behavioral contrast and inhibitory stimulus control.

Transcription

1 Animal Learning &Behavior 1975, Vol. 3 (4), Behavioral contrast and inhibitory stimulus control JAMES V. COVCH Madison College, Harrisonburg, Virginia 2281 Two experiments examined the presumed relationship between behavioral contrast and inhibitory stimulus control. In Experiment I, pigeons were exposed to mult VI I-min VI L-min or mult VI 5-min VI 5-min during baseline training prior to mult VI I-min VI 5-min discrimination training. Half of the subjects received a timeout (TO) component during baseline in order to reduce the degree of contrast during discrimination training. Only 3 of 8 subjects receiving the TO showed contrast while all other subjects showed various degrees of contrast. Postdiscrimination generalization gradients indicated excitatory rather than inhibitory control by the stimulus associated with the VI 5-min schedule. During baseline training in Experiment II, responding to all the generalization stimuli was reinforced. In addition, some subjects received the TO stimulus. The subjects were next exposed to mult VI I-min EXT, mult VI I-min VI 5-min, or just the VI 5-min component. Generalization gradients indicated inhibitory control by the stimulus associated with EXT or VI 5-min for 19 of 2 subjects even though some subjects did not show contrast. These results question the presumed relationship between behavioral contrast and inhibitory stimulus control. It has been suggested (Farthing & Hearst, 1958; Terrace, 1966, 1968, 1971; Weisman, 1969; Yarczower, 197) that the OCCurrence of behavioral contrast is a necessary antecedent for the development of inhibitory stimulus control. For example, Weisman (1969) reported V-shaped postdiscrimination generalization gradients (a presumed correlate of inhibitory stimulus control) around a stimulus associated with a variable-interval (VI) 5-min schedule of reinforcement for only those subjects exhibiting behavioral contrast during multiple (mult) VI l-min VI 5-min discrimination training. If a relationship exists between contrast and inhibitory stimulus control, then manipulations which modulate the degree of behavioral contrast should also modulate the degree of inhibitory stimulus control. One procedure that has been successfully used to produce and modulate behavioral contrast is timeout from reinforcement (TO), which consists of darkening the operant chamber and response key for a specified duration. Reynolds (1961), Sadowsky (1973), Taus and Hearst (197), and Wieth and Rilling (1972) have found that behavioral contrast will occur during one component of a multiple schedule if that component is alternated with a TO component. The TO procedure coupled with the Friedman and Guttman (1965) finding that if behavioral contrast is produced in one discrimination phase then no further contrast is evident in a subsequent discrimination phase, leads to an experimental design which allows for the modulation of behavioral contrast and the later determination of inhibitory stimulus control. Therefore, These studies are based on a dissertation submitted to the Graduate School of the UnivelSity of Massachusetts in partial fulfillment of the requirements for the PhD degree. This reaeareh was supported by an NIMH Experimental PsycholoilY Training Grant awarded to the Department of Psychology and by NIMH Grant MH awarded to John W. Donahoe. The author wishes to acknowledge his indebtedness to John W. Donahoe. under whose supervision this research was conducted. Reprints may be obtained from the author. Department of Psychology. Madison College. Harrisonbuzg, Virginia it was the purpose of Experiment I to employ TOs during baseline training in an attempt to reduce the degree of contrast normally concomitant with the initiation of discrimination training. Inhibitory stimulus control, as measured by a postdiscrimination generalization test, could then be related to the degree of contrast evident in discrimination training. EXPERIMENT I Method Subjects. Sixteen White Carneaux pigeons, at least six months old, were used as SUbjects. After introduction into the laboratory, the subjects were housed individually and were given several days of free food in order to obtain stabilized body weights. All subjects were then deprived offood and reduced to 75% of their free-feeding weight. Each SUbject was maintained at this deprivation level throughout hte experiment. Apparatus. Four identical Lehigh Valley Electronics pigeon operant chambers and accompanying sound attenuating hulls were used. Only the rightmost of two keys mounted on the front wall was operative. A force of approximately 1 g was required to operate the response key. Reinforcement consisted of 4-sec access to a grain hopper through a centrally located aperture in the front wall. White noise at a sound level of 85 db was delivered through a speaker also mounted on the front wall and to the left of the feeder aperture. The different stimuli that transilluminated the response key were generated by an Industrial Electronics Engineers display cell located behand the response key. Six orientationsofa white line (9, -6, -3,, +3, and +6 deg from vertical) and a green background could be projected by the display cell. Line orientations superimposed upon a green background were obtained by illuminating a lamp behind a green Kodak Wratten filter and a lamp for the specific line orientation. Brightness differences due to the illumination of two lamps (line orientation and surround) as compared to illumination of only one lamp (green surround alone) were eliminated by the addition of neutral density filters between each line orientation and its light source. Procedure. On the first day, the SUbjects were habituated to the operant chamber for approximately IS min. On the following 2 days, the SUbjects were trained to eat from the feeder with 3 feeder presentations given daily. The house light 347

