DIFFERENTIATION LEARNING AND AUDITORY GENERALIZATION IN NORMAL AND PREFRONTAL DOGS AFTER EXTENSIVE AVOIDANCE TRAINING

Transcription

1 ACTA NEUROBIOL. EXP. 1981, 41: DIFFERENTIATION LEARNING AND AUDITORY GENERALIZATION IN NORMAL AND PREFRONTAL DOGS AFTER EXTENSIVE AVOIDANCE TRAINING Danuta KOWALSKA, Kazimierz ZIELINSKI and James BRENNAN Department of Neurophysiology, Nencki Institute of Experimental Biology Pasteura 3, Warsaw, Poland Key words: avoidance, CS termination, differentiation learning, auditory generalization, prefrontal lesions, dogs Abstract. The continuation of a study of 24 dogs in the stages of avoidance reacquisition, differentiation training and generalization testing is reported. Subjects had previously received avoidance training with either response contingent CS termination on all trials or on only 50 /o of the training trials, prior to the administration of medial or lateral prefrontal lesions in 16 subjects. The remaining 8 subjects served as nonoperated controls. Following the appropriate surgical treatment, subjects were tested twice for generalization along the tonal frequency dimension of the avoidance CS. Trials to reacquisition criterion were comparable among all dogs, and subjects trained earlier with continuous response contingent CS termination had overlall faster response latencies, except for the medial subject that were slower than subjects trained under the partial schedule. Differentiation performance tended to show more success in the group given the earlier continuous response contingent avoidance training, although data from both training groups were variable. Subjects trained with the partial schedule had more extra- and intertial responses, which are considered as a byproduct of residual fear. Systematic effects from the surgical treatments were not found during differentiation training. Stimulus control by tonal frequency was more pronounced in subjects that had received the partial training schedule, and frequency generalization was disrupted by the lesion effects. The data are considered in light of reinforcement determinates from CS prolongation in the modification of defensive reflex control.

2 INTRODUCTION The question of stimulus control of instrumental defensive reflexes after selective prefrontal lesions in dogs was examined in an investigation reported by Zieliliski, Brennan and Kowalska (22). They used procedures analogous to a similar study of alimentary refle,xes (2), which had resulted in a double dissociation of medial and lateral prefrontal effects on both stimulus generalization functions and sensitivity to extinction influences. In extending this question to defensive reflexes, another variable of critical importance, namely, the reinforcement density during acquisition training, was added (22). One group of dogs was trained to avoid shock with response contingent conditioned stimulus (CS) termination of all trials, whereas a second group received the same training except that on 50 /o of the trials, shock was unavailable and the CS was of a fixed duration. Following lesions of the medial or lateral prefrontal cortex, or after a rest pause for intact control subjects, two generalization tests to varying values of the tonal frequency dimension of the CS were conducted. Sampled generalization involved the insertion of a block of test stimuli within sessions of the respective training contingencies. After attaining the reacquisition criterion, subjects were given a second generalization test, wherein multiple test stimuli were presented in complete extinction. Response latencies during the two generalization tests reflected the extent of extinction influences from the sampled test with reduced shock availability and from the second test with no shock available. However rapid responding to the tonal frequency test values was found in all subjects. These short latency responses were least differentiated in subjects given lateral lesions and trained under both reinforcement contingences. During the test in complete extinction, slower response levels with pronounced stimulus control were obtained from both normal groups, although the normal subjects trained with the partial shock schedule showed greater stimulus control that their normal counterparts trained with continuous shock availability. The response levels of subjects trained with continuous reinforcement and given medial prefrontal lesions were sensitive to the extinction influences of the second generalization test, while the medial, partially reinforced group and both lateral groups retained rapid response latencies and had minimal differentiated responding. The present report describes the continuation of the experimental stages of the Zielinski et al. (22) study. Specifically, the results of differentiation training and subsequent gene~alization testing over a wider range of frequency values are presented. Differentiation training typically produces an increase in stimulus control (e.g. 11). Much of the data on

