This article was downloaded by: [Moskow State Univ Bibliote]On: 04 September 2013, At: 23:52Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number:1072954 Registered office: Mortimer House, 37-41 Mortimer Street,London W1T 3JH, UK

Theory Into PracticePublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/htip20

Basic unresolvedteaching‐machine problemsSidney L. Pressey aa Professor emeritus of psychology, Ohio StateUniversityPublished online: 16 Aug 2010.

To cite this article: Sidney L. Pressey (1962) Basic unresolved teaching‐machineproblems, Theory Into Practice, 1:1, 30-37, DOI: 10.1080/00405846209541773

To link to this article: http://dx.doi.org/10.1080/00405846209541773

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of allthe information (the “Content”) contained in the publications on ourplatform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy,completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views ofthe authors, and are not the views of or endorsed by Taylor & Francis.The accuracy of the Content should not be relied upon and should beindependently verified with primary sources of information. Taylor andFrancis shall not be liable for any losses, actions, claims, proceedings,demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, inrelation to or arising out of the use of the Content.

This article may be used for research, teaching, and private studypurposes. Any substantial or systematic reproduction, redistribution,reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access

and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

BASIC UNRESOLVED

TEACHING-MACHINE PROBLEMS

SIDNEY L. PRESSEY

Mr. Pressey, professor emeritus of psychology at TheOhio State University, devised and exhibited in 1925what is now generally recognized as the first teachingmachine. He raises some thought-provoking questions

about current methods and devices.

E FFORTS to develop "teaching machines" and materials to go withthem have been going on for some forty years; for the past five

years or so such work has been little less than frenetic. As appar-ently the one person who has from the beginning been active in thisfield, the writer sees changes in basic orientations over this periodwhich seem to have gained relatively little notice, but if considered,might much facilitate progress.

Adjunct versus Initial Auto-InstructionIn all the auto-instruction up to about ten years ago, the student

first looked over a reading assignment, laboratory exercise, or othermaterial, and only after some such first contact with the matter tobe learned did the auto-instructional procedure present carefullychosen questions on that matter, immediately appraise each answer,and if it was wrong indicate or guide to the correct answer. Theauto-instruction thus functioned like a good teacher or tutor who,after a student is presumed to have made some effort to deal withan assigned task and as an adjunct to that effort, asks questionspointing up the important and possibly difficult issues, and expli-cates each if difficulty appears.

Thus in a required course in psychology under the writer'sdirection, each student, having read a chapter in the textbook,went down a mimeographed sheet of 30 four-choice questions onthe major points of that chapter, checking his answer to eachquestion in the appropriate answer space of a 3 x 5 "chemo-card." If, for example, he thought the third alternative was the

30

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

TEACHING-MACHINE PROBLEMS 31

correct answer to the first question, he put a check mark in thethird of the first row of four quarter-inch squares printed on thecard. He used a fountain pen filled with a special red ink ; in each"correct" answer square there had been printed an invisiblechemical which instantly turned a mark black. If his mark re-mained red, the student knew that he had chosen a wrong alterna-tive, and tried again until he did find the correct (color-changing)answer. After many questions on the mimeographed sheet werepage numbers so that he could easily look up an issue which theauto-instruction had not sufficiently clarified. And a second formof each "auto-test" made it easy for the student to assure himselfof his understandings and get further auto-instructional aid, if hedesired.1

In startling contrast, teaching machines of the past few yearshave attempted to replace textbooks and initially present what isto be learned. The student is shown this material one bit or frameat a time in the window of a mechanism or space of a programedtextbook. He cannot readily look back at what he has been overor ahead to sense what is to come, or discover any outline or struc-ture in the material, or read or review selectively. His view isconfined to the window or program item and he must look atperhaps several thousand frames each by itself and in the predeter-mined order. For effective reading, for general understanding ofmain ideas, and for adequate study and review, this procedureseems to be about as clumsy as asking a person to apprehend apicture but letting him see, in a set order, only one square inch ata time!

Thus one course may have, instead of the textbook, some twothousand questions on long rolls of paper. Each roll is put in abox with a window just large enough to show one question andspace to write a one-word answer to it. When a student haswritten his answer, he turns a knob which rolls up the paperenough to show him the correct answer and the next question.Another course has a programed textbook which consists of aboutthe same number of such questions so printed that the studentsees on one page a question and space for him to write its answer,then turns the page to find the right answer and next question.Devices such as Crowder's "scrambled books" and some machinespresent more in a frame, but involve similar difficulties of over-view, selective use, and review.

