinrateofreinforcementwillformadiscrimination;itsrateofrespondinginthepresenceofAwillbecomegreaterthanintheabsenceofA.Thismightbecalledapositivesensorysuperstition.If,ontheotherhand, reinforcements happen to occur relatively infrequently in the presence ofA, adiscriminationwilldevelopintheoppositedirection,astheresultofwhichtherateinthepresenceofAwillberelativelylow—asortofnegativesensorysuperstition.Whenanaccidentalcontingencyhasproducedahigherorlowerrateofresponding

inthepresenceofanincidentalstimulus,asecondeffectfollows.Iftheratehasfallenin the presence of A (because reinforcements have been relatively infrequent),responseswillbeevenlesslikelytobereinforcedinthepresenceofA.Inthelimitingcase no responses will be made in the presence of A, and no response, of course,reinforced. Moreover, reinforcements which are made available during A are not

501

obtainedbecauseresponsesarenotmade.ThefirstresponsefollowingthewithdrawalofAisthenreinforced,andthediscriminationisfurtherstrengthened.Similarly,whenthe rate is increased during A because of favorable accidental reinforcements, allreinforcementssetupduringAarelikelytobeobtained,andiftheprecedingconditioncommandsarelativelylowrate,somereinforcementssetupatthattimemayactuallybeobtainedafterAhasappeared,tostrengthenthediscrimination.Bothtypesof“sensorysuperstition”havebeendemonstratedexperimentallyinthe

pigeon.Theapparatus consistedof theusual experimental space13×22×16 in.Asmall translucent plastic plate wasmounted behind a 1-in. circular opening at headheight on onewall. Itwas lighted frombehind by an orange 6-w. bulb.The pigeonpecked this disk to operate the controlling circuit. Food was presented forreinforcementinanopeningbelowthiskey.Waterwasavailable.Three pigeons, two of which had previously been reinforced on other schedules,

wereplacedonavariable-intervalscheduleofreinforcementwithameanintervalof30min. The shortest interval between reinforcements was 1 min., the longest 59 min.Dailyexperimentalsessionsvariedbetween6and20hr. in length.Bodyweightwasmaintainedatapproximately80%oftheadlibweight, anda reinforcementconsistedofaccesstomixedgrainfor5sec.Theresultingperformancewasatalowmeanrateofrespondingwith some local irregularity.Against this baseline, an incidental stimulusconsistingofabluelightprojectedonthekeyinsteadoforangewasintroducedfor4min.onceperhour.Thescheduleofoccurrenceofthisstimuluswasindependentoftheprogrammingschedule.Therateofrespondingwasrecordedcontinuouslyintheusualcumulativecurve.Briefdownwardmovementsofthepenmarkedreinforcements,andthe pen remained down during the 4-min. period of the incidental stimulus, thusslightlydisplacingtherecordmadeinthepresenceofthestimuluswithoutchangingitsslope.

502

FIG.1.Superstitiousdiscriminativecontrolofrespondingbyanincidentalstimulus.Thestimulusispresentedfor4min.onceperhour,andismarkedbythedownwarddisplacementofthe

recordingpen(a,b,c,andd).CurvesAandBaresegmentsofcumulativeresponse-curvesonavariable-intervalschedule(meaninterval=30min.)forapigeonwithalowerratewhenthe

incidentalstimulusispresentthanwhenitisabsent—a“negativesuperstition.”CurvesCandDaresegmentsforanotherpigeononthesameschedulewithahigherrateinthepresenceofthe

incidentalstimulus—a“positivesuperstition.”

Segmentsof records showing superstitiousdifferences in rateunder the control oftheincidentalstimulusareshowninFigure1.Exceptfortheportionsa,b,c,andd,thecurves are characteristic of the baselines obtained on the schedule described above.Curves A and B are for a pigeon which showed sustained periods of negativesuperstition.Theover-allrategeneratedbythescheduleisrelativelyhigh(oftheorderof0.5responsespersec.).Whenevertheincidentalstimulusappears,theratedropstoalowvalue or to zero (a andb).RecordsC andD are for another pigeon showing apositivesuperstition.Thebaserateisrelativelylow,andtherateinthepresenceofanincidentalstimulusatcanddisclearlyofamuchhigherorder.The direction of the superstition is not necessarily stable. In a long experimental

sessionapositive superstitionmaygiveway to anegative form,orviceversa.Suchchangesareusuallyeasilyexplainedintermsofadventitiousreinforcementorfailuretoreceivereinforcementinthepresenceoftheincidentalstimulus.Allthreebirdsshowedperiodsofbothpositiveandnegativesuperstition.

503

There are several arbitrary features of such an experiment. In the present case theincidentalstimuluswaspresent1/15ofthetimeduringasession.Arelativelyshorterperiodwouldbelesslikelytoreceivereinforcementsonagivenschedule,andmightbeexpected to produce negative superstitionmore frequently.At the other extreme, anincidental stimulus which occupied half the experimental sessionwould presumablyshare so nearly equally in the reinforcements that there would be no substantialseparation of rates.The schedule and the performance generated are also relevant indetermining the frequency of adventitious reinforcement. Finally, the nature andintensityoftheincidentalstimulusalsomayhavetheireffect.Pendinganinvestigationoftheseparameters, itmayat leastbesaidthatincidental

stimuli adventitiously related to reinforcement may acquire marked discriminativefunctions.

1Morse,W.H.Ananalysisofrespondinginthepresenceofastimuluscorrelatedwithperiodsofnon-reinforcement.Unpublisheddoctoraldissertation,HarvardUniversity,1955.

504

Two“SyntheticSocialRelations”

About 10 years ago, two demonstration experiments were designed for a GeneralEducation course inHumanBehavior atHarvard.Theywere briefly described in anillustrated weekly and are occasionally referred to in the psychological literature. Itseemsadvisabletopublishasomewhatmoreexplicitaccount.

The“Ping-Pong”PlayingPigeons

Therewereseveralversionsof thisapparatus, inoneofwhichamotor-drivendevicereturned the ping-pong ball to the playing surface so that the apparatus ranwithoutattention.Inalessmechanizedversion,the“ping-pong”tablewasapproximately8in.wide, 16 in. long, and 8 in. high (Fig. 1). A pigeon standing at one end couldconvenientlypeckaballasitarrivedattheedgeofthetable.Iftheballrolledofftheedge,itfellintoatroughandtrippedaswitchwhichoperatedafooddispenserundertheoppositeedgeandthusreinforcedthepigeonwhich“wonthepoint.”Lightmetalrailspreventedtheballfromfallingoffthesidesofthetable.Thesurfacewasslightlycanted,slopingfromacenterlinetowardeachedgesothattheballwouldnotstoponit. Wire barriers prevented the pigeons from jumping up on the table but did notinterferewithplay.Inthefinishedperformance,thedemonstratorwouldstartaballnearthemiddleof

the table. It rolled tooneedgeand thepigeonon that sidepecked it,driving itbackacrossthetable.Attheotheredgeitwaspeckedbytheotherpigeonandthusreturned.The pigeons usually watched the course of the ball as it crossed the table, andmaneuvered into position to meet the return. They developed considerable skill insendingtheballstraightacross.Movingpicturesshowralliesofasmanyasfiveorsixshots before a point was made. There is no evidence, however, that either pigeonreached the stage of placing or changing the pace of its shots so that the opponentwouldmiss.

505

FIG.1.Twopigeons“playingping-pong.”

Conditioningwasbegunwithonepigeonatatime.Astandardtabletennisballwasfastenedattheedgeofthetable,andahungrypigeonwasreinforcedwithfoodwhenitpeckedit.Atthisstagetheballwasnotapowerfulcontrollingstimulus;whenitwasmoved to a different part of the edge, thepigeonsoftenpecked the airwhere it hadbeen.Eventually,however,theypeckedtheballregardlessofitsposition.Theballwasthenmade free to roll away from the pigeonwhen struck.Amechanical reinforcingsystem was set up in which the ball, rolling up a slight grade, struck a cross-baroperating the fooddispenser.The distance to the barwas gradually increased. If theball failed to reach the bar, it rolled back and came to rest against a raisedmoldingalongtheedge.Themoldingwaslaterremoved.As the distance between the pigeon and the reinforcing bar was increased,

reinforcement was more and more delayed, and the behavior occasionally suffered.Eventually,however,mediatingbehaviorarosetobridgethetemporalgap.Evenso,inthefinalgame,inwhichtwopigeonsparticipated,thedelaybetweenstrikingtheballand the successful outcome of getting the ball past the opponent was occasionallytroublesome. A deteriorating performance could be rescued by reinforcing a pigeonwith a hand-switch at the moment it struck the ball. Eventually the behavior wassustainednotonlyforralliesofseveralshotsatatimebutforafull“game.”The demonstration offers a convenient example of competition. One bird is

reinforcedat theexpenseofanother. Ifoneisrepeatedlysuccessful, theothersuffersextinction (“discouragement”). It was possible to maintain a reasonable balance insuccessful play by lowering theweight of the relatively unsuccessful bird or raisingthat of the successful, the principal effect being to sustain attention rather than alteraccuracyorpower.

CooperatingPigeons

Twopigeons inadjacentcompartmentswereseparatedbyapaneofglass.Three red

506

buttonswerearrangedinaverticalrowoneachsideoftheglass,asshowninFig.2.Thebuttonswereapproximately10in.,7½in.,and5in.fromthefloor,respectively.Bypeckingabuttonthepigeonclosedaswitch.Inthefinalperformance,bothpigeonswere reinforcedwith food (Fig.2,below)when theypeckedacorrespondingpairofbuttons so nearly simultaneously that the brief closures of the circuits (each lastingperhapsatenthofasecond)overlapped.Atanygiventime,however,onlyonepairofbuttonswasoperative,andtheeffectivepairwasscheduledinaroughlyrandomway.It was necessary for the pigeons to cooperate in two tasks: (1) discovering the

effectivepairand(2)peckingbothbuttonsatthesametime.Ingeneral,nopatternofexplorationcouldbeobserved.Thepigeonstestedallthreepairsofbuttonsinwhatwasevidentlyanunsystematicway.Ingeneral,therewasadivisionoflaborwithrespecttothetwotasks.Onepigeon(the“leader”)explored—thatis,itstruckthethreebuttonsinsomeorder.Asimilarperformancecouldhavebeengeneratedinonepigeonaloneintheapparatusbyrequiringsimplythatagivenoneofthreebuttonsbestruck.Theotherpigeon (the “follower”) struck the button opposite that being struck by the leader.Similarbehaviorcouldhavebeengeneratedinonepigeonaloneintheapparatusifonebuttonafteranotherhadbeenmarkedbyadiscriminativestimulus.Awell-markedleader-followerrelationcouldbeestablishedorreversedbyaltering

therelativeleveloffooddeprivation,themoredeprivedbirdassumingthepositionofleader bymovingmore alertly to the buttons. However, even a decisive leaderwasprobably to some extent following. A deprived pigeon would usually “wait to befollowed”byonelessdeprivedbeforeexploringthebuttonsvigorously.Underlevelsof deprivation at which both birds responded quickly and without interruption,performancebecame so perfect that it gave the impression of one pigeon seen in amirror.

507

508

FIG.2.Above:Twopigeonscooperatingbypeckingcorrespondingbuttonsatthesamemoment.Below:Pigeonseatingfromfooddispensers.

Theperformancewas establishedby conditioningeachbird separately topeck thethree buttons, reinforcement being roughly randomized. When sustained behavioroccurredonallthreebuttons,twobirdscouldbeputintheadjacentspacesforthefirsttime. The presence of another bird temporarily disturbed the performance, but bothbirds eventually began to respond to the buttons. At this stage responses tocorrespondingbuttonswithin,say,halfasecondofeachotherwouldtriggerbothfood-dispensers.Thesecontingenciessufficedtobuildcooperativebehaviorwithoutfurtherattention.Thevisualstimulationsuppliedbyonepigeonpeckingonabuttonbecameadiscriminativestimuluscontrollingaresponsetothecorrespondingbuttononthepartoftheother.Prolonged exposure to these conditions made pigeons strongly imitative in other

respects. They would often drink from glasses of water in the compartments at thesametime,forexample.Theextenttowhichtheirbehaviorsweremutuallycontrolledwasinformallydemonstratedwhentheexperimentwasshowntoagroupofbiologists,one of whom suggested putting the birds in the opposite compartments. The birdsimmediately lined up alongside the glass plate, facing away from the buttons. Theythusassumedtheirpreviouslyeffectivepositionsrelativetoeachother,butwerenowfacing the audience through the transparent front wall of the apparatus. Though nobuttons were available, they immediately began to cooperate in exploring acorrespondingarea,bobbingupanddowninaperfectmirror-imagepatternunderthecontrol of each other’s behavior. Possibly because the leader-follower relation hadfrequentlybeenshifted,eachbirdwasevidentlylargelycontrolledbythebehavioroftheother.

FromJournaloftheExperimentalAnalysisofBehavior,1962,5,531–533.ThisresearchwassupportedbyNSFGrantG18167.

509

ConcurrentActivityUnderFixed-IntervalReinforcement

(WITHW.H.MORSE)Undermany schedules of reinforcement there are significant periods of time duringwhichtheorganismdoesnotdisplaythebehaviorunderinvestigation.Forexample,ona fixed-interval schedule there is commonly a period of no responding afterreinforcement. The behavior which actually occurs at such times is apparentlyunrelated to any explicit reinforcement, yet the schedule is at least effective incontrolling the time of its occurrence. Interesting examples are found in the field ofpsychotic behavior, where the patient engages in compulsive or other idiosyncraticwaysonlywhenheisnotexecutingthebehaviorunderthecontrolofagivenschedule.What the organism is doing when it is not showing the behavior produced by aschedule of reinforcement is especially importantwhenwe come to set up complexbehaviorinwhichtwoormoreresponsesarestudiedatthesametime.

Method

Unreinforcedbehavioraccompanyingtheperformancegeneratedbyacommonscheduleofreinforcement was studied in the following way. A standard rat lever and foodmagazinewereinstalledalongsidetheopenfaceofanaluminumrunningwheel.Thewheelwas13in.indiameter,hada lowmomentof inertia,and turned inonlyonedirectionagainst frictionwhich was just overcome by a tangential force of 30 gm. Through a gear-reduction andpulleysystem the rotationof thewheeladvancedapenacross themovingpaper tapeofamodifiedcumulativerecorder.Scaleswerechosenso that thecommonestspeedofrunningproduced a slope in the neighborhood of 45°. Responses to the lever operated a standardcumulativerecorder.

Tworatswereconditionedtorespondtotheoperationofthemagazineuntiltheyquicklyseizedandate the .1-gm.pelletof fooddelivered.At thisstage thewheelwas immovable.After three sessions of magazine training all responses to the lever were reinforced forseveral sessions. A fixed-interval schedule was then put into effect, under which areinforcementwassetupevery5min.(FI5)duringasessionof3hr.and20min.Bothratsdevelopedmarked“scallops”duringthefirstsession,andthefixed-intervalperformancewasallowed to develop during several sessions until the pause after reinforcement was wellmarked.AtthebeginningoftheninthexperimentalsessiononFI5thewheelwasunlocked.Theratcouldnowrespondtotheleverorruninthewheel.(Itcouldnotdobothatthesametime because it faced away from the lever when running.) Because of its low inertia thewheelcouldbestoppedandstartedinstantly;henceachangecouldreadilybemadefromoneformofbehaviortotheother.ResponsestothelevercontinuedtobereinforcedonF15;noreinforcementwasexplicitlycontingentuponrunning.

510

Results

FIG.1

The first performancewhen thewheelwas unlocked is shown for one rat in Fig. 1.Segments A and A′ comprise a continuous record of the movement of the wheel.Segments B and B′ show the concurrent activity on the lever. Reinforcements aremarked by small hatches. Reinforcement a in Record A is the same event as thatmarkedata inRecordB, and the samecorrespondence ismaintained throughout thefigure. It will be seen that the movable wheel occupies the rat for nearly 20 min.(beginningofRecordA),duringwhichnoresponsesweremadetothelever(beginningofRecordB).Afirstresponsetotheleverisreinforcedata.TheratcontinuestoshowahighrateofrunningonthewheelbetweenreinforcementsaandcinRecordsAandA′,withcorrespondinglowactivityonthelevershownbetweenreinforcementsaandcinRecordB.Afewresponsesaremadetothelever,however,andreinforcementsarereceivedapproximatelyonschedule.Beforetheendofthesessiontheover-allrateofrunninghasfallenoffconsiderably

(endofA′),andafairlynormalfixed-intervalperformancehasbeenrestored(endofB′). Running during each interval already begins to show a temporal pattern, whichbecameconsistentinlatersessions.Afterreinforcementtheratneitherrunsnorpressestheleverforashortperiodoftime,whichismuchlongerthanthetimerequiredtoeatthepelletoffood.Itthenrunsactivelyfordistancesoftheorderof100to200ft.beforereturning to the lever to resume a fairly standard-interval performance. After thereinforcement marked n, for example, a short horizontal section appears in bothrecords. There is then a short burst of running in thewheel,marked by the verticaldotted lines in A′. The corresponding period in the lever record is indicated in B′.Runninginthewheelceasesduringthesecondhalfoftheinterval,whenfairlyconstantrespondingontheleverisshown,leadingtothereinforcementato.In order to obtain a collateral measure of the rate of unreinforced running, the

followingprogramwasinstituted.Atthebeginningofeachsessiontherathadaccesstothe wheel for approximately ½ hr., during which no responses to the lever werereinforced.A6-w.lamp,previouslyusedforgeneralilluminationintheapparatus,was

511

off during this period, and responses to the lever were quickly extinguished in theabsence of the light. The light was then turned on for 100 min., during which 20responseswerereinforcedonFI5.Thelightwasthenturnedoff,andfurtherrunningintheabsenceofrespondingtotheleverwasrecorded.Theendof sucha session for the rat inFig.1 is shown inFig.2. The curve atB

showsthelastfivesegmentsofthefixed-intervalperformanceprevailingatthisstageof the experiment. The segments at A andA′ comprise a continuous record of theperformanceinthewheelduringthefiveintervalsand,beginningata,afterthelighthasbeenturnedoffandrespondingtotheleverhasceased.Correspondingportionsareconnected by broken lines. The fixed-interval segments in B are now relatively“square”andsuggestthatrespondingisbeingpostponedbytheactivityinthewheel.When pressing begins, however, a fairly stable, uniform rate is assumed untilreinforcement.Forasubstantialperiodoftimeaftereachreinforcement,theleverisnotpressedandrunningdoesnotoccur.Sustainedrunningthentakesplacefor2or3min.,yieldingtothebehaviorontheleverfortherestoftheinterval.

512

FIG.2

ThebeginningofadailysessionfortheotherratatthesamestageasFig.2isshowninFig.3. Segment A and the first part of A′ show the relatively fast spontaneousrunning in the absence of the light, when responses to the lever are not reinforced.When a first response is reinforced in the presence of the light at a, aperformanceappropriatetoFI5isbegun(RecordB).Notethat,althoughthereislittleactivityontheleverduringthefirsttwointervals,runninginthewheelisgreatlysuppressed(comparethesegmentsprecedingbandcwithRecordA).Astableperformancepersistsfor thebalanceof the session,notallofwhich is shown in the figure.Fora shortperiodoftimeafterreinforcement theratdoesnot respondtoeither leverorwheel.Aburstofrunning then follows. This yields to responding to the lever in a somewhat morecontinuousfashionthaninthecaseof theotherrat.Therateofrunningfallsoff ina

513

particularly smoothcurveatd ande.ThesegmentsofRecordB showingconcurrentactivityontheleverhaveacorrespondingpositivecurvaturewhichisabsentinFig.2.AcontinuouschangeasinFig.3wasoccasionallyshownbythefirstrat,althoughthemorerapidshiftfromwheeltoleverseeninFig.2wascharacteristic.

FIG.3

Since responding on the wheel falls off to zero or very nearly zero late in eachinterval,fewifanyadventitiousreinforcementsofrunningcouldoccurastheresultofquickshiftsfromwheeltolever.Wemaythereforesupposethatactivityinthewheelremainsessentiallyunconditionedwithrespecttofoodreinforcement.

Summary

Whenaratisfreetoruninalow-inertiarunningwheelortopressaleverforfoodonafixed-intervalschedule,theresolutionofthecompetitionbetweenrunningandpressingcan be expressed in the following way. When the schedule normally generates asubstantialrateofresponding,runninginthewheelissuppressed.Whenthescheduledoes not generate a substantial rate, running in the wheel occurs. Shortly afterreinforcement,however,bothbehaviorsareabsent.

FromJournalofComparativeandPhysiologicalPsychology,1957,30,279–281.

514

SustainedPerformanceDuringVeryLongExperimentalSessions(WITHW.H.MORSE)

Operantbehaviorisusuallystudiedinexperimentalsessionsseparatedbyperiodswhentheorganismisnotcloselyobservedinitshomecage.Eventhoughtechnicaladvancesare now available for maintaining responding for extended periods of time, thebehavior under observation is seldom followed continuously 24 hours per day. Thepossibilityofdecrementsinbehavior(suchasoccurin“mentalfatigue”)promptedustoseejusthowlongananimalcouldbekeptrespondinginacontinuoussession.The schedule of positive reinforcement best suited to maintain responding

continuously is the differential reinforcement of very low rates. In the pigeon, thisschedulecansustainbehaviorwithafrequencyofreinforcementbelowthatrequiredtomaintain body weight.1 The effect of sustained responding on several differentperformances can be observed by combining other schedules with the differentialreinforcementoflowratesinamultipleprogram,providedthatreinforcementontheseotherschedulesoccursinfrequentlyenoughtopreventsatiation.In the present experiment a multiple schedule was used which had the following

components: a basic schedule,DRL,maintained responding at a fairly constant, lowrateunderthecontrolofonekey-color.Onceeveryhourthecolorofthekeychangedto that designating another schedule, which was then in effect until reinforcementoccurred. The other schedule was, alternately, either FR or FI. In this way, threedifferent performances could be studied during continuous experimental sessionslastingmanydaysorevenweeks.Thebirdremainedcontinuouslyintheexperimentalspace,andtheover-allrateofreinforcementmaintainedthebodyweight,withonlyaslowdrifttowardsatiationoramoreextremedeprivation.

Procedure

Astandardpigeonboxwasmodifiedtoaccommodateaconsiderablequantityofgrainin the food hopper. The magazine presented grain for 5 seconds. Fresh water wassupplied from theoutsideof thebox througha tube.TwoadultmaleWhiteCarneaupigeons were magazine-trained, and all responses to the key were reinforced (crf)during threesessionsfora totalof180reinforcementsperbird.Thecolorof thekeychanged after each reinforcement towhite, red, or green in randomorder.Thebirdswerethenrunfor12dailysessionslastingfrom4to7hourseachonDRL1min.inthepresenceofthewhitekey-light—thatis,aresponsewasreinforcedonlyifitfollowedaperiod of 1 minute without a response. If a response occurred before 1minute hadelapsedsincethepreviousresponse,thetimerprogrammingreinforcementwasresettozero,andthenextresponsewasreinforcedonlyif itoccurredafteranotherminuteormorehadelapsed.When a stable performance had developed on the DRL schedule, the multiple

515

schedulewas introduced.At1-hour intervals the colorof thekeywas changed fromwhite(DRL)toeitherredorgreen:whenitwasred,the50thresponsewasreinforced(FR50);whengreen,thefirstresponseafter10minuteswasreinforced(FI10).Afterreinforcement on either the fixed-ratio or fixed-interval component, the key-colorchangedtowhiteandtheDRLschedulewasagainineffect.Threecumulativerecordswere taken: one of the DRL performance (the recorder stopping when the otherscheduleswereinforce),oneofthealternatingFRandFIperformances(therecorderstopping duringDRL), and one of theDRL performance on a recorderwith greatlyreduced scales (one-twelfth the usual coordinates) to provide a short summarizingrecordofthewholeexperiment.Whenthemultipleschedulewasfirstintroduced,eachpigeonwasrunforseveralsessionsapproximately3dayslong.Later,thesessionwasallowedtocontinueuntilthebirdstoppedrespondingbecauseofsatiation(oruntiltheapparatusfailed).

Results

It was found that a pigeon could sustain a continuous performance indefinitely ifprecautionswere taken to prevent satiation.During a long session the fixed-intervalperformanceofthemultiplescheduleoccasionallyshowedunusualproperties,andallthreeperformanceschangedinexpecteddirectionsifsatiationoccurred.Typicaleffectsareshowninthecumulativerecordswhichfollow.In the transition to DRL from crf for Pigeon 162 (Fig. 1), a first response was

preceded by a pause ofmore than 1minute andwas reinforced (ata).A pattern ofrespondingtypicalofextinctionthenfollowed,asexpectedfromthepreviouscrf.

ACQUISITIONOFTHEDRLPERFORMANCE

516

FIG.1.Pigeon162.Cumulative-responsecurveduringthe1stsessionofDRL1minuteaftercrf.Seetextforexplanationofletters.

A pause sufficiently long to meet the DRL requirement finally occurred, and asecondresponsewasreinforced(atb).Anextinctioncurvebeginningatahigherratefollowed,andtheDRLcontingencywasagainmetatc.Theover-allrategraduallyfellto a fairly steady, low level, but occasional pauses were long enough to set upreinforcements,asatdande.Thelatterpartofthesessionisnotshown.Thedevelopmentofthe1-minuteDRLperformanceforPigeon163isshowninFig.

2.CurveBisthesummarizingrecordfromtherecorderwithreducedscales.RecordsAandC,segmentsfromtheotherDRLrecord,arelocatedasindicatedonRecordB.TheearlyperformanceofthisbirdissimilartotheearlystagesofFig.1.Bytheendofthesessiontheover-allratehasdeclined,but therequirementofa1-minutepauseismetonlyinfrequently.

517

FIG.2.Pigeon163.Performanceduring1stsessionofDRL1min.aftercrf.RecordsAandCareenlargedsegmentsofRecordB.

Further development of theDRL performance is shown in Fig.3. CurveA is forPigeon163,CurveBisforPigeon162.Bothcurvesaretakenfromthemiddleofthethird session, approximately 10 hours after the beginning of DRL. Although theprevailingratefluctuates,fewpausesofgreaterthan1minuteoccur.CurveC,shownonreducedcoordinates,isthelastDRLsessionforPigeon163.Notethattheratehereissouniformlysustainedabove1responseperminute thatonlythreereinforcementsare received during the 5½-hour session. A rate of this order is needed foruninterruptedexperimental sessionsof themultiple schedule since24 reinforcementswillbe receivedeachdayon thehourly ratioand intervalcomponents,andnotmorethananadditional10reinforcementsperday isneeded tomaintain thepigeon in therangeofitsexperimentalbodyweight.

518

FIG.3.PerformanceonDRL1min.RecordA(Pigeon163)andRecordB(Pigeon162)arefromthemiddleofthethirdsession.RecordC(Pigeon163)isthecomplete12thsessiononDRL

greatlyreduced.

PROLONGEDPERFORMANCEDURINGCONTINUOUSSESSIONS

The reduced summary record of the performance on the DRL component of themultiplescheduleforPigeon162duringasinglecontinuoussession154hourslongisshowninFig.4.Thiswas the sixth longsession for thisbird, earlier sessionshavingbeenfrom7 to17days long. (Thesummaryrecord for the first longsession for thisbirdhasbeenpublishedelsewhere.2)InFig.4therecorderwasstoppedwhiletheotherscheduleswereineffect,butallreinforcementsareshown.ThosereceivedonFRorFIcangenerallybeidentifiedbecausetheyoccurredexactly1hourapart.Itwillbeseenthat only an occasional reinforcement is received onDRL. Reinforcements onDRLfrequentlyoccurredeitherafterchangingfromtheFRorFIschedule(asataandb),orin pairs during the DRL component, as at c or d. (It is fairly characteristic ofrespondingonDRLthatreinforcementsarereceivedinpairs.)

519

FIG.4.Pigeon162.PerformanceonDRLcomponentofmultFI10FR50DRL1min.during6thextendedsession(6½dayslong).

InthesessionshowninFig.4 the rate fellbecauseofan increase inbodyweight.Reinforcement occurred more frequently as satiation progressed because the DRLcontingencywasmorefrequentlymet.ThelasttwolinesofFig.4showafrequencyofreinforcement which produced such complete satiation that the experiment wasstopped.Note that evenwhen the over-all frequency of reinforcement is fairly high,reinforcementsstilltendtobegroupedtogether,asate.DuringthelongsessionshowninFig.4,performancesonthefixed-ratioandfixed-

intervalcomponentswere recorded ina singlecumulativecurve.RecordAofFig.5showstheinitialperformanceonthefirstday.Thefirstintervalshowsnoscalloping.AshallowscallopischaracteristicofthestartofthesessiononFI,butthepresentcurveisunusuallystraight.Thismaybedueinparttothefactthatthebirdhadnotbeenintheapparatusfor5days(whiletheotherbirdhadbeenstudied).TheshallowscallopingoflaterintervalsinFig.5Amaybepartlyduetothefactthattheintervalcomponentsareseparatedby2hoursofotherschedules.ThisfactmayalsohaveabearingontheratherirregularperformanceshowninRecordB(asegmentbeginning45hoursafterthestartof the experiment) andRecordC (beginning after 105hours).Manyof the intervalsshowalongpausefollowedbyahighinitialrate,asataandb.“Knees”arecommoninRecords A and B, and are seen later (as at c). In general, the terminal rate of theintervalsisratherlowandthecurvesarefrequentlynegativelyacceleratedjustbeforereinforcement. Some of this irregularity may be due to induction from DRL. InmultipleschedulescontainingDRLcomponents,superstitious“markingtime”behavioris occasionally observed in other interval components. Occasionally, however, themoregradualincreaseinratetypicaloffixed-intervalreinforcementcanbeseen(asatf).

520

FIG.5.Pigeon162.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.during6thextendedsession.RecordAbeginsafter1hour,RecordBafter45hours,andRecordC

after105hoursfromthebeginningoftheexperiment.

Thefixed-ratioperformancescarcelychangedduringthepartof thesessionshowninFig.5.Slightpausessometimesoccuratthebeginningoftheratio(asatdande),buttheterminalrateremainshighthroughout.Both the fixed-ratio and fixed-interval performances change markedly toward the

endofthelongsession.Figure6,recordedimmediatelyafterFig.5C,showsthefixed-ratioandfixed-intervalperformancesformorethanafullday.Prolongedpausesnowoccurinintervalcomponents(asata),andterminalratesintheintervalsvarywidely(compareintervalsatb andc).The ratioperformance frequentlybeginswithapause(as at d and e), but the rate remains high after responding has begun. The alteredperformanceshown inFig.6 is presumably due to satiation; the record is similar tothose obtained during progressive satiation onmultiple fixed-ratio and fixed-intervalschedules.Thebirdhadbeguntheexperimentweighing390grams(80%free-feedingweight)andweighed440grams(about90%)whenremovedfromtheexperimentalbox7dayslater.