3 retention of differentiation tasks after prefrontal lesions has been reported using alimentary reflexes. Briefly, in dogs given asymmetrical differentiation training of alimentary motivated responses, deterioration is found with medial lesions, whereas lateral lesions affect deficits in symmetrically reinforced differentiation (3-5). The results of retention of differentiated defensive tasks after prefrontal lesions in dogs are not as clear. Although the analogous procedures for asymmetrical and symmetrical differentiation training between alimentary and defensively based responding are somewhat ambiguous (cf. 6, 16, 21), appears that prefrontal lesions have no effect on aversively motivated asymmetrical differentiation training (14). However, in symmetrical differentiation training of defensive responses, medial prefrontal lesions produce a shortening of response latencies and an increase in the numbers of extraresponses and intertrial responses, while lateral lesions result in fewer errors, longer response latencies, and less extra responses and intertrial responses (15). Whether prefrontal lesions would disturb acquisition of a differentiation task after prolonged avoidance training with differing reinforcement densities remained open. Moreover, stimulus control of defensive reflexes after differentiation training in lesioned animals was unexplored. METHOD The subjects were 24 male dogs that had received extensive avoidance training under one of two schedules of shock availability. Half of the subjects, Group A, had been trained to avoid shock signalled by a Hz tonal CS potentially associated with shtock on each training trial. Responses made within 5 s of CS onset terminated the tone and precluded the shock; responses elicited after 5 s coterminated both the tone and shock. The remaining 12 dogs, Group B, received avoidance training with the same CS value, but only 50 /o of the daily training trials could involve shock delivery. On nonshock trails, the CS action was prolonged to 5 s, independent of the subject's responding. After reaching the avoidance performance criterion, 8 dogs had bilateral lesions of the medial prefrontal cortex (Subgroups AM and BM), 8 subjects received bilateral removal of the lateral prefrontal cortex (AL and BL), while the remaining subjects served as nonoperated controls (AN and BN). All subjects were then tested for frequency generalization under two successive procedures. Sampled generalization consisted of 20 daily sessions in which one block of 5 trials out of 20 total daily trials presented generalization test values of 600, 800, 1.000, 1.200, and Hz. The order of generalization blocks in one of the four quarters of the daily trials was counterbalanced across days and subjects within each subgroup. After 3 - Acta :iruroblologiae Exp. 5/81

4 avoidance response retraining the second frequency generalization test was given which presented the same values during 10 testing days of 20 daily trials in complete extinction. Each test stimulus occurred four times daily in random order with the constraints that each value was presented once in five trial blocks and the same value did not occur consecutively. Specific details of the subjects experimental histories were reported previously (22). The present effort continued these experimental series through avoidance response retraining, differentiation training and an additional frequency generalization test, outlined in Table I. Outline of the experimental procedure. CS+ indicates the stimulus value potentially associated with shock; CS- refers to the stimulus value never paired with shock. R's denote responses. Stages of experiment Number of daily trials Performance and testing criteria Resumed avoidance training 10 90% avoidance R's for 5 consecutive days Differentiation training 20 90% avoidance R's to CS+ and 90% inhibited R's to CS- for 10 consecutive days, or 100 sessions Postdifferentiation generalization in days complete extinction (7 values) Avoidance retraining. All subjects received daily avoidance reacquisition training sessions of 10 presentations of the Hz tonal CS. Independently on the previous training history all subjects were retrained on the same continuous schedule of shock availability and response contingent CS termination. On each training trial responses within 5 s of CS onset successfully avoided shock and terminated the CS. Responses made after the 5 s CS duration terminated both the tone and the shock unconditioned stimulus (US). Trials were separated by variable intervals of 40, 60 or 80 s, averaging 1 min during a training session. This procedure continued until a criterion of 90 /o avoidance responses on 5 consecutive days was met. Go, no-go differentiation. On succeeding days, training was expanded to 20 trials, consisting of 10 trial presentations each of the Hz tone (CS+) followed by shock if the subject failed to responed within the 5 s CS-US interval, and the 600 Hz tone (CS-) that was never associated with shock. The CS- value remained on for 5 s, independent of responses by the subject. The order of CSf and CS- trials was determined by a Gellerman Series (7). Differentiation training continued until the sub-

