". 3-." ~'§\ 'J .‘ “b.0905 ...«v.‘ ‘ . “ . fl .0. 5.. ’ Je- ‘fflnfi. - EHI.‘ J 1‘ 1 a“ o 7‘ N «m '0 in "h ‘ . o u . A‘ .u v o .. a. CA. 7.1.. .... a . n . A .u «QM... . Hr... - I .\ 0.“.- | ...-v.o\.. m..:.. m n F. \... . 5 . . . . .9 .i I o . «o L o: .n n .\. . n N #0..” ‘ y 7...}. a A « ”No-0.58 by 6H!» taut... m M.“ S‘W.» m. r . fl. .3.“ '1...“ t ml, L. a” 'V I c u I ~ I 1. no n. J .. , q «.1, m . a a V... n .9. . ,. v. L .3. .M ,3 [a Q s. n..- . no 0 o . mi.“ 4, .... yum“ 5.2 «3»... ”Q 9". qt L} a... .3 w. v. 9 \ . 4.1... mt C... q a .. x. n3 1.. u... .33... avg “-3..“ Mr”. *1‘ fl 8).“. DOiIt £40.qu 3;“; ‘f 1. NJ“. 4!.l "’o.?. m”! o .v!.. ‘ . .- H..:¢ $.50... Nov 7ADMII 5‘ »\.l.. .. 1“ 0‘. .I. . fl . x S E ”.1. .. .- Q J I 3-: 1 i U U 8.: t v 1 ‘3},— I. .. 7; a. .u 3...” fl L n .u 'o 0‘ ~.\JJ . .... L . q r J. . .. ‘Q d .\ 9". I“ O.- .. ‘ O m 3 3x. fl... fl . 7... .. 1-“ J. i J b _ "1. -. ”€93 Au (IV a. «u ._, . 4.. «o. 1M“: 1...... P.~.i.. t)...“ 9....-. v9 . \J o). .. a. 4.. lug h \V ‘ .. I!!! I . . ...l‘. 1...... .u .. a?! . p . . “‘2“ A s .1. a a . 6 . I o 11 '. .12. I..3.I a “HR ( "’ U “a .. ”u.- 1.5. «$95!?» ‘3‘ ‘n 1‘ I V m. J w..p “ ..a.-l.0 v! .1 D‘LHUH o m-Hls I'LL-- a . 5...: CG. [I v.‘ 9’ I‘“ ,c J" I I. . . . a _ U-.. I I: w. Do. u». ‘9’. ‘t..t ‘l‘. .3“ 'c I-. in; n!!! 53,; __‘:.:_:_,:_::33.ng1: mmm LIBRARY MiChigan Sum University ABSTRACT STIMULUS DELAY AND THE REDUCTION OF ERRORS IN THE TRANSFER OF STIMULUS CONTROL IN A SIMULTANEOUS DISCRIMINATION By Charles Hawkins Brown Pigeons initially trained on a simultaneous discrimination of line orientation were subsequently transferred to a wave length discrimination. Three transfer procedures were employed. Subjects assigned to the abrupt- transfer procedure were abruptly switched from line orientation to the wave length discrimination. Subjects assigned to the stimulus-compounding pro- cedure were trained on a compound stimulus consisting of the line orienta— tion and the wave length dimensions displayed in superimposition prior to the presentation of the wave length dimension alone. Subjects assigned to the stimulus-delay procedure were trained on a compound stimulus in which the presentation of theline orientation component of this stimulus was de- layed for successively longer intervals as a result of a correct response on the preceding trial. Stimulus-delay subjects transferred by responding to the wave length dimension prior to the presentation of the line orienta- tion component to form the compound. Subjects transferred with the stimulus- delay procedure exhibited superior performance to the wave length dimension as indexed by errors and response latencies. Subjects transferred by the stimulus-compounding and the abrupt-transfer procedure displayed five and ten times as many errors to the wave length dimension respectively as sub- jects receiving the stimulus—delay procedure. . , C’ , Approved: 229$: (g: Eflggfifi Date: { 25% “£1: Z Z 2 2 2 Committee Ch ‘rman STIMULUS DELAY AND THE REDUCTION OF ERRORS IN THE TRANSFER OF STIMULUS CONTROL IN A SIMULTANEOUS DISCRIMINATION By Charles Hawkins Brown A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1973 Q J" ' ACKNOWLEDGMENTS I wish to acknowledge my indebtedness to Dr. Mark E. Billing for his invaluable guidance in the preparation of this thesis. Gratitude is also extended to the other members of my committee, Dr. M.R. Denny and Dr. R. Raisler, for their insightful criticism and assistance in the design and execution of this research. ii List of Figures Introduction . Method . . . Results . . Discussion . List of References TABLE OF CONTENTS iii Page iv 11 22 27 LIST OF FIGURES Figure Page 1 Mean number of errors for the subjects in each group on the 21 trials to the S2 dimension. . . . . . . . . . l4 2 Mean response latency for the subject in each group on the 21 trials to the $2 dimension. . . . . . . . . . 17 3 Response latency during each trial of the transfer session for the best and worst subject in each group. . . . . . 20 iv INTRODUCTION A traditional view of discrimination learning, as represented by Hare low (1959b), has argued that acquisition is primarly a process of error elim- ination. Other theorists have held (see Holland, 1965 for a review) that er- rors are an integral part of the discrimination learning process. These the- ories have been challenged by the phenomenon of errorless learning. In error- less learning a subject acquires a discrimination between two stimuli withv out responding (making errors) to the 3- value. In the typical errorless learning paradigm the subject initially acquires a very easy discrimination in which the stimuli are maximally different. Then the distance between the stimuli along some physical continuum is gradually reduced until the final values are attained. The classic demonstration of errorless learning was performed by Ter- race (1963a). Building on experiments by Skinner (1938), Schlosberg and Solomon (1943), Lawrence (1952), Baker and Osgood (1954), Terrace reinforced the pecking response of a pigeon when the