2 348 COUCH 1r , BlANK PI a LINE Z :f -, :4 z g2 4 P2 After baseline training, all groups received 14 days of discrimination training with a mult VI l-min VI 5-min schedule of reinforcement. The VI l-min schedule was correlated with SI while the VIS-min schedule was correlated with S2. Each session consisted of a random order of 14 Sl and 14 S2 stimuli each presented for one min with a 5-sec TO separating successive: stimuli presentations. Following discrimination training, all subjects received a generalization test in extinction during which the six orientations of the white line, each on a green background, were presented along with the green background alone. The stimuli were each presented 1 times in randomized blocks of seven 3-sec trials, each separated by a 5-sec TO. Four reinforced (VI l-min) one-minute presentations of the SI stimulus preceded the generalization test. Results Baseline training VI I-min conditions. The response rate in the presence of the O-deg and blank green stimulus on each day of baseline training for each SUbject in the mult VI l-min VI l-min condition is shown to the left of the vertical line in Figure 1. Similarly, Figure 2 indicates the response rates for each subject of the mult VI l-rnin VI l-rnin TO condition. The mean response rate for the last three baseline sessions averaged across the S1 and S MUll VI-lVI-I SESSIONS Figure 1. Mean rate of responding per subject for the 51 and 82 stimuli on each session of mult VI I-min VI l-min baseline training and on each session of mult VI I-min VI S-min discrimination training of Experiment I. provided the only illumination during these three sessions. On the fourth day, the subjects were trained to peck the green stimulus key by the method of successive approximations. Following keypeck training, the subjects were given one day of continuous reinforcement with 25 reinforcements given for responses to the blank green stimulus (SI) and 25 reinforcements given for responses to the green O-deg line tilt stimulus (S2). On the following 2 days, a fixed ratio of increasing length (maximum FR 33) was in effect and approximately 4 reinforcements, equally divided between the Sl and S2 stimuli, were obtained. On the next day, the subjects were reinforced for responding to the SI and S2 stimuli. Fourteen SI and S2 stimuli presentations were randomly scheduled each day. The eight subjects which were to receive mult VI l-min VI l-min training were given two mult VI 3-sec VI 3-sec sessions before the introduction of the VI I-min schedules. The eight subjects which were to receive mult VI 5-min VI 5-min baseline training received preliminary training on a progressive series of VI schedules [VI 3-sec(2 sessions), VI l-min (2 sessions), VI 3-min (3 sessions)j. All subjects were then given 2 sessions of baseline training on the appropriate multiple schedule. Half of the subjects from each of the above conditions were presented with a third stimulus, the TO. The TO consisted of darkening the response key and the chamber for a l-min duration. In all conditions a stimulus was presented for I min with a 5-sec TO separating the successive stimuli presentations. That is, for those subjects receiving a TO, each session consisted of a randomized order of 14 presentations of S1, 14 presentations of S2, and 14 TOs I. MUll VI-I VI-1 TO MUlT VI-1 VI-5 SESSIONS Figure 2. Mean rate of responding per subject for the 81 and 82 stimuli on each session of mult VI I-min VI l min TO baseline training and on each session of mult VI I-min VI S-min discrimination training ofexperiment L

3 BEHAVIORAL CONTRAST AND INHIBITORY STIMULUS CONTROL 349 components for all subjects indicated that the VI I-min VI l-min TO condition generated a higher response rate (71.33 responses per min) than the mult VI I-min VI l-min condition (38.37 responses per min). VI 5-min conditions. The response rate in the presence of the O-deg and blank green stimulus on each day of baseline training for each subject in the mult VI 5-min VI 5-min condition is shown to the left of the vertical line in Figure 3, with similar data being shown in Figure 4 for each mult VI 5-min VI 5-min TO trained subject. As had occurred with the mult VI I-min VI l-min baseline training, the mean rate of responding over the last three training days for the mult VI 5-min VI 5-min TO subjects (82.5 responses per min) was higher than the mean rate of responding for the mult VI 5-min VI 5-min subjects (47.46 responses per min). Since the above comparisons between TO trained subjects and non-to trained subjects are between subjects comparisons, the response rates for all subjects over the final three baseline training days were submitted to an analysis of variance. The results of the analysis indicated that the response rate for subjects receiving a TO during training was significantly higher ẕ 15 P13.!tANK LINE 3 z o ISO PIS,OO SO 8 Pl '9 LANk DOG LINE ' 18 2 ' '1 14 MUlTYI-SVI-.5fOSESS,ONS MUlTYI-JVI-5 Figure 4. Mean rate of responding per subject for the 81 and 82 stimuli on each session of mult VI 5-min VI 5-min TO baseline training and on each session of mult VII-miD VI S min discrimination training of ExperimentI. 