5 jects reached the criteria of 90 /o correct responses to CS+ and 90 /o inhibited responses to CS- during 10 consecutive days, or if 100 training sessions elapsed, whichever occurred first. Postdifferentiation generalization testing. After meeting one of the criteria for differentiation training, subjects were tested for frequency generalization in complete extinction. The test stimuli were tones of 400, 600, 800, 1.000, 1.200, and Hz. Each of the 10 days of testing consisted of 21 total trials divided into blocks of 7 trials, and within each block, the test values were ordered randomly with the exception that two consecutive trials did not present the same test frequency. The duration of test stimuli was fixed at 9 s, independent of the subject's behavior. During differentiation training and generalization testing, trial response latencies, the number of correct responses, the number of intertrial responses (ITRs) and the number of extraresponses (ERs) were recorded. This last measure was defined as a response emitted during a test stimuli action after the first response to the value was emitted. Histology. The subjects in the medial and the lateral groups were sacrifised by administration of an overdose of Nembutal. The brains were removed and subjected to histological analyses by Kliiver's and Nissl's techniques. Detailed examination of the extent of cortical damage was made, and the results of reconstructions were transferred to Kreiner's patterns (17) to eliminate the differences in shapes of the dogs' brains and to facilitate comparisons of the lesions. RESULTS It should be recalled that by the time subjects reached the stage of differentiation training, they had extensive acquisition experience with either continuous (Group A) or partial (Group B) schedules of response contingent CS termination. Accordingly, although all subjects shared common treatments in the present experimental series, the statistical analyses of their behavior retained the distinction in training schedules, which was characteristic of earlier experimental stages. Differentiation training. All subjects easily reached the predifferentiation training avoidance criterion during the daily sessions of 10 reacquisition trials: Group A, 5.5, 6.5; Group B, 5.25, 8.5, 5.5 mean sessions for normal, medial and lateral subjects, respectively in each group. These scores did not differ significantly. The distributions of response latencies during this reacquisition stage for the normal, medial and lateral subgroups that ealier had avoidance training with continuous response contingent CS termination (Panel A) or partial response contingent

6 CS termination (Panel B) are shown in F'ig. 1. The distributions of response latencies in Fig. 1, prior to differentiation training, were remarkably similar to the response latency distributions prior to the Fig. 1. Distributions of response latencies during pre-differentiation reacquisition criterion sessions of Group A (Panel A) and Group B (Panel B) for the normal (solid lines), medial (broken lines) and lateral (broken lines with dots) subgroups. massed generalization test (see Fig. 3, 22). The overall comparison of distributions from both panels of Figure 1 indicates that normal and lateral subjects of Group A were faster th'an Group B. This impression was confirmed by comparisons between each surgical subgroup of the two avoidance training schedules (Part 3, Table 11) showing that the earlier reinforcement history distinguished the normal and medial subgroups from each other, and the point of maximum differences between each surgical subgroup occurred at very rapid latency values. Within each reinforcement treatment (Part 2, Table 11), comparisons of the point of maximal difference between surgical subgroups after earlier training with continuous response contingent CS termination (A) showed that the laterals were fastest and the medials were slowest. However, after extensive avoidance training with partial noncontingent reinforcement (B), the laterals were fastest and the normals were slowest during reacquisition. Accordingly, despite reacquisition training under the same avoidance contingencies, group differences resulting from reinforcement histories as well as surgical effects were still evident. The results of differentiation training were most variable. The number of sessions for each subject is shown in Table 111, which indicates that mastery of the differentiation task was not specific to either training history or surgical treatment. To provide a more sensitive index of discriminative performance, Mann-Whitney U tests of response latencies on

7 Mean median performance latencies (1) during five days of ten avoidance training trials prior to differentiation training. Comparisons are indicated within groups (2) previously trained with continuous reinforcement availability (Group A) or partial reinforcement (Group B), and between respective groups (3) of each surgical treatment level. N > M indicates that in the normal subgroup, a greater proportion of responses was emitted with latencies shorter than the point of Dm,, compared to the responses of the medial subgroup Training history Group A (CRF) Group B (PRF) 1. Mean median latencies in seconds Normal (N) Medial (M) Lateral (L) Within reinforcement treatments N vs. M N > M Dmax 0.195** Seconds 2.6 N vs. L N<L Dm, Seconds 0.8 M vs. L M t L Dmax * Seconds Normals Dmax Seconds Medials Dmax Seconds Laterals Dmax Seconds 3. Between reinforcement treatments A>B * 1.o A<B 0.170** 1.o A>B p i ** p < 0.001; Kolmogorov-Smirnov two-tailed test. CS+ and CS- trials were conducted for each differentiation training session for all subjects 1. Table I11 also shows the percent of total sessions 1 This procedure involved the separate calculation of U statistics for each differentiation training session for each subject. A statistically significant U value for a given differentiation training session indicated independence between the distributions of ranked latencies between responses to CS+( and to CS- conversely, a nonsignificant U statistic indicated dependence between ranked response latencies to CS+ and to CS-, reflecting poor differentiation (see also 23).