response key was illuminated with a red light but not when the key was darkened for brief periods of time. Peeks to the key occurred when the key was red, but not when it was dark. Terrace transferred the stimulus control from a brightness to a wave length discrimination by varying the 8- value along two dimensions. First the dura- tion of the S- was gradually increased between successive presentations. Then by gradually reducing the value of a resistance in series with a light source, Terrace introduced a green hue on the previously dark response key. 1 2 By the end of three sessions the subjects had errorlessly acquired a discrim- ination between a green 3- of duration and intensity equal to that of the red 3+. Terrace's procedure is called fading or stimulus shaping in which a stimulus is gradually changed along some physical dimension. This tech— nique forms the basis of most errorless procedures. Errorless learning has also received attention in paradigms requiring the transfer of stimulus control from one dimension to another. The trans- fer paradigm employs pretraining in contrast with conventional acquisition in which no pretraining is employed. Three different transfer procedures are commonly used. In the first procedure, the subject is trained for 1 sessions on the $1 dimension and then the $2 dimension is abruptly presented. In the second procedure, the subject is trained for X sessions on the $1 dimension and then a compound stimulus consisting of the simultaneous combination of the 81 and 82 dimensions is presented for X sessions. And then the $2 dimen- sion is presented alone. In the third procedure, the subject is trained for 1 sessions on the SI dimension and then the compound stimulus is presented for X.sessions. Gradually the 81 component of the compound stimulus is faded out; and finally the $2 dimension is presented alone. The transfer paradigm has two advantages, first, it allows the experimenter to monitor and deter- mine the immediate history of the organism prior to the presentation of the novel stimulus dimension, and second, the dependent variable (eg. pecking) is already an established member of the subject's response repertoire and the association of the novel stimulus with reinforcement may be monitored independently of the acquisition of the response. The errorless transfer of stimulus control was initially reported by Terrace (1963b). Subsequently errorless transfer has been obtained with normal and retardate humans (Moore and Goldiamond, 1964; Touchette, 1968) 3 as well as with a varsity of animal subjects (Schusterman, 1966; 1967; Ter- race, 1966; Westbrook and Miles, 1970) substantiating it as a rather robust phenomena. These studies share the common property of depending on empirical- ly derived programs of systematic stimulus change to maintain errorless per- formance throughout the transfer session. Terrace's (1963b) initial proce- dure consisted of the acquisition of an errorless successive wave length discrimination. After ten sessions of training, the subject was presented with the compound stimulus display consisting of the established dimension, wave length, and the dimension of transfer, line tilt. For five sessions the subject responded in the presence of the superimposed stimulus display. On the sixth session wave length was gradually faded out by systematically re- ducing the intensity of that dimension. Terrace's pigeons transferred to the line tilt dimension without making errors. However, as Reynolds (1961) demonstrated, either feature of a com- pound display could have been exerting stimulus control. Terrace's procedure did not reveal the mechanism by which the second dimension acquired stimulus control. In errorless transfer, the answers to two questions have remained elusive. When does the second dimension in a compound display acquire stim- ulus control? What are the significant parameters of the procedure which re- sult in the transfer of stimulus control? These questions were addressed by Schusterman (1967) who employed probe trials during transfer in an attempt to ascertain the controlling dimension. The training prOgressed on a simultaneous size discrimination through a ser- ‘ ies of reversals on a form discrimination. Transfer was effected by fading out (reducing the size difference of) the SI cue. The results of this study suggest first, that stimulus control transfers gradually, and second, that substantial intrasubject variability exists in the dimension exerting 4 control for subjects receiving identical training. However, these conclusions may have been produced by averaging data across a series of reversals and may not reflect the course of acquisition for any one discrimination. Touchette (1971) has reported a simultaneous discrimination procedure permitting the measurement of the point stimulus control transfers between continua. The point of transfer is used here to signify the trial by which the second dimension acquires stimulus control. Subjects are first trained to a criterion on a simultaneous discrimination. In the next session the dimension of transfer, 82, is presented in conjunction with the dimension exerting control. A correct response to the compound is reinforced and ter- minates the trial. 