15,, ':: 75'" ::.r--.r 75 SO " MUlT VI-5 VI-5 SESSIONS MUlT VI-' VI-,S FiguJe3. Mean rate of respondins per subject for the 81 and 82 stimuli on each session of mult VI 5-min VI S min baiiine training and on each session of mult VI I-min VI S min discrimination trainingof ExperimentI. than the response rate for subjects not receiving the TO, [F{1/12) = 13.63, p <.5]. Discrimination Training The data plotted to the right of the vertical line in Figures I, 2, 3, and 4 indicates the response rate in the presence of the 81 and the 82 stimulus for each subject in each group during the discrimination training sessions. VI l-min conditions. The response rate of subjects shifted from mult VI l-min VI l-min to mult VI l-min VI 5-min increased an average of 1.91 responses per min during 8 I, while a mean decrease of responses per min occurred during the 82 stimulus. Since the response rate increased during SI and decreased during 82 over the course of discrimination training, Reynolds' (1961) definition ofbehavioral contrast was satisfied. When the SI response rates during baseline and discrimination training were compared for the mult VI I-min VI l-min TO condition, it was found that two subjects increased their 81 response rate while two subjects decreased their 81 response rate. Considering the 82 response rates, all four subjects showed a decline, the decline being an average of responses permin. Thus, behavioral contrast was eliminated in two subjects

4 35 COUCH 2 2 "''A "'1 : 5. Ż "' ọ..'j '3 J 1 ;.»:: 5 V -- o 1, ' LINE OIIENTATION Figure 5. Postdiscrimination line orientation gradients fot each subject of the mult VI I-min VI I-min baseline mult VI I-minVI 5-min discjimination conditionof Experiment I. through the use of the TO procedure during baseline training. VI 5-min conditions. The response rate for subjects shifted from mult VI 5-min VI 5-min to mult VI l-min VI 5-min increased during presentations of SI and remained elevated, relative to baseline, throughout discrimination training. Considering the response rate to S2 during discrimination training,a mean increase of responses per minute was observedin three ofthe four subjects. The response rate to S2 for the remaining subject decreased slightly with the introduction of discrimination training. Considering the S1 response rates during baseline and discrimination training for the mult VIS-min VI 5-min TO condition, it was found that the SI response rate increased and remained elevated for three of the four subjects. The Sl response rate for the remaining subject declined and was below the terminal baseline training level by the end of discrimination training. With regard to the S2 response rate, an increase in response rate was evident for two subjects while a decrease in S2 response rate was apparent for the remaining two subjects. However, for only one of the latter two subjects was there a concurrent response rate increase to the Sl stimulus. Therefore, only one subject from the mult VI 5-min VI 5-min TO baseline condition satisfied the requirements for behavioral contrast. In summary, behavioral contrast during discrimination training was evident for all subjects given mult VI l-min VI l-min training, while only two of the mult VI l-rnin VI l-min TO subjects exhibited contrast. For the mult VI. 5-min VI 5-min condition, all of the subjects increased their response rates during SI, with three of the four subjects also increasing their response rate during S2 (positive induction). The S2 response rate for the fourth subject decreased slightly. For the mult VI S-min VI Smin TO condition, one subject exhibited: behavioral contrast. Discrimination Performance An analysis of variance-using Sl percentage scores [Sl responses/(si +S2 responses)] over the last three discrimination sessions indicated a reliable groups effect, [F(3/12} = 4.95, P <.25]. When the groups effect was partitioned, it was found that subjects receivingmult VI l-min VI l-min baseline training differed significantly from subjects receiving mult VI S-min VI S-min baseline training [F(1/12) = 5.6, P <.5], and that the reinforcement schedule during baseline training interacted with the TO [F(1/l2) = 5.42, P <.5]. Considering the group means, 73% of the total responses for the mult VI l-min VI l-min condition were emitted during Sl while only 59%, 59%, and 58% of the 5 HO 15!15 :z: 1 "' so z : A (9" (2I. /\ ('Il ISO. HO 15(f o A IS:.I -' LINE OIIENTATION Figure 6. Postdiscrimination line orientation gradients for each subject of the mult VI t-min VI I-min TO baseline mult VI I-min VI S-mindiscrimination condition of Experiment I.