1 For either student or instructor, the card was a compact yet detailed record.The red checks showed the difficult questions. A count of them was an error score,uniquely adequate, since it was possible to make more than one error on a question.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

32 THEORY INTO PRACTICE

To some psychologists such initial presentation of material tobe learned bit by bit with immediate reinforcement may seem soundtheoretically. However, some half-dozen experiments seem to agreethat, in a given time, no more learning may actually result than ifthat same material were organized in continuous discourse (ques-tions and their answers turned into declarative sentences) andsimply read!2 And most of the materials used were too brief for theadvantages of well-organized continuous discourse over many littlequestions to be realized fully.

For much drill, as in arithmetic or relatively rote learning in aforeign language, special training as in industry or the military,or work with young children, initial presentation of material one"frame" at a time with immediate response called for may be ofmuch value. But study of a complex and structured subject seemsbetter begun by an overview of reading matter to display thestructure and order the complexity. A good book will show itsstructure in the table of contents and catalog its contents in theindex; with such aids the learner can easily move about in its num-bered pages with only the flick of a finger, using page headings andsubheads in the text to guide him. He may turn back and forthfrom table or graph to related text, skip something already known,review selectively for major and difficult points. In the writer'sopinion, only after such first contact with a complex structuredtopic should a student turn to auto-instruction for review anddifferentiation of major points in material just read or perhapsobserved in a demonstration, field trip, or laboratory. The auto-instruction will then assure the student when he is right and identifyand correct any misconceptions—as a good teacher or tutor mightthen do. Auto-instruction as an adjunct to the usual materials and

2 Lumsdaine, A. A., and Glaser, Robert. Teaching Machines and ProgrammedLearning. Washington, D. C. : National Education Association, 1960, pp. 500-501.Silverman, Robert E. "Automated Teaching: a (Review of Theory and Research."Port Washington, 'New York: U. S. Naval Training Device Center, June 8, 1960.Evans, James Lee; Glaser, Robert; and Homme, Lloyd E. "An Investigation of'Teaching Machine' Variables Using Learning Programs in Symbolic Logic." Pitts-burgh, Pennsylvania: Department of Psychology, University of Pittsburgh, December,1960. Goldbeck, Robert A., and Briggs, Leslie J. "An Analysis of Response Modeand Feedback Factors in Automated Instruction." Technical Report No. 2. SantaBarbara, California: American Institute for Research, November, 1960. Goldbeck,Robert A. ; Campbell, Vincent N. ; and Llewellyn, Joan E. "Further ExperimentalEvidence on Response Modes in Automated Instruction." Technical Report No. 3.Santa Barbara, California: American Institute for Research, December, 1960. Roe,Arnold, et al. "Automated Teaching Methods Using Linear Programs." Los Angeles :Department of Engineering, UCLA, Report No. 60-105, December, 1960. Clough,John, et al. "Oberlin College Teaching Machine Project 1959-60: a Report Submittedto the Ford Foundation." Oberlin, Ohio : Oberlin College, February, 1961.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

TEACHING-MACHINE PROBLEMS 33

methods of instruction would seem both more widely useful andmore practicable than current efforts to replace textbooks andmethods with radical initial programing. For example, a teachermight prepare only two or three pages of adjunct questions, supple-menting that many especially difficult chapters in the textbook.These pages could be constructed so as to be useable with any oneof a dozen auto-instructional devices. A teaching-machine companyor professional programer might publish, with good profit, suchinexpensive materials for use with textbooks now widely used.3

Mode of Pupil ResponseUntil about 1950, all efforts at auto-instruction used objective

types of questions; since then, most programs have called forwrite-in answers. For these last, a theoretical justification can bemade, but so also for objective forms. Choice of form should be onthe basis of empirical findings as to comparative efficiency in bring-ing about learning, and "The findings to date do not justify anyconclusion that multiple-choice responses are less effective thanconstructed responses . . . ."4 As the next section of this paper willindicate, objective questions constructed and selected on the basis oftheir actual effectiveness in bringing about learning should be moreeffective for that purpose; no materials to date seem to have beenconstructed on this logical but laborious basis.

The total cost in time and trouble of write-in answers should berecognized. Not only must each answer be written in the teaching-machine window or space in the programed book; the paper mustthen be rolled up or the page turned and comparisons made—andall this must be done over and over and over again. A well-knownprogramed text requires over 2300 page turnings back and forth.The total of eye and hand movements (often including puttingpencil down and picking it up again) may well be cumulativelyfatiguing—and boring. In contrast, objective material requires onlychecking or key-pressing; and the key machine may automaticallykeep a cumulative score, of value to both learner and teacher—sub-stantial evidence indicates that knowledge of progress helps learners.