0n the following trial the $2 dimension is presented alone for 0.5 sec before 81, the controlling dimension, is superimposed on 82. A correct response to the compound increases by 0.5 sec the interval during which the 82 dimension is presented alone prior to the superimposi- tion of the $1 dimension to form the compound. Similarly errors terminate the trial and reduce by a factor of 0.5 see the interval between trial on- set and the formation of the compound stimulus on the subsequent trial. Behavior is under the control of the initial stimulus dimension as long as the subject waits for the presentation of the compound before responding. Stimulus control has transferred to the second dimension when the subject responds before the compound is presented. The efficacy of Touchette's pro- cedure is revealed by the errorless performance of the three retarded boys who served as subjects. His data demonstrate that the subject's latency to response can indicate that trial in the training sequence by which the $2 dimension has acquired control over the response. Furthermore his data in- dicate that individual differences exist in the controlling dimension fol- lowing identical training; however, transfer generally occurred following 5 a surprisingly few number of trials to the compound. The status of Touchette's results may be questioned for two reasons. First, Touchette's subjects had extensive experimental histories. It is un- clear how naive subjects would perform with his procedure. Second, Touchette's procedure does not employ a conventional fading procedure. It is true that Touchette's technique manipulates the latency of $1 onset, however, all of the fading procedures described above employed a program which altered the physical characteristics of the stimulus array (eg. size, intensity, hue etc.) in addition to the temporal parameters. Consequently the question re- mains, would Touchette's subjects have learned as much about the $2 dimen- sion had they been presented with a conventional compound display for the same number of trials that they received before transfer? This research extends Touchette's work first, by employing naive infra- human subjects (pigeons) to control for experimental history; and second, by incorporating two additional groups, the stimulus-compounding group, and the abrupt-transfer group, to measure any advantage of the progressive delay feature. Specifically the research was designed to partition the progressive delay procedure into two of its components, a conventional compound proce- dure, and an abrupt transfer procedure. And to compare the performance of subjects receiving these treatments on the $2 dimension following transfer. Group 1, the stimulus-delay group, was transferred following Touchette's procedure. Group 2, the stimulus—compounding group, received the compound stimulus the same number of trials as presented to the subjects in group 1. And then group 2 was abruptly presented with the $2 dimension. Thus a supe— riority in performance of group 1 over group 2 on the $2 dimension must be due to the progressive delay feature of Touchette's procedure. A significant difference would indicate that Touchette's procedure not only is useful for 6 measuring when the $2 dimension acquires stimulus control, but also that his procedure assists in the deveIOpment of that control. Group 3, the abrupt- transfer group, was trained on the $1 dimension and then abruptly switched to the $2 dimension. The performance of group 3 serves as a control for de- termining the amount learned about the 82 dimension by either compound pro- cedure. Thus if group 1's or group 2's performance was not different from group 3's performance then it may be concluded that that compound procedure failed to transmit information about the $2 dimension. METHOD Subjects: Twenty-four locally obtained White Carneaux pigeons served as subjects. The birds were experimentally naive and were maintained at 80% ad libitum weight throughout the experiment. Apparatus: A standard Lehigh Valley 3-key experimental pigeon chamber was used. Inline Digital Display stimulus projectors transilluminated the cen- ter and left response keys with horizontal and vertical lines, or red and green lights. The centers of these response keys were 8.25 cm apart and 25 cm from the floor of the chamber. The right response key was not used. Con- ventional electromechanical programming apparatus was used to control the experiment and was housed in an adjoining room. A punched paper tape reader was used for program control and the serial order of responses and latencies were recorded on a printout counter. Procedure: Training on the $1 stimulus dimension: Subjects were randomly assigned to one of three groups: stimulus-delay, stim- ulus-compounding, and abrupt-transfer. Each group was comprised of eight birds. All subjects received 