FIG.6.Pigeon162.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.during6thextendedsession.Recordbegins125hoursafterthebeginningoftheexperiment.

521

Satiation can be prevented by changing theDRLvalue or by presenting the otherschedules less frequently. In another extended session the bird shown in Fig. 4–6responded for28days continuously.ThevalueofDRLwasvariedbetween3 and4minutes,asneeded,topreventextremesatiationordeprivation.Duringthefirst2daystherewasaslight increase insatiationand thebirdpaused formore than10minutesuponninedifferentoccasions.Duringthenext6days,however,nopauselastedforaslong as 10minutes. Subsequently, one pause of 11minutes and one of 23minutesappeared,butthebirddidnotpauseforaslongas10minutesduringthenext6days.Neartheendoftheexperimentitagainrespondedfor4dayswithnopauselongerthan12minutes.The fixed-ratio and fixed-interval performances during this 28-day session are

sampled in Fig.7 and 8.Figure 7A begins 56 hours after the start of the session.Occasional “knees” appear in the interval curves (as ata andc), andmany intervalsshowtheabruptshifttoaterminalratefollowingtheinitialpause(asatb).Sustainedrespondingatarateappropriatetotheratiocomponentoccursatd,whichisunusualinan ordinary multiple schedule. The ratio performance, meanwhile, shows acharacteristichighratewithlittleornopausingatthestart.

FIG.7.Pigeon162.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.during7thextendedsession(28dayslong).RecordAbeginsafter56hours,andRecordBafter

264hours.

522

FIG.8.Pigeon162.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.during7thextendedsession.Recordbegins474hoursafterthebeginningoftheexperiment.

Figure7B was recorded after 264 hours. Although prolonged pauses and abruptshiftstoamoderateterminalratestillappear,somefairlycharacteristicintervalcurvesmaybeseen(asateandf).Aseriesofreasonablynormalintervalscallopsappearlaterin the session, as shown in Fig. 8, which begins after 474 hours of continuousresponding. Both the interval and ratio performances are normal except for anoccasionalslightpauseatthebeginningoftheratio,asata,b,andc.TheperformanceinFig.8 shows little or no effect of the intervening hours on DRL or of sustainedrespondingforalmost500hours.The other subject, Pigeon 163, showed more marked irregularities on the other

componentsof themultiple schedule.Figure9A, from the beginningof an extendedsession,showsarelativelylowrateontheratiocomponent,butafairlygoodscallopontheintervalcomponent.TheratiorateisstilllowinFig.9B,175hourslater,althoughitis now usually higher than the terminal rate in many of the intervals. “Knees” arecommon,asatb,andtheintervalcurvestendtoshownegativeacceleration,asataandc.AtthistimethebirdwasrespondingatasteadysustainedrateonDRL.

523

FIG.9.Pigeon163.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.during2ndextendedsession(24dayslong).RecordAbegins1hour,andRecordB175hours

afterthebeginningofthesession.

During another long session with Pigeon 163, the DRL rate was so consistentlyabove that required for reinforcement that at one point in the experiment noreinforcement was received on DRL for over 31 hours. In other words, the pigeonneverpausedforaslongas3minutesduringthatperiod.Thesustainedperformanceonthe DRL component must have been due at least in part to induction from theintervening FR and FI components; no comparable performance has ever beenobservedonasimpleDRLschedule.Aswith theotherbird,however, inductionwasobservedintheotherdirection.Figure10showstheFRandFIperformancesbeginning984hoursafterthestartofthesession.Ingeneral, theratioperformancesaretypical,andmanyoftheintervalcurvesarealsofairlynormal.Aseriesofintervalsegments,however(ata,b,andc,andagainatd ande), showsadrop toa low rateduring theinterval.ThissuggestsinterferencefromtheDRL.

524

FIG.10.Pigeon163.PerformanceonFIandFRcomponentsofmultFI10FR50DRL1min.duringthe3rdextendedsession(65dayslong).Recordbegins984hoursafterthebeginningofthe

session.

Summary

Thebehaviorofpigeonshasbeenstudiedcontinuously for sessionsofmanydaysorweeks by using a multiple schedule in which slow but continuous responding ismaintainedwith few reinforcementsonDRL,whileFIandFRschedulesalternateathourly intervals. To a remarkable degree the performance on the ratio and intervalcomponentsafterhundredsofhoursresemblesthatobtainedusingaconventionaldailysession.Themajordeviations in theratioandintervalperformanceduringthese longsessionsaremostlyattributableeithertochangesindeprivationortoaslightfailureinstimuluscontrol.Possible induction fromDRL toFI isoccasionally shownbya lowrateinFI.InductionfromFRtoFImayexplainanoccasionalunusuallyhighrateinFI.InductionfrombothFRandFIcomponentsappears tosustainperformanceformanyhours on DRL without reinforcement on that schedule, although “superstitious”reinforcements from the change to other key-colors may also be relevant. Thecombined schedule provides a base line of continuous responding which should beuseful in studying the effects of physiological, pharmacological, or other variablesactingoverlongperiodsoftime.

FromJournaloftheExperimentalAnalysisofBehavior,1958,1,235–244.1Herrnstein,R.J.,andMorse,W.H.Someeffectsofresponse-independentpositivereinforcementonmaintainedoperantbehavior,J.Comp.Physiol.Psychol.,1957,50,461–467.

2Skinner,B.F.;seepage132.

525

JohnBroadusWatson,Behaviorist

JohnBroadusWatson,self-styled“thebehaviorist,”diedon25September1958attheageof80.Hisscientificlifehadcometoacloseathirdofacenturyearlier,andhewasunknownpersonally to awhole generation of youngermenwhose field of scientificactivityhehaddefinedandvigorouslydeveloped.Hisplaceinthehistoryofscience,and somethingofhis stature, are indicatedby threenames—Darwin,LloydMorgan,andWatson—whichrepresentthreecriticalchangesinourconceptionofbehavior.Inestablishing thecontinuityofspeciesDarwinhadattributedmentalprocesses to

lowerorganisms.Hewassupportedbyahostofanecdotalnaturalistswhorecountedinstances of reasoning, sympathy, and even artistic enjoyment on the parts of dogs,cats, elephants, and soon.The inevitable reactionwas epitomized in thewritingsofLloydMorgan,whoarguedthatsuchevidencesofmentalprocessescouldbeexplainedinotherways.Athirdstepwasinevitable,anditwasWatsonwhotookit:Iftherewereotherexplanationsofmentalprocessesinlowerorganisms,whynotalsoinman?Indispensingwithmentalisticexplanationsofbehavior,Watsonclearedthewayfor

ascientificanalysis.IndoingsoheacknowledgedhisdebtbothtoLloydMorganandto Thorndike, who, though he remained a mentalist, supplied a classical alternativeexplanation of “reasoning” in his experiments on trial-and-error learning. Theepistemologicalissuewasalsointheair.Watsonnevertooktophilosophy(though,ashe later said, his “milk teethwere cut onmetaphysics”), but it wasGeorgeHerbertMead’s great personal interest in Watson’s animal experiments which supplied animmediate and crucial contact with relevant philosophical issues. A behavioristicinterpretation of mental processes was later adopted by operationists and logicalpositivists,buttheissuewastoremainprimarilyempiricalratherthanlogical.Born inGreenville,SouthCarolina,Watsonwas to rememberhimselfasabelow-

averageandtroublesomeschoolboywithlittletoshowforhisearlyeducationexceptaloveformanualskills.(Helaterbuiltaten-roomhousewithhisownhands.)HisfiveyearsatFurmanUniversity,wherehereceivedanA.M.insteadofanA.B.,werealsoremembered as a bitter disappointment. But his educational luck suddenly improvedwhenaninterestinphilosophytookhimtotheUniversityofChicago.Fewmenhavemade as many fortunate contacts during their graduate careers: with John Dewey(though Watson later complained “I never knew what he was talking about and,unfortunately,stilldon’t”);withAngell(whotaughthimtowrite);withJacquesLoeb(whomAngell thought “unsafe” asWatson’s thesis advisor); and, particularly, withMead. Under the Chicago influence his interests turned to biology, and he alwaysregretted that, in addition to his Ph.D., hewas not able to finishwork for theM.D.degreeatChicago.Attheageof29hewenttoJohnsHopkinsUniversityasprofessorof psychology, where he came into even closer contact with biologists andmedicalmen,particularlyJenningsandAdolphMeyer.AmongpsychologistsheworkedwithKnightDunlap andRobertYerkes (whowere later to formulate their ownvarietyof“psychobiology”) andwithCurtRichter andLashley, the latter fresh fromJennings’

526

laboratory.FromallthisexceptionalstimulationWatsonemergedwithaburningrecognitionof

theneedforascienceofbehavior.In1912,whenhefirstoutlinedhis“behaviorism,”there was no scientific discipline devoting itself to this important aspect of nature.Sociologistsandeconomistsfrequentlyconsideredthebehaviorofmen,butseldomofman as an individual. Psychology, in spite of the early American movement offunctionalism, was dominated by an introspective “science of mind” whichWatsonviewedwithanimpatiencewhichwasnevertobesatisfied.Inhismostimportantbook,PsychologyfromtheStandpointofaBehaviorist,published in1919,Watsondefinedthe field he wanted to see studied and assembled available techniques and facts. Asecond edition in 1924 contained a clearer and bolder programmatic statement. Theemphasiswasnecessarilyontheprogram,fornotmorethanone-thirdofeventhe1924edition contained facts strictly relevant to the science of behavior the author wasproposing. Anatomical and physiological material were used to complete the book.Watson’s own contributions were not great, and he was to have no opportunity toextendthem.Hisstudiesonmazebehaviorandhisconceptof“habit”madeanuneasymarriage with Pavlov’s principle of conditioning, then just beginning to attractattentioninthiscountry.Hisfrequencytheoryoflearningwasshort-lived.Inspiteof itsshortcomings thebookhada tremendouseffect.Thenewmovement

immediately attracted attention and adherents. Dissenters fell into line on the otherside.Inthecontroversywhichfollowed,Watson’stastefor,andskillin,polemicsledhimintoextremepositionsfromwhichheneverescaped.Hecouldnotcontenthimselfwithprosecuting an empirical studyofbehavior simply as such, forhebelieved thatpsychologywasthesciencedestinedtodealwiththatsubjectmatter,andhewantedtoreform it accordingly. He had another reason for crusading against the stronglyentrenchedintrospectionists,sincetheyclaimedtoofferdirectevidenceofthementalprocesseshewantedtodiscard.Watsonseizeduponlaryngealandothercovertverbalactivitiesas the“thoughtprocesses”of the introspectivepsychologistsandrefused toacknowledgesensoryaspectsofbehaviorwhichcouldalsobeobservedbythebehaverhimself. It hasbeen suggested thathemightnothimselfhavehadvisualor auditoryimagery. In any case his sweeping denial of the existence of self-observed sensoryevents(theacknowledgmentofwhichwouldnot,aswenowknow,haveimplied thedualism hewas so anxious to avoid) occupied him inwhat he later described as “acontinualstorm.”Thesametasteforpolemicsledhimintoanextremeenvironmentalisticposition.In

Psychology from the Standpoint of a Behaviorist he had devoted two chapters tohereditarybehavior.Likeall thosewhowanttodosomethingaboutbehavior,hehademphasized the possibility of environmental modification, and this was widelymisunderstood. Under the stress of battle he was led at last to the well-known cry:“Givemeadozenhealthyinfants,well-formed,andmyownspecifiedworldtobringthemupinandI’llguaranteetotakeanyoneatrandomandtrainhimtobecomeanytypeofspecialistImightselect—doctor,lawyer,artist,merchant-chief,andyes,evenbeggar-man and thief, regardless of his talents, penchants, tendencies, abilities,vocations,andraceofhisancestors.IamgoingbeyondmyfactsandIadmitit,butsohavetheadvocatesofthecontraryandtheyhavebeendoingitformanythousandsofyears.”Watsonalsowentbeyondhisfacts,andinthesamecrusadingspirit,inhisviewson

child training.Experimentson thebehaviorof infantshadshownhim thatemotional

527

patternscouldoftenbetracedtoconditionedemotionalreflexes(atermhetookfromPavlovviaLashley).Hethoughthesawtheseedsofmanybehaviorproblemsinearlyhomeexperiences,andinhisPsychologicalCareof theInfantandChild—abookhelater publicly regretted—he cautioned parents against the unconsidered display ofaffection.(Current“motherlove”theoriesaretheotherswingofthatpendulum.)Andso itcameabout thatWatsonwas toberememberedfora long time,byboth

laymenandpsychologistsalike,foratoonarrowinterpretationofself-observation,foranextremeenvironmentalism,andforacoldlydetachedtheoryofchildcare,nooneofwhichwasanecessarypartofhisoriginalprogram.Hisbrilliantglimpseof theneedfor, and the nature and implications of, a science of behaviorwas all but forgotten.Perhapshistoryisreadytoreturnamoreaccurateappraisal.AyearbeforehisdeathhehadthesatisfactionofdedicatingapaperbackeditionofhispopularbookBehaviorismtotheAmericanPsychologicalAssociation,whichon7September1957citedhimasfollows:“ToDr.JohnB.Watson,whoseworkhasbeenoneofthevitaldeterminantsofthe form and substance of modern psychology. He initiated a revolution inpsychologicalthought,andhiswritingshavebeenthepointofdepartureforcontinuinglinesoffruitfulresearch.”

FromScience,1959,129,197–198.

528

HowtoTeachAnimals

The“powerofpositivereinforcement”isprobablyneverfullyappreciateduntilonehas seen an actual demonstration. Several suggestions for casual experiments aremade in this article from the Scientific American (1951, 185, 26-29), which isreprintedherebypermission.

Teaching,itisoftensaid,isanart,butwehaveincreasingreasontohopethatitmayeventuallybecomeascience.Wehavealreadydiscoveredenoughaboutthenatureoflearning to devise training techniques which are much more rapid and give morereliableresultsthantherule-of-thumbmethodsofthepast.Testedonanimals,thenewtechniqueshaveprovedsuperiortotraditionalmethodsofprofessionalanimaltrainers;theyyieldmoreremarkableresultswithmuchlesseffort.Ittakesrathersubtlelaboratoryconditionstotestananimal’sfulllearningcapacity,

butthereadermaybesurprisedathowmuchhecanaccomplishevenunderinformalcircumstances at home. Since nearly everyone at some time or other has tried, orwished he knew how, to train a dog, a cat, or some other animal, perhaps themostuseful way to explain the learning process is to describe some simple experimentswhichthereadercanperformhimself.“Catchyourrabbit”isthefirstiteminawell-knownrecipeforrabbitstew.Yourfirst

move,ofcourse,istochooseanexperimentalsubject.Anyavailableanimal—acat,adog,apigeon,amouse,aparrot,achicken,apig—willdo.(Childrenorothermembersofyourfamilymayalsobeavailable,butitissuggestedthatyousavethemuntilyouhavehadpracticewithlessvaluablematerial.)Supposeyouchooseadog.The second thing you will need is something your subject wants, say food. This

serves as a reward or—to use a term which is less likely to be misunderstood—a“reinforcement”forthedesiredbehavior.Manythingsbesidesfoodarereinforcing—for example, simply letting the dog out for a run—but food is usually the easiest toadministerinthekindofexperimenttobedescribedhere.Ifyouusefood,youmustofcourse perform the experiment when the dog is hungry, perhaps just before hisdinnertime.The reinforcementgivesyouameansofcontrolling thebehaviorof theanimal. It

restson thesimpleprinciple thatwheneversomethingreinforcesaparticular formofbehavior,itincreasesthechancesthattheanimalwillrepeatthatbehavior.Thismakesitpossible toshapeananimal’sbehavioralmostasasculptorshapesa lumpofclay.Thereis,ofcourse,nothingnewinthisprinciple.Whatisnewisabetterunderstandingoftheconditionsunderwhichreinforcementworksbest.To be effective a reinforcement must be given almost simultaneously with the

desiredbehavior;adelayofevenoneseconddestroysmuchoftheeffect.Thismeansthatofferingfoodintheusualwayislikelytobeineffective;itisnotfastenough.Thebestwaytoreinforcethebehaviorwiththenecessaryspeedis tousea“conditioned”reinforcer.Thisisasignalwhichtheanimalhasobservedinassociationwithfood.The

529

animal is always given food immediately after the signal, and the signal itself thenbecomesareinforcer.Thebettertheassociationbetweenthetwoevents,thebettertheresult.Foraconditionedreinforceryouneedaclearsignalwhichcanbegiveninstantlyand

towhichthesubjectissuretorespond.Itmaybeanoiseoraflashoflight.Awhistleisnot recommendedbecauseof the time it takes todrawabreathbeforeblowing it.Avisual signal like a wave of the arm may not always be seen by the animal. Aconvenientsignal isa rapona tablewithasmallhardobjector thenoiseofahigh-pitcheddevicesuchasa“cricket.”You are now ready to start the experiment with your dog.Work in a convenient

placeasfreeaspossiblefromdistraction.Letussaythatyouhavechosena“cricket”asyourconditionerreinforcer.Tobuildupits reinforcingpowerbeginbytossingafewpiecesoffood,oneatatimeandnotoftenerthanonceortwiceaminute,wherethedogmay eat them. Use pieces so small that 30 or 40 will not appreciably reduce theanimal’shunger.Assoonasthedogeatspiecesreadilyandwithoutdelay,begintopairthecricketwiththefood.Soundthecricketandthencossapieceoffood.Waithalfaminute or so and repeat. Sound the cricket suddenly, without any preparatorymovementsuchasreachingforfood.Atthisstageyoursubjectwillprobablyshowwell-markedbeggingbehavior.Itmay

watchyouintently,perhapsjumponyou,andsoon.Youmustbreakupthisbehavior,becauseitwillinterferewithotherpartsoftheexperiment.Neversoundthecricketorgive foodwhen thedog isclose toyouor facingyou.Waituntil it turnsaway, thenreinforce.Your conditioned reinforcerwill beworking properlywhen the dog turnsimmediately andapproaches the spotwhere it receives food.Test this several times.Waituntilthedogisinafairlyunusualposition,thensoundthesignal.Timespentinmakingsurethedogimmediatelyapproachesthefoodwilllaterbesavedmanyfold.Now,havingestablishedthenoiseasareinforcer,youmaybeginteachingthedog.

Togetthefeelofthetechniquestartwithsomesimpletask,suchasgettingthedogtoapproach thehandleona lowcupboarddoor and touch itwith itsnose.At first youreinforceanyactivitywhichwillbepartofthefinalcompletedactofapproachingandtouchingthehandleofthecupboard.Theonlypermissiblecontactbetweenyouandthedogisviathecricketandthefood.Donottouchthedog,talktoit,coaxit,“drawitsattention,”orinterfereinanyotherwaywiththeexperiment.Ifyoursubjectjustsits,youmayhavetobeginbyreinforcinganymovement,howeverslight.Assoonasthedog moves, sound the cricket and give food. Remember that your reaction time isimportant.Trytoreinforceasnearlysimultaneouslywiththemovementaspossible.Afteryoursubjecthasbeguntomoveabout,reinforcewhenitturnstothecupboard.

Almost immediately you will notice a change in its behavior. It will begin to facetoward the cupboard most of the time. Then begin to reinforce only when the dogmovesnearerthecupboard.(Ifyouwithholdreinforcementtoolongatthisstage,youmay lose the facing response. If so, go back and pick it up.) In a very short time—perhaps aminute or two—you should have the dog standing close to the cupboard.Nowbegintopayattentiontoitshead.Reinforceanymovementwhichbringsthenoseclose to the handle. You will have to make special efforts now to reduce the timebetweenthemovementandthereinforcementtotheveryminimum.Presentlythedogwilltouchthehandlewithitsnose,andafterreinforcementitwillrepeatthisbehaviorsolongasitremainshungry.Usuallyittakesnomorethanfiveminutes,evenforabeginner,toteachadogthis

530

behavior. Moreover, the dog does not have to be particularly “smart” to learn it;contrary to the usual view, all normal dogs learnwith about equal facilitywith thisconditioningtechnique.Beforegoingonwithotherexperimentstesttheeffectofyourconditionedreinforcer

againtwoorthreetimes.Ifthedogrespondsquicklyandeatswithoutdelayyoumaysafely continue. You should “extinguish” the response the dog has already learned,however,beforeteachingitanother.Stopreinforcingtheactoftouchingthecupboardhandleuntilthedogabandonsthisactivity.Asasecondtest,letussay,youwanttoteachthedogtoliftitsheadintheairand

turnaround to the right.Thegeneralprocedure is the same,butyoumayneedsomehelpinsharpeningyourobservationofthebehaviortobereinforced.Asaguidetotheheight towhichthedog’sheadis toberaised,sightsomehorizontal lineonthewallacrosstheroom.Wheneverthedog,initsrandommovements,liftsitsheadabovethisline,reinforceimmediately.Youwillsoonseetheheadrisingabovethelinemoreandmorefrequently.Nowraiseyoursightsslightlyandreinforceonlywhentheheadrisesabovethenewlevel.Byaseriesofgradualstepsyoucangetthedogtoholditsheadmuchhigherthanusual.Afterthisyoucanbegintoemphasizeanyturningmovementinaclockwisedirectionwhile thehead ishigh.Eventually thedogshouldexecuteakindofdancestep.Ifyouuseavailablefoodcarefully,asinglesessionshouldsufficeforsettingupthisbehavior.Havingtestedyourabilitytoproducethesesimpleresponses,youmayfeelconfident

enough toapproachamorecomplexassignment.This timesupposeyou tryworkingwith a pigeon. Pigeons do not tame easily. You will probably want a cage to helpcontrolthebird,andforthisyoucanrigupalargecardboardcartonwithascreenorlatticetopandwindowscutinthesideforobservingthebird.Itismuchlessdisturbingto the bird if youwatch it from below its line of vision than if you peer at it fromabove.Ingeneralkeepyourselfoutoftheexperimentalsituationasmuchaspossible.Youmaystilluseacricketasaconditionedreinforcer,andfeedthebirdbydroppingafew grains of pigeon food into a small dish through a hole in thewall. Itmay takeseveraldailyfeedingstogetthebirdtoeatreadilyandtorespondquicklytothecricket.Your assignment is to teach the pigeon to identify the visual patterns on playing

cards.Tobeginwith,hangasinglecardonanailonthewallofthecageafewinchesabovethefloorsothatthepigeoncaneasilypeckit.Afteryouhavetrainedthebirdtopeck the cardby reinforcing themovementswhich lead to that end, change the cardandagainreinforcethepeck.Ifyoushufflethecardsandpresentthematrandom,thepigeonwilllearntopeckanycardoffered.Now begin to teach it to discriminate among the cards. Let us say you are using

diamonds and clubs (excluding face cards and aces) and want the bird to selectdiamonds.Reinforce onlywhen the card presented is a diamond, neverwhen it is aclub.Almostimmediatelythebirdwillbegintoshowapreferencefordiamonds.Youcan speed up its progress toward complete rejection of clubs by discontinuing theexperiment for amoment (amild form of punishment)whenever it pecks a club.Agoodconditionedpunishmentissimplytoturnoffthelight[“blackingout”]orcoverorremovethecard.Afterhalfaminutereplacethecardorturnonthelightandcontinuetheexperiment.Undertheseconditionstheresponsewhichispositivelyreinforcedwithfood remains part of the repertoire of the bird,while the responsewhich leads to ablackoutquicklydisappears.Thereisanamusingvariationofthisexperimentbywhichyoucanmakeitappear

531

thatapigeoncanbetaughttoread.YousimplyusetwoprintedcardsbearingthewordsPECKandDON’TPECK, respectively.Byreinforcingresponses toPECKandblackingoutwhen the bird pecksDON’TPECK, it is quite easy to train the bird to obey thecommandsonthecards.The pigeon can also be taught the somewhat more “intellectual” performance of

matchingasampleobject.Letussaythesampletobematchedisacertaincard.Fastenthreecardstoaboard,withoneaboveandthetwootherssidebysidejustbelowit.Theboardisplacedsothatthebirdcanreachallthecardsthroughwindowscutinthesideofthecage.After trainingthebirdtopeckacardofanykindimpartially inall threepositions,present the threechosencards.Thesample tobematched,say the threeofdiamonds,isatthetop,andbelowitputathreeofdiamondsandathreeofclubs.Ifthebirdpecksthesamplethreeofdiamondsatthetop,donothing.Ifitpecksthematchingthreeofdiamondsbelow, reinforce it; if it pecks the threeof clubs,blackout.Aftereachcorrectresponseandreinforcement,switchthepositionsofthetwolowercards.Thepigeonshouldsoonmatchthesampleeachtime.Conversely,itcanalsobetaughttoselectthecardwhichdoesnotmatchthesample.Itisimportanttoreinforcecorrectchoices immediately.Your own behaviormust be letter-perfect if you are to expectperfectionfromyoursubject.Thetaskcanbemadeeasierifthepigeonisconditionedtopeckthesamplecardbeforeyoubegintotrainittomatchthesample.Inamoreelaboratevariationofthisexperimentwehavefounditpossibletomakea

pigeonchooseamong fourwords so that it appears to“name the suit”of the samplecard.Youpreparefourcardsaboutthesizeofsmallcallingcards,eachbearinginblockletters the name of a suit: SPADES,HEARTS,DIAMONDS, andCLUBS. Fasten thesesidebysideinarowandteachthepigeontopeckthembyreinforcingintheusualway.Now arrange a sample playing card just above them. Cover the name cards andreinforcethepigeonafewtimesforpeckingthesample.Nowpresent,say,thethreeofdiamonds as the sample.When the pigeon pecks it, immediately uncover the namecards. If thepigeonpecksDIAMONDS, reinforce instantly. If it pecks awrongnameinstead,blackoutforhalfaminuteandthenresumetheexperimentwiththethreeofdiamondsstillinplaceandthenamecardscovered.Afteracorrectchoice,changethesample card to a different suit while the pigeon is eating. Always keep the namescovereduntilthesamplecardhasbeenpecked.Withinashorttimeyoushouldhavethebirdfollowingthefullsequenceofpeckingthesampleandthentheappropriatenamecard.Astimepassesthecorrectnamewillbepeckedmoreandmorefrequentlyand,ifyoudonottoooftenreinforcewrongresponsesorneglecttoreinforcerightones,thepigeonshouldsoonbecomeletter-perfect.Atoypianooffersinterestingpossibilitiesforperformancesofamoreartisticnature.

Reinforceanymovementofthepigeonthatleadstowarditspressingakey.Then,byusingreinforcementsandblackoutsappropriately,narrowtheresponsetoagivenkey.Thenbuildup a two-note sequenceby reinforcingonlywhen the sequencehasbeencompletedandbyblackingoutwhenanyothercombinationofkeysisstruck.Thetwo-notesequencewillquicklyemerge.Othernotesmaythenbeadded.Pigeons,chickens,smalldogs,andcatshavebeentaught in thiswaytoplaytunesoffourorfivenotes.The situation soon becomes too complicated, however, for the casual experimenter.You will find it difficult to control the tempo, and the reinforcing contingenciesbecomeverycomplex.Thelimitofsuchanexperimentisdeterminedasmuchbytheexperimenter’s skill asby thatof theanimal. In the laboratorywehavebeenable to

532

provideassistancetotheexperimenterbysettingupcomplicateddeviceswhichalwaysreinforceconsistentlyandavoidexhaustionoftheexperimenter’spatience.The increased precision of the laboratory also makes it possible to guarantee

performance up to the point of almost complete certainty.When relevant conditionshavebeencontrolled,thebehavioroftheorganismisfullydetermined.Behaviormaybe sustained in full strength for many hours by utilizing different schedules ofreinforcement.Someofthesecorrespondtothecontingenciesestablishedinindustryindaily wages or in piece-work pay; others resemble the subtle but powerfulcontingenciesofgamblingdevices,whicharenotorious for their ability tocommandsustainedbehavior.Thehumanbabyisanexcellentsubject inexperimentsof thekinddescribedhere.

You will not need to interfere with feeding schedules or create any other state ofdeprivation,becausethehumaninfantcanbereinforcedbyverytrivialenvironmentalevents; it does not need such a reward as food. Almost any “feedback” from theenvironmentisreinforcingifitisnottoointense.Acrumplednewspaper,apanandaspoon, or any convenient noisemaker quickly generates appropriate behavior, oftenamusinginitsviolence.Thebaby’srattleisbaseduponthisprinciple.One reinforcer towhichbabiesoften respond is the flashingonandoffof a table

lamp. Select some arbitrary response—for example, lifting the hand. Whenever thebaby lifts its hand, flash the light. In a short time a well-defined response will begenerated. (Human babies are just as “smart” as dogs or pigeons in this respect.)Incidentally,thebabywillenjoytheexperiment.Thesameprincipleisatworkinthebehaviorofolderchildrenandadults.Important

amonghuman reinforcements are those aspects of the behavior of others, often verysubtle, which we call “attention,” “approval” and “affection.” Behavior which issuccessful inachieving these reinforcementsmaycome todominate the repertoireoftheindividual.Allthismaybeeasilyused—andjustaseasilymisused—inourrelationswithother

people.Tothereaderwhoisanxioustoadvancetothehumansubjectawordofcautionis in order. Reinforcement is only one of the procedures through which we alterbehavior.Touseit,wemustbuildupsomedegreeofdeprivationoratleastpermitadeprivationtoprevailwhichitiswithinourpowertoreduce.Wemustembarkuponaprograminwhichwesometimesapplyrelevantreinforcementandsometimeswithholdit. In doing this,we are quite likely to generate emotional effects.Unfortunately thescienceofbehavior isnotyetassuccessful incontrollingemotionas it is inshapingpracticalbehavior.A scientific analysis can, however, bring about a better understanding of personal

relations.Wearealmostalwaysreinforcingthebehaviorofothers,whetherwemeantobe or not. A familiar problem is that of the child who seems to take an almostpathological delight in annoying its parents. In many cases this is the result ofconditioning which is very similar to the animal training we have discussed. Theattention, approval, and affection which a mother gives a child are all extremelypowerful reinforcements. Any behavior of the child which produces theseconsequences is likely to be strengthened. Themothermay unwittingly promote theverybehavior shedoesnotwant.Forexample,whenshe isbusyshe is likelynot torespondtoacallor requestmade inaquiet toneofvoice.Shemayanswer thechildonly when it raises its voice. The average intensity of the child’s vocal behaviortherefore moves up to another level—precisely as the head of the dog in our

533

experimentwas raised toanewheight.Eventually themothergetsused to this levelandagainreinforcesonlylouderinstances.Thisviciouscirclebringsaboutlouderandlouderbehavior.Thechild’svoicemayalsovaryinintonation,andanychangeinthedirection of unpleasantness is more likely to get the attention of the mother and istherefore strengthened. One might even say that “annoying” behavior is just thatbehaviorwhichisespeciallyeffectiveinarousinganotherpersontoaction.Themotherbehaves, in fact, as if shehadbeengiven the assignmentof teaching the child tobeannoying!Theremedy insuchacase issimplyfor themother tomakesure that sherespondswithattentionandaffectiontomostifnotalltheresponsesofthechildwhichareofacceptableintensityandtoneofvoiceandthatsheneverreinforcestheannoyingformsofbehavior.