5 BEHAVIORAL CONTRAST AND INHIBITORY STIMULUS CONTROL 351 responses were during SI for themult VI I-min VI I-min TO, mult VI 5-min VI 5-min TO, and mult VI 5-min VI 5-min conditions, respectively. Therefore, the discrimination performance was retarded for subjects who received either a TO or VI 5-min during baseline sessions. In addition, the discrimination results of the mult VI l-min VI l-min condition are in exact accord with the results reported by Weisman (I969). Stimulus Control Figures 5, 6, 7, and 8 indicate the generalization gradients for all subjects of the mult VI l-rnin VI l-min, mult VI l-mm VI l-rnin TO, mult VI 5-min VI 5-min, and mult VI 5-min VI 5-min TO conditions, respectively. From inspection of these figures it is evident that the postdiscrimination generalization gradients were not U-shaped, as would have been expected from Weisman's (I969) results. Similarly, with respect to the presence or absence of behavioral contrast, there was no relationship between the degree of behavioral contrast and the shape of the postdiscrimination gradient. Discussion From the generalization gradients, it is clear that none of the conditions generated gradients that would be taken as indicative of inhibitory stimulus control. Rather, all of the gradients obtained were decremental '/\ 2o ISOI 1 'A. ż o (72) 3 :z: v 15 ọ 'A 3 :: 2 ":,,. 1 ọ -- 'J\ 3 2 o,:,) PO I LINE O.lunATIO'" Figme 7. Postdiscrlmination line orientation gradients for each subject of the mult VI 5-OOn VI 5-min baseline mult VI I-min VI S min discrimination condition of Experiment I :::')\.:. o 1 '1\':' 6'/\{S,.! 4 :z: 2. ll..--, Figure 8. Postdiscrlmination line orientation jladients for each subject of the mult VI S-min VI S-minTO baseline mult VI I-min VI S-mindiscrimination condition of ExperimentI. with the gradient peak at or near the S2 stimulus. It is similarly clear that the gradients reported by Weisman (1969) obtained after mult VI l-min VI I-min changed to mult VI l-min VI 5-min training were U-shaped, indicating inhibitory stimulus control. Since some of the experimental conditions from Experiment I were a replication of Weisman's investigation and since the baseline and discrimination performance from the present experiment closely approximates comparable data from the Weisman investigation, the reason for the absence of V-shaped gradients in the present case is unclear. Since behavioral contrast was observed in all of the mult VI I-min VI l-min subjects and for other individual subjects in the remaining conditions, and since inhibitory gradients were not obtained for any of these subjects, the presence of behavioral contrast as a sufficient condition for inhibitory stimulus control must be denied. EXPERIMENT II The generalization gradients obtained in Experiment I indicated that the frequency of responding occasioned by the VI 5-min reinforcement schedule in effect during

6 352 COVCH CJ r r - 3 BlAN", D line :: ' VI-l V 1-5 ' ' SESSIONS F;gure 9. Mean rate of responding per subject for the 81 and 82 stimuli on each session of VI l-min baseline training and for the 82 on each session of VI s-min single stimulus training of Experiment Il, presentations of the O-deg line tilt stimulus was greater than the generalized responding to the other line tilt stimuli. this suggests one possible procedure which could be employed to increase the probability of observing V-shaped generalization gradients. That is, the response rate for each line tilt stimulus could be elevated above the baseline level through the use of reinforcement instead of through the weaker process of generalization from the orthogonal dimension. Though this procedure, the occurrence of a decrease in the response strength for the S2 stimulus might be more easily detected. Specifically, if during baseline training responding was reinforced in the presence of each of the generalization test stimuli, then the response tendency at all points of the line tilt dimension would be greater as compared to baseline training with only the SI and S2 stimuli presented. When discrimination training is then instituted, any decrease in responding produced by the reinforcement schedule associated with the 82 stimulus should be observed when a postdiscrimination generalization test is administered. The above procedure has been used by Honig (1961) and Weisman and Palmer (1968). In the latter investigation, extremely steep V-shaped generalization gradients were obtained following mult VI l-min ext discrimination training. It was, therefore, the purpose of Experiment II to examine further the relationship between the occurrence of behavioral contrast and the development of inhibitory stimulus control. As in Experiment I, the TO stimulus was employed in baseline training for some subjects so as to.reduce the degree of behavioral contrast during discrimination training. Method Subjects. Twenty White Carneaux pigeons, at least six months old, Were used as subjects. The subjects were housed individually and were placed on a maintenance schedule that was identical to the maintenance schedule for the SUbjects of Experiment I. Apparatus. The apparatus was identical to that employed in. Experiment I. Procedure. PrelimiTlllry training. As in Experiment I, the subjects were habituated to the operant chamber, feeder trained, trained to peck the response key which was illuminated with the green stimulus, and finally given experience with a fixed ratio schedule of reinforcement. All procedural details are common with those of Experiment I. Baseline training. Following two VI 3-sec sessions, all subjects were given 2 daily VI l-min baseline sessions during which four presentations of each of the seven stimuli (, ±3, ±6, and 9 deg, and green alone) were scheduled. All subjects were exposed to a random order of the seven generalization test stimuli each day. Two random orders of stimuli presentations were used. In addition, eight subjects (TO trained) also received 14 TO stimuli which were inserted in the same ordinal position of the daily sessions as in Experiment I P P 22 BLANK D O LIN! VI! a 2 SfSIONS Figure 1. Mean rate ofresponding per subject for the 81 and 82 stimuli on each session of VI I-min baseline trainins and on each session of mult VI I-min ext discrimination training of Experiment 11.