20 sessions of simultaneous discrimination train- ing on the $1 dimension. A vertical and horizontal line served as 5+ and S- respectively. A trial began with the presentation of a stimulus on each of the two keys and was terminated when a peck occurred to one of the stimuli. No maximum limit was imposed on the duration of a trial. Correct responses were reinforced by producing 3 sec access to.mixed grain. Responses to the 7 8 incorrect stimulus terminated the trial without reinforcement. Each session was terminated following the delivery of 50 reinforcements. A 5 sec inter- stimulus interval preceded each trial. Responses during this interval pre- vented the onset of the subsequent trial by an additional 5 sec. The exper- imental chamber was dark during the intertrial interval, while during the trial both response keys and the house light were illuminated. The onset of the house light preceded the illumination of the key lights by 340 msec. The discriminative stimuli were presented on the response keys according to a modified Gellerman (1933) series. A repeated trials procedure followed an error. This prevented a change in the stimuli on the keys until a correct response occurred. The presentation of the $2 stimulus dimension: After the completion of 20 sessions on the $1 dimension the subject in all three groups were pre- sented with a simultaneous wave length discrimination in which a red and green light served as the 8+, and 8- respectively. The transfer session con- sisted of 60 trials. The first 10 trials for all three groups were on the original line tilt dimension and provided a warmdup series of trials. The subsequent 50 trials presented the S2 stimulus dimension in a manner specif- ic to each group. Group 1, the stimulus-delay group, was presented the 82 dimension fol- lowing Touchette's procedure. 0n the eleventh trial of the transfer session the SI dimension was superimposed on the $2 dimension in order to form a compound stimulus display. A peck to the correct key increased the interval from trial onset to the presentation of the SI dimension on the next trial by 340 msec. If no errors occurred, on the fourth transfer trial the 82 di- mension was presented for 1.02 sec prior to the superimposition of $1 to form the compound. Thus the interval in which the subject waited for the 9 presentation of the compound was titrated by the sequence of errors and cor- rect responses emitted by the subject. In this procedure, transfer was de- fined by that trial subsequent to the subject's last response to the com— pound. When transfer occurred, the latency from the trial onset to the res- ponse was less than the scheduled duration from trial onset to the formation of the compound; thus the compound was not presented. The session was ter- minated upon the completion of 60 trials. Group 2, the stimulus-compounding group, was presented with the trans- fer session in the following manner. On trials 11 through 39 group 2 re- ceived a conventional compound stimulus display. On trials 40 through 60 group 2 received the $2 stimulus dimension alone. The 29 trials that the compound stimulus was presented to group 2 was determined by the fact that this was the maximum number of trials required for transfer by any subject in the stimulus-delay group. Thus the subjects in group 2 received the com- pound display atleast as many times as the subjects in group 1; the central difference between the procedures was that the compound stimulus for group 2 was not modified by the progressive delay feature. Group 3, the abrupt-transfer group, received the transfer session as follows. 81, the line tilt dimension was presented for 10 trials. 82, the wave length dimension, was presented for the subsequent 50 trials. This group was not presented with a superimposed or compound stimulus display. Training on the $2 stimulus dimension: 0n the session following the transfer session the subjects in all three groups were presented the 82 discrimination following the identical contingencies as the subjects re- ceived during training on the SI discrimination. In the 20 training sessions preceding transfer, in the transfer session, and in the session following transfer the performance of all subjects was 10 assessed by two dependent variables: the latency of response and the per- centage of errors. The latency was recorded up to a hundred multiples of a 340 msec time constant. Thus the maximum latency included in the ensuing analysis was 34 sec. A repeated trials procedure was in effect throughout the transfer session for the subjects in all three groups. RESULTS Performance on the 81 dimension: The subjects in all three groups re- ceived identical training on the $1 dimension. 0n the session preceding trans- fer (the twentieth session of training) one subject responded to the cor- rect key 92% of the time, three subjects were 96% correct, three subjects were 93% correct and the remaining 17 subjects were 100% correct. The ave- rage latencies for this session ranged from 1.4 to 2.7 sec. The subjects were randomly assigned to the three grpous. An analysis of variance of er- rors and of latencies for the twentieth session revealed that there were no significant differences between groups. (F/errors (2, 21)=0.26; F/latencies (2. 