534

BabyinaBox

Since publication of this article in the Ladies Home Journal in October, 1945,severalhundredbabieshavebeenrearedinwhatisnowknownasan“Air-Crib.”Theadvantagesreportedherehavebeengenerouslyconfirmed.Althoughculturalinertiaisperhapsnowheremorepowerfulthaninchild-raisingpractices,andinspiteofthefactthatthedeviceisnoteasytobuild,itsusehassteadilyspread.Theadvantagestothechild and parent alike seem to be too great to be resisted.One early user, JohnM.Gray,sentaquestionnaire to73coupleswhohadusedAir-Cribs for130babies.Allbutthreedescribedthedeviceas“wonderful.”Thephysicalandpsychologicalbenefitsreportedbytheseusersseemtowarrantextensiveresearch.

Inthatbravenewworldwhichscienceispreparingforthehousewifeofthefuture,theyoungmotherhasapparentlybeenforgotten.Almostnothinghasbeendonetoeaseherlotbysimplifyingandimprovingthecareofbabies.Whenwedecidedtohaveanotherchild,mywifeandIfeltthatitwastimetoapplya

little labor-saving inventionanddesign to theproblemsof thenursery.Webeganbygoing over the disheartening schedule of the youngmother, step by step.We askedonlyonequestion:Isthispracticeimportantforthephysicalandpsychologicalhealthofthebaby?Whenitwasnot,wemarkeditforelimination.Thenthe“gadgeteering”began.Theresultwasaninexpensiveapparatusinwhichourbabydaughterhasnowbeen

living for elevenmonths.Her remarkable good health and happiness andmywife’swelcomeleisurehaveexceededourmostoptimisticpredictions,andweareconvincedthatanewdealforbothmotherandbabyisathand.Wetackledfirst theproblemofwarmth.Theusualsolution is towrap thebaby in

half-a-dozen layers of cloth—shirt, nightdress, sheet, blankets. This is nevercompletely successful. The baby is likely to be found steaming in its own fluids orlyingcoldanduncovered.Schemes topreventuncoveringmaybedangerous, and infact theyhavesometimesevenprovedfatal.Clothingandbeddingalsointerferewithnormal exercise and growth and keep the baby from taking comfortable postures orchangingpostureduringsleep.Theyalsoencouragerashesandsores.Nothingcanbesaid for the system on the score of convenience, because frequent changes andlaunderingsarenecessary.

535

FIG.1

Why not, we thought, dispense with clothing altogether—except for the diaper,which serves another purpose—and warm the space in which the baby lives? Thisshould be a simple technical problem in themodern home.Our solution is a closedcompartment about as spacious as a standard crib (Figure 1). The walls are well

536

insulated,andoneside,whichcanberaised likeawindow, isa largepaneof safetyglass. The heating is electrical, and special precautions have been taken to insureaccuratecontrol.After a little experimentation we found that our baby, when first home from the

hospital,wascompletelycomfortableandrelaxedwithoutbenefitofclothingatabout86°F.Asshegrewolder,itwaspossibletolowerthetemperaturebyeasystages.Now,atelevenmonths,weareoperatingatabout78°,witharelativehumidityof50percent.Raising or lowering the temperature by more than a degree or two produces a

surprisingchange in thebaby’sconditionandbehavior.This response is so sensitivethat we wonder how a comfortable temperature is ever reached with clothing andblankets.Thediscoverywhichpleasedusmostwas thatcryingandfussingcouldalwaysbe

stoppedbyslightlyloweringthetemperature.Duringthefirstthreemonths,itistrue,the babywould also crywhenwet or hungry, but in that case shewould stopwhenchangedorfed.Duringthepastsixmonthsshehasnotcriedatallexceptforamomentor twowhen injuredorsharplydistressed—forexample,when inoculated.The“lungexercise”whichsooften isappealed to toreassure themotherofababywhocriesagooddealtakesthemuchpleasanterformofshoutsandgurgles.Howmuchof this sustained cheerfulness is due to the temperature is hard to say,

becausethebabyenjoysmanyotherkindsofcomfort.Shesleepsincuriouspostures,nothalfofwhichwouldbepossibleundersecurelyfastenedblankets.Whenawake,sheexercisesalmostconstantlyandoftenwithsurprisingviolence.Her

leg,stomach,andbackmusclesareespeciallyactiveandhavebecomestrongandhard.Itisnecessarytowatchthisperformanceforonlyafewminutestorealizehowseverelyrestrained the averagebaby is, andhowmuchenergymustbediverted into theonlyremainingchannel—crying.Awider range and variety of behavior are also encouraged by the freedom from

clothing.Forexample,ourbabyacquiredanamusing,almostapelikeskillintheuseofherfeet.Wehavedevisedanumberoftoyswhichareoccasionallysuspendedfromtheceilingofthecompartment.Sheoftenplayswiththesewithherfeetaloneandwithherhandsandfeetincloseco-operation.One toy isa ringsuspendedfromamodifiedmusicbox.Anotecanbeplayedby

pullingtheringdownward,andaseriesofrapidjerkswillproduceThreeBlindMice.Atsevenmonthsourbabywouldgrasptheringinhertoes,stretchoutherlegandplaythetunewitharhythmicmovementofherfoot.We are not especially interested in developing skills of this sort, but they are

valuableforthebabybecausetheyarouseandholdherinterest.Manybabiesseemtocryfromsheerboredom—theirbehaviorisrestrainedandtheyhavenothingelsetodo.Inourcompartment,thewakinghoursareinvariablyactiveandhappyones.Freedom fromclothes andbedding is especially important for theolderbabywho

playsandfallsasleepoffandonduringtheday.Unlessthemotherisconstantlyonthealert, it is hard to cover the baby promptly when it falls asleep and to remove andarrangesheetsandblanketsassoonasitisreadytoplay.Allthisisnowunnecessary.Remember that these advantages for the baby do not mean additional labor or

attentionon the part of themother.On the contrary, there is an almost unbelievablesaving in timeandeffort.Forone thing, there isnobed tobemadeorchanged.The“mattress”isatightlystretchedcanvas,whichiskeptdrybywarmair.Asinglebottom

537

sheetoperateslikearoller towel.1 It isstoredonaspooloutside thecompartmentatoneendandpassesintoawirehamperattheother.Itistenyardslongandlastsaweek.Acleansectioncanbelockedintoplaceinafewseconds.Thetimewhichisusuallyspent in changing clothes is also saved. This is especially important in the earlymonths.When we take the baby up for feeding or play, she is wrapped in a smallblanketorasimplenightdress.Occasionallysheisdressedup“forfun”orforherplayperiod.Butthatisall.Thewrappingblanket,rollersheet,andtheusualdiapersaretheonlylaundryactuallyrequired.Timeandlaborarealsosavedbecausetheairwhichpassesthroughthecompartment

is thoroughly filtered. The baby’s eyes, ears, and nostrils remain fresh and clean.Aweekly bath is enough, provided the face and diaper region are frequently washed.Theselittleattentionsareeasybecausethecompartmentisatwaistlevel.Ittakesaboutoneandone-halfhourseachdaytofeed,change,andotherwisecare

forthebaby.Thisincludeseverythingexceptwashingdiapersandpreparingformula.Wearenotinterestedinreducingthetimeanyfurther.Asababygrowsolder,itneedsa certain amount of social stimulation.And after all,when unnecessary chores havebeeneliminated,takingcareofababyisfun.An unforeseen dividend has been the contribution to the baby’s good health.Our

pediatricianreadilyapprovedtheplanbeforethebabywasborn,andhehasfollowedthe resultsenthusiastically frommonth tomonth.Herearesomepointson thehealthscore:Whenthebabywasonlytendaysold,wecouldplaceherinthepreferredface-down positionwithout danger of smothering, and she has slept thatway ever since,with theusualadvantages.Shehasalwaysenjoyeddeepandextendedsleep,andherfeedingandeliminativehabitshavebeenextraordinarilyregular.Shehasneverhadastomachupset,andshehasnevermissedadailybowelmovement.Thecompartmentisrelativelyfreeofsprayandair-borneinfection,aswellasdust

andallergicsubstances.Althoughtherehavebeencoldsinthefamily,ithasbeeneasytoavoidcontagion,andthebabyhascompletelyescaped.Theneighborhoodchildrentroopintoseeher,buttheyseeherthroughglassandkeeptheirschool-agediseasestothemselves.Shehasneverhadadiaperrash.Wehavealsoenjoyedtheadvantagesofafixeddailyroutine.Childspecialistsare

stillnotagreedastowhetherthemothershouldwatchthebabyortheclock,butnoonedeniesthatastrictschedulesavestime,forthemothercanplanherdayinadvanceandfind time for relaxation or freedom for other activities. The trouble is that a routineacceptable to the baby often conflicts with the schedule of the household. Ourcompartment helps out here in twoways.Even in crowded living quarters it can bekeptfreeofunwantedlightsandsounds.Theinsulatedwallsmuffleallordinarynoises,andacurtaincanbedrawndownoverthewindow.Theresultisthat,inthespacetakenbyastandardcrib,thebabyhasineffectaseparateroom.Weareneverconcernedlestthedoorbell,telephone,piano,orchildrenatplaywakethebaby,andwecanthereforelethersetupanyroutineshelikes.But a more interesting possibility is that her routine may be changed to suit our

convenience.A good example of this occurredwhenwe dropped her schedule fromfourtothreemealsperday.Thebabybegantowakeupinthemorningaboutanhourbeforewewanted to feedher.Thisannoyinghabit,onceestablished,maypersist formonths.However,byslightlyraisingthetemperatureduringthenightwewereabletopostpone her demand for breakfast. The explanation is simple. The eveningmeal is

538

usedbythebabymainlytokeepherselfwarmduringthenight.Howlongitlastswilldependinpartuponhowfastheatisabsorbedbythesurroundingair.One advantage not to be overlooked is that the soundproofing also protects the

familyfromthebaby!Ourintentionsinthisdirectionweremisunderstoodbysomeofourfriends.Wewereneverputtothetest,becausetherewasnocryingtocontendwith,butitwasneverourpolicytousethecompartmentinordertoletthebaby“cryitout.”Everyeffortshouldbemadetodiscoverjustwhyababycries.Butifthecondition

cannotberemedied,thereisnoreasonwhythefamily,andperhapstheneighborhoodas well, must suffer. (Such a compartment, by the way, might persuade many alandlordtodropa“nobabies”rule,sinceothertenantscanbecompletelyprotected.)Before thebabywasborn,whenwewerestillbuilding theapparatus, someof the

friends andacquaintanceswhohadheard aboutwhatweproposed todowere rathershocked. Mechanical dish-washers, garbage disposers, air cleaners, and otherlaborsaving devices were all very fine, but a mechanical baby tender—that wascarryingsciencetoofar!However,allthespecificobjectionswhichwereraisedagainsttheplanhavefadedawayinthebright lightofourresults.Averybriefacquaintancewith the scheme in operation is enough to resolve all doubts. Some of the toughestskepticshavebecomeourmostenthusiasticsupporters.Oneof the commonest objectionswas thatwewere going to raise a “softie”who

wouldbeunpreparedfor the realworld.But insteadofbecominghypersensitive,ourbabyhasacquiredasurprisinglyserenetoleranceforannoyances.Sheisnotbotheredbytheclothesshewearsatplaytime,sheisnotfrightenedbyloudorsuddennoises,sheisnotfrustratedbytoysoutofreach,andshetakesalotofpommelingfromheroldersister like a good sport. It is possible that shewill have to learn to sleep in a noisyroom,butadjustmentsofthatsortarealwaysnecessary.Atoleranceforanyannoyancecan be built up by administering it in controlled dosages, rather than in the usualaccidentalway.Certainlythereisnoreasontoannoythechildthroughoutthewholeofitsinfancy,merelytoprepareitforlaterchildhood.Itisnot,ofcourse,thefavorableconditionstowhichpeopleobject,butthefactthat

inourcompartmenttheyare“artificial.”Allofthemoccurnaturallyinonefavorableenvironmentoranother,wherethesameobjectionshouldapplybutisneverraised.Itisquite in thespiritof the“worldof thefuture” tomakefavorableconditionsavailableeverywherethroughsimplemechanicalmeans.Afewcriticshaveobjectedthat theywouldnot liketoliveinsuchacompartment

themselves—theyfeelthatitwouldstiflethemorgivethemclaustrophobia.Thebabyobviouslydoesnotshareinthisopinion.Thecompartmentiswellventilatedandmuchmore spacious than aPullmanberth, considering the size of the occupant.Thebabycannotgetout,ofcourse,butthatistrueofacribaswell.Thereislessactualrestraintinthecompartmentbecausethebabyisfreertomoveabout.Theplainfactisthatsheisperfectlyhappy.Shehasnevertriedtogetoutnorresistedbeingputbackin,andthatseemstobethefinaltest.Anotherearlyobjectionwasthatthebabywouldbesociallystarvedandrobbedof

the affection andmother lovewhich she needs. This has simply not been true. Thecompartment does not ostracize the baby. The largewindow is nomore of a socialbarrierthanthebarsofacrib.Thebabyfollowswhatisgoingonintheroom,smilesatpassers-by,plays“peek-a-boo”games,andobviouslydelightsincompany.Andsheishandled,talkedto,andplayedwithwheneversheischangedorfed,andeachafternoonduringaplayperiodwhichisbecominglongerasshegrowsolder.

539

Thefact is thatababywillprobablygetmore loveandaffectionwhen it iseasilycaredfor,becausethemother isnotsolikelytofeeloverworkedandresentfulof thedemandsmadeuponher.Shewillexpressherloveinapracticalwayandgivethebabygenuinelyaffectionatecare.Itiscommonpracticetoadvisethetroubledmothertobepatientandtenderandto

enjoyherbaby.And,ofcourse,thatiswhatanybabyneeds.Butitistheexceptionalmother who can fill this prescription upon demand, especially if there are otherchildreninthefamilyandshehasnohelp.Weneedtogoonestepfurtherandtreatthemotherwithaffectionalso.Simplifiedchildcarewillgivemotherloveachance.Asimilarcomplaintwasthatsuchanapparatuswouldencourageneglect.Buteasier

careissuretobebettercare.Themotherwillresistthetemptationtoputthebabybackintoadampbed if shecanconjureupadryone in fiveseconds.Shemayverywellspendlesstimewithherbaby,butbabiesdonotsufferfrombeingleftalonebutonlyfromthediscomfortswhicharisefrombeingleftaloneintheordinarycrib.Howlongdoweintendtokeepthebabyinthecompartment?Thebabywillanswer

thatintime,butalmostcertainlyuntilsheistwoyearsold,orperhapsthree.Afterthefirstyear,ofcourse,shewillspendafairpartofeachdayinaplay-penorout-of-doors.Thecompartmenttakestheplaceofacribandwillgetaboutthesameuse.Eventuallyitwillserveassleepingquartersonly.We cannot, of course, guarantee that every baby raised in thiswaywill thrive so

successfully.Butthereisaplausibleconnectionbetweenhealthandhappinessandthesurroundingswehaveprovided,andIamquitesurethatoursuccessisnotanaccident.The experiment should, of course, be repeated again and againwith different babiesanddifferentparents.Onecaseisenough,however,todisprovetheflatassertionthatitcan’tbedone.Atleastwehaveshownthatamoderateandinexpensivemechanizationofbabycarewillyielda tremendoussaving in timeand trouble,withoutharmto thechildandprobablytoitslastingadvantage.

AWordAboutBoxes

Thetitle“BabyinaBox”wasnotmine;itwasinventedbytheeditorsoftheJournal.Nevertheless,theAir-Cribisasortofbox,andthisisalsotrueoftheapparatusknownas the“SkinnerBox”—anexpressionwhichIhaveneverusedandwhichmy friendsacceptasverboten.(IbelievethetermwasfirstpopularizedbyHullandhisstudentsintheform“ModifiedSkinnerBox.”)Helplessly,Ihavewatchedtheteachingmachinesgraduallyassumetheformofboxes.(Thereisconsolationinthefactthatinthiscasetheorganismremainsontheoutside).Iconfesstooneextensionoftheterm,thoughIpleadirresistiblecircumstances.At

thededicationoftheRenardHospitalinSt.Louis,atwhichthepaperonpage303wasgiven, there was a summarizing round-table discussion. Someone mentioned the“Skinnerbox”andadiscussionensued.1tooknopart init,andAlanGregg,sittingnext tome,evidently sharedmydiscomfiture.Finallyhe leanedoverandwhispered,“Box et praeterea nihil.” Itwas doubly appropriate: (1)we had been hearing about“nothingbutboxes”forseveralminutes,and(2)thegistoftheargumenthadbeenthat

540

one could have an effective science about the facts observed in the box andnothingmore.Atthatpointsomeoneaskedmetocommentonthewayinwhichmyutopiannovel,

Walden Two, had been received. I could not resist the variables, intraverbal andotherwise, which were playing upon my behavior and therefore replied that I hadevidentlybeenaccusedofwanting tobuildaboxpopuli.Idonotbelieve the remarkhas seriously damaged the notion of an experimental community. Nevertheless, it issurprisinghowoftenaproposal toremedysomedefect incurrentculturaldesigns isbrushedasidewithareferencetoboxes.

1Thecanvasand“endless”sheetarrangementwassoonreplacedwithasinglelayerofwovenplastic,whichcouldbecleanedandinstantlywipeddry.

541

ThePsychologyofDesign

IntheSpringterm,IagaintaughtacourseinthepsychologyofaestheticswhichIhad given two years before, and I began to be interested in the field. An instantauthority,IwasinvitedtodiscussPicassoontheradioandtowriteanarticleon“ThePsychologyofDesign” forArtEducationToday [1941].GuyBuswell had publishedrecordsshowinghowpeoplelookedatapicture,theireyesdartingfromonepointtoanother,and itoccurred tome thatsuccessive fixationsshouldproduce theapparentmovement called thePhiPhenomenon.Pictureswith linesat different angles shouldappearlively,andthosewithparallellinesquiet.Themovementmightevenberelatedtothesubject.…[fromSkinner,B.F.(1979).TheShapingofaBehaviorist:PartTwoofanAutobiography.NewYork:Knopf,p.238.]

Any design, considered as a simple objective thing, can be described in physicalmathematicalterms,butthiswillnotsufficeforanunderstandingoftheplaceofdesigninart.Theartistisnotsomuchinterestedinthephysicalstructureofadesignasintheeffectithasupontheonewholooksat it.Thishappensalsotobetheconcernofthepsychologist.Wemightsaythatadesigniswhatitisonlybecauseofwhatitdoes,and,if this is true, some knowledge of the various kinds of effectswhich visual patternshaveshouldhelpusinunderstandingthenatureofdesign.

TheBehaviorofLooking

Onekindof effect of a visual pattern is thatwhich it hason the simplebehavior oflooking.A uniform surface (so large that some central part need not be regarded asbounded by a frame or border) does not cause the observer to behave in any verydefiniteway.However, ifweplaceuponitasingledotofcontrastingcolororvalue,wemay be said to have established a rudimentary design, formore or less uniformbehaviorwillbeinducedwhenthesurfaceisplacedbeforeanobserver.Itisnotlikelythatthedotwillfirststriketheregionofclearestvision,buttheeyewillturn,almostinevitably,inthedirectionofthedotandstopthere,atleastmomentarily.Thisisaveryelementarydesign,foritspossibilitiesarelimited.Butifweintroduce

aseconddot,weimmediatelyenhancetheeffect.Theeyenowtendstomovefromonedottotheother,notsoinevitablyasinthefirstcase,butwithconsiderableuniformity.Thismovementhascertainproperties(ofwhichtheobservermayormaynotbeaware)whichcontributetothetotaleffect:itestablishesa“direction”anda“distance.”Ifweadd a third or fourth dot, we greatly increase the variety of the resulting activity,althoughwearelessabletopredicthoworwheretheeyewillmove.And,ingeneral,anincreaseinthenumberofdotsreducesthepredictabilityoftheeffectandweakensthedesign,althoughtherearemanyexceptionstothisrule.Anumberofdotsarrangedinalineoratthecornersofanimaginaryregularpolygonwillyieldamorepredictableeffectthanthesamenumberscatteredatrandom.

542

Whenwe reach theorderof complexity exhibitedbymostof thedesigns actuallyused in art, we can say very little about a uniform behavior of looking. Byphotographingtheactivityoftheeyeinthepresenceofvariouspictures,Buswell1hasdemonstrated this lackofuniformity inaconvincingway.Wecanno longerexplainthe effect of a design by contending that the eye follows contours, or falls with arhythmic beat upon the repeated elements of a border, or performs any of the otherfeatswhichused tobeattributed to it fromself-observation.Thisdoesnotmeanthatthere are no gross differences in the movement of the eye in looking at differentpictures.Themovementmaybelocalizedorscattered,directedoraimless,andsoon;andthesepropertieshaveanobviousbearinguponthecharacterofthepicture.Butnowthat something isknownabout them, eyemovementshave lostmuchof their earlierimportanceintheexplanationofdesign.What many people think of as eye movements are apparently movements of

attention, and this brings us to another kind of effectwhich a designmay have.Aseveryoneknows,itispossibletoholdtheeyefixeduponsomeobjectandatthesametimeto“attendto”objectsinneighboringpartsofthefieldofvision.Onemaytestthiswithasimpleexperiment.Ifthereaderwilllookattheletterattheleftendofthisline,hewillfinditpossibletoexaminenearbyletters(say,thoseaboveorbelowit)withoutnoticeablymovingtheeye.Thiskindofbehaviorisnotclearlyunderstood,butitseemsto involve very slight (perhaps merely potential) movements of looking toward orreachingtowardotherpartsofthefield.Onemaybeawareofthiswhileperforming theexperiment.The tendency towardmovementmaybesostrongas tobreak intoa fullshift of the center of vision, or into amodifiedposture of thewholebody.Wemaysuppose,then,thatwhiletheeyedoesnotactuallyfollowacurvedline,theremaybeatendency for various other muscles of the body to move as if the line were beingfollowed,andthiskindofbehaviorwillcontributetothetotaleffect.

ThePhiPhenomenon

Ourmodernknowledgeofeyemovements,whileweakening theirexplanatoryvalue,compensates for this to some extent by setting up certain new and interestingpossibilities.When one is looking at a picture, the eyemay be roughly likened to acamerawhichistakinganumberofstillphotographsofdifferentpartsofthecanvas.These pictures are thrown upon the eye in rapid succession. Unlike the successiveframes of a film, the pictures are usually quite diverse. Under exceptionalcircumstances, as in looking at a strongly contrasted checkerboard, there may be anoticeable“flicker,”butthisisfortunatelynottheusualcase.Therapidsuccessionofpictures has, however, a number of effectswhichmake an important contribution todesign.Theprincipalexampleistheproductionof“apparentmovement”—thekindofmovementwhichmakesmovingpicturesmoveandwhich is calledbypsychologiststhe“Phiphenomenon.”The simplest demonstration of apparentmovement is provided by two successive

neighboringspotsoflightprojecteduponascreen.Ifonespotisremovedjustbeforethesecondoneisthrownonthescreen,itappearsasifasinglespothadmovedfromonepositiontotheotherthroughtheinterveningspace.Trafficlights,ifsatisfactorilytimed,givetheeffect inspiteof thedifferenceincolor.Now,asimilareffectcanbe

543

obtainedwhenthesuccessivepositionsof the lightareprovidedbythemovementoftheeyeitself.InFigure1,akindofapparentmovementisobtainedifonelooksquicklyfrom one of the birds to the other. The effect is best if the eye falls upon the samerelativeposition(nearthebeak)inbothpictures.Thebirdontheleftappearsasifitsbeakwere just being closed, and the one on the right as if its beakwere just beingopened.Bylookingveryrapidlyfromonetoheother,alively“movement”isseen.Manycompositionsaredesigned,eitherintentionallyorotherwise,toencouragean

effectof thissort.Themovement ismostobvious if thecompositioncontains twoormoreprincipalmasseshavingapproximately thesamenaturebutdifferingslightly inattitude or in the position of minor appendages. In such case the required alternatefixations of the eye are likely to be evoked without explicit instructions, and aconsiderablesimilaritytorealmovementortoamovingpicturemayresult.InFigure2,when the eyemoves from the figure of Sancho Panza,with its inclination to theupperright,tothatofDonQuixote,whichisinclinedtotheupperleft,arealmovementissuggested.ThepictureappearsasifSanchoPanzahadjustdrawnhimselfbackintothepositionhenowholdsandasifDonQuixotehadjust thrownhimselfforwardonhishorse. Ifeitherhalfof thepicture iscovered, thiseffect isdestroyed.Eachfigurestillrepresents a kind ofmovement; we recognize each figure as inmotion but thelively and forceful contributionwhich is providedby thePhi phenomenonwhen thefiguresareviewedalternatelyismissing.

FIG.1

This picture also exemplifies the related effect of an apparent changeof size. If aspotoflightisthrownuponthescreenandthenreplacedbyalargerspot,itappearsasifasinglespothadsuddenlygrownlarger.Ifitisreplacedbyasmallerspot,thereisanapparent shrinking in size. InFigure2,when the figures are viewed alternately, thiseffectmakesSanchoPanzaappearasifheweremovingtowardtheobserverand(whatismoreimportantfortheeffectofthepictureasawhole)itmakesDonQuixoteappearas if he were riding rapidly away. All of this occurs in addition to the regularrepresentedmovementofthepicture.

544

The possibility of apparent movement in a still picture has not tomy knowledgebeenpointedoutinitsbearingupondesign.Butitwillaccountforagreatdealoftheanimationofmanyworksofart.SomeofthedancinggirlsofDegas(whenatleasttwofigures appear) offer excellent examples.Portraits (of twoormorepeople)generallyprovideinterestingcases. If theeyebrowsofonefacearea littlehigher than thoseofanother,apeculiarliftingandloweringmovementmaybegeneratedasonelooksfromonefacetotheother.Figure3,GrantWood’s“DaughtersofRevolution,”makesuseofthis,aswellasofasimilarapparentstretchingandshrinkingofthenecksofthewomenportrayed.

FIG.2.DonQuixotewithSanchoPanzaWringingHisHandsArtist:HonoréDaumier.CollectionofMrs.CharlesS.Payson.CourtesyofM.Knoedler&Co.,NY

545

FIG.3.DaughtersofRevolutionArtist:GrantWood.CollectionofEdwardG.Robinson.CourtesyofMuseumofModernArt

Eventhoughapparentmovementisnotalwaysaspronouncedasintheseexamples,it may nevertheless contribute a good deal to the total effect of a picture.When itsupplementsarepresentedmovement,thenatureofthegainisclear;butadesignneednot dealwith amoving object, and itmay even be abstract,while at the same timereceiving a considerable contribution from the Phi phenomenon. In any canvas inwhichalltheprincipallineslieinapproximatelythesamedirection,itisnotlikelythatthesuccessivestillpicturestakenastheeyemovesfromonepointtoanotherwilldifferin direction in sufficientmagnitude to produce any considerable “Phi.”Other thingsbeingequal,thepaintingwillbequiet.Ontheotherhand,inacanvasinwhichthereare pronounced curves or an assortment of directions, it is very probable that twosuccessive“stillphotographs”willcontainlinesdifferinginslopeinsuchawayastoproducesomemovementof thissort.Thecanvaswillbeanimated—quiteapart fromanyanimationinherentinitssubjectmatter.

EffectofLearnedReactions

One other kind of effect which a design may have arises from various learned oracquiredreactions.Theseconstitutethegreaterpartofthe“meaning”ofthedesign.Itisquiteunlikelythatanyarrangementoflinesorareasiswhollyunfamiliartotheadultobserver,andhencewhollywithoutmeaning.No“abstraction”iscompletelyfreefroma resemblance to previously observed forms, and representative designs are franklybased upon such resemblances. In the latter case, the relation to our previousexperience is usually easily identified, but some contribution to the total effect of adesignmust not be overlooked in the former as well. For example, an architecturalpainting may, with its converging lines and overlapping surfaces, give an excellentrepresentationofathree-dimensionalworld.Wehavemovedaboutinsuchaworldallour lives, andwe have come to know that parallel lines convergewhen they stretchawayfromus,thatthesmalleroftwosimilarobjectsisprobablythemoredistant,that

546

distant colors are usually of lower saturation, and so on. Our response to such apainting cannot escape the influence of this training. This will also be true in anabstraction, since converging lines cannot wholly free themselves from theircommonestmeaningintheeverydayworldaboutus,brightanddullcolorsarelikelytotake their characteristic relative positions, and so on. Although some psychologistsdoubt that our perception of the third dimension is learned, the heightening of thisperceptionthroughourexperienceswithsimilardesignsinthepracticalworldaboutuscanscarcelybedenied.

EffectofEmotionalReactions

Ofgreaterimportanceinthefieldofartthananyofthesereactionsaretheemotionalreactionswithwhichwerespondtomanyvisualpatterns.Tomuchoftheworldaboutusweareperhapsrelativelyindifferent,buttherearemanysituationsinwhichmoreorlessvigorousemotionalreactionsarecommon.Adesignmayacquiresomeofitseffectthrough a resemblance to an emotional stimulus, and it is the contention of manypsychologists that thismay be true of pure design aswell, even thoughwe are notawareoftheresemblanceordonotunderstandtheoriginofourreactions.

FIG.4.NightSnakePlate26fromSnakesoftheWorldbyRaymondL.Ditmars.TheMacmillanCompany,1932.Courtesyofthepublisher.