8 in whlch responding to CSf was significantly (i.e., at least P(0.05) faster than to CS-. An analysis of the arc sine transformed percentages of these data revealed only an effect from training history (Fills = 4.50, P < 0.05), which confirmed that Group A had better overall differentiation than The total number of differentiation sessions and the percent of sessions in which responding to CS+ was significantly faster than to CS-, for normal 0, medial (M) and lateral (L) subjects of Group A and Group B. Specification Group A (CRF) Group B (Pw Total Percent of Total Percent of sessions faster CS+ sessions faster CS+ Group B. However, neither the effect of surgical treatment nor the interaction of surgery X training history were significant. In addition, an analysis of the numbers of errors during differentiation training showed that the overall number of errors to CS+ was significantly less than to CS- in Group A (Fin = 31.42, P<0.001) as well as in Group B (Fl19 = =33.92, P<0.001). The effect of surgery and the interaction of surgery X type of error were not significant in either group. During the predifferentiation testing of tonal frequency generalization, the numbers of responses emitted after the initial trial response during the remaining stimulus action (ERs), as well as during the intertrial intervals (ITRs) were noticeably high. To determine if differentiation training affected the levels of ERs and ITRs, the numbers of such responses were compared between the generalization sessions administered before differentiation training and testing sessions given after differentiation training. Group mean numbers of ERs and ITRs are indicated in

9 Table IV for each of the three surgical subgroups under the earlier training schedules of continuous (A) or partial (B) response contingent CS termination. Analysis of individual ER levels confirmed that subjects with the hist,ory of partial response contingent CS termination had more The mean number of extra responses (ERs) per sec and the mean number of intertrial responses (ITRs) per trial during generalization testing before (Pre-DT) and after (Post-DT) differentiation training Group A (CRF) Group B (PRW Specification ERs during generalization ITRs during generalization Pre-DT Post-DT Pre-DT Post-DT N M L N M L ERs than subjects in Group A (Fln8 = 6.33, P < 0.05). Differences between pre- and post-differentiation levels of ERs were not found, and neither the effect of surgical treatment nor any of the interactions reached acceptable significance. The analysis of IT% also indicated a trend toward more responses in Group B than in Group A, but this difference only approached significance (Fln8 = 3.82, P < 0.10). As with the measure of ERs, none of the other main effects or interactions attained significance for ITRs. Thus, the levels of responding after the initial trial responses during generalization testing remained constant following differentiation training. That Gmup B tended to have more ERs and ITRs than Group A proved resistant to any intermediary effects from differentiation training. All of this indicate that diff'erentiation braining af defensive reflexes in this experiment was affected by the residual influences of previous training. In particular, the effect of partial response contingent CS termination and accompanying CS prolongation appears to have frustrated differentiation between CS+ and CS- more than the continuous, response contingent schedule. In addition, the lack of any clearly discernable surgical influence on discriminative performance was apparent. Generalization testing. Mean response latencies to the 7 frequency stimuli over the 10 days of testing in extinction were submitted to a surgery X training history X testing days X stimuli, mixed design analysis of variance. This overall analysis showed significant effects of progressive

10 testing days (F9/162 = 2.66, P < 0.01), generalization stimuli (Fuloe = 26.11, P < 0.001) and the interaction of these two variables (F5,/,, = 1.90, P < < 0.01). The effects of surgical treatment, training history and associated interactions were all nonsignificant. Separate analyses of surgery X testing days X stimuli were conducted for Group A and for Group B. The analysis for Group A indicated only the significant effect of generalization stimuli (F6154 = 15.66, P < 0.001). The same analysis for Group B showed significant effects of testing days (F9181 = 3.93, P < 0.001) and generalization stimuli (FBIs4 = 4.24, P < 0.01). An additional overall ana- lysis was done on the absolute number of trials with responses to each stimulus during generalization testing. Only the effect of stimuli (FsllOB = = 13.98, P < 0.001) was significant. Neither the effects of surgery, training history nor any of the interactions were significant. Thus, overall response latencies during this generalization test suggested comparable performance within each group, regardless of surgical treatment. The differences from the avoidance training history, found in the separate analysis of Group B, showed that the response levels of these subjects decreased over testing days with stimulus control retained. Further examination of this finding indicated that subjects in Group B tended toward longer response latencies over days of generalization testing in extinction, which did not occur in Group A. To gain a more sensitive appraisal of differential responding during this last generalization test, distributions of response latencies to the 7 test values were constructed for each subgroup. Since the levels of ERs and ITRs were high and variable, construction of the distributions of response latencies were adjusted, so that the respective levels of ITRs were subtracted from the distribution of each test value. Briefly, this procedure involved calculation of individual ITRs across the total time of response opportunity. These estimations of the levels of individual expected spontaneous responses were grouped into cummulative distributions and then subtracted from the levels of responding to test values over the 9 s total stimulus duration. For comparison, the same procedure was applied to the distributions obtained in predifferentiation generalization in complete extinction, reported earlier (22). These functions are presented for each reinforcement training condition (Groups A and B) in Figs Similarly, the adjusted distributions of response latencies from postdifferentiation generalization testing appear in Figs The removal of the level of responding that might be expected on the basis of intertial responding reduced the overall vigor of responding shown in Figs. 2-4, while retaining the same shape of the functions depicted in the nonadjusted distributions reported earlier (22). Specifically, the overall level of responding, across the 5 test stimuli, was still grea-

