JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

KEY PECKING UNDER RESPONSE-INDEPENDENTFOOD PRESENTATION AFTER LONGSIMPLE AND COMPOUND STIMULI'

JOHN A. Ricci

KANSAS STATE UNIVERSITY

Sixteen pigeons were trained to peck a key using a response-independent (auto-shaping)procedure of food presentation. The 4-sec grain presentations wvere independent of respond-ing but a keylight stimulus preceded each, with a 4-min interval betveen the grain pre-sentation and the next stimulus. Subjects were divided into four groups, with two durationsof the keylight (30 or 120 sec) and either one or four successive colors on the response keypreceding food delivery. In Phase 2, the birds were continued with the saiiie keylightduration but were presented the alternative number of key colors. All pigeons pecked thekey during the stimuluis. Birds in the twvo groups wvith the 30-sec stimulus duration beganto respond significantly sooner than birds wvith the 120-sec duration. There were no sig-nificant differences in rate of pecking between groups by the last five days of Phase 1. InPhase 1, the pigeons exposed to the four stimultus comiiponents showed an increase in rateof pecking over the four comiiponents as grain presentation approached. The pigeons withone stimulus conmponent did not exhibit this regularity. Analogotis conditions in Phase 2had similar results except for one group. The implications of the occurrence of key peckingdue to response-independent food delivery for multiple and chained schedules were pointedout.

Two procedures are commonly used in"shaping" pigeons to key peck. The first, andmost widely known, is the method of reinforc-ing successive approximations of the key-peckresponse (Ferster and Skinner, 1957). In thisprocedure, grain presentations are responsedependent. The second method, often calledauto-shaping, has only recently received muchattention (Brown and Jenkins, 1968; Gamzuand Williams, 1971; Rachlin, 1969; Williamsand Williams, 1969). This latter procedurediffers from the first in that food presentationsare scheduled independently of respondingand follow keylight presentations. Using thisresponse-independent method, pigeons willpeck at a stimulus that is intermittently pre-sented on the response key in a well-lighted

'These data are based on a thesis submitted to theDepartment of Psychology, Kansas State University, inpartial fulfillment of the requirements for the M.S.degree and were obtained during the author's tenure asa NDEA Title IV Fellow. The research was supportedby NSF grant GB-27595 to Kansas State University,Charles C. Perkins, Jr., principal investigator, and byNational Institutes of Health Biomedical Sciences Sup-port Grant FR-07036 to Kansas State University. I wishto thank Drs. C. C. Perkins, J. Frieman, and E. J.Phares for their support and advice. Reprints may beobtained from the author, Department of Psychology,Kansas State University, Manhattan, Kansas 66506.

box if the presentation of that stimulus regu-larly precedes access to food.Brown and Jenkins (1968) showed that the

presence of the stimulus preceding the grainpresentation is necessary for the occurrence ofthe directed response-independent key peck-ing. Williams and Williams (1969) presentedconvincing evidence that response-indepen-dent key pecking is not primarily the resultof adventitious reinforcement of the key-peckresponse. Several investigators (e.g., Gamzuand Williams, 1971) pointed out that pairingsof stimuli and grain presentations that weresufficient to produce response-independent(non-adventitious) pecking, were also presentin some more conventional response-depen-dent procedures. They suggested that perhapspurely response-independent effects can influ-ence the results of response-dependent proce-dures where the presence of a keylight is cor-related with access to grain.Up to the present, however, durations of the

keylight that have been reported to produceresponse-independent key pecking have rangedfrom 3 to 8 sec only, and show little evidencethat this effect is not confined to a brief periodjust before grain presentation. The presentexperiment used durations of the keylightstimulus of 30 and 120 sec. Durations of this

509

1973, 19, 509-516 NUMBER 3 (MAY)

JOHN RICCI

order justifiably allow comparisons with stan-dard response-dependent procedures wheregrain presentations are correlated with keycolor, such as in some multiple schedules.