Figure 4 is a picture of an object to which very few people remain whollyindifferent.Whetherthereisaninheritedfearofsnakesisperhapsquestionable,butitisnevertheless true that thisparticularpatternfrequentlyarousesastrongreaction.Asnakeisnotacommonsubjectmatter inWesternart,perhapsbecausethereactionis

547

too obvious, but it is an effective subject in the sense that few people would beindifferenttoarealisticpaintingofasnakeonthewallsofamuseum.Thereactiontoapictureofthissortiseasilyattributedtothereactionwhichwould

bemadetoitssubjectmatterinreallife.Butwhataretheprecisecharacteristicsofthepatternwhichevoke the reaction?Theheadof thesnakecouldbeconcealedwithoutmuchloss,thescalesneednotbevisible,andasnakeofadifferentmarkingwouldgivethesameeffect.The“snake-ness”of thepicture isapparentlysimply thecoilingandcurvingofthemainlines.Butapictureofacoilofrope,ofacrackingwhip,orofthetendrilofagrapevinewouldalso satisfy thiscondition. If it is true thatweneednotrecognize the snake as a snake in order to obtain the emotional effect, as manypsychologistscontend,wemust suppose that the reactionmade to snakes in real lifecarriesovertomanysimilardesignsinart,evenofanabstractnature.Thisexampleis,of course, an extremeone,but someat least of theotherwiseunexplainedemotionaleffectofgooddesigncanreasonablybeattributedtoanoverlapwithvisualpatternstowhichwehaveacquired,oratleasttowhichwepossessemotionalreactions.

Thisisnotanexhaustivelistofthekindsofeffectswhichadesignmayhave,butitmust suffice. The list is long enough to suggest that complete interpretation of anygiven design is perhaps impossible. The design itself, as an objective geometricalfactor, is only a small part of the story, and we can look for little help from the“formalists,”who seek a solution in that direction.A design is a psychological andculturalobjectofenormouscomplexity.Butifitisidletoattempttoaccountfullyforany one example it is at least important to know the kinds of effects which areinvolved,andthisisall thathasbeenattemptedhere.It isnoparadoxthatwehaveabetterchanceofunderstanding thenatureofdesign ingeneral thanofaccounting fortheeffectsofanygivenexample.There is nothing inwhat has been said that is not familiar to the practicing artist,

even though he does not put it intowords.Nor is it likely that an understanding ofprocessesofthissortwillsupplanttheartistbygeneratingacalculatedorscientificart.Eventhoughwemayeventuallyachieveanexhaustivelistoftheprocessesinvolvedinthepracticeofdesign, theproductionofaworkofartwillprobablystill require thatmass of unverbalized knowledgewhich arises from individual experience. Similarly,on the side of the observer, it would be a mistake to identify the understanding ofdesignwithitsenjoyment.Theprocessesinvolvedinreactingtodesignseemtooperatefreely without being recognized; and, although recognition may sharpen theireffectiveness,itcanhardlyactasasubstituteforartisticsensitivity.

1Buswell,GuyT.HowPeopleLookatPictures.UniversityofChicagoPress,1935.

548

PigeonsinaPelican

ThispaperwaspresentedatameetingoftheAmericanPsychologicalAssociationatCincinnati,Ohio,September,1959,andwaspublishedintheAmericanPsychologistinJanuary,1960.Itisreprintedwithpermission.

Thisisthehistoryofacrackpotidea,bornonthewrongsideofthetracksintellectuallyspeaking, but eventually vindicated in a sort ofmiddle-class respectability. It is thestoryofaproposal touse livingorganisms toguidemissiles—ofa researchprogramduringWorldWarIIcalled“ProjectPigeon”andapeacetimecontinuationattheNavalResearchLaboratorycalled“ORCON”fromthewords“organiccontrol.”Bothoftheseprogramshavenowbeendeclassified.Manhasalwaysmadeuseofthesensorycapacitiesofanimals,eitherbecausethey

aremoreacutethanhisownormoreconvenient.Thewatchdogprobablyhearsbetterthanhismasterand,inanycase,listenswhilehismastersleeps.Asadetectingsystemthedog’searcomessuppliedwithanalarm(thedogneednotbetaughttoannouncethepresence of an intruder), but special forms of reporting are sometimes set up. Thetrackingbehaviorof thebloodhoundand thepointingof thehuntingdogareusuallymodifiedtomakethemmoreuseful.Trainingissometimesquiteexplicit.ItissaidthatseagullswereusedtodetectsubmarinesintheEnglishChannelduringWorldWarI.The British sent their own submarines through the Channel releasing food to thesurface. Gulls could see the submarines from the air and learned to follow them,whethertheywereBritishorGerman.Aflockofgulls,spottedfromtheshore,tookonspecial significance. In the seeing-eye dog the repertoire of artificial signalingresponsesissoelaboratethatithastheconventionalcharacteroftheverbalinterchangebetweenmanandman.The detecting and signaling systems of lower organisms have a special advantage

whenusedwithexplosivedeviceswhichcanbeguidedtowardtheobjectstheyaretodestroy,whether by land, sea, or air.Homing systems for guidedmissiles havenowbeendevelopedwhichsenseandsignalthepositionofatargetbyrespondingtovisibleorinvisibleradiation,noise,radarreflections,andsoon.Thesehavenotalwaysbeenavailable, and in any case a living organism has certain advantages. It is almostcertainlycheaperandmorecompactand,inparticular,isespeciallygoodatrespondingtopatternsandthoseclassesofpatternscalled“concepts.”Thelowerorganismisnotusedbecauseitismoresensitivethanman—afterall,thekamikazedidverywell—butbecauseitisreadilyexpendable.

ProjectPelican

Theethicalquestionofourrighttoconvertalowercreatureintoanunwittingheroisapeacetime luxury.Therewere bigger questions to be answered in the late thirties.Agroupofmenhadcome intopowerwhopromised,andeventuallyaccomplished, the

549

greatest mass murder in history. In 1939 the city ofWarsaw was laid waste in anunprovokedbombing, and the airplane emerged as a new andhorrible instrument ofwar against which only the feeblest defenses were available. Project Pigeon wasconceivedagainstthatbackground.Itbeganasasearchforahomingdevicetobeusedinasurface-to-airguidedmissileasadefenseagainstaircraft.Asthebalancebetweenoffensive and defensiveweapons shifted, the directionwas reversed, and the systemwas to be tested first in an air-to-groundmissile called the “Pelican.” Its name is ausefulreminderofthestateofthemissileartinAmericaatthattime.Itsdetectingandservomechanismstookupsomuchspacethattherewasnoroomforexplosives:hencetheresemblancetothepelican“whosebeakcanholdmorethanitsbellycan.”Mytitleisperhapsnowclear.Figure1showsthepigeons,jacketedforduty.Figure2showsthebeakofthePelican.At theUniversityofMinnesota in thespringof1940thecapacityof thepigeon to

steertowardatargetwastestedwithamovinghoist.Thepigeon,heldinajacketandharnessedtoablock,wasimmobilizedexceptforitsneckandhead.Itcouldeatgrainfromadishandoperateacontrolsystembymovingitsheadinappropriatedirections.Movement of the head operated the motors of the hoist. The bird could ascend bylifting its head, descend by lowering it, and travel from side to side by movingappropriately. The whole system, mounted on wheels, was pushed across a roomtowardabull’s-eyeonthefarwall.Duringtheapproachthepigeonraisedorlowereditselfandmovedfromsidetosideinsuchawayastoreachthewallinpositiontoeatgrainfromthecenterofthebull’s-eye.Thepigeonlearnedtoreachanytargetwithinreach of the hoist, no matter what the starting position and during fairly rapidapproaches.

FIG.1.Thirty-twopigeons,jacketedfortesting.

550

FIG.2.NoseofthePelican,showinglenses.

TheexperimentwasshowntoJohnT.Tate,aphysicist,thenDeanoftheGraduateSchoolattheUniversityofMinnesota,whobroughtittotheattentionofR.C.Tolman,oneofagroupofscientistsengagedinearlydefenseactivities.Theresultwasthefirstofalongseriesofrejections.Theproposal“didnotwarrantfurtherdevelopmentatthetime.” The project was accordingly allowed to lapse. On December 7, 1941, thesituationwassuddenlyrestructured;and,onthefollowingday,withthehelpofKellerBreland, then a graduate student atMinnesota, furtherworkwas planned.A simplerharnessingsystemcouldbeusedifthebombweretorotateslowlyduringitsdescent,whenthepigeonwouldneedtosteerinonlyonedimension:fromsidetoside.Webuiltan apparatus in which a harnessed pigeon was lowered toward a large revolvingturntable across which a target was driven according to contacts made by the birdduring its descent. It was not difficult to train a pigeon to “hit” small ship modelsduring fairly rapiddescents.Wemade ademonstration film showinghitsonvariouskindsoftargets,andtwopsychologiststhenengagedinthewareffortinWashington,Charles Bray and Leonard Carmichael, undertook to look for government support.Tolman,thenattheOfficeofScientificResearchandDevelopment,againfeltthattheprojectdidnotwarrantsupport, inpartbecause theUnitedStateshadat that timeno

551

missilecapableofbeingguided towarda target.Commander (nowAdmiral)LuisdeFlorez,thenintheSpecialDevicesSectionoftheNavy,tookasympatheticview.Hedismissed the objection that there was no available vehicle by suggesting that thepigeon be connected with an automatic pilot mounted in a small plane loaded withexplosives.Buthewasunabletotakeontheprojectbecauseofothercommitmentsandbecause,asheexplained,hehad recentlybetononeor twootherequally longshotswhichhadnotcomein.Theprojectlapsedagainandwouldprobablyhavebeenabandonedifithadnotbeen

forayoungmanwhoselastnameIhaveungratefullyforgotten,butwhosefirstname—Victor—wehailedasapropitioussign.HissubsequenthistoryledustorefertohimasVanquished;and this,as it turnedout,wasamore reliableomen.Victorwalked intotheDepartmentofPsychologyatMinnesotaonedayinthesummerof1942lookingforan animal psychologist. He had a scheme for installing dogs in antisubmarinetorpedoes.Thedogsweretorespondtofaintacousticsignalsfromthesubmarineandtosteerthetorpedotowarditsgoal.Hewantedastatementfromananimalpsychologistastoitsfeasibility.Hewasunderstandablysurprisedtolearnofourworkwithpigeonsbut seized upon it eagerly; citing it in support of his contention that dogs could betrained to steer torpedoes, he went to a number of companies in Minneapolis. Hisprojectwasrejectedbyeveryoneheapproached;butonecompany,GeneralMills,Inc.,askedformoreinformationaboutourworkwithpigeons.WedescribedtheprojectandpresentedtheavailabledatatoArthurD.Hyde,Vice-PresidentinChargeofResearch.Thecompanywasnot lookingfornewproducts,butHydethoughtthat itmight,asapublicservice,developthepigeonsystemtothepointatwhichagovernmentalagencycouldbepersuadedtotakeover.BrelandandImovedintothetopfloorofaflourmillinMinneapolisandwiththe

help of Norman Guttman, who had joined the project, set to work on furtherimprovements.Ithadbeendifficulttoinducethepigeontorespondtothesmallangulardisplacementofadistanttarget.Itwouldstartworkingdangerouslylateinthedescent.Itsnaturalpursuitbehaviorwasnotappropriatetothecharacteristicsofalikelymissile.A new system was therefore designed. An image of the target was projected on atranslucent screen as in a camera obscura. The pigeon, held near the screen, wasreinforcedforpeckingattheimageonthescreen.Theguidingsignalwastobepickedupfromthepointofcontactofscreenandbeak.Inanearlyarrangementthescreenwasatranslucentplasticplateformingthelarger

endofatruncatedconebearingalensatthesmallerend.Theconewasmounted,lensdown,inagimbalbearing.Anobjectwithinrangethrewitsimageonthetranslucentscreen;andthepigeon,heldverticallyjustabovetheplate,peckedtheimage.Whenatargetwasmovedaboutwithinrangeofthelens,theconecontinuedtopointtoit.Inanother apparatus a translucent disk, free to tilt slightly on gimbal bearings, closedcontacts operating motors which altered the position of a large field beneath theapparatus.Smallcutoutsofshipsandotherobjectswereplacedonthefield.Thefieldwas constantly in motion, and a target would go out of range unless the pigeoncontinuedtocontrolit.Withthisapparatuswebegantostudythepigeon’sreactionstovariouspatternsandtodevelopsustainedsteadyratesofrespondingthroughtheuseofappropriate schedules of reinforcement, the reinforcement being a few grainsoccasionally released onto the plate. By building up large extinction curves a targetcould be tracked continuously for a matter of minutes without reinforcement. Wetrained pigeons to follow a variety of land and sea targets, to neglect large patches

552

intendedtorepresentcloudsorflak,toconcentrateononetargetwhileanotherwasinview,andsoon.Wefoundthatapigeoncouldholdthemissileonaparticularstreetintersectioninanaerialmapofacity.Themapwhichcamemosteasilytohandwasofacitywhich,intheinterestsofinternationalrelations,neednotbeidentified.Throughappropriateschedulesofreinforcementitwaspossibletomaintainlongeruninterruptedrunsthancouldconceivablyberequiredbyamissile.Wealsoundertookamoreseriousstudyofthepigeon’sbehavior,withthehelpof

W.K.EstesandMarionBreland,whojoinedtheprojectat thistime.Weascertainedoptimalconditionsofdeprivation,investigatedotherkindsofdeprivations,studiedtheeffect of special reinforcements (for example, pigeons were said to find hemp seedparticularly delectable), tested the effects of energizing drugs and increased oxygenpressures,andsoon.Wedifferentiallyreinforcedtheforceofthepeckingresponseandfoundthatpigeonscouldbeinducedtopecksoenergeticallythatthebaseofthebeakbecameinflamed.Weinvestigatedtheeffectsofextremesoftemperature,ofchangesinatmospheric pressure, of accelerations produced by an improvised centrifuge, ofincreasedcarbondioxidepressure,ofincreasedandprolongedvibration,andofnoisessuch as pistol shots. (The birds could, of course, have been deafened to eliminateauditory distractions, but we found it easy to maintain steady behavior in spite ofintense noises and many other distracting conditions using the simple process ofadaptation.) We investigated optimal conditions for the quick development ofdiscriminations and began to study the pigeon’s reactions to patterns, testing forinductionfromatestfiguretothesamefigureinverted,tofiguresofdifferentsizesandcolors,andtofiguresagainstdifferentgrounds.Asimpledeviceusingcarbonpapertorecordthepointsatwhichapigeonpecksafigureshowedapromisewhichhasneverbeenproperlyexploited.Wemade another demonstration film and renewed our contactwith theOffice of

ScientificResearchandDevelopment.AnobserverwassenttoMinneapolis,andonthestrengthofhisreportweweregivenanopportunitytopresentourcaseinWashingtoninFebruary,1943.Atthattimewewereofferingahomingdevicecapableofreportingwithanon-off signal theorientationofamissile towardvariousvisualpatterns.Thecapacitytorespondtopatternwas,wefelt,ourstrongestargument,butthefactthatthedeviceusedonlyvisibleradiation(thesameformofinformationavailabletothehumanbombardier)madeitsuperiortotheradio-controlledmissilesthenunderdevelopmentbecause itwas resistant to jamming.Our filmhadsomeeffect.Otherobserversweresent toMinneapolis to see the demonstration itself. The pigeons, as usual, behavedbeautifully. One of them held the supposed missile on a particular intersection ofstreetsintheaerialmapforfiveminutesalthoughthetargetwouldhavebeenlostifthepigeonhadpausedforasecondortwo.TheobserversreturnedtoWashington,andtwoweekslaterwewereaskedtosupplydataon(a)thepopulationofpigeonsintheUnitedStates (fortunately, the census bureau had some figures) and (b) the accuracy withwhichpigeons struckapointon aplate.Thereweremanyarbitrary conditions tobetakenintoaccount inmeasuringthe latter,butwesuppliedpossiblyrelevantdata.Atlonglast,inJune,1943,theOfficeofScientificResearchandDevelopmentawardedamodestcontracttoGeneralMills,Inc.to“developahomingdevice.”Atthattimeweweregivensomeinformationaboutthemissilethepigeonswereto

steer. The Pelican was a wing-steered glider, still under development and not yetsuccessfully steered by any homing device. It was being tested on a target in NewJerseyconsistingof a stirrup-shapedpatternbulldozedoutof the sandy soil near the

553

coast.Thewhite linesof the target stoodoutclearlyagainstbrownandgreencover.Coloredphotographsweretakenfromvariousdistancesandatvariousangles,andtheverisimilitudeofthereproductionwascheckedbyflyingoverthetargetandlookingatitsimageinaportablecameraobscura.Becauseofsecurityrestrictionsweweregivenonlyveryroughspecificationsofthe

signal tobe supplied to thecontrolling system in thePelican. Itwasno longer tobesimplyon-off;ifthemissilewasbadlyofftarget,anespeciallystrongcorrectingsignalwasneeded.Thismeantthatthequadrant-contactsystemwouldnolongersuffice.Butfurtherrequirementswereleftmainlytoourimagination.TheGeneralMillsengineerswereequaltothisdifficultassignment.Withwhatnowseemslikeunbelievablespeed,theydesigned and constructed a pneumatic pickup systemgiving a graded signal.Alensinthenoseofthemissilethrewanimageonatranslucentplatewithinreachofthepigeon inapressure-sealedchamber.Fourairvalves restingagainst theedgesof theplatewerejarredopenmomentarilyasthepigeonpecked.Thevalvesattherightandleftadmittedairtochambersonoppositesidesofonetambour,whilethevalvesatthetopandbottomadmittedairtooppositesidesofanother.AironallsideswasexhaustedbyaVenturiconeonthesideofthemissile.Whenthemissilewasontarget,thepigeonpecked the center of the plate, all valves admitted equal amounts of air, and thetamboursremainedinneutralpositions.Butiftheimagemovedaslittleasaquarterofan inchoff-center, corresponding to avery small angulardisplacementof the target,moreairwasadmittedbythevalvesononeside,andtheresultingdisplacementofthetambourssentappropriatecorrectingordersdirectlytotheservosystem.The device required no materials in short supply, was relatively foolproof, and

delivered a graded signal. It had another advantage. By this time we had begun torealizethatapigeonwasmoreeasilycontrolledthanaphysicalscientistservingonacommittee. Itwasverydifficult toconvince the latter that the formerwasanorderlysystem.Wethereforemultipliedtheprobabilityofsuccessbydesigningamultiple-birdunit.TherewasadequatespaceinthenoseofthePelicanforthreepigeons,eachwithits own lens and plate.A net signal could easily be generated.Themajority vote ofthreepigeonsofferedanexcellentguaranteeagainstmomentarypausesandaberrations.(Welaterworkedoutasysteminwhichthemajoritytookonamorecharacteristicallydemocratic function.Whenamissile is falling toward two ships at sea, for example,thereisnoguaranteethatallthreepigeonswillsteertowardthesameship.Butatleasttwomust agree, and the third can then be punished for hisminority opinion.Underproper contingencies of reinforcement a punished bird will shift immediately to themajorityview.Whenall threeareworkingononeship,anydefectionisimmediatelypunishedandcorrected.)

554

FIG.3.Demonstrationmodelofthethree-pigeonguidancesystem.

ThearrangementinthenoseofthePelicanisshowninFigure3.Three systemsoflenses and mirrors, shown at the left, throw images of the target area on the threetranslucentplatesshowninthecenter.Theballisticvalvesrestingagainsttheedgesoftheseplatesandthetubesconnectingthemwiththemanifoldsleadingtothecontrollingtambours may be seen. A pigeon is being placed in the pressurized chamber at theright.TheGeneralMillsengineersalsobuiltasimulator(Figure4)—asortofLinktrainer

forpigeons—designed tohave thesteeringcharacteristicsof thePelican, inso farasthesehadbeencommunicatedtous.Likethewing-steeredPelican,thesimulatortiltedandturnedfromsidetoside.Whenthethree-birdnosewasattachedtoit,thepigeonscould be put in full control—the “loop could be closed”—and the adequacy of thesignal testedunderpursuitconditions.Targetsweremovedbackandforthacross thefarwall of a room at prescribed speeds and in given patterns of oscillation, and thetrackingresponseofthewholeunitwasstudiedquantitatively.

555

FIG.4.Simulatorfortestingtheadequacyofthepigeonsignal.

Meanwhilewecontinuedourintensivestudyofthebehaviorofthepigeon.Lookingaheadtocombatusewedesignedmethodsforthemassproductionoftrainedbirdsandforhandlinglargegroupsoftrainedsubjects.Wewereproposingtotraincertainbirdsfor certain classes of targets, such as ships at sea, while special squads were to betrained on special targets, photographs of which were to be obtained throughreconnaissance.Alargecrewofpigeonswouldthenbewaitingforassignment,butwedevelopedharnessingandtrainingtechniqueswhichshouldhavesolvedsuchproblemsquiteeasily.Amultiple-unittrainerisshowninFigure5.Eachboxcontaineda jacketedpigeon

heldat anangleof45° to thehorizontalandperpendicular toan8″×8″ translucentscreen.Atargetarea isprojectedoneachscreen.Twobeamsof light intersectat thepointtobestruck.Allon-targetresponsesofthepigeonarereportedbytheinterruptionofthecrossedbeamsandbycontactwiththetranslucentscreen.Onlyafour-inch,disk-shaped portion of the field is visible to the pigeon at any time, but the boxesmoveslowlyaboutthefield,givingthepigeonanopportunitytorespondtothetargetinallpositions.Thepositionsofallreinforcementsarerecordedtorevealanyweakareas.Avariable-ratioscheduleisusedtobuildsustained,rapidresponding.

556

FIG.5.Atrainerforfourpigeons.

ByDecember,1943,lessthansixmonthsafterthecontractwasawarded,wewerereadytoreporttotheOfficeofScientificResearchandDevelopment.Observersvisitedthelaboratoryandwatchedthesimulatorfollowatargetaboutaroomunderthecontrolof a team of three birds. They also reviewed our tracking data. The only questionswhich arose were the inevitable consequence of our lack of information about thesignalrequiredtosteerthePelican.Forexample,wehadhadtomakecertainarbitrarydecisions in compromising between sensitivity of signal and its integration orsmoothness.Ahighvacuumproducedquick,rathererraticmovementsofthetambours,whilealowervacuumgaveasluggishbutsmoothsignal.Asitturnedout,wehadnotchosen the best values in collecting our data, and in January, 1944, the Office ofScientificResearchandDevelopmentrefusedtoextendtheGeneralMillscontract.Thereasons given seemed to be due to misunderstandings or, rather, to lack ofcommunication. We had already collected further data with new settings of theinstruments,andtheseweresubmittedinarequestforreconsideration.Weweregivenonemorechance.Wetookournewdatatotheradiationlabat the

Massachusetts Institute of Technology where they were examined by theservospecialistsworking on thePelican controls.To our surprise the scientistwhosetask it was to predict the usefulness of the pigeon signal argued that our datawereinconsistentwith respect to phase lag and certain other characteristics of the signal.Accordingtohisequations,ourdevicecouldnotpossiblyyieldthesignalswereported.Weknew,ofcourse,thatithaddoneso.Weexaminedthesupposedinconsistencyandtracedit,orsowethought,toacertainnonlinearityinoursystem.Inpeckinganimagenear the edge of the plate, the pigeon strikes a more glancing blow; hence the airadmittedatthevalvesisnotlinearlyproportionaltothedisplacementofthetarget.Thiscould be corrected in several ways: for example, by using a lens to distort radial

557

distances.ItwasourunderstandingthatinanycasethesignalwasadequatetocontrolthePelican.Indeed,oneservoauthority,uponlookingatgraphsoftheperformanceofthesimulator,exclaimed:“Thisisbetterthanradar!”Two days later, encouraged by ourmeeting atMIT,we reached the summit.We

were to present our case briefly to a committee of the country’s top scientists. Thehearing began with a brief report by the scientist who had discovered the“inconsistency”inourdata,and tooursurprisehestill regardeditasunresolved.Hepredictedthatthesignalwereportedwouldcausethemissileto“hunt”wildlyandlosethetarget.Buthispredictionshouldhaveappliedaswelltotheclosedloopsimulator.Fortunatelyanotherscientistwaspresentwhohadseenthesimulatorperformingunderexcellent control andwho could confirm our report of the facts. But realitywas nomatchformathematics.The basic difficulty, of course, lay in convincing a dozen distinguished physical

scientiststhatthebehaviorofapigeoncouldbeadequatelycontrolled.Wehadhopedto scoreon thispointbybringingwithusademonstration.Asmallblackboxhadaround translucentwindow in one end.A slide projector placed some distance awaythrewonthewindowanimageoftheNewJerseytarget.Inthebox,ofcourse,wasapigeon—which, incidentally, had at that time been harnessed for 35 hours. Ourintentionwastoleteachmemberofthecommitteeobservetheresponsetothetargetbylookingdownasmalltube;buttimewasnotavailableforindividualobservation,andwewereaskedtotakethetopoffthebox.Thetranslucentscreenwasfloodedwithsomuch light that the target was barely visible, and the peering scientists offeredconditionsmuchmoreunfamiliarandthreateningthanthoselikelytobeencounteredina missile. In spite of this the pigeon behaved perfectly, pecking steadily andenergetically at the imageof the target as itmovedabouton theplate.One scientistwithanexperimentalturnofmindinterceptedthebeamfromtheprojector.Thepigeonstoppedinstantly.Whentheimageagainappeared,peckingbeganwithinafractionofasecondandcontinuedatasteadyrate.Itwas a perfect performance, but it had just thewrong effect.One can talk about

phaselaginpursuitbehavioranddiscussmathematicalpredictionsofhuntingwithoutreflectingtoocloselyuponwhat is insidetheblackbox.Butthespectacleofa livingpigeon carrying out its assignment, nomatter how beautifully, simply reminded thecommitteeofhowutterlyfantasticourproposalwas.Iwillnotsaythatthemeetingwasmarkedbyunrestrainedmerriment,forthemerrimentwasrestrained.Butitwasthere,anditwasobviousthatourcasewaslost.Hyde closed our presentation with a brief summary: we were offering a homing

device,unusuallyresistanttojamming,capableofreactingtoawidevarietyoftargetpatterns,requiringnomaterialsinshortsupply,andsosimpletobuildthatproductioncouldbestartedin30days.Hethankedthecommittee,andweleft.Asthedoorclosedbehindus,hesaidtome:“Whydon’tyougooutandgetdrunk!”Officialwordsooncame:“Furtherprosecutionofthisprojectwouldseriouslydelay

others which in the minds of the Division would have more immediate promise ofcombatapplication.”Possibly the referencewas toaparticularcombatapplicationatHiroshimaayearandahalflater,whenitlookedforawhileasiftheneedforaccuratebombing had been eliminated for all time. In any case we had to show, for all ourtrouble,onlyaloftfulofcuriouslyuselessequipmentandafewdozenpigeonswithastrangeinterestinafeatureoftheNewJerseycoast.Theequipmentwasscrapped,but30 of the pigeons were kept to see how long they would retain the appropriate

558

behavior.In the years which followed there were faint signs of life. Winston Churchill’s

personal scientific advisor, Lord Cherwell, learned of the project and “regretted itsdemise.”A scientistwhohadhad somecontactwith theprojectduring thewar, andwhoevidentlyassumedthatitsclassifiedstatuswasnottobetakenseriously,madeagood story out of it for the AtlanticMonthly, names being changed to protect theinnocent. Other uses of animals began to be described. The author of the AtlanticMonthly story alsopublishedan accountof the “incendiarybats.”Thousandsofbatsweretobereleasedoveranenemycity,eachcarryingasmall incendiarytimebomb.Thebatswouldtakerefuge,asistheircustom,undereavesandinotherout-of-the-wayplaces; and shortly afterwards thousands of small fires would break out practicallysimultaneously. The schemewas never used because it was feared that it would bemistakenforgermwarfareandmightleadtoretaliationinkind.AnotherstorycirculatingatthetimetoldhowtheRussianstraineddogstoblowup

tanks.Ihavedescribedthetechniqueelsewhere.1ASwedishproposal touseseals toachievethesameendwithsubmarineswasnotsuccessful.Thesealsweretobetrainedto approach submarines to obtain fish attached to the sides. They were then to bereleased carryingmagneticmines in the vicinity of hostile submarines.The requiredtrainingwasapparentlyneverachieved.Icannotvouchfortheauthenticityofprobablythemostfantasticstoryofthissort,butitoughttoberecorded.TheRussiansweresaidtohavetrainedsealionstocutminecables.Acomplicateddeviceattachedtothesealionincludedamotor-drivencablecutter,atankfullofsmallfish,andadevicewhichreleasedafewfishintoamuzzlecoveringthesealion’shead.Inordertoeat,thesealionhadtofindaminecableandswimalongsideitsothatthecutterwasautomaticallytriggered,atwhichpointafewfishwerereleasedfromthetankintothemuzzle.Whenagivennumberofcableshadbeencut,boththeenergyofthecuttingmechanismandthe supplyof fishwere exhausted, and the sea lion receiveda special stimulusuponwhichitreturnedtoitshomebaseforspecialreinforcementandreloading.