11 ter in Group B subjects, even after the adjusted removal of spontaneous response levels. The best differentiation of stimuli occurred in Group B normal subjects (Fig. 2B). The normal dogs from Group A (Fig. 2 A) showed some stimulus frequency control, as did the laterals of Group A 1 Ttrne In seconds Time In seconds Fig. 2. Adjusted cummulative frequency distributions of response latencies to the test stimuli during pre-differentiation generalization in complete extinction (data from 22) in the normal subjects of Group A (Panel A) and Group B (Panel B):O denote responses to the 600 Hz test value; A, 800 Hz; 0, 1,000 Hz; V 1,200 Hz; 0, 1,400 Hz tones. Fig. 3. Adjusted cummulative frequency distributions of response latencies to the test stimuli during the pre-differentiation generalization test in complete extinction (data from 22) in the medial subjects of Group A (Panel A) and Group B (Panel B). Denotations as in Fig. 2.

12 (Fig. 4A). Generalization levels in the medial subjects of both reinforcement training groups (Fig. 3A, B) were complete and nondifferential, but the two medial subgroups differed from each other in their overall response vigor. The lateral subjects of Group B,(Fig. 4B) showed nondiffe Tlrne in seconds Tlrne in seconds Fig. 4. Adjusted cummulative frequency distributions of response latencies to test stimuli during the pre-differentiation generalization test in complete extinction (data from 22) for the lateral subjects of Group A (Panel A) and Group B (Panel B). Denotations as in Fig. 2. Tlrne In u ~onds Time In seconds Fig. 5. Adjusted cummulative frequency - distributions of response latencies to test stimuli during post-differentiation generalization in the normal subjects of Group A (Panel A) and Group B (Panel B): denotes responses to the 400 Hz test value; Q, 600 Hz; A, 800 Hz; 0, Hz; V, Hz; 0, Hz; X, Hz tones.

13 rentiated generalization of an overall vigorous level, comparable to the level of medial responding in Group B (Fig. 3B). The adjusted distributions of response latencies from postdifferentiation generalization (Figs. 5, 6 and 7) support the statistical analyses cited above. Agai,n, as in previous stages of this experiment, the overall level Ttrne In seconds Fig. 6. Adjusted cummulative frequency distributions of response latencies to the test stimuli during post-differentiation generalization in the medial subgroups of Group A (Panel A) and Group B (Panel B). Denotations as in Fig. 5. Ttrne In seconds T~me tn seconds Fig. 7. Adjusted cummulative frequency distributions of response latencies to the test stimuli during post-differentiation generalization in the lateral subjects of Group A (Panel A) and Group B (Panel B). Denotations as in Fig. 5.

14 of responding of the B subgroups, across test stimuli, was greater than in the A subgroups. In terms of stimulus control, the normal subjects of Group B (Fig. 5B) clearly discriminated both the values of the CS' (1.000 Hz) and the CS- (600 Hz) from the other 5 test stimuli. Differential responding of an almost hierarchical order from CS' to CS- emerged in the medial B subgroup (Fig. 6B), and the lateral B subgroup (Fig. 7B) responded in a similar manner, especially at longer latency values. The normal A subgroup appears to have responded in a bimodal manner. In Fig. 5A, the higher frequency values are nondifferentially grouped around the value of CS+, while the 600- (CS-) and 400-Hz values are together at a lower level. In the medial A subgroup (Fig. 6A), the bimodal tendency is also present, but by the 4th s of test stimuli action, the value of CSf emerged separately. In the lateral A subgroup (Fig. 7A), the Hz value was discriminated, while responding to the remaining test values was nondifferential. Thus, while the measures of differentiation learning were variable and did not show clear, compelling evidence of successful discriminative acquisition, the postdifferentiation distributions of response latencies demonstrate better differentiation and stimulus control than found in the generalization test given prior to differentiation training, suggesting some mastery of the differentiation task. A final assessment of postdifferentiation generalization involved the numbers of extraresponses performed during the action of the test stimuli. Analysis of the absolute number of extraresponses to each test frequency value for Group A revealed the presense of a systematic effect of stimuli (F6/54 = 5.08, P < 0.001). A similar analysis of the number of extraresponses to the seven test stimuli for Group B showed a significant effects of stimuli (F6/54 = 12.43, P < 0.001) and progressive stages of testing (FlI5, = 10.31, P < 0.01), as well as the interaction of stimuli X X stages (Fern4 = 3.98, P < 0.005). To assess the relationship between the number of extraresponses and the opportunity to emit extraresponses during test stimuli action, a ratio measure was used. Specifically, the numbers of ERs emitted during the remaining action of each test stimulus value were calculated. Then, the time opportunities for extraresponses to occur, which consisted of the durations of residual test frequency action following the emission of the first response, were also calculated for each test value during the 10 testing sessions. The ratio of the number of ERs to the temporal opportunity to make ERs were analyzed for both groups. The effects of stimuli were significant for both the continuous (F8/42 = = 5.21, P < 0.001) and the partial (F6/,, = 4.73, P < 0.005) training conditions. Although this measure of extraresponses did not attain significance for either the surgical or the acquisition training schedule varia-