7 BEHAVIORAL CONTRAST AND INHIBITORY STIMULUS CONTROL 353 Single stimulus trammg. Following baseline training, four subjects which had experienced VI l-min baseline training were given 14 sessions of single stimulus training in the presence of the O-{ieg line imposed on the green background. Each daily session consisted of 14 one-min stimulus presentations separated by a 5-sec TO, during which the VIS-min reinforcement schedule was operative. This condition served as a comparison oondition against which the effects of the mult VI I-min VI 5-min schedule could be evaluated. Discrimination training. The remaining 16 subjects were given discrimination training with either a mult VI I-min ext or a mult VI l-rnin VI 5-min schedule. Of the eight subjects who received VI l-min baseline training, four subjects were assigned to the mult VI I-min Ext discrimination condition while four subjects were assigned to the muit VI l-min VI 5-min condition. Similarly, four of the previously trained VI l-min TO subjects received the mult VI I-min ext discrimination, while the remaining four subjects received the mult VI l-min VI 5-min discrimination. Discrimination training was conducted for 14 sessions with the same daily stimulus order as employed in Experiment I. Following discrimination training, all SUbjects received a generalization test which was in all ways identical to the generalization test administered at the termination of Experiment I. Results P 3 p ' M 2 Baseline Training The mean daily response rate in the presence of the 81 and the 82 stimuli for each subject of the five 12 6 Pl J8 2 '2 4 6 a JO v i-: TO MUll VI-j EXT SESSIONS Figure II. Mean rate of responding per subject for the SI and 82 stimuli on each session of VI I-min TO baseline training and on each session of mult VI L-min ext discrimination training of Experiment II. '2 4 6 a Vl-l MUll VI-}vl-5 sessions Flgure 12. Mean rate of responding per subject for the SI and 82 stimuli on each session of VI L-min baseline training and on each sesslon of mult VI I-min VI S-min discrimination training of Experiment 11. conditions is shown to the left of the vertical line in Figures 9,1,11,12, and 13. In contrast to the findings of Experiment I, the terminal response rate for the subjects receiving the TO was not significantly elevated relative to the non-to trained subjects (F < 1). Though the difference between the TO and non-to conditions did not approach statistical reliability, the mean response rate for the TO condition (73.25 responses per min) was higher than the mean response rate for the non-to condition (64.9 responses per min). When a comparison of the mean response rate for the VI l-min TO conditions of Experiment I and the TO conditions of Experiment II was made, it was found that the mean rate was essentially the same for the two experiments (71.33 and responses per min, respectively). When a similar comparison of the mean response rate for the non-to trained subjects of the two experiments was made, it was found that the mean response rate from Experiment I was lower than the mean response rate of Experiment II (38.37 and 64.9 responses per min). The conclusion becomes, then, that with the addition of either the TO stimulus or the occurence of multiple stimuli during baseline training, the response rate is elevated above that level exhibited by subjects who receive only the 81 and 82 stimuli.