ORCON

The story of our own venture has a happy ending. With the discovery of Germanaccomplishments in the field of guided missiles, feasible homing systems suddenlybecame very important. Franklin V. Taylor of the Naval Research Laboratory inWashington, D. C., heard about our project and asked for further details. As apsychologistTaylorappreciatedthespecialcapacityoflivingorganismstorespondtovisual patterns and was aware of recent advances in the control of behavior.Moreimportant, he was a skillful practitioner in a kind of control which our project hadconspicuously lacked: he knew how to approach the people who determine thedirection of research. He showed our demonstration film so often that it wascompletely worn out—but to good effect, for support was eventually found for athoroughinvestigationof“organiccontrol”underthegeneraltitleORCON.Tayloralsoenlisted the supportofengineers inobtainingamoreeffective reportof thepigeon’sbehavior.Thetranslucentplateuponwhichtheimageofthetargetwasthrownhadasemiconducting surface, and the tip of the bird’s beak was covered with a goldelectrode.Asinglecontactwiththeplatesentanimmediatereportofthelocationofthe

559

targettothecontrollingmechanism.Theworkwhichwentintothissystemcontributedto the so-called Pick-off Display Converter developed as part of the Naval DataHandlingSystemforhumanobservers.Itisnolongernecessaryfortheradaroperatortogiveaverbalreportofthelocationofapiponthescreen.Likethepigeon,hehasonlytotouchthepipwithaspecialcontact.(Heholdsthecontactinhishand.)At the Naval Research Laboratory inWashington the responses of pigeons were

studiedindetail.Averagepeckrate,averageerrorrate,averagehitrate,andsoon,wererecordedundervariousconditions.Thetrackingbehaviorof thepigeonwasanalyzedwith methods similar to those employed with human operators (Figure 6). Patternperception was studied, including generalization from one pattern to another. Asimulator was constructed in which the pigeon controlled an image projected by amoving-picturefilmofanactualtarget:forexample,ashipatseaasseenfromaplaneapproachingat600milesperhour.A few framesof amovingpictureof thepigeoncontrollingtheorientationtowardashipduringanapproachareshowninFigure7.The publications from the Naval Research Laboratory which report this work2

provide a serious evaluation of the possibilities of organic control. Although insimulated testsa singlepigeonoccasionally losesa target, its trackingcharacteristicsaresurprisinglygood.Athree-orseven-birdunitwiththesameindividualconsistencyshouldyieldasignalwithareliabilitywhichisatleastoftheorderofmagnitudeshownbyotherphasesofguidedmissilesintheirpresentstageofdevelopment.Moreover,inthesevenyearswhichhavefollowedthelastofthesereports,agreatdealofrelevantinformation has been acquired. The color vision of the pigeon is now thoroughlyunderstood; itsgeneralizationalongsinglepropertiesofastimulushasbeenrecordedand analyzed; and themaintenance of behavior through scheduling of reinforcementhasbeendrasticallyimproved,particularlyinthedevelopmentoftechniquesforpacingresponsesforlesserraticandsteadiersignals.3Testsmadewiththebirdssalvagedfromthe old Project Pigeon showed that even after six years of inactivity a pigeon willimmediately and correctly strike a target to which it has been conditioned and willcontinuetorespondforsometimewithoutreinforcement.

560

FIG.6.Arrangementforstudyingpursuitmovements.

Theuseoflivingorganismsinguidingmissilesis,itseemsfairtosay,nolongeracrackpotidea.Apigeonisanextraordinarilysubtleandcomplexmechanismcapableofperformanceswhichat themoment canbe equalledbyelectronic equipmentonlyofvastlygreaterweightandsize,anditcanbeputtoreliableusethroughtheprincipleswhichhaveemergedfromanexperimentalanalysisofitsbehavior.Butthisvindicationof our original proposal is perhaps the least important result. Something happenedduringthebrieflifeofProjectPigeonwhichithastakenalongtimetoappreciate.Thepractical taskbeforeuscreatedanewattitude toward thebehavioroforganisms.Wehadtomaximizetheprobabilitythatagivenformofbehaviorwouldoccuratagiventime.Wecouldnotenjoytheluxuryofobservingonevariablewhileallowingotherstochange in what we hoped was a random fashion. We had to discover all relevantvariables and submit them to experimental control whenever possible.We were nodoubt under exceptional pressure, but vigorous scientific research usually makescomparable demands. Psychologists have too often yielded to the temptation to becontent with hypothetical processes and intervening variables rather than press forrigorous experimental control. It is often intellectual laziness rather than necessitywhichrecommendstheaposterioristatisticaltreatmentofvariation.Ourtaskforcedus

561

toemphasizepriorexperimentalcontrol,anditssuccessinrevealingorderlyprocessesgave us an exciting glimpse of the superiority of laboratory practice over verbal(includingsomekindsofmathematical)explanation.

562

563

FIG.7.Framesfromasimulatedapproach.

TheCrackpotIdea

If Iwere toconclude thatcrackpot ideasare tobeencouraged, I shouldprobablybetoldthatpsychologyhasalreadyhadmorethanitsshareofthem.Ifithas,theyhavebeenentertainedbythewrongpeople.Reactingagainsttheexcessesofpsychologicalquackery, psychologists have developed an enormous concern for scientificrespectability. They constantly warn their students against questionable facts andunsupported theories. As a result the usual PhD thesis is a model of compulsivecautiousness,advancingonlythemosttimidconclusionsthoroughlyhedgedaboutwithqualifications.Butitisjustthemancapableofdisplayingsuchadmirablecautionwhoneeds a touch of uncontrolled speculation. Possibly a generous exposure topsychologicalsciencefictionwouldhelp.ProjectPigeonmightbesaidtosupportthatview.Exceptwithrespecttoitsavowedgoal,itwas,asIseeit,highlyproductive;andthiswas in largemeasurebecausemycolleaguesand Iknew that, in theeyesof theworld,wewerecrazy.One virtue in crackpot ideas is that they breed rapidly and their progeny show

extraordinarymutations.Everyone is talking about teachingmachinesnowadays, butSidneyPresseycan tellyouwhat itwas like tohaveacrackpot idea in that field 40yearsago.Hisself-testingdevicesandself-scoringtestformsnowneednodefense,andpsychomotortrainingdeviceshavealsoachievedasubstantialrespectability.Thisdidnot,however,preparethewayfordevicestobeusedinverbalinstruction—thatis,inthekindsofteachingwhicharetheprincipalconcernofourschoolsandcolleges.Evenfive short years ago that kind of instruction by machine was still in the crackpotcategory.(Icanquoteofficialopiniontothateffectfromhighplaces.)Now,thereisadirectgeneticconnectionbetweenteachingmachinesandProjectPigeon.Wehadbeenforced to consider the mass education of pigeons. True, the scrap of wisdom weimpartedtoeachwasindeedsmall,buttherequiredchangesinbehaviorweresimilartothose which must be brought about in vaster quantities in human students. Thetechniquesofshapingbehaviorandofbringingitunderstimuluscontrolwhichcanbetraced,asIhavesuggestedelsewhere,4toamemorableepisodeonthetopfloorofthatflourmillinMinneapolisneededonlyadetailedreformulationofverbalbehaviortobedirectlyapplicabletoeducation.I am sure there is more to come. In the year which followed the termination of

ProjectPigeonIwroteWaldenTwo,autopianpictureofaproperlyengineeredsociety.Some psychotherapistsmight argue that Iwas suffering from personal rejection andsimplyretreatedtoafantasiedworldwhereeverythingwentaccordingtoplan,wherethereneverwasheardadiscouragingword.Butanotherexplanationis,Ithink,equallyplausible.Thatpieceofsciencefictionwasadeclarationofconfidenceinatechnologyofbehavior.Callitacrackpotideaifyouwill;itisoneinwhichIhaveneverlostfaith.I still believe that the same kind of wide-ranging speculation about human affairs,supported by studies of compensating rigor, will make a substantial contributiontowardthatworldofthefutureinwhich,amongotherthings,therewillbenoneedforguidedmissiles.

564

1Seepage121.2Chernikoff,R.,andNewlin,E.P.ORCON.PartIII.Investigationsoftargetacquisitionbythepigeon.NavalRes.Lab.lett.Rep.,1951,No.S–3600–629a/51(Sept.10).Conklin,J.E.,Newlin,E.P.,Jr.,Taylor,F.V.,andTipton,C.L.ORCON.PartIV.Simulatedflighttests.NavalRes.Lab.lett.Rep.,1953,No.4105.Searle,L.V.,andStafford,B.H.ORCON.PartII.ReportofphaseIresearchandbandpassstudy.NavalRes.Lab.lett.Rep.,1950,No.S–3600–157/50(May1).Taylor,F.V.ORCON.PartI.Outlineofproposedresearch.NavalRes.Lab.lett.Rep.,1949,No.S–43600–157/50(June17).White,C.F.DevelopmentoftheNRLORCONtactilemissilesimulator.NavalRes.Lab.Rep.,1952,No.3917.

3See,forexample,thepaperbeginningonpage132.4Seepage165.

565

SomeResponsestotheStimulus“Pavlov”

InopeningthefirstsessionthismorningDr.Reesesaidthathefelttheprogramshowedsignsofanecumenicalspirit.Idon’tknowwhethertheremarkwasareferencetomyProtestantactivities,butifso,ImustsayitwasextraordinarilygenerousoftheCollegeofCardinalstohavemademePope.Ishalltrytolearntheproperrituals.Inthemovie,ZorbatheGreek, theFrenchwomanwasneveracceptedbytheislandersbecauseshecrossedherselffromleft toright, rather thanright to left. Ishall trynot tomakethatkindofmistake.Fromnowonitwillbeconditionalreflexes.ItseemsappropriateonthisoccasiontotrytosaywhatPavlovhasmeant tome.I

havetostrainabittogetbacktopersonalreminiscences,butIcanjustmakeit.In1929Pavlov was enthusiastically received in Boston as President of the InternationalCongress of Physiology. I was in my first year of graduate study in psychology atHarvard, and I turned up hero-worshipping wherever Pavlov could be expected toappear.Aphotographerwastakingordersforanofficialportrait.HehadaskedPavlovto write his name on a sheet of paper, and he assured prospective buyers that thesignaturewouldappearoneachphotograph. Ioffered tobuyone if I couldhave thesheet of paperwhen hewas throughwith it. I got it, and still have it. It is the onlyautographIhaveevercollected.ItwasmybiologyteacheratHamiltonCollege,AlbroMorrill,whofirstcalledmy

attentiontoAnrep’stranslationofPavlov.Ihadgoneback,ayearaftergraduation,totalkwithhimaboutgraduatestudyinpsychology.HehadalwayshopedIwouldbeabiologist,andperhapsthatwaswhyheshowedmePavlov’sbook,whichhehadjustreceived. Inoneofhiscourseshehadalreadycalledmyattention to JacquesLoeb’sThePhysiologyoftheBrainandComparativePsychology.IboughtPavlov’sbookandtookitwithmetoGreenwichVillage,whereIspentseveralBohemianmonthsbeforegoingon toHarvard. I readPavlovbyday and sowedwildoats bynight. I am surePavlovhimselfwouldhaveapprovedof thispairingofstimuli.Eventodayapageofhisbookelicitsmanywarm,ifsomewhatfaded,autonomicresponses.PavlovwasparticularlyrelevanttoaratherdrasticchangewhichIhadmadeinmy

plansforacareerandaboutwhichIwasnotyettoosecure.IhadmajoredinEnglishandhadplannedtobeawriter.IttookmeayearortwotodiscoverthatalthoughIhadlearnedhowtowrite,Ihadlearnednothingworthwritingabout.IfoundIhadnothingtosay.Idecidedtogointopsychologytoremedythedefect.AtaboutthattimeH.G.Wellswroteanarticle for, Ibelieve, theSundayNewYorkTimes. In ithecomparedPavlov and George Bernard Shaw. They looked rather alike, with their great whitebeards,anditwaseasyforWellstocontrastthewittypropagandistwiththelaboratoryscientist.Heposedahypotheticalquestion:Ifthesetwomenweredrowning,andyouhadonlyonelifepreserver,towhichwouldyouthrowit?WellsenormouslyreassuredmeinthedecisionIhadmadebythrowingittoPavlov.Several years later I worked for a time in the laboratory of Professor Walter B.

Cannon at theHarvardMedical School.Cannon and Pavlovwere close friends, and

566

when Pavlov came toAmerica, he stayedwith the Cannons in a house onDivinityAvenueabout60yards(Ipaceditoffthismorning)fromtheauditoriuminwhichwehavebeenmeeting.CannontoldmanyamusingstoriesaboutPavlov.Theyweregentlestories,scarcelyworthtellingaboutanyone,evenofPavlov’sdistinction,butherearetwoof them.Pavlovknew littleEnglish.HeandCannonconversed inGerman.OnemorningatbreakfastPavlovwas trying to read theheadlineson thesportspageofaBostonpaper.ItappearedthataninthinningrallyoftheRedSoxhadfizzled.PavlovcalledacrossthetabletoCannon,“WasmeintdasWort‘fizzle’?”—pronouncingit“fit—zell.”AnotherstoryhadtodowiththeepisodeinwhichPavlovwasrobbeduponhisarrivalinNewYork.HeandhissonwereinGrandCentralTerminalintendingtotakea train forBoston.Heevidentlydisplayedhiswallet a little tooconspicuously ashepaid for his tickets. Two young thugs bumped against him, grabbed the wallet,containingnearly$2,000,andran.(IbelieveitwastheRockefellerFoundationwhichcame to Pavlov’s rescue, making him a grant of $1,000 to permit him “to studyphysiologyintheUnitedStates.”)ThenextdayPavlovwassittingonCannon’sfrontporchwhenCannon suggested theywalk toHarvard Square for a soda.AsCannonstarteddown the steps, Pavlov said, “But you haven’t locked your door.” Cannonassuredhimitwouldbeallright.“Butthere’snooneinyourhouse,andyourdoorisunlocked,”insistedPavlov.“Itdoesn’tmatter,”Cannonsaid,“we’llbebackinalittlewhile.” Pavlov shook his head uncertainly. “My,” he said, “what a great differencebetweenNewYorkandCambridge!”Pavlov’sbookprovedtobeenormouslyhelpfulinmygraduatestudies.Possiblythe

mostimportantlessonI learnedfromit,andoneeasilyoverlooked,wasrespectforafact.OnDecember15,1911,atexactly1:55intheafternoon,adogsecretedninedropsofsaliva.Totakethatfactseriously,andtomakeone’sreaderstakeitseriously,wasnomean achievement. It was important too that it was a fact about a single organism.Animal psychology at that time was primarily concerned with the behavior of theaveragerat.Thelearningcurveswhichappearedintextbooksweregeneratedbylargegroupsoforganisms.Pavlovwastalkingaboutthebehaviorofoneorganismatatime.Healsoemphasizedcontrolledconditions.Hissoundproofedlaboratory,apictureof

which appears in the book, impressed me greatly, and the first apparatus I builtconsistedofasoundproofedchamberandasilentrelease-box.IsuspectthatthecontroloftheenvironmentinPavlov’slaboratorywouldseemratherinadequatetoday.Ihavealwaysbeensuspiciousofthatexperimentinwhichadog,givenfoodevery30minuteswithoutanysignal,beginstosalivatepromptly29minutesafterthepreviousdelivery.Ihaveoftenwonderedwhattheexperimenterdidduringthose30minutes.Myguessisthathelefttheroomtoattendtoothermatters,perhapstohaveasmoke.At29minutesbyhiswatch,ratherthanbyanytemporalconditioning,I imaginehimtiptoeingbackandnotingwithsatisfactionthattheredfluidintheglasstubeshortlythereafterbeganto move. But whether or not the control was adequate, it was held to be of firstimportance.The motto of this society is taken from Pavlov: “Observation and observation.”

Pavlovmeant,ofcourse, theobservationofnature,notofwhatsomeonehadwrittenaboutnature.Hewasopposedtodogmaandwouldbeopposedtocurrentdogmaabouthimself. It is now50 years since hewasmost active, and that is a long time in thehistory of science. His position in that history is secure, for he made extraordinarycontributions; but he was not free of certain limitations. His influence upon thesubsequenthistoryof thestudyofbehaviorhasnotalwaysbeenhappy.I thinkIcan

567

expressmyrespectforPavlovinawaywhichhewouldbemostlikelytoapproveifIindicatecertainpointsonwhichIthinkhewaswrong.He turned too quickly to inferences about the nervous system.The subtitle of the

Anrep translation is “An Investigation of the Physiological Activity of the CerebralCortex.”Pavlovneversawanyofthatactivity;hewasstudyingmerelywhathetooktobeitsproducts.Hisfactswereaboutbehavior,andhisefforttorepresentthemasfactsaboutthenervoussysteminterferedwithhisreportsandmusthaveaffectedthedesignofhisexperiments.Pavlovprobablytookthislineasaproductofanineteenthcenturymaterialism.Sherringtondidthesamethingataboutthesametime.AdifferentbrandofmaterialismcameintothestorywhentheSovietsmadePavlova

nationalhero.Thereisnodoubtthatthenervoussystemismaterial;whenitdecays,itsmells,andcouldoneaskforbetterproof?Behaviorontheotherhandisevanescent.Intalking about it without mentioning the nervous system, one runs the risk of beingcalledanidealist.TherehasneverbeenaseparateRussianscienceofbehavior.Perhapsthatisonereasonwhycyberneticshasbeentakenupsoenergetically.MathematicsandthemachineanalogyhaveatlastpermittedtheRussianstotalkaboutbehaviorwithoutmentioning the nervous system.The fear of being called an idealist has led to someabsurdpractices.Inateacher’scollegeinTashkentthedirectortoldusthatthecollegewasinterestedin“highernervousactivities.”Hemeantsimplythattheywereteachingteachers.Pavlov’s physiological metaphors encouraged him to speculate about processes

supposedtobegoingonbehindhisfactsratherthanaboutthefactsthemselves.Freudhaddonethesamethingbutmuchmoredramatically.ThevariouskindsofinhibitionwhichPavlovthoughthesawinhisdatawerelogicallyunnecessary.Aresponsemay,for many reasons, grow weak; it is not necessarily suffering extinction. But themetaphorofacentral,probablycortical,processisattractive.Aprestigeattachestothestatementthatinhibitionhasspreadacrossthecortex,aprestigewhichislackinginamererecitalofthefactsuponwhichthestatementisbased.Divertedfromastrictformulationofbehavioralfactsassuch,itwaseasyforPavlov

tobelievethatconditionedreflexescomprisedthewholefieldoflearnedbehavior,andtooverlookdifferencesevenamongthekindsofbehaviortowhichtheprincipleseemsto apply. It was extraordinarily lucky that he began with the salivary reflex. Thereseemstobenootherresponsequitesosimple.Otherglandularsecretions,forexample,tearsorsweat,arebynomeanssoeasytocontrol,andwehaveheardtodaysomethingof the enormous complexity of conditioned cardiac responses. The extension of thePavlovianformulationtoskeletalmusculatureraisesespeciallydifficultquestions.Toinsist that the Pavlovian experiment is a useful prototype in formulating all learnedbehaviorisnotreallyveryhelpful.TwoorthreeyearsagotheMoscowCircuscametoBoston,andthebeartrainer,Mr.

Filatov,expressedaninterestintalkingwithmeaboutanimal-trainingprocedures.Mywife invitedhim todinner, togetherwith a charming interpreter andherdate for theevening, the ringmasterof the circus.Whenwegot around to shop talk,Mr.Filatovannouncedthegroundrules:“Now,itisallamatterofconditionedreflexes,isn’tit?”he said. I replied that inAmericawe tried tomakeadistinctionbetween thecase inwhichthereinforcingstimulusaccompaniedanotherstimulusandthecaseinwhichitfollowedaresponse.“Thatdoesn’tmatter,”heinsisted.“Whetherthereinforcementcomesbeforeorafter

theresponse,itisstillaconditionedreflex.Right?”Isawthatwecouldnototherwise

568

getontothetrainingofbears,andsoIagreed.ButofcourseIdonotagree.Acarefulanalysisofcontingenciesofreinforcementin

bothoperantandrespondentbehaviorseemstomeanabsolutelyessentialfirststep.Itisnotaquestionofdifferencesintheory,itisamatterofreachingaformulationwhichfits the known facts. This is a task to which Pavlov, if he were alive today, woulddevotehimselfwithhischaracteristicenthusiasm.Factsandformulationsoffactschangeasscienceprogresses.Theexperimentalspirit

andtheintegrityofthescientistdonotchange.IntheabidingaspectsofthelifeofascientistwestillhavemuchtolearnfromIvanPetrovitchPavlov.

FromConditionalReflex,1966,1,74–78.TheseremarkswereoriginallydeliveredatadinnerofthePavlovianSocietyheldonSaturday,February12,1966.

569

SquirrelintheYard

Fortwenty-fiveyearstheendofthespringtermatHarvardhasbeenheraldedbythesoundofhammers ringingout in theYard,asworkmenreconstruct theTercentenaryTheater for another Commencement. A solid platform rises along the south side ofMemorialChurchandispaintedafreshgray,inashowofpermanencethatdeceiveseventhegrassgrowingbeneathit,whichmustlaterbereplacedwithfreshsod.Fifteenthousand folding chairs are set in place in staked-out rows, fanning out from theplatform to the stepsofWidenerLibrary.On thedaybeforeCommencementagreatcanopy is stretched above the platform, and loud-speakers are lashed to treesthroughouttheYard.ThisisalldoneintheconfidentexpectationthatCommencementDay will be fair, and, indeed, no Commencement has been driven indoors by badweathersincetheTercentenaryTheaterwasfirstbuiltin1936.Thusreassured,Harvardhasreplacedthewoodenplatformwithapermanentstone

terrace,tobeusedthisyearforthefirsttime.NothingelseaboutCommencementwillbegreatlychanged.Thedaywillbeginwithanearlyswellofexcitementoutside thewalls of theYard. Someof the iron gateswill be closed and locked.Otherswill bedefendedbypolicemenwith instructions toadmitalumniand thoseabout tobecomealumni,theirfamiliesandguests,membersoftheHarvardCorporationandtheBoardofOverseers in their frock coats and silk hats, and Facultymembers in their brightgowns and hoods, some in the strange regalia of foreign universities. TheGovernorwillarrivefromtheStateHouseinBoston,nolongercomingonhorsebackhimselfbutstillescortedbytheNationalLancersontheirspiritedmounts.Bynine-thirtyIshallhavejoinedmycolleaguesnearMassachusettsHall.Itwillbe

mydaytowearthecrimsongownownedjointlybythreemembersofmydepartment(andunfortunatelytailoredtoaccommodatethetallestofthem).Weshallstandabout,greeting friends whom we have possibly not seen since last year, glancingunobtrusivelyatacolleague’snewhoodandtryingtoguessfromitscoloredliningthenameoftheuniversitythatgaveittohim.WeshallpeeralittlelesscautiouslyatthoseabouttoreceiveHarvard’shonorarydegrees,whoseidentityhasbeenkeptsecretuntilthismoment,butwhowillnowbestandingaboutinfrontofMassachusettsHall.Eventually, the line will move. Two by two we shall shuffle forward, through a

sombergauntletofblack-gownedseniors,manyofthemlookingontheirprofessorsforthelasttime.(TheremaybeatwingeofguiltasIthinkoftheseniorwhoisnotherebecausehe flunkedmycourse—who isnotonlynot therebut cannotgoon that all-summerhoneymoonoraccept thecommissionwaitingforhiminoneof theservicesbecausehemustcomebacktosummerschooltopassanothercourseforhisdegree.)WeshallmovepastUniversityHallundertheflagswhichstirgentlyabovethestatueof JohnHarvard.We shall take off ourmortarboards out of respect for the flag, orpossiblyJohnHarvard,andputthemonagaincarefully,thetasslestickling.Crossingin front of the steps ofWidener, we shall turn slowly down the center aisle of theTheatertothenewterraceunderthecanopyandmovecarefullyupthesteps(carefully,

570

becausenothing ismore awkward than stumblingonan academicgown;youdonotfall,yousimplywalkup the frontof thegown, forcingyourself intoan ignominiousgenuflection).Weshallstandbeforeourseatsontheterraceforalongtimeasthecandidatesfor

degrees file into theirdesignatedplaces.Then thebell inMemorialChurchwillpealvigorously.TheMarshalwillcallout,“Mr.Sheriff,praygiveusorder!”andtheSheriffof Middlesex County, rapping on the platform three times, will perform his verbalmagic:“Themeetingwillbeinorder.”Nothing that followswill be unpredictable, or evenofmuch interest.Godwill be

toldofourconcernsandhopes.TheLatinSalutatoryDisquisitionwillbeunderstoodby some of the Faculty; the rest of us will simply wait for references to pullaepulcherrimae Radcliffensis and laugh heartily. The English Dissertation and theEnglishDisquisitionwill,inspiteofeveryeffortofthespeakersandtheiradvisors,beperiodpieces, fornooneorates in that styleanymore.Thosewhocame toget theirdegreeswillgetthemandbeadmittedintothesocietyofeducatedmen(andwomen).The citations for the honorary degreesmay be amusing. Later that day some of therecipients of honorary degrees will make speeches from that terrace, and once in awhileoneofthemmakeshistory.Itishardtoexplainwhythisissoexcitingandglamorous.InanycaseIhavelostall

hopeofmakingaproperevaluationmyself,forthebellwhichringsoutsojoyouslytoproclaimtheemancipationofthesethousandsofstudentsawakensinmeacompulsivememory.The bell figured prominently in a ceremonywhich I attended on the samespotseveralyearsbeforetheTercentenaryTheaterwasbuilt.Asquirrelstoletheshow.Initswayitdeservestomakehistory,too.

I owe a special debt to squirrels.When I began graduate work at Harvard, anotherstudentandIordered threeof themfromananimalsupplyhouse.Somehoworothertheycameintomyexclusivecharge,andforaboutayearIgavethemtherunofmylaboratory.Whennotfencingwithmyrats throughthescreenedwallsof theircages,theywouldsitonmydeskorlieinwaitontopofabookcase,leapingtotheshouldersofunsuspectingvisitors.Withintheyeartheygrewtoodestructivetobeleftfree,butImadetheircageaslargeaspossibleandaddedawheel(theclassicalsquirrelcage)toprovideasemblanceoffreedomintheopportunitytorunwithoutinterruption.Thefollowingsummerwasunusuallywarm,andIincreasedtheverisimilitudeofthe

wheelasaninfinitelylongrunningspacebyarrangingafansothatitblewintothefaceofthesquirrel.Itwasanoldelectricfan.Themotorhadgone,butIattachedapulleytotheshaftofthebladeandranabelttoalargerpulleyontheshaftoftherunningwheel.Whenever thewheel turned, the fan turned, at a higher rate. In a sense, the squirrelfanneditselfasitran.Therewas a graduate student fromKorea in our department, aMr.Kim.He had

found it difficult to adjust to the mechanized life in America, and the sight of anAmericansquirrelautomaticallyfanningitselfwastoomuch.Hewouldcomeintomylaboratory day after day and stand beside the squirrel cage, laughing quietly as hewatchedasquirrel runninghead-on into itsself-generatedbreeze.Not longafterwardMr.KimquietlyreturnedtoKoreawithouthisdegree.IalsoexperimentedonthesquirrelsintheYard.Atthattimetherewasconsiderable

interestinthewaychimpanzeesandotherapessolvedproblems—byrakinginbananasthroughthebarsoftheircageswithlongsticksorpilingboxesinordertoreachthem

571

wheretheyhungneartheceiling.Wehadoriginallyboughtoursquirrelsbecausetheyhadhandswhichmightpermitthemtodosimilarthings.IntheopenlaboratoryoftheYardIwould tieapeanut toa two-or three-foot lengthofstringandhang it fromabranchonayoungelm.Asquirrelwouldsquatonthebranchabovethestringandpullitin,handoverhand,asdeftasanysailorpullinginacast-offline.Thissciurinephaseofmystudiesofanimalbehaviorcametoanend in thespring

vacationof1932.TheoldAppletonChapelhadbeentorndown,andMemorialChurchwasjustbeingcompletedinitsplace.Thebelfrywasstillempty,butthebellwhichhadhunginHarvardHallacrosstheYardhadbeenbroughttothebaseofthetower,whereitrestedonanimprovisedplatform.Cablesandpulleyshadbeenrigged,anditwastobe hoisted into place at one o’clock in the afternoon. A little ceremony had beenplanned.PresidentLowell,whohadgiventhebelltoHarvard,wouldbepresent.

Onthemorningofthatday,workmenhadbeguntotakedownalargeelmnearthebell.TheDutchElmDiseasehadnotyetstrucktheYard,butthetreestoodtooclosetothechurchandhadbeencondemned.Bynoonworkmenhadcutoffallthebranchessothatnothing remained but a heavy block Y. As they unpacked their lunches, sittingalongside thewall of the church, quiet settled over theYard.A fewminutes later afrightened squirrel lookedout froma hole in the crotchof theY.Shehad evidentlybeencoweringinthetreeallmorning,throughtheshockandvibrationofaxeandsaw.Acceptingthecease-fire,sheemergedfromthehole,flickedhertailnervously,spiraledhead-firstdownthetreeinsuddenstarts,andsetouttoexplorethe.Yard.Aquarterofanhour later she returned, slipped into thehole in the tree, andcameout carryingastrugglingyoungsquirrel.Itwaswellgrown,probablymorethanamonthandahalfold.Itstailwasstillrat-like,buttherestofitsbodywaswellcoveredwithfur.Ithadalmostcertainlybeenweaned,butasquirrelfamilystaystogetherallsummer,andthisonewastobenoexception.Theyoung squirrel resisted strenuously as itsmother packed it into amanageable

mouthful, patting it rapidlywith her forepaws.With the squirming youngster in hermouthshecameawkwardlydownthetreeandsetoffforEmersonHallahundredandfiftyyardstotheEast.Sheclimbedupthewell-grownivyontherightsideofthedoor,twisting, turning, and occasionally moving backward to get her struggling burdenthroughthebranchesandleavestoaledgeabovethesecondfloor.Afewminuteslatershecamedownempty-mouthed,andwentstraightbacktothetree.Shebroughtoutasecond squirrel of the same vintage, also stoutly protesting, and started off againtowardEmerson.Itwasnownearlyoneo’clockandpeoplehadbeguntogatherfortheceremony.Mr.

Lowellarrivedwithhisredcockerspaniel—toooldtochasesquirrels(andsodeafthatMr.Lowell couldcall itonlybywavingbotharms inabroadsemaphore).Whenhewas told what was happening, Mr. Lowell gave orders that the bell was not to behoisteduntilthesquirrelhadfinished.Therewasashoutof“Hereshecomes!”,andthecrowddividedintotwothinlines,

stretchingfromthebaseof the treewell towardEmersonHall.Completelyabsorbed,thesquirrelenteredthisformidablegauntletandproceededtothetreeandupandintothehole.Shecameoutwrestlingwithanotheryoungster,packedit intohermouthaswell as she could, spiraled down the tree, and set off between the lines, a small legkickingfromhermouthinprotestasshewent.