Anotlher variable examined here is the num-ber of discrete stimultus components in se-quence that make up the keylight stimulus.The use of sequential cues with a response-independent procedture will allow comparisonsof the present procedure with respondingmaintained on response-dependent scheduleswhere a sequence of cues is temporally relatedto the occturrence of grain presentations, suclas in fixed-interval (FI) chain sche(dtules.

METHOD

SubjectsSixteen experimentally naive Whl1ite King

pigeons, obtained from a local stupplier, weremaintaine(l at 75% of their free-feeding weiglhtby daily sessions in the apparatus and, whennecessary, by supplementary feeding in theirhome cages. One otlher subject died on thetlhird day of the experiment and was replaced.

Appara tus

Two idlentical test clhambers (Grason-Stad-ler, E6446CA) were eaclh equipped witlh atranspairent response key 5 in. (12.5 cm) abovea solenoid-actuated food magazine. Each hada 10-W lamp mounted above the magazinewhich illuminated the opening on every grainpresentation, and a 10-W lamp mounted onthe far right of the panel, at approximatelythe same heiglht as the response key, whichprovided general illumination at all times ex-cept (luring magazine presentation. Any offotur colors (red, blue, green, or yellow) froma Miultiple Stimulus projector (Grason-Stadler,#45801) could be presented on the key. Arielay motunited inside the chamber providedauditory feedback whenever a response wasmade in the presence of the keylight but notduring the intertrial intervals.

Wlhite noise, whichi ranged between 62 and78 dB inside the test chambers (as measured onthe A scale of a General Radio Co. sound levelmeter), masked extraneotus noise. In an ad-jacent room, relay operatecl switching circuits,steppers, and clocks controlled both boxes;counters and print-out counters recorded re-sponses. With this arrangement, two birds

weie run simtultaneouisly, prestumably witlhoutcuLes from the other box or control room.

Procedure

The pigeons, after being randomly dividedinto four groups, were all given magazinetraining. The experimenter lheld the deprivedbird over the raised and filled magazine untilthe animal began to eat. He then carefully re-leased the bird while the subject was still eat-ing( and closed the experimental chamber.After the pigeon had eaten for about 30 sec,the experimenter lowered the magazine andthen quiickly r-aised it again. Tlhen, by present-ing successively slhorter perio(ds of access tograin at suiccessively longer intervals of time,the bii-ds were traine(l to eat from the grainmagazine witlhin 4 sec of its presentation. Thisentir-e process took b)etween five andl 15 grain)resentations. Special care was taken to avoidslhaping the birds to peck the key. To assurethlat the birds wotldl contintue to eat, the firstfive presentations of foodl in the first condition-ing sessioin were of 10-sec (Iluration, andc thenext five, of 8-sec duration; all subsequentmagazine presentations were 4 sec. Immedi-ately after- magazine training, the birds weregiven the first 30 trials.The experiment consisted of two phases.

Table 1 suimmarizes the main feattures of theexperiment. The first number in all groupnames refers to the total keylight duration inseconds; the second, to the ntumber of com-ponents in Phase 1, and the tlhird, to the num-ber of components in Phase 2.

Plhase I of the experiment was a simple 2by 2 design. Two grotups were trained under-a 120-sec total stimtultus (luration. Group 120:1-4 was presenited a single color for 120 sec onevery trial. Group 120: 4-1 was presented aseries of fouir (liffereilt colors in sequience eachof 30-sec (Iuration on every trial. Two othergrouips were trained uinder a 30-sec total stim-uilus duiration. Grotup 30: 1-4 had only onecolor presented for the entire 30 sec; Group30: 4-1 hia(1 four 7.5-sec colored stimuli pre-s'2nted in seqtuence on every trial. The ordersof stimuiltus presentation use(d for the groupswitlh fouir stimuli wei-e ABCD, DCBA, BDAC,and CADB (A = yellow, B = green, C = red,D = blute), witlh one suibject in eaclh group hav-ing one of the or(lers. In eaclh grotup with asingle stimulus, eaclh bird hadl a different color

510

RESPONSE-INDEPENDENT KEY PECKING

Table 1

Sumnmary of Procedure

Procedure

Phase 1 Phase 2

Stimulus StimulusDuration Number Duration Number

of ofGrouip N'ame Total Component Components Total Component Components

Group 30: 4-1 30 sec 7.5 sec 4 30 sec 30 sec 1

Group 30: 1-4 30 30 1 30 7.5 4

Group 120: 4-1 120 30 4 120 120 1

Group 120: 1-4 120 120 1 120 30 4

stimulus. Figure 1 presents a sclhematic repre-sentation of these stimuli.