15 bles, it is intriguing that reliable generalization gradients did emerge, based upon the number of ERs. Histology. As seen from Fig. 8, the prefrontal lesions situated in the medial aspect of the hemispheres were large. In all dogs the pregenual areas (PC I, PG I1 and PG 111, according to Kreiner, 17) were completely, or almost completely, removed bilaterally. The medial parts of the proreal gyri were bilaterally ablated in all dogs except for subject BM 2 that had only partial lesioning of this region in the left hemisphere. The polar areas (POL) were completely removed in 3 subjects (AM 1, AM 4 and BM 3) and were intact bilaterally in subject AM 3 and unilaterally in subject BM 1. In the remaining subjects, this cortex was only partially removed. Lesions in the subproreal gyri (SPR I and SPR 11) were not complete. The widest damage of these regions were made in subject BM 3, and in subject BM 1 this region was intact. The subgenual area (SG) was totally or almost completely removed in all animals. Only subject BM 1 had a unilateral lesion of this cortex. As was intended, partial bilateral removals of the anterior part of the medial precruciate cortex (XM I and XM I1 areas) were made in all subjects. There were also incisions to the anterior part of the genual gyrus (G). The largest damage of the genual cortex was found in subjects BM 2 and BM 4, which also touched some fibers of the corpus callosum. Incisions to the pregenual fibers were done in subjects AM 1, AM 2 and AM 3. Small damage to proreal fibers was made in subjects AM 3 and BM 3. The largest cortical areas were ablated in subjects AM 1 and AM 4. The smallest lesion was found In subject BM 1. The prefrontal lesions situated in the lateral aspect of the hemispheres are shown in Fig. 9. There were small differences between damaged areas among the lateral subjects. In all subjects, the lateral aspect of the proreal gyrus (PR, PRL I and PRL 11) and a larger part of the orbital cortex (ORB I' and ORB I") were completely removed. Polar areas (POL) were partially ablated. There were only some minor differences among damaged areas of the orbital cortex in the depth of the presilvian sulcus. Lesioned areas in ORB I1 were only partially destroyed in three subjects (AL 1, BL 3 and BL 4) and unilaterally in subject BL 1. There wete also limited incisions in areas PORD' and PORD". In some dogs, the white matter was touched. Under the proreal gyrus, the fibers were bilaterally damaged lightly in five subjects (AL 1, AL 2, BL 2, BL 3 and BL 4). The orbital fibers were touched bilaterally in subjects AL 3, BL 2, BL 3 and BL 4, and unilaterally in subjects AL 2 and BL 1. As may be seen from both Figs. 8 and 9, there were some individual, but no differences between respective A and B surgical groups in the lesions made in the medial and lateral aspects of the prefrontal cortex.

16 Fig. 8. Reconstructions of the medial prefrontal lesions in dogs from Group A and Group B Fig. 9. Reconstructions of the lateral prefrontal lesions in dogs from Group A and Group B. DISCUSSION The stages of reacquisition, differentiation training and the final test of generalization described presently, in conjunction with the earlier stages of this study reported before (22), may be considered along several interrelated dimensions. The first concerns the manner of reinforcement delivery during initial avoidance training, which has been described in terms of response contingent CS offset vs. partial contingency of CS offset with the response. The second is related to the dissociative effects of medial and lateral prefrontal lesions. Finally, the generalization performance may be compared to earlier tests of stimulus control, as well as to the similar extended examination of stimulus control in dogs, but based upon alimentary motivation (2).