572

Itwasawarmday,andshewasgrowingtired.Shewouldstoptoflattenherselfagainstthecoolground,thenrallyherforces,recomposeheruncooperativeburden,andsetoff.As she reached EmersonHall and lost herself in the ivy, the crowd began tomoveuncertainly about. It was well past one. The workmen tested the ropes and cablesimpatiently,lookingateachotherandatMr.Lowell.ButMr.Lowellhadspoken,andthebellstayedinitsplace.The squirrel appeared again, and the lines, densernow, reformed.She cameback,

pausingfromtimetotimeingrowingexhaustionbutlookingneithertorightnortoleftinherdedication.Shetookafourthyoungsterfromthetreeandsetofftowardhernewquarters.Speculationaroseastothesizeofsquirrellitters.Unreliablerumorsspreadastohow

manyyoungsters shehadalready transported, and smallwagerswere laid as tohowmanywerestilltocome.Thebookssay“threetosix”inalitterforthegraysquirrel,buttheremustbeexceptions.Possiblythissquirrelhadhadthebadlucktobetakingcareofaneighbor’schildren.Inanycase,shewastocarrysevenyoungsquirrelsfromonenesttotheotherthatday.Shecarriedeachofthemnearlyfivehundredfeet,plustwotallflightsofivyandoneflightofelm.Another speculation ran through the growing crowd. Would she, arriving on the

ledge of Emerson Hall, know that she had rescued her last child? How would sheknow?Didsheknowhowmanychildrenshehad?Couldshe,infact,count?Orwouldshebeabletorememberthatshehadleftanemptynestbehindheronherlasttrip?Itwasnearlytwoo’clockwhen,forthelasttime,someonecried,“Hereshecomes!”

Thecrowd,nowswelledfarbeyondthosewhocaredaboutthemerehoistingofabell,broke intowell-disciplined lines.The exhausted squirrel dragged herself back to theelm, inchby inch,andstrugglingslowly to the topof the trunkslippedwithobviousreliefintothehole.Wewereallsilent.Shereappearedempty-mouthed.Shehadcomeback once too often. Shemade a brief effort to cover hermistake, flicking her taillightlyas if shehadknownall along that thenestwasempty,and thengave in.Shesettledherselflimply,likeatinyfurpiece,ontheroughbarkoftheelmalongsidethehole,lookingtimidly,andalittlesheepishly,outoverheraudienceforthefirsttime.Mr. Lowell nodded curtly to theworkmen, ropeswere shaken clear, and the bell

movedslowlyupwardandwasswungintoplaceinthetowerofMemorialChurch.

FromHarvardAlumniBulletin,May1962.

573

PARTX

Coda

CanPsychologyBeaScienceofMind?

574

‘Faith,Imustleavethee,love,andshortlytoo;Myoperantpowerstheirfunctionsleave

[cease]todo.HAMLET,III,ii

575

CanPsychologyBeaScienceofMind?

[This article was completed August 17, 1990, the evening before its author died. Itappeared in the November 1990 issue of the American Psychologist.The followingaccountofB.F.Skinner’slastdayswaswrittenbyhisdaughter,JulieS.Vargas.]

My father always spoke admiringly of his grandfather’s death. “He diedwith his bootson,”he’dsay,approvingly.Sometimeshewouldadd,“That’sthewayIwanttogo.”Well,hecameclose.…Eight days before his death,… my father received, from the American Psychological

Association,thefirstAPACitationforOutstandingLifetimeContributiontoPsychology.Theassociationofficialshadassured the family that theywouldkeepmy father fromcrowds—importantbecauseofhisheightenedsusceptibilitytoinfectionfromleukemia—andtheykepttheirword.At 1:00o’clockonAugust 10th, a limousineappearedat theSkinnerhome todriveourpartytotheconventionhotel.Thereweweremetandusheredupstairsinourownelevatortoahotelroom,“likemoviestars,”myfatherremarked.Afewminutesbeforetheopeningsessionwastobegin,wewereusheredbackdownstairsandtakenbyabackwaytothesidedooroftheauditorium.Iwasholdingmyfather’sarmasweentered.Theroomwaspacked.Asecondroomtothesidehadbeenopenedandit,too,wasoverflowing.Whenwehadtakentwostepsintotheroomeveryonestoodupandbegantoapplaud.Myfathermadeanawkwardnodof his head in acknowledgment as he continuedwalking—I could tell hehadn’texpectedsuchareception.Theapplausewasthunderous.Itcontinuedaswemadeourwaytothebottomofthestepsatthemiddleofthestage.Itcontinued,undiminished,asmyfatherwasescortedacross the stage tohis chair.He turnedaroundandmadeagraciousbow of his head, but there was no sign of the applause letting up. Finally, APA officialsinterruptedtheapplauseandstartedtheprogram.…Iwasgladthathehaddecidednottouseatextorevennotes,fortheglaringlightfromthe

TVcameracrewswouldhavemadeitimpossibleforhimtohaveseenanything.Hebegan.“President Graham, past-President Matarazzo, distinguished guests, ladies and

gentlemen….”1Hetalkedsmoothly,thewayIhadheardhimtalkatdozensofconventions,completewithnamesanddatesIwouldhavehadtroubleremembering.Thetalkturnedtothesplitinpsychology,“onepartgoinginthedirectionoffindingouttheessenceofthefeeling,the essence of the cognitive process, and the other going in the direction of references tocontingenciesofreinforcement.”HedrewananalogybetweenthedifficultyofacceptanceofDarwin’s natural selection and the difficulty of acceptance of Skinner’s own selection byconsequences,culminatingin thestatement,“SofarasI’mconcerned,cognitivescienceisthe creationism of psychology.” The whole audience gasped audibly. A sprinkling ofclappingcouldbeheardhereandthere.(Clearlythesplitwasnot50–50.)Skinnercontinued,endinginjustover15minutes—ashehadplanned.…Overtheweekend,myfatherworkedonthepaperfromwhichhisremarksweretaken.It

wastogointheAmericanPsychologistandhewasanxioustofinishit.Wetalkedaboutthefuture—whathewouldworkonnext.Hedidn’tfeelhewouldhavetimetocompletethework

576

hehaddoneonthederivationsofwords,ortoputintoarticleformmaterialfromabookonethicshehadbegunbutgivenup.Asitturnedout,hewasrightabouttime.Mondaymorningmy fatherhadanhour interviewwithDanBjork,ahistorianwhohad

been working on a biography of him. In the afternoon, I took him for his usual platelettransfusion.ThateveningIwasworkinginthedenIhadcreatedoutofastoreroominthebasementwhenIgotacall.Itwasmyfather.“Julie,couldyoucomehere?”Irushedtohisstudy to find him shivering under several blankets in his reclining chair. I panicked. Thesymptomswerethesameasthosethathadputhimintoacomaonetimebefore.Unabletoreach a doctor, I called 911. By the time the ambulance arrived he was feeling better,althoughhisheartwasstillracing.Theambulanceteamgavehimoxygenandtookallsortsof measures, conferring by phone with medical personnel from his infirmary. My fatherdidn’twant to go back to the hospital, so, because he appeared stable, the team left. Theoxygenhadmadehimfeelbetter,soIhookedupanoldtankhehadsavedfromyearsago.Iencouragedhimtostaysittingupbecausehisheartwasstillracing,andasittingpositionislessstrainontheheart.Heagreed,andtookoutabooktoread.Isetupacotinhisstudyandbroughtinmyguitar.ForanhourIplayedforhim—alloftheclassicalpiecesIcouldplayreasonablywell.Itpleasedhim.Hehadn’theardmeplayinsometimeandcommentedonthe“richness”ofthesound.TheoxygenandIbothranoutaboutthesametime.Later,inhisbed,aJapanese-designedsleepingcubicleinthefarcornerofhisstudy,wetalked.Isatontheedge,holdinghishand,assomanytimes,dewy-eyed,hehadheldminewhenputtingmetobedasachild.Onlythistimethereweretearsinbothofoureyes.Iawokethenextmorningtofindhimawake,butweak.Inspiteofmyurging,herefusedto

cancelaTVcrewscheduledtotakefootageforthatnight’snewscast.Wednesdaymorning,another interview. That afternoon hewent into the hospital for the last time. But the daybeforehedied,inthehospital,heworkedonthelastchangesinhispaperfortheAmericanPsychologist.B.F.SkinnerwasamemberoftheHemlockSocietyandbelievedintherighttotakeone’s

ownlife.Hehadmadealivingwilland, in thehospital,againrefused“heroic”lifesavingeffortsthatcouldhaveprolongedthefunctioningofhisorgans.Neartheend,hismouthwasdry.Uponreceivingabitofwaterhesaidhislastword,“Marvelous.”[FromVargas, J. S. (1990). B. F. Skinner—The last few days. Journal of the Applied

AnalysisofBehavior,23,409–410.]

Manypsychologists,likethephilosophersbeforethem,havelookedinsidethemselvesfor explanations of their behavior. They have felt feelings and observed mentalprocesses through introspection. Introspection has never been very satisfactory,however. Philosophers have acknowledged its inadequacies while insisting that it isneverthelesstheonlymeansofself-knowledge.PsychologistsoncetriedtoimproveitbyusingtrainedobserversandthebrassinstrumentsofwhichWilliamJameshadsuchalowopinion.Introspectionisnolongermuchused.Cognitivepsychologistsmayseerepresentationsandmayevenarguethattheyaretheonlythingsthatcanbeseen,buttheydonotclaimtoseethemselvesprocessingthem.Instead,likepsychoanalystswhoface the same problem with processes that cannot be seen because they areunconscious,theyhaveturnedtotheory.Theoriesneedconfirmation,however,andforthatmanyhave turnedtobrainscience,whereprocessesmaybesaid tobe inspectedratherthanintrospected.Ifthemindis“whatthebraindoes,”thebraincanbestudiedas any other organ is studied. Eventually, then, brain science should tell us what itmeans to construct a representation of reality, store a representation in memory,

577

convertanintentionintoaction,feeljoyorsorrow,drawalogicalconclusion,andsoon.Butdoesthebraininitiatebehaviorasthemindorselfissaidtodo?Thebrainispart

ofthebody,andwhatitdoesispartofwhatthebodydoes.Whatthebraindoesispartofwhatmustbeexplained.Wherehasthebody-cum-braincomefrom,andwhydoesitchangeinsubtlewaysfrommomenttomoment?Wecannotfindanswerstoquestionsof that sort in the body-cum-brain itself, observed either introspectively or with theinstrumentsandmethodsofphysiology.

Thebehavioroftheorganismasawholeistheproductofthreetypesofvariationandselection.Thefirst,naturalselection,isresponsiblefortheevolutionofthespeciesandhenceforspeciesbehavior.Alltypesofvariationandselectionhavecertainfaults,andoneofthemisespeciallycriticalfornaturalselection:Itpreparesaspeciesonlyforafuture thatresembles theselectingpast.Speciesbehavior iseffectiveonly inaworldthatfairlycloselyresemblestheworldinwhichthespeciesevolved.Thatfaultwascorrectedbytheevolutionofasecondtypeofvariationandselection,

operant conditioning, throughwhich variations in the behavior of the individual areselectedbyfeaturesof theenvironmentthatarenotstableenoughtoplayanypart inevolution.Inoperantconditioning,behaviorisreinforced,inthesenseofstrengthenedormademore likely tooccur,bycertainkindsofconsequences,whichfirstacquiredthepowertoreinforcethroughnaturalselection.A second fault in variation and selection is critical for operant conditioning:

Selectionmustwaituponvariation.Theprocessisthereforeusuallyslow.Thatwasnotaproblem fornatural selectionbecause evolution could takemillionsofyears, but arepertoire of operant behavior must be constructed during a lifetime. Operantconditioningmustsolvethe“problemofthefirstinstance”:Howandwhydoresponsesoccurbeforetheyhavebeenreinforced?The problem was solved in part by the evolution of processes through which

individuals take advantage of behavior already acquired by others. Imitation is anexample. It often brings the imitator into contact with the reinforcing consequencesresponsiblefor thebehavior imitated.Thebehaviorof the imitator is“primed”in thesenseofmadetooccurforthefirsttimeandusuallywhenitislikelytobereinforced.At thispoint thehuman species appears tohave takenaunique evolutionary step.

Other species imitate, but if they model behavior to be imitated, it is only as theproductofnaturalselection.Theconsequenceofmodelingthebehavioroftheimitatoris too remote to serve as an operant reinforcer.Only in the human species does thebehavioroftheimitatorreinforcemodeling.The species underwent another unique evolutionary change when its vocal

musculaturecameunderoperantcontrolandwhenvocalbehaviorbegantobeshapedandmaintainedbyitsreinforcingconsequences.Peoplecouldthenprimethebehaviorofothersbytellingthemwhattodoaswellasbyshowingthem.(Inapresumablylaterstep,temporaryreinforcingconsequenceswereaddedtomakethebehaviormorelikelytoremain instrengthuntil theconsequenceforwhich itwasprimedcouldcomeintoplay.Addingtemporaryreinforcementsinthiswayisteaching.)Advicecanbeusefulonmorethanoneoccasion,anditisthenoftengivenortaught

in such a way that it is passed on from person to person or from generation togeneration.Maxims(“greatsayings”)andproverbs(“sayingsputforth”)areexamples.

578

They describe rather general contingencies of reinforcement—a penny (as well asmanyotherthings)savedisapenny(aswellasmanyotherthings)earned.Rulesaresayings transmitted by groups, usually with stronger reinforcing consequences. Thelaws of governments and religions describe the contingencies of (usually negative)reinforcementmaintainedby those institutions.Theyhave theeffectofwarnings:Byobeyingthelawapersonavoidsbehavinginwaysthatwouldbepunished.Thelawsofphysics and chemistry (“rules for effective action”) describe the contingencies ofreinforcementmaintainedbythephysicalenvironment.

Modeling, telling, and teaching are the functions of the social environments calledcultures. Different cultures emerge from different contingencies of variation andselection and differ in the extent to which they help their members solve theirproblems.Memberswhosolvethemaremorelikelytosurvive,andwiththemsurvivethepracticesoftheculture.Inotherwordsculturesevolve,inathirdkindofvariationand selection. (Cultures that shape and maintain operant behavior are exclusivelyhuman. Animal societies have many similar features, but only as the product ofcontingenciesofsurvival.)Culturalevolutionisnotabiologicalprocess,butasakindofvariationandselectionithasthesamefaults.Thefactthataculturepreparesagrouponlyforaworldthatresemblestheworldinwhichthecultureevolvedisthesourceofourpresentconcernforthefutureofahabitableearth.Theprocessofvariationandselectionhasa thirdfault:Variationsarerandomand

contingenciesofselectionaccidental.Whatevolvedisnotasingleslowlydevelopingspeciesbutmillionsofdifferentspecies,competingwitheachotherforaplaceintheworld. The product of operant conditioning is not a single coherent repertoire butthousands of smaller repertoires, conflicts amongwhichmust somehowbe resolved.Theevolutionofsocialenvironmentshasproducednotasingleculturebutmanyoftenconflictingones.

Althoughoperantcontrolofthevocalmusculatureisexclusivetothehumanspecies,itis seldom if ever cited as its distinguishing feature. The presence or absence of“consciousness”or“consciousintelligence”ismorelikelytobecited.Theroleplayedbythemind/brainhasalwaysbeenaproblemincomparingspecies.Descartesexcepted“man”fromhismechanicalmodelofanorganism,andWallace,unlikeDarwin,didnotbelievethatevolutioncouldexplainthehumanmind.Brainscientistshaveexpressedsimilar reservations. Evolutionary theorists have suggested that “consciousintelligence” is an evolved trait, but they have never shown how a nonphysicalvariation could arise to be selected by physical contingencies of survival. (Thesuggestion simply moves the bothersome physical-metaphysical distinction a stepfurtheroutofsight.)Ithasbeensaidthatwemayneverknowhowaconsciousmindevolved because nothing would survive for palaeontologists to discover, but theoperant control of the vocalmusculature and the showing, telling, and teaching thatfollowhavesurvived,anditispossiblethattheyexplainintrospectionandalsowhatis“seen”withitshelp.The root spect suggests vision. We say that we “look at” and “see” what is

happeningwithinourselves,butnoinnereyehaseverbeendiscovered.Wecanavoidspecifyingakindoforganbysayingobserve,notice,ornoteratherthansee, and it is

579

significantthatobserve,notice, andnote, and lesscommonlyremark,meanbothsayandsee.Muchdependsonwhatitmeanstosenseanypartoftheworldwithanykindof organ. Input-output theories, as in stimulus-response or information-processingmodels,makeasharpdistinctionbetweensensinganddoing.Wearesaidtosensetheworld before acting upon it. The experimental analysis of behavior assigns a verydifferentroletothestimulus,however.Anoperantresponseismorelikelytooccurinthepresenceofastimulusthatwaspresentwhenitwasreinforced.Sensingisasmuchaproductofvariationandselectionasdoing.Itisapartofdoing.Forsimilarreasons,natural selection explains the readiness with which animals respond instantly tofeaturesoftheenvironmentthathavebeencrucialtothesurvivaloftheirspecies,suchas the sight, sound, or smell of food or sexual opportunity, or a threat of danger,including thedangerof theunfamiliar.Animalspresumably “receive” all the stimulithat impinge upon them, but it is possible that they only respond to those that haveplayed a part in contingencies of selection. (We cannot know whether nonverbalanimals see stimuli that have never played such a part, because we should have toarrangecontingenciescontainingsuchstimuliinordertofindout.)Weourselvesmayseethingswithrespecttowhichwehavetakennopracticalaction(weseethingsthatareoutofreach,forexample)butpossiblyonlybecausewehavetalkedaboutthem.Toseethingswithouttakingfurtheractionistobeawareofthem.(Therootinaware isalsofoundinwary;wearewaryofthingsthathavebeenpartofnegativecontingenciesofselection.)Thewordconscious,usedmoreoftenthanaware,meansco-knowledge(Latin:con-science)or“knowingwithothers”—anallusiontotheverbalcontingenciesneededforbeingconscious.Allthisisparticularlyimportantwhenwhatweseeiswithinourbody,thekindof

seeing towhichweusually reserve theword introspection.Butwhat dowe actuallysee?Psychologistswhoareuneasyaboutthemetaphysicalnatureofmentallifeoftensaythatwhatweseethroughintrospectionmustbethebrain,butthatisunlikely.Wehaveno sensorynerves going to important parts of the brain; a surgeon canoperatewithoutanesthesia.Nocontingenciesofselectionwouldhavepromotedtheevolutionof such nerves before the advent of verbal behavior, and that was very late in theevolutionofthespecies.Itismorelikelythatwhatweseethroughintrospectionaretheearly stages of our behavior, the stages that occur before the behavior begins to actupontheenvironment.Sensingissuchastage;weseethingsbeforewerespondtotheminanyotherway,

andwe see thatwe are seeing themwhenwe are doingnothing else.Thenecessarycontingencies are supplied by people who ask us whether we see things. The verybeginning of action is another early stage. It does not raise any question of theavailabilityofsensorynervesbecauseweshouldbeable toseeearlystageswith thenervesneededfor thecompleteaction. (It isalsopossible thatwearesometimesnotintrospecting at all, but are responding to the external setting, as if “I am going togo…”meant“InsituationslikethisIhaveusuallygone….”)TheGreeksaresaidtohavediscoveredthemind,butitismorelikelythattheywere

the first to talk at great length about what they saw within themselves and thusconstruct the contingencies needed for introspection. The “Great Conversation” ofPlato’sAcademywouldhavecreatedcontingenciesunderwhichmoreandmoreofthebeginningsofbehaviorwouldbeseen.Itmusthavebeenapuzzlingworld.Weseethepublicworldaboutus,butwealsofeel,hear,taste,andsmellit.Wedonothingwithaninnerworldbut“see”it.ItisnotsurprisingthattheGreekscalleditmetaphysical.

580

Unfortunately,whattheysawoccurredatjustthetimeandplacetobemistakenforacause of what they then did, and it was therefore easy to suppose that they haddiscoveredanoriginatingselformind. Ifwhat theysawwassimplyanearlypartofwhattheythendid,however,itwasnomoreacauseoftherestofwhattheydidthanthebackswingofagolferisthecauseofthestrokethatstrikestheball.Earlypartsofbehavior affect later parts, but it is the behavior as a whole that is the product ofvariationandselection.Such an analysis of introspection and of the “consciousness” introspected needs

carefulconsideration,ofcourse,buteveryeffortshouldbemadetopreserveitbecauseitdispenseswithanyneedtoappealtoaspecialkindofknowingoraspecialkindofstuff known. It stayswithin theworld of physics and chemistry and the sciences ofvariationandselection.Itavoidsanysuggestionofabreakintheprocessesofvariationandselection.

Twoestablishedsciences,eachwithaclearlydefinedsubjectmatter,haveabearingonhuman behavior. One is the physiology of the body-cum-brain—amatter of organs,tissues,andcells,andtheelectricalandchemicalchangesthatoccurwithinthem.Theother is a group of three sciences concerned with the variation and selection thatdeterminetheconditionofthatbody-cum-brainatanymoment:thenaturalselectionofthe behavior of species (ethology), the operant conditioning of the behavior of theindividual(behavioranalysis),andtheevolutionofthesocialenvironmentsthatprimeoperantbehaviorandgreatlyexpanditsrange(apartofanthropology).Thethreecouldbesaidtoberelatedinthisway:Physiologystudiestheproductofwhichthesciencesofvariationandselectionstudytheproduction.Thebodyworksas itdoesbecauseofthe laws of physics and chemistry; it doeswhat it does because of its exposure tocontingenciesofvariationandselection.Physiologytellsushow thebodyworks, thesciencesofvariationandselectiontelluswhyitisabodythatworksthatway.The two sciences observe very different causal principles. The body-cum-brain

obeysthelawsofphysicsandchemistry.Ithasnofreedomandmakesnochoices.Noothervisionof“manamachine”(inthiscaseabiochemicalmachine)haseverbeensowellsupported.Somebrainscientistshaveargued that thebrainmusthavestructuralfeaturesthatallowforfreedomofchoice,creativity,andthelike,butindoingsotheyarguefromwhatthebraindoesratherthanfromitsstructure.Ithasalsobeensaidthatvariationandselectionmayoccur in thebrain,butalthough thebrain, likeanyotherpart of the body, undergoes variations, the selecting contingencies are in theenvironment.Themorewe know about the body-cum-brain as a biochemicalmachine, the less

interestingitbecomesinitsbearingonbehavior.Ifthereisfreedom,itistobefoundintherandomnessofvariations.Ifnewformsofbehaviorarecreatedtheyarecreatedbyselection.Thefaultsinvariationandselectionareasourceoffascinatingproblems.Wemust adapt to new situations, resolve conflicts, find quick solutions. A lawfulbiochemicalstructuredoesnothingofthesort.Computer simulations of human behavior are electronic machines designed to

behave as the biochemical machine of the body behaves.We know how they weredesignedandbuilt,andhenceweasknoquestionsaboutorigin.Forthesamereason,however,simulationsareofnoparticularinteresttobehavioranalysts.Theinterestingthingsinlifecomefromthevagariesofvariationandselection,intheconstructionof

581

themachine.Behavioranalysisistheonlyoneofthethreesciencesofvariationandselectionto

be studied at length in the laboratory. Ethologists observe behavior in the field andreconstruct evolution from evidence that survives from earlier times. Ethology issupportedbyalaboratoryscience,genetics,butnoonehasyetproducedanewspecieswith a repertoire of innate behavior under laboratory conditions. The evolution of acultureisalsoprimarilyamatterofinferencesfromhistory.Itisspeedthatmakesthedifference; only operant conditioning occurs quickly enough to be observed frombeginning to end. For the same reason it is the only one of the three sciences to bemuchusedforpracticalpurposesindailylife.It is thereforehard tounderstandwhyoperantconditioninghasnotattractedmore

attention.Theroleofvariationandselectioninthebehavioroftheindividualisoftensimplyignored.Sociobiology,forexample,leapsfromsocio-tobio-,passingoverthelinking individual.Many of the psychologists who have studied behavior have alsoneglected variation and selection. Thorndike’s Law of Effect came close, but hisexperiment suggested that variations were trials and consequences errors. Watson,Lashley, andHull appealed to habit formation and stimulus and response. Tolman’spurpose, like goal orientation or subjective expected utility, projected copies of pastconsequencesintothefutureasattractionsthatseemedtopullbehavior.Behavioranalysisistheyoungestofthethreesciences(theoriesofnaturalselection

andtheevolutionofculturesdatingfromthemiddleofthe19thcenturyandbehavioranalysisonlyfromtheendofthefirstthirdofthe20th),butimmaturitywillnotexplainwhyithassooftenbeenneglected.Abetterexplanationmaybethatitsfieldhadbeenoccupiedforsolongbythatextraordinarilyintriguingtheoryofaninternaloriginatingmindorself.

We do not speak the languages of brain science and behavior analysis in our dailylives.Wecannotsee thebrainandweknowvery littleabout thehistoryofvariationandselectionresponsibleforagiveninstanceofbehavior.Instead,weusealanguagethatcameintoexistencelongbeforetherewerephilosophersorscientistsofanykind.Itisproperlycalledavernacular.Thewordmeans,asitsrootmeantfortheRomans,the language of the household, of daily life. We all speak it. It is the language ofnewspapers,magazines,books,radio,andtelevision.Whenspeakingofthebehaviorofthe individual, it is the languageofbehavioralscientists—psychologists,sociologists,anthropologists,politicalscientists,andeconomists.WilliamJameswrotePrinciplesofPsychology in the vernacular. Behaviorists speak it in their daily lives (and youngbehavioristsmustlearntodosowithoutembarrassment).Thevernacularreferstomanyfeelingsandstatesofmind.InEnglish,forexample,

wesaythatwedowhatwefeellikedoingorwhatweneedtodotosatisfyourdesires.Wesaythatwearehungryandarethinkingofgetting something toeat. It is easy tosupposethat thereferencesaretoaninitiatingmind,but,aswehaveseen, theusefulallusions are toprior contingenciesof selectionor to thebeginningsof action.From“I’mhungry”weinferthatapersonhasnoteatenforsometimeandwillprobablyeatwhen food isavailable.From“Iam thinkingofgetting something toeat”we inferaprobabilityofdoingsomethingthatwillmakefoodavailable.Through the use of the vernacular with its allusions to personal history and

probability of action, psychology has emerged as an effective, essential, and highly

582

respectedprofession.Theattempt touse theapparentreferences toaninitiatingmindandtoconvertthevernacularintothelanguageofasciencewas,however,amistake.Watson and other early behaviorists thought the mistake lay in using introspection.How well could feelings be felt or mental processes seen? Anticipating logicalpositivism,theyarguedthataneventseenbyonlyonepersonhadnoplaceinascience.Theproblemwasnotintrospection,however.Itwastheinitiatingselformindtowhichintrospectionseemedtogainaccess.In face-to-face contact with another person, references to an initiating self are

unavoidable.There isa ‘you,’and there isan‘I.’ Iseewhat ‘you’doandhearwhat‘you’sayandyouseewhat‘I’doandhearwhat‘I’say.Wedonotseethehistoriesofselection responsible forwhat is done and therefore infer an internalorigination, butthesuccessfuluseofthevernacularinthepracticeofpsychologyoffersnosupportforitsuseinascience.Inascientificanalysis,historiesofvariationandselectionplaytheroleoftheinitiator.Thereisnoplaceinascientificanalysisofbehaviorforamindorself.What,then,arewetomakeofthefactthatfor100yearspsychologistshavetriedto

build just such a science ofmind?What about the brilliant analyses that have beenmadeofintelligenceortheclaimsforthevalueoftheconceptofsubjectiveexpectedutilityor theequations thathavebeenwritten todescribepsychologicalspace?Havethesebeenpartsofasearchforsomethingthatdoesnotexist?Itappearsthatwemustsayso,butallisnotlost.Intelligence,neverintrospectible,isclearlyaninferencefromthe behavior sampled in intelligence tests, and an analysis of different kinds ofintelligenceisananalysisofdifferentkindsofbehavior.Expectation,anotherkindof“spection,” cannot possiblymean seeing the future andmust be the product of pastcontingencies of reinforcement.Utilitymeans usefulness or use, the act ormeans ofdoingsomethinginsuchawaythatconsequencesfollow.Psychologicalspaceisrealspace as it enters into the control of contingencies of reinforcement; at issue is theextenttowhichastimuluspresentwhenaresponseisreinforcedgeneralizesinsuchawaythatsimilarstimulithatwerenotpresentexertcontrol.Inshort,psychologistshaveunwittingly been analyzing contingencies of reinforcement, the very contingenciesresponsibleforthebehaviormistakenlyattributedtoaninternaloriginator.Butwhatabouttheillustriousphilosopherswhothroughoutthecenturieshavetried

to follow the injunction of the Delphic Oracle and to know themselves throughintrospection? Is there a similar justification or have they been uselessly pursuing awill-o’-the-wisp?Tosaysowouldseemlittleshortofarrogance if therewerenotanilluminatingparallel.Equally illustriousmenandwomenhavesearchedmuch longerandwith greater dedication for anotherCreator, spelled this timewith a capital ‘C,’whosereportedachievementsarealsobeingquestionedbyscience.ItwasDarwin,ofcourse,whomadethedifference.Itholdsfortheoriginofbehavioraswellasfortheorigin of species. After almost a century and a half, evolution is still not widelyunderstood. It isvigorouslyopposedbydefendersof a creator.Asa result, it is stillimpossibletoteachbiologyproperlyinmanyAmericanschools.Acreationsciencehasbeen proposed to be taught in its place. The role of variation and selection in thebehaviorof the individual suffers from the sameopposition.Cognitive science is thecreation science of psychology, as it struggles tomaintain the position of amind orself.The history of psychology is informative. It began, 100 years ago, with an

introspective search for mind. Watson attacked introspection in his behavioristic

583

manifesto of 1913, and for that or other reasons introspection was essentiallyabandoned. Behaviorists turned to the study of behavior for its own sake, andnonbehavioristicpsychologiststurnedtothebehaviorofteachers,students,therapists,clients,childrengrowingolderyearbyyear,peopleingroups,andsoon.Cognitivepsychologists triedtorestorethestatusquo.Behaviorism,theydeclared,

was dead. They could not have meant that psychologists were no longer studyingbehavior,ofanimalsinlaboratoriesandofteachers,students,therapists,clients,andsoon.What they hoped was dead was the appeal to selection by consequences in theexplanation of behavior. Themind or, failing that, the brainmust be restored to itsrightfulposition.Because of its similarity to the vernacular, cognitive psychology was easy to

understandandtheso-calledcognitiverevolutionwasforatimesuccessful.Thatmayhave accelerated the speed with which behavior analysts drew away from thepsychological establishment, founding their own associations, holding their ownmeetings, publishing their own journals. They were accused of building their ownghetto, but theywere simply accepting the fact that they had little to gain from thestudyofacreativemind.Cognitivepsychologywasleftasthescientificcompanionofaprofessionandasthe

scientificunderpinningofeducational,clinical,developmental,social,andmanyotherfieldsofpsychology.Thehelpithasgiventhemhasnotbeenconspicuous.Aversionofthevernacularrefinedforthestudyofmentallifeisscarcelymorehelpfulthanthelayversion,especiallywhentheorybegantoreplaceintrospection.Muchmoreusefulwouldhavebeenbehavioranalysis.Itwouldhavehelpedintwoways,byclarifyingthecontingencies of reinforcement to which the vernacular alludes, and by making itpossible to design better environments—personal environments that would solveexistingproblemsandlargerenvironmentsorculturesinwhichtherewouldbefewerproblems. A better understanding of variation and selection will mean a moresuccessful profession, but whether behavior analysis will be called psychology is amatterforthefuturetodecide.