Sessions of 30 trials with intertrial intervalsof 4 min were condLucted for 20 consecutivedays. The magazine was always presented atthe offset of the keylight. Pecking in no wayinfluenced eitlher the stimuli or magazine pre-sentation. The only effect of pecking was topiroduce the relay click during the stimulusperiods. Responses were recorded during eachquarter of the stimulus and during the inter-trial intervals.At the conclusion of Phase 1, 16 additional

sessions were given for a second phase. DuringPhase 2, the subjects received a keylight of thesame duration as in Plhase 1 but were presentedwitlh the alternative number of stimulus com-ponents; i.e., the groups with one color forthe stimulus during Phase 1 had four colorsfor the stimulus in Phase 2, and vice versa.The color that had been the single stimulus inPhase 1 now became the initial stimulus for

30:4 R YI B0 G6

30:1 Redg

Tjie & 30 Seo

120:4

120:1

Time

Red I Yeellow I Ilue I Green

Red

10 120 Sec

Fig. 1. Schematic diagram of the stimuli presentedto the four groups. The specific color sequence is an

example of the four sequences actually used for differ-ent subjects.

tlhe four-component sequence in Phase 2. Anal-ogously, the initial stimulus in four-compo-nent conditions of Phase 1 became the singlestimulus used in Plhase 2; e.g., pigeons that re-ceived single stimulus D in Plhase 1 receivedthe sequence DCBA in Plhase 2, and pigeonsthat received the sequence BDAC in Phase 1received the single stimulus B in the secondplhase. Phase 2 was simply an interclhange ofnumber of stimulus components witlhin dura-tion conditions. In all other respects, the pro-cedure was the same as Phase 1.

RESULTSFigure 2 presents rates of responding during

the stimulus on each day of the experimentfor the four birds in each group. Each pointrepresents mean responses per second for eachsubject on one day of training. The fourpanels in Figure 2 divide the subjects intotheir respective groups as indicated. This fig-ure slhows that all subjects pecked the key dur-ing the keylight presentation. In addition,considerable overlap can be seen in the re-sponse rates of individual subjects from onegroup to another. All but two subjectsachieved a response rate of at least 0.1 responseper second during Phase 1. These two birds,both in Group 120: 1-4, increased their ratesof responding during Phlase 2. All birds, ex-cept A-13, pecked during the stimulus at arate of 0.3 response per second or better on atleast one day of the experiment.Although there were obvious differences in

group rates for the first several days of Phase1 (where all subjects in Group 120: 1-4 had es-sentially a zero rate) these differences were not

JOHN RICCI

Bi r dsA-13A-is.....C-7 -0-7

V'4.S.,,,,, ,

. "'i--.E..

Group 120: 1-4

4. A

14. ......I ~~1.

* _~ ~~~~~~~~~ Al.

5 10 iS 20 25 30 35 5 10 15 20 25 30 35

DAYS

Fig. 2. Mean responses per second during the stimulus on each day of the experiment for the four birds in eachgroup. Each panel shows response rates for the individual subjects in each separate group as labelled.

maintained throughout the experiment. Differ-ences in the mean rates of pecking for the lastfive days of Phase 1 do not approach statisticalsignificance (F (3,12) = 0.69).