8 354 COUCH eo P 3". &lank a OOLINE J4 '11-1 TO MUl T Vl-1 '11-,5 SESSIONS Figure 13. Mean rate of responding per subject for the 81 and 82 stimuli on each session of VI I-min TO baseline training and on each session of mutt VI I-min VI 5-min discrimination training of Experiment II. Single Stimulus Training To the right of the vertical line in Figure 9 is plotted the mean daily response rate for each subject during presentations of the O-deg line tilt stimulus during which a VI 5-min schedule of reinforcement was operative. Over the first several VI 5-min sessions, the effect of a reduction in reinforcement density was to lower the response rate. However, by the end of the 14 daily sessions, the response rate for each subject had increased to a level similar to that evident at the termination of VI l-min baseline training. When the terminal VI I-min response rate was compared with the terminal VI 5-min response rate, it was found that the response rate had decreased 7.5 responses per min by the end of the VI 5 min training. This decrease was not statistically significant, t(3) = Discrimination Training mutt VI I-min ext conditions. The daily mean response rate for the Sl and S2 stimuli for each subject during mult VI l-min ext discrimination training is depicted to the right of the vertical line in Figure 1. Relative to the VI I-min baseline, the response rate of all subjects both increased during presentations of the Sl stimulus and decreased during presentations of the S2 stimulus. Subtraction of the mean response rate of the last three sessions of VI l-min training from the mean response rate of the last three sessionsof mult VI I-min ext. training for both the SI and S2 stimulus indicated that the mean response rate during the SI stimulus increased 33.5 responses per min while the mean response rate during the S2 stimulus decreased responses per min. It is apparent, then, that behavioral contrast was evident in this condition. During the mult VI l-min ext discrimination sessions following VI I-min TO baseline training (see Figure 11), the 82 response rate decreased for all subjects while the S2 response rate for three of the four subjects increased an average of responses per min relative to baseline. For the remaining subject, the mean 81 response rate across the last three discrimination sessions had decreased very slightly relative to the terminal baseline response rate. Considering the S2 stimulus, it was found that the response rate for all subjects decreased during the 82 stimulus over the course of discrimination sessions. The mean S2 response rate relative to baseline decreased responses per min. Therefore, during the mult VI I-min ext discrimination, all of the non-to trained subjects and three of the four TO trained subjects developed behavioral contrast. mult VI I-min VI 5-min conditions. The daily mean response rate for the 81 and 82 stimuli for each subject during mult VI I-min VI 5-min discrimination training following VI I-min training is shown to the right of the vertical line in Figure 12. In the analysis for behavior contrast, it was found that two of the four subjects showed evidence of behavioral contrast, e.g., a increase in mean response rate during SI and a decrease in mean response rate during S2. Of the remaining two subjects, one subject showed a decrease in both the 81 and 82 response rate, while the other subject increased both the SI and 82 response rate. During mult VI I-min VI 5-min discrimination training following VI l-min TO baseline training (see Figure 13), only one subject showed behavioral contrast. Of the remaining three subjects, one subject increased both the 81 and 82 response rate, while two subjects decreased both the 81 and 82 response rate. Therefore, for the mult VI I-min VI 5 rnin discrimination condition, two of the non-to trained subjects and one of the TO trained subjects developed a moderate degree of behavioral contrast, while the remaining subjects either increased or decreased their response rate during both the 81 and 82 stimuli. Stimulus Control Figures 14, 15, 16, 17, and 18 depict the generalization gradients for all subjects from each of the five experimental conditions of Experiment II. It is

9 BEHAVIORAL CONTRAST AND INHIBITORY STIMULUS CONTROL 355 evident from inspection of the figures that 18 ofthe 2 subjects employed in Experiment II produced U-shaped generalization gradients during the post-discrimination generalization test. In regard to the relationship between behavioral contrast and inhibitory stimulus control, it should be remembered that one subject of the mult VI l-min ext condition and five subjects of the mult VI l-min VI 5 min condition failed to show behavioral contrast during discrimination training. However, the generalization gradients for the mult VI l-rnin ext subject and four of the five mult VI l-min VI S-min subjects! were in no fundamental way different from the gradients exhibited by subjects who did show behavioral contrast. This result is taken as further evidence against the hypothesis relating behavioral contrast and inhibitory stimulus control. The occurrence of a. If-shaped generalization gradient does not necessarily mean inhibitory stimulus control. For instance, it is possible that the subject would respond in such a way as to produce a U-shapedgradient prior to discrimination training. The present experimental design allows for a test of this hypothesis. That is, since the baseline training procedure entailed presentation of the same stimulus order as was presented o 2 '" zo 8 '" Z V'" P 17 <»: PI8 V":,, 1 P19 (H 1) ----v P I LINE ORIENTATION Figure 14. Postdiscrimination line orientation gradients for each subject of the VI I min baseline VI S-min single stimulus training condition of Experiment D. P 21 9,v Z :::; 6 % U o (t 1J :: O '" (372) ISO '" GI: 1 -< o v.roo', 2 1 SO O Io:H (33) 2)J 