1AllquotesinVargas’sintroductionarefromextemporaneouscommentsmadebyB.F.SkinneruponreceivingAPA’slifetimeawardattheopeningsessionoftheAmericanPsychologicalAssociation’s98thAnnualConvention,August10,1990,inBoston.

584

APPENDIX

IntroductionsRevisedforthe1961EditionTheintroductionsusedinthisvolumearefromtheFirstEditionofCumulativeRecord,publishedin1959.Inpreparingthe1961EnlargedEdition,Skinnermodifiedmanyofthem but none more than the two presented here. Additional information about hisrevisionsisprovidedintheForeword.

1.HasGertrudeSteinaSecret?Therevisedintroductorycommentsgivenbelowfirstappearedonpage261intheEnlargedEdition(1961)andthenwereusedasapostscripton page 369 of the Third Edition. The original introduction from the First Edition(1959)appearsonpages405–406ofthisvolume.IfirstheardabouttheAutobiographyofAliceB.Toklas fromMaryLouiseWhite

(Aswell).IthadreachedherdeskintheeditorialofficesoftheAtlanticMonthlyasabonafideautobiography,butthelastparagraphhadcomeasnosurprise:“AboutsixweeksagoGertrudeSteinsaid,itdoesnotlooktomeasifyouwereevergoingtowritethatautobiography.YouknowwhatIamgoingtodo.Iamgoingtowriteitforyou.IamgoingtowriteitassimplyasDefoedidtheautobiographyofRobinsonCrusoe.Andshehasandthisisit.”MissWhitewastobetheonlyAtlanticreadertoenjoyGertrudeStein’s little joke forwhenpartsof thebookwerepublished in thatmagazineduringthesummerof1933,theauthor’snameappearedonthetitlepage.MyarticleappearedintheJanuary,1934numberofAtlanticMonthlyandisreprintedwithpermission.Literary critics are by no means agreed on the significance of Gertrude Stein’s

psychologicalexperiments.ConradAiken,writingin theNewRepublic,reviewed theprevailingattitude towardMissStein’s literary revolution—“Like the splittingof theatom, or the theory of relativity,Miss Stein’s destruction ofmeaningwas inevitablygoing tochange, ifnot theworld,atanyrate theword”—andexpressed theopinionthatthepresentarticlemadeofthewholethingaverycrueljoke.“Whatbecomesofallthispreciseanddetachedandscientificexperimentationwithrhythmandmeaning if,after all, it has been nothing on earth but automaticwriting? Is itmerely onemoreinstance of the emperor’s new clothes? Have we been duped, and has Miss Steinherself,perhaps,beenduped?Itlooksverylikeit—thoughofcourseitisnotimpossiblethatMissSteinhasbeenpullingourlegs.”AmongthosewhocametothedefensewasSherwoodAnderson,whoquiteproperlyrejoinedthat“allgoodwritingis,inasense,automatic.”ButpartisansofGertrudeSteinhavebeenafraidofthisanswerandhavegenerallydeniedthatherpublishedworkwaswrittenautomatically.

2.TheConceptof theReflex in theDescriptionofBehavior.This revisionof theoriginal introductionappearedonpages319–321of theEnlargedEdition (1961)andwas retained in that form on pages 429–431 in the Third Edition (1972). TheintroductionfromtheFirstEdition(1959)appearsonpages475–477of thisvolume.

585

[Pagenumbersbelowrefertothepresentvolume.]Thispaper,writteninthesummerof1930,wasanearlyexampleoftheoperational

analysis of terms describing behavior.The clue to the definition of reflexmay havecomefromBertrandRussell.Somewhere,possiblyinaseriesofarticlesintheDialinthe late20’s,Russellpointedout that theconceptof thereflex inphysiologyhad thesame status as the concept of force in physics. Add that toBridgman’s treatment offorceinTheLogicofModernPhysicsandyouhave thepresentpoint.TheargumentcouldbesupportedwithaMachiananalysisofthehistoryofthereflextoexplainthetraditional definition as unconscious, involuntary, and unlearned behavior. TheoperationalanalysisofSherrington’ssynapse,andthemoregeneralizedstatement inChapter12ofTheBehaviorofOrganismswhereitwassuggestedthatC.N.S.mightbetaken tostand for theConceptualNervousSystem,havebeen interpretedasshowingananti-physiologicaloranti-neurologicalbias.Theobjection,however,wasmerelytotheuseofinferencesfrombehaviortoexplainbehavior.Thepaperarguedforthesolidstatus of behavioral facts apart from imagined physiological counterparts. Animportantpointisthatoperationaldefinitionsaresuggestednotonlyforreflex,butfordrive,emotion,conditioningandothertermsappropriatetotheintactorganism.The article also insisted upon the appropriateness of the concept of the reflex in

describing behavior.One consequence is a statement of behavioral facts in a formwhichmost readily makes contact with physiological concepts andmethods. In thissensethepaperis,Ibelieve,apositivecontributiontophysiology.True,thetermreflexproved too rigid. It implieda complete specificationofpropertiesof behaviorwhichwas not in fact supported by the data either from whole or surgically subdividedorganisms (a pointwhichwasmade subsequently in the paper on page504),and itinsisted upon a demonstrable and presumably controllable stimulus as the principalindependentvariable(apointeventuallyrejectedinthepaperonpage535andinTheBehaviorofOrganisms).Similarlysharpreference tobehavior inphysical termswasmaintainedby the concept of the“operant” introduced to remedy these defects (seepage535).A further elaboration of this concept in interpreting, for example, verbalbehaviorpresents thebehavioral facts tobeaccounted forby thephysiologist in thesimplest possible way. Psychological facts remain on the plane of the physical andbiological.Thepaperalsoofferedaprogramandageneralformulationofascientificanalysis

of behavior. As to program, the last paragraph on page 501 describes withconsiderableaccuracya subsequent investigationof“conditioning,”“emotion,”and“drive,” some results of which are sketched in the paper on page 132. As toformulation,theobservablefactsunderlyingtheconceptofthereflexpermitustowritetheequation:

R=f(S)whereR=responseandS=stimulus.Changes in this function provide another sort ofdatum.Forexample,ifwerepeatedlyelicitaresponse,thereflexundergoes“fatigue.”Wemayrewritetheequation:

R=f(S,A).InthispapertheletterAwasnoncommittallyreferredtoasa“thirdvariable,”apointwhichisrelevanttocurrentcontroversiesaboutinterveningvariables.ThevariableAismerelyanothervariabletobetakenintoaccount;itdoesnotinterveneeventhough

586

thetraditionalpracticeistoassignthechangeinthefunctionwhichitaccountsfortosome inner state or condition. Some experimental studies on rate of eating (offered,togetherwiththepresentpaper,asmydoctoralthesis)exemplifiedthisdistinction.Theconcept of “hunger” as an inner statemay be useful for certain purposes, but in astrictly operational definition we must confine ourselves to changes in “reflexstrength”asafunctionofdeprivation,satiation,andsimilaroperations.Edward C. Tolman missed this point when he set up his own version of an

“operationalbehaviorism.”1Tolman’sequation:

B=f(SPHTA)is patterned after the equations above (with B for “behavior” in place of R for“response,”andS for “environmental stimulus condition” rather than “stimulus”).But what about P, H, T, and A? Tolman is sufficiently operational in saying thatgroups of these variables are all he finds in the way of mental processes (thusdisposing of the mentalistic nature of such processes), but he is still looking forsubstitutes and hence (and for no other reason) calls the additional variables“intervening.”Thepresent paperwas published in the Journal ofGeneral Psychology, (1931,5,

427–458)andisreprintedherebypermissionoftheeditor.Note: The introduction to “The Processes Involved in the Repeated Guessing ofAlternatives” was modified slightly by its author for reasons unrelated to thosegoverningchangestootherpapers.ThefollowingsentenceappearedinboththeFirstEditionandtheEnlargedEditionbutwasomittedfromtheversionused in theThirdEdition:“Thepresentpaperisanefforttoaccountfortheobservedpatternsofguessingintermsoftendenciestoalternatecalls.”

1Tolman,E.C.Operationalbehaviorismandcurrenttrendsinpsychology.InProc.25thAnniv.Celebr.Inaug.Grad.Stud.LosAngeles:Univ.South.Calif.Press,1936.ReprintedinMarx,MelvinH.Psychologicaltheory:ContemporaryReadings.NewYork:Macmillan,1951.

587

ACKNOWLEDGMENTS

pp.3–18:AppearedinAmericanScholar,Winter1955–1956.pp.19–24:ReprintedwithpermissionfromTransactionsoftheNewYorkAcademyofSciences,SeriesII,1955,17:7,547–551.pp.25–38:ReprintedwithpermissionfromScience,1956,124,1057–1066.Copyright©1956bytheAmericanAssociationfortheAdvancementofScience.pp.39–50:ReprintedbypermissionfromDaedalus,JournaloftheAmericanAcademyofArtsandSciences,fromtheissueentitled“EvolutionandMan’sProgress,”Summer1961, Vol. 90, No. 3. Copyright © 1961 by the American Academy of Arts andSciences.pp.51–57:FromProceedingsoftheAmericanPhilosophicalSociety,1964,108,482–485. Copyright © 1964 by the American Philosophical Society. Reprinted withpermissionofthepublisher.pp.58–65:FromInternationalEncyclopediaoftheSocialSciences,DavidL.Sills,ed.,NewYork:Crowell,CollierandMacmillan,1968.Vol.16,pp.271–275.Copyright©1968 by Crowell, Collier and Macmillan, Inc. Reprinted by permission of thepublisher.pp.69–100:FromPsychologicalReview,1950,57,193–216.Copyright©1950bytheAmericanPsychologicalAssociation.Reprintedwithpermission.pp.101–107:AppearedwithadaptationsinAmericanPsychologist,1953,8,69–79.pp.108–131:FromAmericanPsychologist,1956,11,221–223.Copyright©1956bytheAmericanPsychologicalAssociation.Reprintedwithpermission.pp.132–164:FromAmericanScientist,1957,45,343–371.Reprintedbypermissionofthepublisher.pp.165–175:FromAmericanPsychologist,1958,13,94–99.Copyright©1958bytheAmericanPsychologicalAssociation.Reprintedwithpermission.pp. 179–191: Appeared in Psychology and the Behavioral Sciences, University ofPittsburghPress,1955.pp.192–216:AppearedinScience,1958,128,969–977.pp.217–239:Reprintedbypermission fromHarvardEducationalReview,31:4 (Fall1961), pp. 377–398. Copyright © 1961 by the President and Fellows of HarvardCollege.Allrightsreserved.pp.240–253:FromTeachersCollegeRecord,November1963.Copyright©1963byColumbiaUniversityPress.Reprintedbypermissionofthepublisher.pp.254–270:ReprintedwithpermissionfromScience,1968,159,704–710.Copyright©1968byAmericanAssociationfortheAdvancementofScience.pp.271–281:ReprintedbypermissionfromEducation,December1969,90,93–100.pp.285–294: Reprinted with permission from Scientific Monthly, November 1964.Copyright©1964bytheAmericanAssociationfortheAdvancementofScience.pp.295–302:AppearedinIntegratingtheApproachestoMentalDiseaseeditedbyH.D.Kruse(NewYork:PaulB.Hoeber,Inc.,1957).

588

pp.303–321:FromTheory and Treatment of the Psychoses: Some New Aspects,F.Gildea, ed. Copyright © 1956 by Washington University Studies. Reprinted bypermissionofthepublisher.pp. 322–328: Appeared in Cumulative Record (3rd ed.). (New York: Appleton-Century-Crofts,1972,pp.276–282).pp. 329–337: “Compassion and Ethics in the Care of the Retardate” reprinted bypermissionoftheJosephP.Kennedy,Jr.,FoundationforMentalRetardation.pp.341–359:“ExperimentalPsychology”byB.F.SkinnerisfromCurrentTrendsinPsychology, byWayneDennis, et al.Copyright© 1947 byUniversity of PittsburghPress.ReprintedbypermissionoftheUniversityofPittsburghPress.pp.360–376:FromCurrentTrendsinPsychologicalTheory:ABicentennialProgramby Wayne Dennis, et al. Copyright © 1961 by University of Pittsburgh Press.ReprintedbypermissionoftheUniversityofPittsburghPress.pp. 379–390: From On the Future of Art by Solomon R. Guggenheim Museum.Copyright © 1970 by the Solomon R. Guggenheim Foundation. Reprinted with thepermissionofTheVikingPress,Inc.pp.391–401:LecturegivenatthePoetryCenterinNewYorkCity,October13,1971.p.393:SelectedlinesfromSoMuchNearerbyI.A.Richards.Copyright©1960byI.A.Richards.ReprintedbypermissionofHarcourtBraceJovanovich,Inc.pp.394–395:“VerbalBehaviour”and“ForIvorRichards”fromEncounter,November1962.pp.405–415:AppearedinAtlanticMonthly,January1934.pp.416–430:FromPsychologicalReview,1945,52,270–277.Copyright©1945bytheAmericanPsychologicalAssociation.Reprintedwithpermission.pp.431–436:FromPsychologicalRecord, 1929, 3, 186–192. Copyright© 1929 byPsychologicalRecord.Reprintedwithpermissionofthepublisher.pp.437–453:FromAmericanJournalofPsychology.Copyright©1941bytheBoardof Trustees of the University of Illinois. Used with permission of the University ofIllinoisPress.pp. 454–463: From The Journal of Experimental Psychology, 1942, 30, 495–503.Copyright © 1942 by the American Psychological Association. Reprinted withpermission.pp.467–474:AppearedinTheListener,September30,1971.pp.475–503:FromTheJournalofGeneralPsychology,1931,5,427–458.Copyright©1931.ReprintedwithpermissionoftheHelenDwightReidEducationalFoundation.Published by Heldref Publications, 1319 Eighteenth Street NW, Washington, DC20036-1802.pp.504–524:FromTheJournalofGeneralPsychology,1935,12,40–65.Copyright©1935.Reprintedwith permission of theHelenDwightReidEducational Foundation.Published by Heldref Publications, 1319 Eighteenth Street NW, Washington, DC20036-1802.pp.525–534:FromTheJournalofGeneralPsychology,1935,12,66–77.Copyright©1935.Reprintedwith permission of theHelenDwightReidEducational Foundation.Published by Heldref Publications, 1319 Eighteenth Street NW, Washington, DC20036-1802.pp.535–543:FromTheJournalofGeneralPsychology,1937,16,272–279.Copyright©1937.ReprintedwithpermissionoftheHelenDwightReidEducationalFoundation.Published by Heldref Publications, 1319 Eighteenth Street NW, Washington, DC

589

20036-1802.pp. 544–549: From The American Journal of Psychology, 1944, 57, 276–281.Copyright© 1944 by theBoard ofTrustees of theUniversity of Illinois.UsedwithpermissionoftheUniversityofIllinoisPress.pp.550–554:FromContemporaryPsychology,1956,1,101–103.Copyright©1956bytheAmericanPsychologicalAssociation.Reprintedwithpermission.pp.556–557:FromTheJournalofGeneralPsychology,1932,7,481–482.Copyright©1932.ReprintedwithpermissionoftheHelenDwightReidEducationalFoundation.Published by Heldref Publications, 1319 Eighteenth Street NW, Washington, DC20036-1802.pp. 558–569: From The Journal of Experimental Psychology, 1941, 29, 390–400.Copyright © 1941 by the American Psychological Association. Reprinted withpermission.pp. 570–574: From The Journal of Experimental Psychology, 1948, 38, 168–172.Copyright © 1948 by the American Psychological Association. Reprinted withpermission.pp. 575–578: From The American Journal of Psychology, 1957, 70, 308–311.Copyright© 1957 by theBoard ofTrustees of theUniversity of Illinois.UsedwithpermissionoftheUniversityofIllinoisPress.pp.579–583:Reprintedwithpermission fromJournal of Experimental Analysis andBehavior,1962,5, 531–533.Copyright©1962 by theSociety for theExperimentalAnalysisofBehavior,Inc.pp. 584–588: Appeared in Journal of Comparative and Physiological Psychology,1957,30:3,279–281.pp.589–600:ReprintedwithpermissionfromJournaloftheExperimentalAnalysisofBehavior,1958,1, 235–244.Copyright©1958 by theSociety for theExperimentalAnalysisofBehavior,Inc.pp.601–604:ReprintedwithpermissionfromScience,1959,129,197–198.Copyright©1959byAmericanAssociationfortheAdvancementofScience.pp.605–612:Reprintedwithpermission.Copyright©1951byScientificAmerican,Inc.pp.613–620:AppearedinLadies’HomeJournal,October1945.pp. 621–629: Reprinted by permission of the publisher from B. F. Skinner, “ThePsychologyofDesign”whichappearedinArtEducationToday,(NewYork:TeachersCollegePress[formerlyBureauofPublications],copyright©1941,TeachersCollege,ColumbiaUniversity.Allrightsreserved.).pp.630–647:FromAmericanPsychologist, January1960.Copyright©1960by theAmericanPsychologicalAssociation.Reprintedwithpermission.pp.648–652:AppearedinConditionalReflex,1966,1,74–78.pp. 653–658: From Harvard Alumni Bulletin, May 1962. Copyright © 1962 byHarvardAlumniBulletin(nowHarvardMagazine).Reprintedwithpermission.pp.661–673:FromAmericanPsychologist,1990,45,1206–1210.Copyright©1990bytheAmericanPsychologicalAssociation.Reprintedwithpermission.

590

INDEX

AAbility,346,354TheAbolitionofMan,52,56Acheson,Dean,16,17TheActofCreation,385Acton,Lord,6Admiration,27–28,30,55–57“AdvisoryCommitteeonGovernmentProgramsintheBehavioralSciences,”327After-discharge,539Aggression,55,313AgriculturalAdjustmentAct,31Aiken,Conrad,405Air-Crib,613,620Alexander,Franz,300,316Alliteration,431–436,437,440–441,444,449,453Alternatives,repeatedguessingof,454–463AmericanAcademyofArtsandSciences,39AmericanAssociationfortheAdvancementofScience,285TheAmericanJournalofPsychology,437,544,575AmericanPsychologist,74,82,101,108,122,165,248,250,630,662,663AmericanPsychologicalAssociation,25,74,80,108,110,119,123,165,181,604,630,659

AmericanScholar,3AmericanScientist,132,550Anderson,Sherwood,406Anesthesia,296,542Angell,James,602Anger,Douglas,140AnimalIntelligence,373Animals,howtoteach,605–612Animals,obesityin,149–151Anliker,JamesE.,125,128,149AnnualProceedingsoftheNewYorkAcademyofSciences,152Anrep,G.V.,534,648,651Anticipation,559,560,562AntiochReview,26Anxiety,127,324,558–569Anxiety,extinctionof,567–569Aquinas,Thomas,374Archimedes,416

591

Aristotle,374ArmedServices,265Art,64,379–390,621Art,historyof,383Art,reasonsfor,380ArtEducationToday,621Artists,achievementsof,380,384Assonance,431,437,446,453AStudyofaScience,108“AtalantainCalydon,”438AtlanticMonthly,396–407,642Atomicphysics,317,348,364AtomicPhysicsandHumanKnowledge,364Attention,157,236–237,258–259,348,406,611Aurelius,Marcus,4,5TheAutobiographyofAliceB.Toklas,405,408,410,411,414–415Automaticreading,407,409Automaticwriting,406–407Aversivecontrol,184–186,267,277,297Avoidance,147–149Ayllon,Teodoro,274Azrin,Nathan,207,274

B“BabyinaBox,”613–620Baby-tender,618–619Bacon,Francis,58,60,65,256Baglivi,Giorgio,482BarnesandAffiliatedHospitals,303BBC,393Beals,RalphL.,327Beecher,Henry,125Behavior,290,307,352,372,504Behavior,cooperative,583Behavior,defective,296Behavior,emotional,74,296Behavior,experimentalanalysisof,132–164,374Behavior,literary,431–436Behavior,maintenanceof,168–175,324Behavior,nonverbal,278Behavior,physiologicaltheoryof,349Behavior,probabilityof,41Behavior,psychotic,158–162,303–321Behavior,samplingof,277Behavior,sexual,382Behavior,technologyof,122,473Behavior,verbal,231–232,305,363,416,420,424–426,431Behavior,vocabularyof,544,661–673

592

Behavioralobjectives,275Behavioralsciences,effectivenessof,467–474Behaviorism,70,111,288,418,425,427,428,544,601–604,672Behaviorism,111,604Behaviorism,methodological,429Behaviorism,radical,429Behaviormodification,322–328,335,472TheBehaviorofOrganisms,70,119,123,135,160,323,431,475,508,510,525,526,529,535,559

BehaviorResearchLaboratories,158Behaviortheory,304Bell,Charles,489Benzedrine,124Berkeley,George,350BeyondFreedomandDignity,391,396,398,401Biderman,AlbertD.,327Binet,Alfred,322Biology,315,398,399,661–673Bishop,K.Clayton,69Black,JohnnyS.,369Blough,DonaldS.,122,127,145,146,182Bohr,Neils,363–364Boren,John,157Boring,EdwinGarrigues,416,428,429,430BostonPsychopathicHospital,318Boston,Massachusetts,375,648,662“Boxetpraetereanihil,”620Box,puzzle,72Box,silentrelease,111Box,sound-shielded,133Brady,Joseph,124,154,155,157,163BraveNewWorld,15,20,38Bray,Charles,633Breland,Keller,69,122,165,159,324,633,634Breland,Marion,69,122,635Bribery,278Bridgman,PercyWilliams,417,426,427,475,477Britishempiricism,301BrothersKaramazov,38Brownlaboratories,118Bruner,Jerome,244BulletinoftheAmericanAssociationofUniversityProfessors,30Bush,RobertR.,550–554Buswell,Guy,621Butler,Samuel,396

CCambridgeUniversity,393,405

593

Cannon,WalterB.,119,649–650Carbondioxidepressure,635Cardmatchingexperiment,609Cardozo,Justice32Carlson,Bengt,548,549Carmichael,Leonard,633Carnegie,Dale,21Casehistory,367Cats,126,157,373Causalrelations,296,307,314–316,399Centralnervoussystem,522Centripetalmovement,309Cerebellum,496Chance,386,454,455Chernikoff,R.,643Cherwell,Lord,641Chesterfield,Lord,21Chimpanzees,137,142,323China,474Choice,28,92,99,352Chomsky,Noam,391–393Christianity,5,6Chronicallyill,careof,332Churchill,Winston,641Cincinnati,Ohio,630ClarkUniversity,111Class,506,507,520,522,523Classicalconditioning,542,559Classroomexperience,256–257Classroommanagement,265–269,280,472Clinicalpsychology,327“Coefficientofalliteration,”451Cohen,Harold,471Cohen,S.,250Coleman,Claude,30Coleridge,SamuelT.,393CollegeofCardinals,648Coloreffect,paradoxical,557ColumbiaJester,129ColumbiaUniversity,129,397Comenius,JohnAmos,250CommitteeonPublicHealthoftheNewYorkAcademyofMedicine,295ComparativePsychologyMonographs,323Compassion,329–337Competition,180–181TheCompleteWorksofSt.ThomasMore,58Complexlearning,92–99Computer-aidedinstruction,267

594

Conant,JamesBryant,243,405“TheConceptoftheReflexintheDescriptionofBehavior,”426ConceptualNervousSystem,475ConditionalReflex,648ConditionedReflexes,534–535Conditionedreflexes,117,507,512,525–534,651Conditionedreinforcers,219Conditionedsuppression,154Conditioning,53,79,82,297,532,536,540,545,548,553,571,580Conditioning,cultural,15Conditioning,emotional,20Conditioning,operant—seeOperantconditioning“ConditioningandtheVoluntaryControlofthePupillaryReflex,”543TheConductionoftheNervousImpulse,492Conduction,492–493Conklin,J.E.,643Conrad,D.G.,157Consciouswishes,312Consciousness,293,425,661–673Consequences,329,330,470Consequences,long-term,46Consequences,natural,267Consonants,432,435–436,441–442,444,447,449,450ConstitutionalLaw:CasesandOtherProblems,31Constructs,418Consumer,roleofthe,386–388ContemporaryPsychology,550Contingencies,contrived,278Contingencies,managing,248–249Contingencies,mechanical,324Contingencies,social,277Contingenciesofreinforcement,41–43,63,117,179,195,273,274,278,325,365,420,426,538,575–576,652

ContingenciesofReinforcement:ATheoreticalAnalysis,325Contingenciesofsurvival,383,665Contingencymanagement,271–281,336Continuousperformance,590Control,3–38,45,345,346Control,aversive,31,46,261–262Control,counter-control,20,21–22,32,33,45Control,environmental,558Control,experimental,125,558Control,personal,27–29Control,sensory,575Controlofhumanbehavior,19–38Cook,StuartW.,431Cooper,Devra,208Cooperation,teachingpigeons,581–583

595

Copernicantheory,351Copying,385Correctiveinstitutions,345Correlation,344,540,549,559CosmopolitanasExplanation,415Countercontrol,332–334Cowles,J.T.,323Creativeartists,creating,379–390Criminology,31Crozier,WilliamJohn,111Crusoe,Robinson,405,429Culturalpractices,48Culture,389–390,401,665Cumulativerecord,137CurrentTrendsinPsychology,341Curtis,CharlesP.,405

DDaedalus,39Darkadaptation,127,146Darwin,Charles,385,399,601,666,667Data,consistencyof,516,518“DaughtersofRevolution,”625–626Daumier,Honoré,625deCaus,Salomon,479DeadSeaScrolls,58Defoe,Daniel,405Degas,625DelphicOracle,671Democracy,3,6,8–9,15,16–18,23,26DepartmentofPharmacology,124,152Deprivation,140,149,297,371,570,581,611Descartes,René,347,392,399,478–480,488,502,666Dewey,John,119,211,272,602Dews,Peter,124,152Dial,475Diana,479Dickens,Charles,332Differentialreinforcement,140,424–425Digestiveglands,532DiesensorischeFunctionendesRückenmarksderWirbelthierenebsteinenneuenLehreüberdieLeitungsgesetzederReflexionen,486

Discovery,259–261“Discoverymethod,”384Discrimination,86,92,96,220–222,310,360,529–531,533,538,576,608Discriminationapparatus,127,369Discriminationtraining,127Disinhibition,79

596

Disorders,psychopathic,437DistributionofResponseProbabilities,552Dogs,121,126,541,606Dog,training,606Dostoevsky,Fyodor,5,38“Drive,”528,536,548,354“DriveandReflexStrength:I,”528DRL(differentialreinforcementoflowrate),589,590–600DRLperformancecurves,591–593Drugs,21,55,151–157,531,635DukeUniversity,143Dunlap,Knight,602

EEasternPsychologicalAssociation,108Ebbinghaus,Hermann,369Economics,25,47,305,326,361,467Education,9,10,12–15,23,27,29–31,47,164,169–282,250,271,281,305,328,343,345,346,361,363

Education,271EducationandFreedom,244TheEducationofAmericanTeachers,243Educationalgoals,246–249Educationalinstitutions,47Educationalpolicy,262Educationalpractices,356Educationalpsychologists,363EffectiveReactionPotential,547Efferentnerve,536Elephants,rogue,373Elicitation,effectof,515Eliot,George,398Eliot,ThomasStearns,14AnElucidation,411–413ÉmileoudeL’Education,53–54,259,272Emotion,20,310,315,319,371,423,548,536,559–560,627–628Encounter,395Energy,204,506England,280EnglishChannel,630Environment,10,42,310,347,352Environment,changing,519Environmentalforces,308Environmentalhistory,326Environmentalvariables,289,299Environmentalist,53Erasmus,Desiderius,218Eroféeva,540

597

Escape,332Esquire,15“Anessayonthevitalandotherinvoluntarymotionsofanimals,”482Estes,WilliamK.,69,120,122,553,558,635Ethics,329–337Ethologists,665,668Evolution,4,40,42–43,330,348,661–673“EvolutionaryTheoryandHumanProgress,”39excitabilité,483Experimentalanalysis,326,327Experimentalcontrol,344Experimentaldesign,366“Experimentalinhibition,”78Experimentalmethod,343Experimentalpsychologist,360,362Experimentalpsychology,296–297,298,341–359,361,363,364,372Experimenter,behaviorof,551Extinction,74,78,79–80,84–85,90–91,297,527,531,532,538,569,572Extinctioncurve,81–84,87–88,120,139,564,634Extinctionofanxiety,effectsof,564–568

FFacilitation,527Faraday,Michael,109,259,359Fatigue,flexionreflex,509Fatigue,measuringof,514Fechner,Gustav,322Fearing,Franklin,481FederalAidStudents,432Feelings,329,334Feigl,Herbert,426Ferster,CharlesB.,117,123,125,139,141,142,152,158,169,181,207,209Filatov(Russiancircustrainer),652Finn,J.D.,250Fisher,R.A.,120,130Fixed-intervalfixed-ratioschedule,multiple,152,181Fixed-intervalperformance,591Fixed-intervalscheduleofreinforcement,136,141,153,181,572,584–589,600Fixed-intervalperformance,162,591Fixed-ratioscheduleofreinforcement,137–138,590,600Florez,Luisde,633Flourens,M.,483Ford,Henry,395Formalpatterning,459Formalstrengthening,437,443,444,449,452–453Foster,George,478,480Foster,SirMichael,478FoundationforMentalRetardation,329

598

FragmentsofanAnalysiswithFreud,173France,479Frank,PhilippG.,285Frazier,Edward,32,130“Freeschool,”271Freedom,3,10,22,23,29,53–54,56FrenchRevolution,272Frenkel-Brunswik,E.P.,286Frequencies,expected,450Frequency,134,194Frequencyofresponse,304Freud,Sigmund,130,173,285–294,301,312,322,371,651Freudian,413Freudiandynamism,170,320,362Freudianpsychology,51,361Freudianwish,101,304Freund,PaulA.,31Froebel,Freidrich,272Fulton,J.F.,481Functionalanalysis,344,347,545,546FundfortheAdvancementofEducation,192,207FurmanUniversity,602