Differences in rates of responding betweengroups during the initial portion of the ex-

periment can be attributed to differences inwhen the birds began to respond. Two-tailedMann-Whitney U-tests (1947) were used todetermine if the groups with 30- and 120-seckeylight durations differed in the trial numberof the first, fifth, and tenth pecks. Althoughdifferences in the trial number of the first peckwere not significant (U (8,8) = 29, p < 0.40),differences in both the trial number of thefifth peck (U (8,8 = 11, p < 0.028) and thetrial number of the tenth peck (U (8,8) = 5,p < 0.002) were statistically significant. Thebirds exposed to the 30-sec stimulus made theirfifth and tenth pecks reliably sooner than thebirds exposed to the 120-sec stimulus.

There were no clear effects of number ofcomponents on the initiation of responding(first peck U (8,8) = 21, p < 0.14; fifth peckU (8,8) = 14.5, p < 0.04; tenth peck U (8,8) =20, p < 0.12). However, the birds receiving afour-component stimulus tended to pecksooner than the birds receiving a one-compo-nent stimulus for both the 30- and 120-sec stim-tilus conditions.

Figure 3 presents separately for each group

the mean number of stimulus periods duringwhich there was at least one peck on each dayof the experiment. Number of trials up to a

possible maximum of 30 per day is plotted asa function of days of training. During Phase 1,the two groups with 30-sec stimuli approxi-mated asymptotic performance on this mea-

sture after the first day, while Group 120: 4-1took four days to reach a comparable level.These three groups pecked on more than two-thirds of the trials for the remainder of the

az0

CoLuan

00.anLuI

2.20

2.00-

1.80

1.60

1.40

1.20

1.00

.80

.60

.40

.20

512

RESPONSE-INDEPENDENT KEY PECKING

25'.%

2-

CW AGfROUP 30:4-1

I c o p 3 :-

,. 0 U P 120:1-

5 l 1u 20 25 30 35

Fig. 3. Mean number of stimulus periods duringwhich there wvas at least onie peck oni each day of theexperinient, plotted separately for each group.

experiment (excepting two days in Phase 2 forGroip 30: 4-1). Tl0e lower means in Phase Ifor Grotip 120: 1-4 are due almost entirely totie performance of two pigeons with low rates

0.roup 30: Trn4-1 E

i:1oo Group1 30:4-I

80-70

Z 60-

50l40 A

20-

La

~ 10. 3

in that group. During Phase 2, the perform-ance of this latter group was more in line withthe performance of the other three groups.

The data presented in this figure clearly showthe degree of consistency with which the sub-jects were pecking the key, even though therewere variations in individual rates of respond-ing.The distributions of responses within the

stimulus are presented in Figure 4. Days are

plotted along the abscissa and the ordinaterepresents the mean per cent of total respond-ing. Eachi subject's daily responses were

counted separately during four equal timeintervals, each corresponding to one quarterof the stimulus. Per cent responding in eachqtuarter was calculated for each subject, andthen mean per cent was calculated for thefour subjects in each group. Each curve in thisfigure represents the mean per cent of the re-

sponding during successive quarters of thestimultus and is labelled from I to IV in orderof distance from the grain presentation. That

DAYSFig. 4. Per cent distributions of responses during the stimulus quarters on each day of the experiment, plotted

separately for each group. Each panel shows distributions for each separate group as labelled.

513

JOHN RICCI

is, under four-component conditions these in-tervals correspond to the length of time eachcolored component stimulus was present, andunder the one-component conditions these in-tervals simply divide the stimulus into fourequal parts. For example, in the upper left-hand panel of Figure 4, curve I presents themean per cent of responding in the 7.5 secjust before magazine presentation for Group30: 4-1. Curve IV in the same panel presentsthe mean per cent of responding for the groupin the first 7.5 sec after the keylight stimulusonset.The remaining three panels present the

comparable data for the other groups, aslabelled. All points in each panel are basedon the four birds in each group, except forGroup 120: 1-4 where the Phase 1 results areon the performances of only two birds, D-7and A-18. The other two birds in this group,C-7 and A-13, responded on only eight and twodays, respectively, in the first phase. Both ofthese latter birds responded on all days in thesecond phase, and thus the Phase 2 results areplotted for all four birds.The left-hand panels show response dis-