1 P2A Figure 15. Postdiscrimination line orientation gndients for each subject of the VI I-min baseline mwt VI I-min ext discrimination condition of Experiment It during generalization testing, a gradient of response tendency can be obtained from the terminal baseline sessions. The gradient of response tendency can then be compared with the postdiscrimination generalization gradient and the effect of the discrimination procedure ascertained. In Figure 19 the mean postdiscrimination gradient is presented along with the mean gradient obtained from the last two baseline sessions. In order to test for the reliability ofthe observed differences, the mean response rate emitted during each line tilt stimulus for both the last two baseline sessions and the generalization test was submitted to an analysis of variance. Considering the gradients from the mult VI I-min ext and mult VI I-min VI 5-min conditions, the analysis indicated that the gradients obtained during the generalization test differed significantly from the baseline gradients [F(5/6) = 13.87, p <.1]. Furthermore, this difference was modulated by whether the reinforcement schedule during S2 was extinction or VI 5-min. That is, the postdiscrimination generalization gradients showed a greater degree of depression around the S2 stimulus for subjects receiving extinction during S2 presentations than for SUbjects receiving VI 5-min during S2 presentations [F(5/75) = 3.3, P <.25]. The

10 356 COUCH y":" 1 5 P25 -: 2 Z ::E: 1 ụ. P26 ("i) 8 Z 6 _. AOO. 2 2V (.:" 1 P28 reinforcement was delivered more densely relative to the S2 schedule. GENERAL DISCUSSION The major finding of the present experiments can be summarized as follows: Behavioral contrast is neither a necessary nor a sufficient condition for the development of inhibitory stimulus control.' That contrast is not a necessary condition is seen from the results of Experiment II, where steep incremental U-shaped generalization gradients were obtained from subjects showing little, if any, behavioral contrast during discrimination training. Likewise, the results of Experiment I indicate that behavioral contrast cannot be taken as a sufficient condition for inhibitory stimulus control, since the mult VI l-min VIS-min condition, while showing behavioral contrast during discrimination. training, showed no inhibitory control when a generalization test was conducted. The present findings would necessarily call into question those analyses that depend upon an invariant relationship between behavioral contrast and the development of inhibitory stimulus control. A further conclusion of the present experiments concerns the usefulness of the TO as an aid in y,,:. Figure 16. Postdiscrimination line orientation gradients for each subject of the VI I-min to baseline mutt VI I-min ext discrimination condition of Experiment II. TO effect did not interact significantly with the above 82 schedule difference (F < 1). Of particular interest in regard to the antecedents of U-shaped generalization gradients is the change in the gradient shape for subjects receiving single stimulus VI 5-min training as compared to subjects receiving muit VI I-min VI 5-min discrimination training. Considering the single stimulus VI 5-min condition, it was found that the gradient following single stimulus training did not differ significantly from the gradient obtained for the final two VI I-min baseline sessions [F( 5/75) =2.8]. In contrast, the postdiscrimination generalization gradient for subjects receiving the mult VI l-min VI S-min discrimination did differ significantly from the baseline gradient [F(S/7S) = 2.71, p <.5]. Since the gradients for the single stimulus VI S-min subjects did not change significantly due to the VI 5-min training, the effect ofa reduction in reinforcement density from VI l-min to VI 5-min during the subjects' stimulus was not the sole antecedent for the gradient change observed in the mult VI l-min VI 5 min conditions. It would appear, then, that the occurence of U-shaped generalization gradients around the 82 stimulus following mult VI I-min VI 5 min discrimination training was due primarily to the presence of the orthogonal SI stimulus during which 2 1 P29 Z 5 '.('till r u.. '" P 3 '"Z '" 2. 1 V,,: P 31 \f!":' 6. 2 P a LINE ORIENTATION Figure 17. Postdiscrimination line orientation gradien1li for each SUbject of the VI I-min baseline mutt VI I min VI S-min discrimination condition of Experiment II.

11 BEHAVIORAL CONTRAST AND INHIBITORY STIMULUS CONTROL 357 investigating behavioral contrast and other byproducts of discrimination learning. Since the TO effect did not significantly interact with the measure of stimulus control, the TO effect appears to be specific to the training phase. Through the use ofthe TO in a variety of experimental arrangements, the antecedents of behavioral contrast might be more meaningfully determined. In considering the results from the two experiments presented here, an obvious question arises: Why were Uehaped gradients obtained in Experiment II and not in Experiment 11 The answer to this question certainly involves the baseline procedures which differentiated Experiment II from Experiment I. But, in considering the baseline procedures, the question also arises whether U-shaped generalization gradients should be taken as indicative of an inhibitory process or should be viewed simply as the result of a reduction in excitation or response rate, a suggestion originally made by Skinner (1938, p. 17). Briefly, an argument could be made that the antecedent of a U-shaped generalization gradient lies in the conditioning of a response rate or an inter-response time (IRT) in the presence of the S2 stimulus. If the rate of the conditioned operant emitted during the S2 is lower than the conditioned (or generalized) rate emitted AS 3 Z x 15 u -c '" 2 '"z '" 5 AOO "- IS P 33 I'" v"" P34 