17 The residual effects of the earlier avoidance training under conditions of either continuous response contingent CS termination (Group A) or partial, 50 /o response contingent CS termination (Group B) continued to exert an overriding Influence during the last stages of differentiation training and generalization testing. The persistent effects of partial response contingent CS termination and CS prolongation in Group B of this study have been intriguing. The partial shock schedule also included the fixed duration of the CS during half of the avoidance trials, which did not permit the dissipation of fear associated with the CS, as predicted from previous studies, as well as from traditional two-factor theory (13, 19, 20). A study by Zieliriski and Plewako (24) on the effects of CS prolongation on avoidance training in cats is relevant at the present results from partial reinforcement training in dogs. Cats were given either response contingent CS termination or CS prolongation on each trial from the very beginning of avoidance training. Subjects trained with CS prolongation failed to develop short latency avoidance responses and showed heightened rate of ITRs. Both findings were interpreted as the result of the subject's inability to terminate the CS elicited fear and to avoid the fear state. Additionally, cats with CS prolongation during training were much more resistant to extinction of the avoidance response, since in this group training and extinction conditions differed only in abolition of shock application. Thus, CS prolongation on avoidance training has pronounced effect on two response characteristics, latency and resistance to extinction, which were found to influence shape of generalization function. Typically, sharper generalization gradients were obtained from long-latency than from short-latency responses (1, 12, 18), and gradient sharpened during the extinction test (10). In light of these factors contributing to enhanced stimulus control, four points are critical to adequately understand the findings of Group B in the present study. First, the comparison of alimentary and defensive motivation typically indicates longer response latencies and better stimulus control with alimentary motivated than in defensively based responding (8, 9, 22). In the present study, with two types of defensive reflex acquisition procedures, the training condition of Group B that involved CS prolongation, interfered with the emergence of short latency responses. In turn, the longer latencies, as the result of the ineffectiveness of response-produced CS termination, contributed to better stimulus control in the postdifferentiation generalization test. The second point relates to the finding that increased extinction should result in better stimulus control. Certainly, the comparison between the sampled generalization test and the first test in extinction, re-

18 ported earlier (22), showed that the influence of extinction during the latter test resulted in longer latency responses and better differentiated responses. However, the similarities between CS prolongation during original training and the CS prolongation during testing produced less extinction in Group B than in Group A. Accordingly, given the postdifferentiation generalization performance in Group B, it appears that the effects of generally longer latencies from CS prolongation produced enhanced stimulus control in Group B than did the opposite influence of greater resistance to extinction in these training conditions. The third point also deals with extinction differences between the acquisition training conditions and may be seen in the differentiation training data. Differentiation training was more successful in Group A than in Group B. Differentiation training was based on contrasting two tonal frequencies and related to them training contingencies. Fixed duration of the CS- was not a novel training contingency for either group, since training history of all dogs included two generalization testings with fixed duration of tonal frequencies. However, for subjects in Group B, prolongation of CS- presumably operated in an opposite manner than for Group A. Specifically, these subjects with a history of CS prolongation on 50 /o of training trials were reinforced for longer response latencies, which allowed less aversive effects from continue'd CS action after response emission. Accordingly, longer latencies to all stimuli during differentiation training resulted from their earlier experience with partial CS prolongation. With this strategy interferred with acquisition of the differentiation, the longer latency responses facilitated the emergence of stimulus control during postdifferentiation generalization testing. The fourth point relates to the increased ERs and ITRs in Group B, in comparison with Group A, which persisted as behavioral byproducts of residual fear. Indeed, the heightened rate of ERs and ITRs continued to the final generalization test, since systematic variation in these measures did not change from pre- to postdiffe~entiation generalization. In the first experiment of this series performed on dogs trained in instrumental alimentary reflexes the dissociation between medial and lateral effects on succeptibility to extinction and stimulus control has been discovered. Specifically, dogs with medial prefrontal lesions were the most sensitive to reinforcement density (2). Similar statement is valid also for dogs trained in avoidance response. As expected, extinction of the response in the generalization tests was greater in dogs which had acquisition experience with continuous (Group A) than with partial (Group B) schedules of response contingent CS termination. During the predifferentiation massed generalization test the greatest difference in