21)=2.50). The transfer session: The subjects in all three groups exhibited stim- ulus control on the $2 dimension by the termination of the transfer session, but the manner of this acquisition was quite different between groups. The stimulus-delay subjects transferred to the S2 dimension after receiving 17 to 29 trials to the compound stimulus, with a mean of 22 trials. The num- ber of trials to the compound for this group was determined by each subject and its shift in response latency, while the number of trials to the com- pound for the stimulus-compounding group had to be determined in advance by the experimenter. The position adopted was to present the compound stimulus to the subjects in the stimulus-compounding group the maximum number of trials needed to produce transfer by any subject in the stimulus-delay group. Thus the stimulus compounding group received the compound display 11 12 for 29 trials and was then presented with the $2 dimension for the 21 trials remaining in the session. With these parameters the performance of all sub- jects was compared for their initial 21 trials on the $2 dimension. Figure 1 presents the mean number of errors on the $2 dimension for the subjects in each group. The mean of 0.75 errors for the stimulus-delay group demonstrate that the subjects receiving Touchette's procedure had virtually acquired the S2 discrimination while responding to the compound display. 0f the 8 subjects in this group, 5 made no errors to the $2 dimension, 2 sub- jects made 1 error, and 1 subject made 4 errors. The stimulus-compounding subjects averaged 4.3 errors to the 82 dimen- sion. Only 1 subject in this group was errorless; the remaining 7 subjects made 2. 3. 3. 5, 7, 7, and 8 errors respectively. Thus the progressive delay modification of the compound stimulus as experienced by the stimulus-delay group was successful not only in indicating the point by which the 82 di- mension had acquired stimulus control but in addition had transmitted more information about the $2 dimension than had the regular compound display. The conventional compound stimulus display also facilitated learning the S2 discrimination. This was evidenced by the performance of the abrupt- transfer subjects, who averaged nearly twice as many errors as the subjects in the stimulus-compounding group. Of the 8 subjects in the abrupt-transfer group, 1 subject made no errors, while the remaining 7 subjects made 2, 3, 6, 9, ll, 15, and 18 errors respectively. An analysis of variance for errors in all three groups was significant beyond the 0.01 level. (F/errors (2, 21)=6.12). The develOpment of stimulus control during transfer may also be expres- sed by the latency of the response. Figure 2 presents the mean latency for the 21 trials on the S2 dimension for the subjects in each group. The 13 Figure 1. Mean number of errors for the subjects in each group on the 21 trials to the 32 dimension. l4 8.:- F0 (1! .a\ ,or New. 5“ ‘I I: [w b'wihm'ltzr‘m. . r .‘i. " ff": {1" “5U OF f‘ It 9; £1.- ‘5 \ Moreen-recur! 4-1 -~“*H‘-A Mu. I~l'-I DJ -~—-m-- (35:: (I "if? ‘2 E. 3 'r. (I t. ’L... n *0- cm. ue- MI] I V"..- u-.o\ J-c-m' . .-‘\t' F‘ ' 83: .y by“ .. j m A, P:- 21‘: :J a... .. t 2,. Pg, ’ l in ml 5(- VK‘L' 15 stimulus-delay subjects had the most homogenous distribution of latencies. The mean latency for this group was 2.90 sec. The latency pattern reaffirms the analysis of errors, demonstrating that the stimulus-delay group had ac- quired the $2 discrimination while responding to the compound display. The stimulusccompounding subjects displayed intermediate latencies with a mean value of 5.87 sec. Latencies for this group gradually decreased over trials and by the fourteenth trial the latencies had reached asymptote, ap- proximating the stimulus-delay group's performance. The longer latencies for the first 13 trials reflect that the stimulus-compounding subjects had learned less about the $2 dimension embedded in the compound than had the subjects in the stimulus-delay group. I The latencies for the abrupt-transfer group decreased over trials from 25 sec to the 8 sec range with a mean value of 14.3 sec. The very long lat- encies at the onset of the session again indicate the disruption in stimulus control produced by abrupt transfer. The violation of homoscedasticity in the latency data called for a non- parametric statistical treatment. A KruskaléWallis analysis of variance de- monstrated a significant difference in latencies due to the main effect of groups (12 (2)=l7.12 p<0.001). The corresponding analysis by trials demon- strated that the abrupt-transfer and the stimulus-compounding procedures were significant; (XZ/abrupt (20)=44.55 p<0.001; XZ/compound (20)=35.62 p<0.02) while the latencies for the stimulus-delay procedure did not change as a function of trials (XZ/delay (20)-_-12.44 p