tributions for the groups with a four-compo-nent stimulus initially. In these panels, thePhase 1 results are distributed in an orderedfashion with the greatest number of pecksoccurring in the interval just before magazineonset (I), and the smallest number of pecks oc-curring in the interval furthest away frommagazine onset (IV). Every subject in Group30: 4-1 showed this exact order on at least 15of the 20 days in Phase 1; that is to say, thisgroup effect was also obtained for each in-dividual subject. In Group 120: 4-1, two birds,C-l 1 and A-20, showed this exact ordering onat least 17 of the 20 days, while the other twosubjects in that group showed less-markedtendencies in the same direction.The right-hand panels of Figure 4 present

the response distribution data under condi-tions where a single component constitutedthe stimulus in Phase 1. Although individualsubjects were fairly consistent in the mannerin which their pecks were distributed fromday to day, there were no clear group trendsin response distributions during Phase 1.

Phase 2 results show once again responsesdistributed in an ordered fashion, with thegreatest amount of pecking occurring duringthe interval just before magazine onset (I) for

Group 120: 1-4. Group 30: 1-4 was the onlygroup where the four-component conditionfailed to produce an ordering of response dis-tributions over the stimulus period at any timein the experiment. No group trends were ap-parent for this group.

Phase 2 one-component conditions, Group30: 4-1 and Group 120: 4-1, yielded resultssimilar to the initial one-component condi-tions, i.e., no apparent group trends.There was little or no pecking during the

intertrial intervals for any bird. After thesecond day of training, every subject made atleast 10 pecks during the stimulus periods forevery one peck in the intertrial intervals. Itis noteworthy that the intertrial intervals weretwo or eight times as long as the stimulusperiods for the 120- and 30-sec groups respec-tively.

DISCUSSIONThe main finding is that keylight durations

of 30 and 120 sec produce and maintain keypecking in response-independent procedures.This is a considerable extension of the rangeof stimulus durations that have been studiedwith this type of procedure.Williams and Williams (1969) demonstrated

that pigeons tend to peck the key in response-independent procedures, even if the responseserves to eliminate grain presentations. Thepecking they observed cannot be dismissed asa "superstitious" behavior because their pro-cedure precluded "accidental" reinforcementof key pecking and, in fact reinforced "not keypecking". Although the present study made nospecific attempt to control for the effects of ac-cidental reinforcement of pecking, in the lightof the results of Williams and Williams itseems safe to maintain that the present resultsalso cannot be attributed entirely to adventi-tious reinforcement.The present data, along with previous re-

lated work, imply that whenever there are in-cidental pairings of keylight and food in pro-cedures using pigeon subjects, there may alsobe an increased pecking tendency that is notthe result of adventitious reinforcement.Gamzu and Williams (1971) showed this to beso even if the extent of the "pairing" is simplydifferential probability of grain presentationduring two discriminable stimulus periods.The present experiment demonstrated that

514

RESPONSE-INDEPENDENT KEY PECKING 515

stimuli that precede grain presentations by asmuch as 2 min may produce response-inde-pendent key pecking. Considering both ofthese findings, it seems plausible that theresults of response-dependent procedures, suchas discriminated operant schedules, whichhave incidental pairings between keylight andfood, are confounded with an additional key-pecking tendency that is not the result of theresponse-reinforcer relationship. Discriminatedoperant procedures regularly employ discrim-inative stimuili of about the same duration asthe stimuli used in the present experiment andfood presentations occur on temporal sched-ules similar to the one used by Gamzu andWilliams (1971). The positive behavioral con-trast often observed in these procedures (e.g.,Reynolds, 1961) might be due, at least in part,to the effect of a response-independent key-pecking tendency during the positive discrim-inative stimulus.Up to the present, response-independent