19 the overall level of responding was observed within dogs with medial prefrontal lesions. Similar result was obtained also during the postdifferentiation test, however, in this case difference between Group A alnd Group B emerged also between dogs with lateral prefrontal lesions. It seems that the dissociation between medial and lateral effects both on extinction and on stimulus control attenuated after prolonged differentiation training. The data from this experiment provide a necessary compliment to studies based upon alimentary reflexes. Specifically, the nature of the aversive motivation ifn defensive reflexes accounts for the critical determinants that dramatically affect other variables in the experimental space. The present evidence showing continual, persistent effects throughout later stages of the experiment from the earlier manipulations of response contingent CS termination underscores the importance of perceived control over aversive stimulation. Accordingly, these data, with others from an aversively motivated context, emphasize the importance of secondary drives from fear states elicited by the CS. This experiment was designed and the training of some dogs started when the third author was supported by a research fellowship from the International Research Exchanges Board of New York. Bis participation in the preparation of this manuscript was supported by a grant from the National Academy of Sciences, USA. This investigation was supported by Project of the Polish Academy of Sciences. REFERENCES 1. BACKER, W. M. and HOLLAND M. K Spectral generalization gradients from pigeons as a function of response latency. Psychon. Sci. 10: BRENNAN, J., KOWALSKA, D. and ZIELINSKI, K Auditory frequency generalization with differing extinction influences in normal and prefrontal dogs trained in instrumental alimentary reflexes. Acta Neurobiol. Exp. 36: BRUTKOWSKI, S. and DABROWSKA, J Prefrontal cortex control of differentiation behavior in dogs. Acta Biol. Exp. 26: DABROWSKA, J Dissociation of impairment after lateral and medial prefrontal lesions in dogs. Science 171: DABROWSKA, J On the mechanisms of go-no go symmetrically reinforced tasks in dogs. Acta Neurobiol. Exp. 32: DABROWSKA, J Prefrontal lesions and avoidance reflex discrimination in dogs. Acta Neurobiol. Exp. 35: GELLERMAN, L. N Chance orders of alternating stimuli in visual discrimination experiments. J. Gen. Psychol. 42: HEARST, E Concurrent generalization gradients for food-controlled and shock-controlled behavior. J. Exp. Anal. Behav. 5: HEARST, E Approach, avoidance, and stimulus generalization. In D. I. 4 - ACta Neuroblologlae Exp. 5/81

20 Mostofsky (ed.), Stimulus generalization. Stanford Univ. Press, Stanford, p HOFFMAN, H. S Stimulus generalization versus discrimination failure in conditioned suppression. In R. M. Gilbert, N. S. Sutherland (ed.), Animal discrimination learning. Academic Press, London, p JENKINS, H. M. and HARRISON, R. H Effect of discrimination training on auditory generalization. J. Exp. Psychol. 59: KAKIGI, S Cardiovascular generalization and differentiation: the relationship between heart rate and blood pressure. Cond. Reflex: 6: KAMIN, L. J The effects of termination of the CS and avoidance of the US on avoidance learning. J. Comp. Physiol. Psychol. 49: KOWALSKA, D. DABROWSKA, J. and ZIELINSKI, K Effects of partial prefrontal lesions in dogs on go-no go avoidance reflex differentiation and reversal learning. Acta Neurobiol. Exp. 35: KOWALSKA, D., DABROWSKA, J. and ZIELIRSKI, K Retention of symmetrical go-no go avcnidance differentiation after prefrontal lesions in dogs. Bull. Acad. Pol. Sci. Ser. Sci. Biol. 23: KOWALSKA, D. and ZIELINSKI, K Avoidance responding in dogs trained in symmetrical or asymmetrical go, no-go differentiation. Acta Neurobiol. Exp. 40: KREINER, J Reconstruction of neocortical lesions within the dog's brain. Acta Biol. Exp. 22: MIGLER, B Effects of averaging data during stimulus generalization. J. Exp. Anal. Behav. 7: MOWRER, 0. H. and LAMOREAUX, R. R Fear as an intervening variable in avoidance conditioning. J. Comp. Physiol. Psychol. 39: RESCORLA, R. A. and SOLOMON, R. L Two-process learning theory: relationships between pavlovian conditioning and instrumental learning. Psychol. Rev. 74: ZIELINSKI, K Extinction, inhibition, and differentiation learning. In A. Dickinson, R. A. Boakes (ed.), Mechanisms of learning and motivation. Lawrence Erlbaum Ass. Publ., Hillsdate, New Jersey, p ZIELIfiSKI, K., BRENNAN, J. and KOWALSKA, D Auditory frequency generalization in normal and prefrontal dogs trained in varieties of active avoidance reflexes. Acta Neurobiol. Exp. 39: ZIELINSKI, K. and CZARKOWSKA, J Quality of stimuli and prefrontal lesions effects on reversal learning in go-no go avoidance reflex differentiation in cats. Acta Neurobiol. Exp. 34: ZIELIRSKI, K. and PLEWAKO, M The role of conditioned stimulus termination in short-latency avoidance responding in cats. Behav. Brain Res. 1: Accepted 20 May 1981 James F. BRENNAN, Department of Psychology, University of Massachusetts-Boston, Boston, Massachusetts USA.