v"'" P 35,,:" P o 3 6 9<1' LINE ORIENTATION Figure 18. PostdUcrimination line orientation gradients for each subject of the VII-min to baseline mutt VlI min VIS-min discrimination condition of Experiment IL VI-l-VI-5 5 VI-)-VI-l EXT v VI-l TO. VI-l EXT e 2. ż 5 VI-I- VI-l VI-S I \j VI-I TO-VI-I VI-S S ' I LINE ORIENTATiON Figure 19. Relative generalization gradients for all five conditions on the last two baseline sesions (triangles) and on the generalization test session (circles) of ExperimentIL' to surrounding continuum stimuli, au-shaped generalization gradient results. If the rate of the conditioned operant emitted during the S2 is higher than the conditioned (or generalized) rate emitted to surrounding continuum stimuli, a gradient peaked at S2 results. For example, after baseline training in Experiment II, the response rate was high (short IRTs) across the entire line tilt continuum. With the introduction of the VI 5-min schedule during S2 and the extinction effects due to the decreased reinforcement density, the IRTs become longer and the response rate decreases. With a larger percentage of long IRTs being emitted, the probability of a long IRT immediately preceding reinforcement delivery is increased. Since an IRT has been shown to have the characteristics of an operant (Wilkie & Pear, 1972; but see Reynolds & McLeod, 1971), reinforcement delivered following the emission of a long IRT would further increase the probability of long IRTs in the presence of the S2 stimulus. When a generalization test is given, an observed U-shaped gradient might be the result ofthe generalized tendency to emit an IRT that was conditioned during S2 presentations. That is, as the stimulus distance from the S2 increases, the frequency of the IRT associated with the S2 decreases, and the response rate increases. Since

12 358 COUCH the IRT conditioned during the 82 stimulus is longer than the IRTs characteristic of the adjacent stimuli, the resulting generalization gradient is If-shaped. This account of an incremental If-shaped gradient makes no assumptions concerning a separate inhibitory process. Instead, the account is based solely in terms of the response rate or IRT that is conditioned during presentations of the 82 stimulus and relationship between the 82 response rate and the response rate associated with the other stimuli. While the data of the present experiment are suggestive and could be interpreted according to this model, the data do not necessarily rule out an inhibition account. Further investigations are needed which analyze the IRT distributions generated when the different generalization test stimuli are presented. These IRT distributions could then be compared to IRT distributions obtained following the initial training phase. The results of such comparisons could possibly lead to the ultimate rejection of the concept of inhibition as an explanatory term for the occurrence of U-shaped generalization gradients. REFERENCES Farlhina, G. W., " Hearst. E. Generalization cradients of inhibition after different amounts of tmininc. Jo:unal of the Experimental Analysis of Behavior, 191;8, Friedman, H., " Guttman, N. Further analysis of the various effects of discrimination traininc on stimulus ceneralization gradients. In D. I. Mostofsky (Ed.), Stimulus generalization. Stanford, Calif.: Stanford Univemty Press PP Honig, W. K. Generalization of extinction on the spectral dimension. PsYcholoeical Record. 1961, Reynolds., G. S. Behavioral CODtrast.Journal of the Experimental Analysis of Behavior Reynolds. G. S., " McLeod. A. On the theory of interresponse-time reinforcement. In G. Bower" J. T. Spenc<:! (Eds.). The PlYcholoffY o( learning and motillqtion. Vol. 6. New York: Academic Press PP Sadowsky. S. Behavioral contrast with multiple VI - time-,",t trainin. Paper presented at Eastern Psycholoeical Associatic",. Atlantlc City Skinner. B. F. The behavior ot oreantsm», New Yor:,c: ApPleton-CenturY-Crofts, Taus, S. E. " Hearst. E. Effects of intertrial (blackout) durati"n on response rate to a positive stimulus. Psychonomic Science Terrace. H. S. Stimulus control. In W. K. Honia. Opera"t behavior: Areas o( research and appucation. New York: Appleton-Century-Crofts pp Terrace, H. S. DiscJimination leamina. the peak shift and behavioral contrast. Journal of the Experimental Analysis of Behavior, , Terrace, H. S. Escape from S. Leaminc and Motivation , Terrace, H. S. By-products of discjimination learnidj. In J. Spence and G. Bower (Eds.), The PlYchology of learning and motivation; VoL 5. New York: Academic Press P;, Weisman, R. G. Some deerminants of inhibitory stimul\-'s control. Journal of the Experimental Analysis of Behavior Weisl;nan. R. G., " Palmer. J. A. Factors influencinc inhibitory stimulus control: DiscJimination trainina and pric. -: non-differential reinforcement. Journal of the Experimentr: Analysis of Behavior Weith. A., " Rillinc. M. Comparison of time-out and extinction as determinants of behavioral contrast: An analysis of sequential effects. Psvchonomic Science , Wilkie. D. M., " Pear, J. J. Intermittent reinforcement of a::l interresponse time. Journal of the Experimental Analysis c Behavior, 1972, 17, Yarczower, M. Behavioral contrast and inhibitive stimulus control. Psychonomic Science NOTES 1. The fifth subject of the mult VI I-min VI 5-min condition failed to respond differentially to the Sl and S2 stimuli durin& discjimination training and therefore responded equally to all stimuli during the postdiscjimination generalization test. 2. It should be mentioned that the definition of behavioral contrast used in the present experiments only identifies gross occurrences of the phenomena. Further analyses of behavioral contrast would benefit by a more refined definition which might take into account possible ceiling and floor effects. (Received for publication July (Revision accepted May )