Easy versus Difficulty-Resolving QuestionsAuto-instructional research of the past decade has been domi-

nated by the concept of error-avoidance: the learner should neithersee wrong answers (in a multiple-choice question) nor be given

3 A next step might be to issue revised or new textbooks especially planned foruse with adjunct programs, as will be mentioned later.

4 Silverman, op. cit., p. 22. Most of the other citations come to the same con-clusion.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

34 THEORY INTO PRACTICE

questions sufficiently difficult that he is likely to make any mistake.In improving a program, the effort is to make it yet more easy (see,for instance, Holland in Lumsdaine and Glaser, pages 225-28).Inferences from experiments with animals and in special laboratoryprojects may indicate that this is desirable. But there are dataindicating that auto-instructional human learning of meaningfulmatter proceeds somewhat differently (see Pressey and Stephensin Lumsdaine and Glaser, pages 73-93). The right response is notsimply the reinforced response, but the one that is recognized asmost meaningful or logical. A wrong alternative may help thatprocess if it needs to be differentiated as such to delimit or clarifythe meaning. Moreover, if a human learner is "ready" for a topic,he already has some ideas about it; and if he is efficient in studyhabits, he may begin with a rapid overview of it, the programingbecoming "adjunct." An important, if not major, task of the auto-instruction will then be the explication and co-ordination of theseprior learnings. And the major task in any new learning may be torelate it discriminatingly to the old.

The writer of such a program will not so much seek artfullyto shape the student's responses so that, without his quite knowingwhat is happening, he is cued, reinforced, and faded into his learn-ing; rather he will attempt to present a series of lively challenges,each resolved by the auto-instruction so as progressively to enlargeunderstandings. And revisions will eliminate not only easy itemsbut even easy parts of items. For example, using an auto-instruc-tional device similar to the "chemo-card" described early in thispaper, 30 four-choice questions were given near the beginning of afreshman course in psychology, and then included in an objectivemid-term examination given over a month later. Several of the 30questions showed no gain, mostly because they were so easy thatalmost every student initially knew the answer and there was noroom for gain. But 12 of the questions showed gains of 20 per centor more; on all of these, a control group which did not have theinitial auto-instruction did less well on the mid-term, thus indicatingthat the usual readings and instructional procedures had not takencare of these issues. Further, at least one wrong alternative, in eachof these 12 four-choice questions was so rarely chosen that it couldbe omitted as nonfunctional. It was concluded that the revisedprogram should be made up entirely of such items of proven instruc-tional efficacy, cut to three choices.5 After going over material with

5 Jones, Robert Stewart. "Integration of Instructional with Self-Scoring Measur-ing Procedures." Doctoral dissertation, The Ohio State University, 1950.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

TEACHING-MACHINE PROBLEMS 35

an auto-instructional device, a student should feel, as after a sessionwith a very able tutor, tha t every question involving likelydifficulties has been elucidated.

Might the difficulties thus resolved have been prevented byrevising the textbook or other reading to make it clearer? Probablysomewhat, but not entirely without making it so prolix as to weakenits effectiveness. Might a many-easy-questions program have fore-stalled the difficulties? Not, this writer believes, without a plethoraof question-answering that fails to differentiate the larger structureof the subject and the major ideas.

As regards in-co-ordination with current learning theory, it mustbe considered that human learning of meaningful matter, involvinglanguage and concepts and helpful skills as in reading and studymethod, in all these respects differs from what is mostly rote or ani-mal learning upon which current theory is based. As a well-knowntheorist has declared, "There are probably a number of differentkinds of learning . . . . I t is quite probable that these differentkinds of learning follow different laws . . . ."6

DevicesMany current teaching machines are expensive, space-taking,

clumsy to look after, and difficult to service. The programs may alsobe bulky and, the answer roll at least, useable only once. The latteris true also of a programed book. If only objective questions maybe used, then at one stroke great advantages may be gained in allthese respects. If a separate answer card or strip may also beincluded (as in the first experiment mentioned in this paper) theprogramed book or other material may be used over and over again;and the only "machine" is some cards. If there is "hardware," itcan be a key machine, perhaps only two inches in each direction.

This type of key-driven device (which the writer has usedextensively in classroom experimentation) turns up the number ofthe next question to be answered only when the correct answer tothe question under consideration has been found, and keeps beforethe learner a cumulative count of his errors. If desired, the studentcan very easily set the little machine back to zero and go throughthe lesson again; in successive trials he thus has knowledge ofprogress in terms of decreasing errors, toward an obvious goal—none. Moreover, by a simple movement of one little shift-lever,the device can be changed into an automatically scoring testingmachine; it turns up the next question regardless of which key is

6 Hilgard, Ernest R. Theories of Learning. Second edition. New York : Appleton-Century-Crofts, Inc., 1956, p. 461.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

36 THEORY INTO PRACTICE

pressed but counts right answers—and a simple attachment auto-matically rewards the pupil with a candy lozenge if he scores high!