procedures have generally employed only twostimulus conditions, one that preceded thegrain presentation, and the other that waspresent at all other times. Variations of thisprocedure have included illumination of thekey as the stimulus paired with grain pre-sentation (Brown and Jenkins, 1968, Experi-ment IV); offset of key illumination as thepaired stimulus (Brown and Jenkins, 1968,Experiment III); and constant key illumina-tion where color change served as the pairedstimulus (Gardner, 1969). In the present study,the single-stimulus conditions were similar tothese former studies; and as in these studies,substantial pecking was observed only duringthe stimulus that preceded grain presentationsand not during the intertrial intervals.The present study differed from others in

that in the four-component conditions a se-quence of keylights preceded grain presenta-tions. The use of four stimulus components insequence resulted in an increasing rate ofpecking over the components as the time ofgrain presentation approached. The use ofstimuli in fixed sequence is similar to at leastone response-dependent procedure, i.e.,chained Fl schedules of reinforcement.

In such schedules, stimuli are incidentallypaired with grain presentations. In this case,one stimulus always precedes food presenta-tion, and the others never precede food pre-sentation. An effect often obtained with these

procedures is that pigeons fail to maintainpecking in the initial components of longerchains (e.g., Gollub, 1958). The procedure em-ployed with four-component groups in thepresent study (with the addition of an inter-trial interval) resembles an Fl chain, and here,as in the response-dependent procedure, fewresponses are emitted during the initial com-ponents of the sequence. It is conceivable thatthe stimulus relationships present in the chainschedules may influence pecking indepen-dently of the reinforcing operation.The influence of sequences of stimuli upon

responses that are independent of their conse-quences is not without precedent. Hendry,et al., (1969) presented response-independentshocks to rats at 6-min intervals while barpressing was being maintained on a variable-interval schedule of reinforcement. For somerats, a single stimulus condition was main-tained between shocks while for others a se-quence of three stimuli each lasting 2 min wasinterspersed between shocks. When three stim-uli were used, the rate of bar pressing de-creased with each successive stimulus compo-nent, the lowest rate being in the componentjust preceding shock presentation. Althoughthere was an immediate increase in bar press-ing after shock with either one or three cues,with one stimulus there were no apparent dif-ferences in bar-pressing rates for the two 2-minintervals preceding shock. These results arequite comparable to the present results andseem to imply a commonality between the re-sults of aversive and appetitive response-inde-pendent procedures.

REFERENCESBrown, P. L. and Jenkins, H. M. Auto-shaping of

the pigeon's key-peck. Journal of the ExperimentalAnalysis of Behavior, 1968, 11, 1-8.

Ferster, C. B. and Skinner, B. F. Schedules of rein-forcement. New York: Appleton-Century-Crofts,1957.

Gamzu, E. and Williams, D. R. Classical conditioningof a complex skeletal response. Science, 1971, 171,923-925.

Gardner, W. M. Auto-shaping in bobwhite quail.Journal of the Experimental Analysis of Behavior,1969, 12, 279-281.

Gollub, L. R. The chaining of fixed-interval schedules.Unpublished doctoral dissertation, Harvard Uni-versity, 1958.

Hendry, D. P., Yarczower, M., and Switalski, R. C.Periodic shock with added clock. Journal of theExperimental Analysis of Behavior, 1969, 12, 159-166.

516 JOHN RICCI

Mann, H. B. and WVIhitney, D. R. On a test of wvhetherone of two rand(lomii variables is stocliastically largerthani the othier. The Annals of MIatheiatical Statis-tics, 1947, 18, 50-60.

Rachlin, H. Auto-shaping of keypecking in pigeonswvith negative reiniforcement. Joirnal of the Experi-mental Analysis of Behavior, 1969, 12, 521-531.

Reynolds, G. S. Behavioral contrast. Journal of the

Experimlental Analysis of Behavior, 1961, 4, 57-71.X\'illiamXs, 1). R. anid WVilliams, fI. Auto-maintenanice

in the pigCoIn: snlstainie(d peckinig (lespite conitinigenltnoni-imeiniforceimnenit. Journal of time ExperimentalA nalysis of Behiavior, 1969, 12, 511-520.

Receive(d 24 Jan noa my 1972.(Fimmal Acceptanzce 7 Decemrlber 1972.)