GGalatea,369Galen,488Galileo,353,416Gambling,172–173,279,298,382,388GeneralElectric,359GeneralMills,69,165,634,636,638,640Generalization,331“Generalizedreinforcer,”277“TheGenericNatureoftheConceptsofStimulusandResponse,”504–524Geneticendowment,331Geneticfactors,299–300Genetichistory,326Genetics,40–41,343,348,351,355Genius,295Gestalttheories,350Gilbert,ThomasF.,208–209Glandularactivities,527Glisson,Francis,481–483Goodfellow,LouisD.,454–456,463Government,23,25,26,27,31–34,46,47,60,305,326,345,357,361Grades,267,272“GreatConversation”ofPlato’sAcademy,668Gray,JohnM.,613Gregg,Alan,620

599

Greenville,SouthCarolina,602GreenwichVillage,649Groton,12Groupliving,330Guessing,454–463Guttman,Norman,69,143,144,145,165,170,324,634

HHabits,308,310,311Hacker,Andrew,26,37–38Hales,Stephen,483Hall,Marshall,480,481,483–489,495,496,503Hallucinations,296HamiltonCollege,648Handler,Philip,467Harris,Roy,387HartfordInstitute,125HarvardAlumniBulletin,653HarvardEducationalReview,179,208,217,241HarvardGraduateSchoolofEducation,208HarvardLaboratories,140HarvardMedicalSchool,124,152,649HarvardPsychologicalLaboratories,138,182HarvardPsychologicalReview,406HarvardSchoolofPublicHealth,125HarvardSquare,649HarvardUniversity,12,111,122,208,395,405,427,579,648,653Hayden,Hiram,3Helwig,J.,243Henderson,LawrenceJ.,405Hercules,15Heredity,298,307–310,603Heron,W.T.,120,558Herrnstein,RichardJ.,124,138,173,589Hilton,P.J.,258Hitler,Adolf,16Hoeber,PaulB.,295Holland,James,157,158,207–208Holmes,Justice,16Homme,LloydE.,207Honors,272,278HowtoSolveIt,260Hudgins,C.V.,543Hull,ClarkL.,544–549,620,670Humanbehavior,controlof,19–38“TheHumanSituationToday,”3Humansubjects,158,274Hume,David,350

600

Hunt,Howard,124,154Hunter,Walter,111,418Hutchins,RobertK.,246Huxley,Aldous,15,20,38Huxley,Thomas.H.,14Hyde,ArthurD.,634,641Hypothesis,scientific,550

IIndependentvariables,307,308,312,345Individuality,56–57,354,380Induction,510,514–516,600“Industrialrevolutionineducation,”194Infantbehavior,603,610Innerevents,315–316Innerman,370–372AnInquiryintoMeaningandTruth,417Instincts,308,310,312,315InstituteforBehavioralResearch,471InstitutefortheUnityofScience,285Instruction,language,272Instructionalprograms,248IntegratingtheApproachestoMentalDisease,295TheIntegrativeActionoftheNervousSystem,490Intelligencetesting,297,322Intelligibility,413InternationalCongressofPhysiology,648TheInternationalEncyclopediaoftheSocialSciences,58Intervalperformance,139,152–153,571–572Introspectivepsychology,301–302,316,318,667Involuntaryaction,484Irritability,482,488,502Irritability:APhysiologicalAnalysisoftheGeneralEffectofStimuliinLivingSubstances,481

JJakobson,Roman,392James,William,243,364,375,663,670Jennings,HerbertS.,602Jewitt,J.E.,208JohnsHopkinsUniversity,602JohnLockeLectures,393Johnson,HarryM.,426Jones,HowardMumford,81Jones,Janette,432Jones,Lawrence,392JosephP.Kennedy,Jr.,FoundationforMentalRetardation,329JournalofAppliedPsychology,149

601

JournalofComparativeandPhysiologicalPsychology,140,145,154,584,589JournaloftheExperimentalAnalysisofBehavior,392,535,579,589TheJournalofExperimentalPsychology,81,454,455,559,570TheJournalofGeneralPsychology,146,477,504,525,528,535,543,557JournalofPsychology,431Juvenileoffenders,471

KKateb,George,396,401Keller,FredS.,111,120,124,140Kennedy,JosephP.,Jr.,329“Knees,”595,596,598Koch,Robert,359Koch,Sigmund,108Knoedler,M.,&Co.,625Knowledge,300–301,316–317Koestler,Arthur,385Konorski,Jerzy,535–537,540,542Körperstellung,485Kruse,HeinrichD.,295Kruse,Marian,432Krutch,JosephWood,11,13–14,22–23,32,54,56Kubie,Lawrence,316Kymograph,113,114

LLascienceetl’hypothèse,499Laboratoryresearch,325Laboratory,flightfromthe,360–376Ladies’HomeJournal,613Language,391,417,419Language,391Lanier,LyleH.,426Lapicque,Louis,492Lashley,Karl,369,418,602,603,670Latency,73,74,76–77,539,547,548,553Latentspeech,431Latinsquare,120TheLawofEffect,78,179,399,525,670Leader-followerrelation,581,583Learning,69–100,126,275,297,298–299,310,324,353,363,551,552Learning,complex,92Learning,mathematicaltheoriesof,551Learning,theoriesof,69–100,369,370Learningcurves,369LecturesontheHistoryofPhysiology,478Legalsystem,357Lehmann-Haupt,Christopher,391

602

Leisure,63–64,335,336,379,381–382Lesraisonsdesforcesmouvantesavecdiversemachines…degrottesetdefontaines,479

“Lettinggo,”323Leverpress,508–510Levi,Wendell,141Levi-Strauss,Claude,392Lewis,CliveS.,52–53,56LifeandWorkofSigmundFreud,285Lincoln,Abraham,12Lindsley,OgdenR.,125,158,160,182,266,TheListener,467Livingorganism,modelof,307Locke,John,301,392Loeb,Jacques,111,602,648TheLogicofModernPhysics,426,475,477Logic,426Logicalempiricism,545Lohr,Thomas,125Lowell,A.Lawrence,405,656,657,658Lowes,JohnLivingston,405Lucas,Keith,492LucyChurchAmiably,410

MMacaulay,26MacCorquodale,Kenneth,392Mach,Ernst,426,477,494,499,502Machiavelli,26MacLeish,Archibald,397MacmillanCompany,627Magazinetraining,158,159,166,167,590Magnus,Rudolph,111,132,485,487,489,496Maintainedresponding,589TheMakingofAmericans,406ManualofDiscipline,58Marbles,323Marijuana,382Marks,267,272,278Marshall,Clare,124MarshallPlan,6Marx,MelvinH.,476MassachusettsGeneralHospital,125MassachusettsInstituteofTechnology,397,640MassachusettsStateHospital,125Matching,97–99Mathematico-deductiveTheory,549Mathematics,368–370Matisse,Henri,413

603

Maxwell,JamesClerk,359Mayer,Jean,125,149Mead,GeorgeHerbert,601,602TheMeaningofMeaning,393TheMeasureofMan,11,22,32,54“MechanizationoftheLatinSquare,”558Medicine,328MemoirsontheNervousSystem.MemoirII.OntheTrueSpinalMarrowandtheExito-MotorySystemoftheNerves,483

Memories,308,310,315Memorydrum,194Mendel,Gregor,359Menger,K.,211Mentaldisease,285–337Mentalprocesses,92,311,313,318,371,601,661–673Mentalism,51–52,427Metaphor,287,422MetropolitanStateHospital,158,161,318Meyer,Adolph,602Meyer,Susan,199,207Mice,126,128–129,133,149–150MidwesternPsychologicalAssociation,69,82Mill,JohnStuart,119Miller,S.,535–537,540,542Minneapolis,Minnesota,165,634Missile,pigeonguided,630–647Mistreatment,331–333Modeling,665Models,mathematical,368–370ModernLanguageNotes,453MonheganIsland,Maine,405Monkeys,157Montaigne,Michel,60Montessori,Maria,272Moraldiscourse,10,13–14,23Morality,14More,SirThomas,58Morgan,C.Lloyd,601Morgan,ThomasHunt,359Morgantown,WestVirginia,471Morrill,Albro,648Morse,WilliamH.,124,138,148,173,575,584,589MoscowCircus,652Mosteller,Frederick,550–554Motivation,80,125,149–151,275–276,297,319,423,545,548Motorbehavior,296motorischeKraft,483Movabletype,474

604

Movement,478,481–482,494,503Mowrer,O.Hobart,81,560Müller,Johannes,483Multipleschedule,142,589,591Münsterberg,Hugo,406Murdock,KennethB.,405Muscle,contractionsof,481–482,484Muscularaction,modesof,484MuscularContractionandtheReflexControlofMovement,481Music,64,388Mutations,385–386MyFather’sHouse:AnOneidaBoyhood,58

NNationalAcademyofSciences,327,467NationalInstitutesofHealth,157NationalResearchCouncil,327NationalScienceFoundation,108,136NationalTrainingSchool,WashingtonD.C.,471Naturalsciences,361Naturalselection,4,22,661–673NavalDataHandlingSystem,643NavalRes.Lab.lett.Rep.,643NavalResearchLaboratories,157,630,642,643Negley,G.,32Neill,272Neptune,479Nero,305Nervoussystem,355Neurologist,362Neurology,156,350,546Neurophysiology,70Neuroticbehavior,296,300NewAtlantis,58,65NewJersey,636,641Newlin,E.P.,643NewRepublic,405NewYorkAcademyofSciences,19TheNewYorkTimes,391,649Newsweek,392,393Newton,Issac,308,348“NightSnake,”627,6281984,37Nonverbalbehavior,278“NormalMotorAutomatism,”406Novelty,399Noyes,Pierpont,58Nursinghomes,332

605

OObservabledata,320–321Observation,70,495OfficeofNavalResearch,136OfficeofScientificResearchandDevelopment,69,633,635,636,639,640Olds,James,156–157“OnaParticularFormofConditionedReflex,”535OnEducation,246OntheFutureofArt,379Operant,538,539,540Operantbehavior,73,76–77,273,274,538,539,575,589,652Operantconditioning,135–143,158,274,322,323,548,570,661–673Operantreinforcement,275,276Operantresearch,325“TheOperationalAnalysisofPsychologicalTerms,”416–430Operationalism,315,416–417,426,428–429OrangePark,Florida,142ORCON,642–646OrganicControl,643–646TheOriginofSpecies,385Orwell,George,15,37“Overlearning,”324Oxford,393

PPacificSpectator,10Panza,Sancho,624,625Passionsdel’âme,478Pascal,Blaise,259Pasteur,Louis,359Patrick,J.M.,32ThePatternofResponsibility,16Pauling,Linus,550Pavlov,IvanP.,72,111,117,133,362,486,487,496,507,525,532–534,539,545,603,648–652

Pavlovianconditioning,133,299,602Payson,Mrs.CharlesS.,625Pedagogy,243–246Pelican—seeProjectPigeonPeriodicreconditioning,560Periodicreinforcement,567Perrin,D.G.,250Perry,RalphBarton,10Personalities,309,410Personnelpsychologists,346PerspectivesU.S.A.,3Pestalozzi,J.H.,272Pflüger,E.F.W.,486

606

Phaedrus,30,56Pharmacologicalfactors,300PhiPhenomenon,621,623–626Philistinism,389–390Philosophy,317,426PhilosophyofScienceAssociation,285Phil.Trans.Roy.Soc.,483Physics,301,343,347,427Physiologicalpsychology,69,322ThePhysiologyoftheBrainandComparativePsychology,111,648–649Physiology,51,300,343,355,362,372,543,668Picasso,Pablo,385,413,621Pick-OffDisplayConverter,643Pictures,381–388ThePigeon,141“PigeonsinaPelican,”630–647Pigeons,74,78–91,92–100,125,127,133–138,141,145,146,152,165–169,274,324,570–578,590,608–610,630–647

Pillmachine,117Ping-pong,teachingpigeonstoplay,181,579–581Plato,4,7,30,53,56,260,668Poetry,391–401,431,432,437–453Poincaré,Henri,477,499,502PointFourProgram,6Points,277,518Politicalscience,356,467PoliticsofSocialResearch,327Polya,G.,260PostulateofAfferentNeuralInteraction,547PostulateofBehavioralOscillation,547Postulates,70,546,547Practice,massedandspaced,194Predictions,301–302Preference,95–96Pressey,SidneyJ.,192,233,242,646Pressey’sself-testingmachine,193Pribram,Karl,125PrinciplesofBehavior,544–549PrinciplesofPsychology,670Prisons,332,334Privateevents,317–318Prizes,267,272,278Probability,101–102,432–433,548,551,552Probabilityestimates,553Probabilityofaction,305Probabilityofbehavior,41Probabilityofreinforcement,93Probabilityofresponse,76–78,134

607

Proc.25thAnniv.Celebr.Inaug.Grad.Stud.,476TheProcessofEducation,244Programing,241–243,250Programmedinstruction,179–253,263–265,268,274–279ProjectA,110ProjectPigeon,630–647“Prosthetic”environment,274Pseudo-conditionedreflex,529PsychiatricResearchReports,125,158Psychiatry,125,311,318,328,349Psychoanalysis,293,361,362Psychoanalyticconcepts,285–294Psychoanalytictheory,322Psychodynamics,298PsychologicalBulletin,80,417PsychologicalCareoftheInfantandChild,603PsychologicalLaboratoriesatHarvard,124PsychologicalMonographs,560PsychologicalRecord,120,431,558PsychologicalReports,143PsychologicalReview,69,405,411,416,426Psychologicalterms,416–430,661–673PsychologicalTheory:ContemporaryReadings,476Psychologists,316,342,357,364,366–367,369,426Psychology,51,60,295–300,355,357–358,364,427,428,467,602Psychology,experimental—seeExperimentalpsychologyPsychologyandtheBehavioralSciences,179Psychology:AStudyofaScience,Vol.II,108PsychologyDepartment,UniversityofMinnesota,633PsychologyfromtheStandpointofaBehaviorist,602,603Psychologyofdesign,621–629Psychotherapy,274,305,326Psychoticbehavior,300,303–321,462Psychoticsubjects,318–319Psychotics,296,297,333,334,461Publicaffairs,343,345Punishment,9,14,26,45,55,61,135,273,297,326,608Punishment,conditioned,608Public-privatedistinction,428–430Pupillaryreflex,543Pythagoras,249,260Pythagoras’GoldenTheorem,249

QTheQuestforUtopia,32Quixote,Don,624,625QumranCommunity,58

608

RRadiobroadcast,454Rats,73,111–126,147–148,153,156,273,552,560,563–569,585Rateofoccurrence,539Rateofreinforcement,141,571,576Rateofresponding,74–79,91,95,135,553,562–563,565Ratioschedules,139Ratliff,Floyd,122,182Reactioninhibition,78,81–82Reactiontime,342,607Reese,Thomas,648Reflex,132–133,475–503,517–519,525–534,539Reflex,conditioned,13,525–543Reflex,definitionof,517Reflex,describingbehaviorof,475–503Reflex,elicitationof,504,508Reflex,experimentalstudyof,499–503Reflex,flexion,497,506,507,512,527Reflex,historyof,493,497Reflex,physiologyof,494,495Reflex,pseudo,525,531,539ReflexAction:AStudyintheHistoryofPhysiologicalPsychology,481Reflexarc,479,489,490,498Reflexconduction,characteristicsof,490,491Reflexfatigue,500,501Reflexstrength,501–503Reinforcement,21,27,28,35–36,55,61–64,74,78,79,88–91,135–143,158,165,278,279,299,320,329,366–368,420,423,424,528,531,536,548,570,606

Reinforcement,aperiodic,88Reinforcement,automatic,183,184,187Reinforcement,contingenciesof—seeContingenciesofreinforcementReinforcement,continuous,139Reinforcement,definitionof,280Reinforcement,differential,140Reinforcement,DRLscheduleof,594Reinforcement,effectsof,326Reinforcement,fixed-ratioscheduleof,96,119Reinforcement,frequencyof,576Reinforcement,intermittent,135–136,138,162,279Reinforcement,negative,61–62Reinforcement,periodic,80,82,85–86,93–95,118–119Reinforcement,positive,62–64Reinforcement,programof,88Reinforcement,withholdingof,564Reinforcer,conditioned,267,606,607Reinforcers,335,381Reinforcers,contrived,267Reinforcers,mediating,279

609

Reinforcers,synthetic,382Reinforcers,weak,388Reinforcingsystem,mechanical,580Relationships,functional,347Religion,26,47,305,361,467Religiousinstitutions,47“RemarksonSomeAspectsofReflexInhibition,”492RenardHospital,303,620Repetition,435–436,438,440,459–460Research,71,251–253ResearchEngineersSocietyofAmerica,132Resistancetoextinction,547Respondentbehavior,537,538,540,652Response,352,418–419,421,504–534,536,540,541,546,548Response,covert,421Response,difficultyof,80–81Response,discriminationof,538Response,magnitudeof,539Response,overt,421,425Response,probabilityof—seeProbabilityofresponseResponse,rateof—seeRateofrespondingResponse,spontaneous,541Response,stepping,572Response,strengthened,574Response,verbal,426Response,walking,573Responsestrength,560Responsibility,personal,28,54Retardates,careof,329–337Retardates,institutionalized,333Rewards,62–64,365Rhyme,220,432,437,453Richards,IvorA.,393,394–395Richter,CurtP.,110,602Rickover,HymanGeorge,243–244,266Riksdaghuset,101RobertF.KennedyCenter,471Roberts,Justice32Robinson,EdwardG.,625Rogers,Carl,25,34–38,316Romanes,G.J.,373Roosevelt,FranklinDelano,12Ross,Denman,383Rousseau,JeanJacques,53–54,259,272,330Rückenmarkseele,486Runningexperiment,587Russell,Bertrand,110,417,475Russia,121

610

SSalivation,526,527,531,542Sapon,StanleyM.,207,209Satiation,297,596SaturdayReview,56SchoolandSociety,242Schedule,geometric,89–90SchedulesofReinforcement,101,136,139Schedulesofreinforcement,87,91,136,139,141,160–162,182,324,634–635Schedulesofreinforcement—alsoseenamesofspecificschedulesSchizophrenia,461Schlosberg,Harold,118SchoolandSociety,193Schools,271–281,345Schools,European,266Schweitzer,Albert,368ScienceandHumanBehavior,35,170,208,292,362,416TheScienceofMechanics,426,477Science,3,5,8,129,132,259,322,365Science,25,110,147,151,154,157,192,247,254,258,327,601ScientificAmerican,156,605Scientificmethod,108–131ScientificMonthly,285,286Scientifictheory,351,357Seals,useofinwar,642Searle,L.V.,643SecretaryofAgriculture,31Sedgwick,Ellery,405–406Self-actualization,36Self-control,35,54–55,292Self-instructionroom,197,207Self-management,46Self-observation,292,374Self-realization,358Sensorypsychology,296,342,343Sex,298Shakespeare’salliterativespan,444,445Shakespeare’sVerbalArtin“Th’ExpenceofSpirit,”392Shakespeare,William,68,392,399,431–433,435,436,440,441,446,452,453,660Shakespeariansonnets,431–436,438Shaltung,486Shaping,195,322,324TheShapingofaBehaviorist,621Shaw,GeorgeBernard,267,649Sheehan,J.G.,171Sherrington,SirCharlesScott,111,112,475,484,490,491,492,496,651Sherrington’ssynapse,475

611

Shock,558–569Sidman,Murray,35,124,147,148,151,157Sidmantechnique,147–148SigmaXiNationalLecture,132SilverSpring,Maryland,471Skinner,B.F.,241,247,248,251,323,324,325,391,528,621,663SkinnerBox,620SnakesoftheWorld,627Socialcontingencies,330Socialinteraction,45Socialpsychology,360Socialreinforcers,330Socialscience,51SocietyofFellows,405Sociology,356,467,602Socrates,249,257,260Solomon,Harry,318Solomons,LeonM.,406“SomeContributionsofanExperimentalAnalysisofBehaviortoPsychologyasaWhole,”101

SomeApostlesofPhysiology,484Sonnets,431–436,444SpecialDevicesSectionoftheNavy,633Spender,Stephen,395“Spontaneousrecovery,”85Squirrels,653–658St.Augustine,58St.Benedict,58St.John,554St.Louis,303,620Stafford,B.H.,643Statistics,127,365–366,553Statisticians,109,366Stein,Gertrude,405–415,431Stensen,Nicolas,478Stevens,S.Smith,417,418,428Stimuli,correlationsof,501,514Stimuli,private,420–425Stimuli,public,420–425Stimulus,181,308,419–420,488,490,494,495,499,501,503–543Stimulus,discriminationsof,538–539Stimulusclass,514Stimuluscontrol,143–147,297Stimulusgeneralizationgradient,144Stimulusinduction,422Stirling,W.,484StochasticModelsforLearning,550–554Stockholm,Sweden,101

612

Stoll,ElmerEdgar,453TheStudyofInstinct,133Successiveapproximation,324Summerhill,272Sumter,S.C.,141Superstition,570,573,575–578Superstitiousbehavior,595“Surreflex,”517Survival,contingenciesof,331Sustainedperformanceinlongexperiments,589–600Swammerdam,Jan,481,482Swinburne,AlgernonCharles,438,439,443,446,448,451–453Synapse,491,492“Syntheticsocialrelations,”579–583

TTmaze,551Tahiti,59TalkstoTeachers,243Tate,JohnT.,633Taylor,FranklinV.,642–643Teachers,training,279–280TeachersCollegeRecord,25,240Teaching,256,262–263,276,279,328,605Teaching,improvementof,269–270Teaching,science,254–270Teachinganimals,605–612Teachingmachines,192–253,647TeachingMachinesandProgrammedLearning:ASurveyoftheIndustry,250Technology,12,48–49,56,55Technologyofbehavior,47TheTechnologyofTeaching,225,251,256,324Teitelbaum,Philip,151Telepathy,experimentin,454,455TenderButtons,408–409,411–413Terms,292,416–430,520Terrace,HerbertS.,233Theories,69–100,348,351,353,358TheoryandTreatmentofPsychosis:SomeNewerAspects,303TheoryofPureDesign,383Thesis,doctoral,111ThirteenthInternationalCongressofPsychology,101Thorndike,EdwardL.,250,251,369,399,525,542,601,670Thorndikebox,369Thorndlike’sLawofEffect,78,179,399,525,670ThreeLives,406,409,410Tinbergen,Niko,133Tipton,C.L.,643

613

Tokeneconomies,336TheTokenEconomy,274Toklas,AliceB.,405Tolman,EdwardC.,418,476,546,548,670Tolman,R.C.,633Traitédel’homme,478,479Traits,346,354TransactionsoftheNewYorkAcademyofSciences,19Treatment,inhumane,331–333Trial-and-errorlearning,601TypeIconditioning,525–534TypeIconditioning—alsoseeReflexes,Respondent,TypeRTypeIIconditioning,526–534TypeIIconditioning—alsoseeReinforcement,Operant,TypeSTypeRconditioning,536–543TypeRconditioning—alsoseeTypeII,Operant,ReinforcementTypeSconditioning,536–543TypeSconditioning—alsoseeTypeI,Reflexes,Respondent

UUnconsciouswishes,312UniversityofChicago,602UniversityofDijon,405UniversityofIndiana,69UniversityofGeorgia,208UniversityofKentucky,208UniversityofMinnesota,120,432,631,633UniversityofPittsburgh,179,341UniversityofPittsburghPress,179,341,360Utopia,58Utopia,59

VTheValidationofScientificTheories,285Variability,sourcesof,125Variable-intervalscheduleofreinforcement,137–139,141,576Variable-ratioscheduleofreinforcement,138,268,279Verbalbehavior,231–232,305,363,417,420,424–426,431VerbalBehavior,391,392,393,397Verworn,Max,481,482Victor,633vismotoria,483visnervosa,483Viteles,Morris,32Vocabulary,popular,520–521,523–524Volition,484,485,503vonHaller,Albrecht,482,483vonUexküll,BaronJakob,486

614

WWald,George,146WaldenTwo,13,22,23,32,33,37,38,130,620,647Wallace,AlfredR.,666Walpole,Horace,119WalterReedArmyInstituteofResearch,157Waltham,Massachusetts,158,161,303WashingtonUniversitySchoolofMedicine,303Watson,JohnBroadus,111,418,601–604,670,672Webster’sInternationalDictionary,447Weisner,Jerome,397Weiss,Paul,418Welfarestate,382Wells,HerbertGeorge,649Westinghouse,359White,C.F.,643Whitehead,AlfredNorth,405White,MaryLouise,405,415Whytt,Robert,482,483,488,489,503Williams,D.R.,151Wilson,M.P.,140Wood,Grant,625–626WorldWarI,630WorldWarII,360,386,630Wortis,J.,173Wundt,Wilhelm,322,350Wyckoff,Jr.,L.B.,182

YYacorzynski,G.K.,460,461,463Yerkes,RobertM.,369,602YerkesLaboratoriesforPrimateBiology,142YoungCommittee,327Youngpeople,468–472

ZZenithexperiments,455,460,463ZenithFoundation,454Zenith-Goodfellowdata,457,459,461ZenithRadioCorporation,454ZorbatheGreek,648

615

THEFOUREDITIONSOFCUMULATIVERECORD:INDEXTOCONTENTS

ThefollowingsummarizesthecontentsofCumulativeRecordby listing the inclusivepage numbers of each article in each of its editions (First, Revised, Third, and thisDefinitiveEdition,respectivelyabbreviatedinbold-faceas1,2,3and4).Wherepagesareidenticalintwoormoreeditions,theeditionsarelistedtogether.Forexample,3/451–57 indicates that the article appears on pages51 through 57 in both the ThirdEditionandthisDefinitiveEdition.ArticlesnewtotheRevisedEditionwereinsertedinto the First Edition pagination; for example, pages 36.01 through .12 are twelvepages thatappearbetweenpages36and37 in theRevisedEdition.Sectionheadingsareunnumberedbecausethenumberofsectionsvariesacrosseditions.

TheImplicationsofaScienceofBehaviorforHumanAffairs

1/2/3/43–18 FreedomandtheControlofMen1/218–23,3/419–24 TheControlofHumanBehavior

(Abstract)1/223–36,3/425–38 SomeIssuesConcerningtheControlof

HumanBehavior236.01–.12,3/439–50 TheDesignofCultures3/451–57 “Man”3/458–65 TheDesignofExperimental

Communities

AMethodfortheExperimentalAnalysisofBehavior1/239–69,3/469–100 AreTheoriesofLearningNecessary?1/270–76,4101–107 TheAnalysisofBehavior1/276–100,3101–124,4108–131 ACaseHistoryinScientificMethod1/2100–131,3125–157,4132–164 TheExperimentalAnalysisofBehavior1/2131–141,3158–1684165–175 ReinforcementToday

TheTechnologyofEducation1/2145–157,4179–191 TheScienceofLearningandtheArtof

Teaching

1/2158–182,4192–216 TeachingMachines

616

2182.01–.22,3171–193,4217–239 WhyWeNeedTeachingMachines3194–207,4240–253 ReflectionsonaDecadeofTeaching

Machines3208–224,4254–270 TeachingScienceinHighSchool—

WhatIsWrong?3225–235,4271–281 ContingencyManagementinthe

Classroom

TheAnalysisandManagementofNeurotic,Psychotic,andRetardedBehavior1/2185–194,3239–248,4285–294 ACritiqueofPsychoanalyticConcepts

andTheories1/2194–202,3249–256,4295–302 PsychologyintheUnderstandingof

MentalDisease1/2202–219,3257–275,4303–321 WhatisPsychoticBehavior?3276–282,4322–328 SomeRelationsBetweenBehavior

ModificationandBasicResearch3283–291,4329–337 CompassionandEthicsintheCareof

theRetardate

ForExperimentalPsychologistsOnly1/2223–241,3295–313,4341–359 CurrentTrendsinExperimental

Psychology1/2242–257,3314–330,4360–376 TheFlightfromtheLaboratory

CreativeBehavior3333–344,4379–390 CreatingtheCreativeArtist3345–355,4391–401 ALectureon“Having”aPoem

LiteraryandVerbalBehavior1/2261–271,3359–369,4405–415 HasGertrudeSteinaSecret?1/2272–286,3370–384,4416–430 TheOperationalAnalysisof

PsychologicalTerms1/2286–292,3385–390,4431–436 TheAlliterationinShakespeare’s

Sonnets:AStudyinLiteraryBehavior1/2292–307,3391–407,4437–453 AQuantitativeEstimateofCertain

TypesofSound-PatterninginPoetry1/2307–316,3408–417,4454–463 TheProcessesInvolvedintheRepeated

GuessingofAlternatives

617

TheoreticalConsiderations

3421–428,4467–474 WhyAretheBehavioralSciencesNotMoreEffective?

1/2319–346,3429–457,4475–503 TheConceptoftheReflexintheDescriptionofBehavior

1/2347–366,3458–478,4504–524 TheGenericNatureoftheConceptsofStimulusandResponse

1/2367–376,3479–488,4525–534 TwoTypesofConditionedReflexandaPseudo-type

1/2376–383,3489–497,4535–543 TwoTypesofConditionedReflex:AReplytoKonorskiandMiller

1/2384–389,3498–503,4544–549 AReviewofHull’sPrinciplesofBehavior

3504–508,4550–554 AReviewofBushandMosteller’sStochasticModelsforLearning

AMiscellany3510–511,4556–557 AParadoxicalColorEffect1/2393–404,3512–523,4558–569 SomeQuantitativePropertiesof

Anxiety1/2404–409,3524–528,4570–574 “Superstition”inthePigeon1/2409–412,3529–532,4575–578 ASecondTypeof“Superstition”inthe

Pigeon3533–537,4579–583 Two“SyntheticSocialRelations”3538–542,4584–588 ConcurrentActivityUnderFixed-

IntervalReinforcement3543–554,4589–600 SustainedPerformanceDuringVery

LongExperimentalSessions3555–558,4601–604 JohnBroadusWatson,Behaviorist1/2412–419,3559–566,4605–612 HowtoTeachAnimals1/2419–426,3567–573,4613–620 BabyinaBox1427,4620 AWordAboutBoxes4621–629 ThePsychologyofDesign2426.01–.18,3574–591,4630–647 PigeonsinaPelican3592–596,4648–652 SomeResponsestotheStimulus

“Pavlov”3597–602,4653–658 SquirrelintheYard

618

Coda

4661–673 CanPsychologyBeaScienceofMind?

619