Not only are such cards and devices very convenient to use;they may be more instructionally effective. Both card and keymachine reinforce instantly; there is no delay in rolling up thepaper or turning the page—and no uncertainty as to whether avariant "constructed" response is correct. Obviously, writing aresponse is slower than checking or key-pressing. With the card,review of difficult points is easy; the student need only note thered checks and look at those questions. Machines may proportionreview to individual need: in 1927 this writer reported a drill devicewhich, in successive runs through a "lesson," returned only thoseitems on which a mistake had been made the previous time through.A Briggs device will prompt, Crowder's machines "branch," andPask's literally "fade" (Lumsdaine and Glaser, pages 42-46, 286-304, 336-66). These last devices, unfortunately, remain largely inan experimental stage because of the current preoccupation with"constructed" responses and problems of "initial" programing.

EvaluationIt has been proposed by some that programs be considered

good insofar as they adequately cover the content of the subjectdealt with. But many subjects have parts that can hardly be pro-gramed (for example, laboratory and field work in science); andadjunct programing expressly aims to aid, not replace, other teach-ing materials like textbooks, and so can properly be appraised onlyin combination with them. That final examination grades are raiseda bit when some auto-instruction replaces a few laboratory orlecture hours may be little justification for that replacement; somevigorous extra reading might have brought the same gain with lesstrouble! In this period of crowded schools and overworked teachers,it is important (as this writer and others experimenting from theelementary school through college have found) that auto-instruc-tion may substantially reduce the teaching time in a course fromperhaps five hours to two—three hours of relatively routine teachingtasks being taken care of by auto-instructional devices. But pro-visions for independent study without auto-instruction may servesimilarly.7 And for such use, adjunct programing supplemental tousual instructional materials may be much more feasible andpossibly more effective than the radical innovation of initial andcomplete change-over in instructional matter.

7 Baskin, Samuel. "Quest for Quality : Some Models and Means." Winslow R.Hatch, editor. Washington, D. C. : U. S. Department of Health, Education, andWelfare, 1960.

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013

TEACHING-MACHINE PROBLEMS 37

SummaryTypically, enthusiasts carry a new idea to extremes; soon these

are eliminated, and the proven residual becomes part of educationalpractice. The writer believes that certain kinds of changes willoccur in future programed learning, (a) Enthusiastic programerswill soon give up trying to replace most textbooks and other corematerial with thousands of "frames" viewed seriatim. Drill androte learning may be so handled. But the larger usefulness of auto-instruction will be found in co-ordination with, not replacement of,other materials and methods. However, the programers' effortsmay bring improvement in these other materials. The experimentscited on page 32 found that very direct and simple expositorystatements were better than the more formal textbook prose. It ishoped that programers, textbook writers, and teachers may combineefforts to produce co-ordinated texts, programs, and methods ofusing them far better than anything now available, (b) Once it isrealized that most auto-instruction can use objective rather thanwrite-in responses, new potentials as to both convenience andeffectiveness will be reached, (c) Abandonment of the "many-easy-questions" concept will both greatly decrease bulk and increaseincisiveness. (d) Devices can be far more simple and inexpensivethan those now on the market—perhaps simply a 3 x 5 card. Ma-chines can also serve each learner's needs far more closely. Useshould then increase greatly, (e) Evaluation should realisticallytake account of both values and costs in time and trouble as wellas in money under everyday conditions. Such evaluations will showthat auto-instruction can be made both far more convenient andpractical, and more useful, than is now generally recognized.8

And the experimental psychologist may find a disturbing needto revise his theories of learning. He may need to admit that acompetent human learner goes at meaningful material looking forits significance somewhat as he might scan a picture, not simplyattempting a series of bit-learnings; he may find good "wrong"alternatives more like contrasting elements in a picture than blindalleys in a maze; his progress may be more by insights than condi-tionings, with the final phase better described as closure than rein-forcement. This is not to urge gestalt theory, but only to suggestthat terms and concepts other than those which have guided theprogramers seem more apt.

8The efficiency may be embarrassing! A student may deserve five hours creditthough taught in a class meeting only two hours; he may finish a semester course inthe first six weeks—and as an outcome finish a four-year college course in three years,or twelve years of public-school work in ten!

Dow

nloa

ded

by [

Mos

kow

Sta

te U

niv

Bib

liote

] at

23:

52 0

4 Se

ptem

ber

2013