Disinhibition of an operant response

LEARNING

AND

MOTIVATION

Disinhibition

(1970)

1, 346-371

of an Operant

Response

C. J. BRIMER Dalhousie

University

A total of 320 rats were employed in six experiments analyzing the phenomenon of disinhibition of an operant lever-pressing response. The parameters explored were type of inhibitory operation preceding the test for disinhibition, and modality, duration, directionality, intensity, and prior exposure to the disinhibiting stimulus. The disinhibition phenomenon was found to be highly general, occurring under almost all test conditions. The phenomenon, however, could be produced during extinction only when the level of responding lay within a critical low range of values. When animals with the same low response rates were tested during acquisition, before lever-pressing had reached a higher level, disinhibition never occurred.

The present series of experiments grew directly out of an earlier investigation (Brimer & Kamin, 1963) concerned with the effects of unsignalled electric shock on the acquisition of a conditioned emotional response. In this earlier work, rats were first trained to a stable rate of lever-pressing under a variable interval schedule of food reinforcement. Animals then received unsignalled electric shocks which produced a radical reduction in the baseline response rate. At this stage, the presentation of white noise increased the rate of lever-pressing in virtually all of the subjects. This effect seemed reminiscent of the phenomenon of “disinhibition,” first described by Pavlov (1927). Although Pavlov studied classically conditioned responses, and our own procedure was that of operant conditioning, the descriptive similarity of the effect we noted to that of Pavlovian disinhibition was striking. Although Pavlov frequently observed the occurrence of disinhibition, he did not do any rigorous experimental work on the parameters which ’ This report is based on a dissertation presented in partial fulfillment of the requirements for the Ph.D. degree at McMaster University. The author is grateful to L. J. Kamin for his assistance, guidance, and encouragement throughout all phases of the research. The author is also indebted to Ronald Schaub for his technical assistance with the apparatus. The research was supported by a grant to L. J. Kamin from the National Research Council of Canada, and the preparation of the paper, by N.R.C. Grant APAto the author. Requests for reprints should be sent to the author, Department of Psychology, Dalhousie University, Halifax, Nova Scotia.

346

347

DlSlNHlBlTlON

control the effect. Similarly, a number of North American investigators have published reports demonstrating a disinhibition effect (e.g., Gagne, 1941; Horns & Heron, 1940; Hunter, 1935; Razran, 1939; Switzer, 1933; Wenger, 1936; Winnick & Hunt, 1951; Yamaguchi & Ladioray, 1962), but detailed parametric investigations are lacking. The feasibility of carrying out such disinhibition studies in the Skinner box was suggested by the great reliability across subjects reported by Brimer and Kamin (1963) who concluded that the main contribution of their experiment might be “the development of a simple and stable preparation for the study of inhibitory and disinhibitory phenomena” (p. 5 15). Disinhibition may be defined as the temporary reappearance of a suppressed or inhibited conditioned response due to the presentation of an extraneous stimulus. Thus, there are two major variables involved in the disinhibition phenomenon, viz., the extraneous (or disinhibiting) stimulus and the inhibition (or suppression) of the conditioned response. Six experiments were carried out to investigate these two areas. Disinhibition was investigated with extraneous stimuli of different sensory modalities. different durations, and different intensities. The effects of different types of inhibitory operations were examined, although for most of the studies a standardized experimental extinction procedure was eventually employed. Finally, a series of experiments sought to determine whether response suppression or only a low level of responding was the sufficient condition for the disinhibition phenomenon. EXPERIMENT

1

Method Subjects

and Apparatus

The Ss in all experiments were experimentally naive male hooded rats, supplied by Canadian Research Animal Farms, ranging in weight from approximately 250 to 300 gm. There were 64 Ss in Expt. 1, haphazardly assigned to eight experimental groups of eight Ss each. The apparatus consisted of eight standard Grason-Stadler Skinner boxes, individually housed in sand-filled “ice-chest” We sound-attenuating wooden boxes. The two experimental stimuli employed were white noise and light. The noise was produced by a Model 901A Grason-Stadler noise generator, which fed into the loudspeakers attached to the outside walls of the Skinner boxes. An eight channel audio-splitter manufactured by Ashman Electronics Limited aliowed the noise intensity delivered to each Skinner box to be independently

348

C. J. BRIMER

adjusted. The mean ambient noise level for the eight Skinner boxes, with the exhaust fans operating, was 55.9 db as measured by a Type 1555-A General Radio sound survey meter. The noise stimulus employed in Expt. 1 increased the sound level, at the rat’s normal location in the box, to 80 db. The light stimulus was produced by a 120-V 6-W electric light bulb attached to the outside wall of the Skinner box above the loudspeaker. The normal condition inside the Skinner box was complete darkness. When the light stimulus was employed, the mean reflected intensity inside the Skinner box at the rat’s normal location was 1.98 ft-c as measured by a Leeds and Northrup Model 1046889 Macbeth illuminometer. The entire experimental procedure was automatically programmed by relay and timer circuits. Responses were recorded on print-out counters. The programming and recording equipment was contained in a room adjacent to the Skinner boxes. Experimental

Design

The first experiment was designed to extend the generality of the Brimer and Kamin observation of stimulus-produced response acceleration, by employing stimuli other than a 3-min white noise, and inhibitory operations other than unsignalled electric shock. The design was a 2 x 2 x 2 factorial, with eight animals run under each of the eight experimental conditions. The three factors were: (a) type of inhibitory operation; (b) modality of extraneous stimulus; (c) duration of stimulus.

and

The inhibitory operation was either experimental extinction of the lever-pressing response, or food satiation. The extraneous stimulus was either an 80-db white noise or house illumination of approximately 2 ft-c. The stimulus duration was either 1.5 set or 3 min, although the test measure always consisted of the number of responses in the 3-min interval after stimulus onset. Preliminary

Training

The animals were initially reduced to 75% of free-feeding weight and maintained at this level throughout the preliminary phase of the experiment, being fed once daily, approximately 1 hr after each experimental session. The rats were first trained to lever-press for food in the Skinner box and then given eight I-hr daily lever-pressing sessions with a 2.5-min variable interval (VI) food reinforcement schedule. With the exception of the first VI day, when the house light was on throughout the hour, all

DISINHIBITION

349

training for all animals was carried out in complete darkness. On VI days 7 and 8, the stimulus, which was later to be employed as a disinhibitor, was presented twice during each session as a pretest. In order to index the reaction to the stimulus on pretest, an inflection ratio identical to the suppression ratio employed by Kamin (1965) was calculated for each animal. This ratio is B/(A + B) where B represents the number of responses occurring during the 3 min after stimulus onset, and A represents the number of responses made in an identical period of time immediately preceding the presentation of the stimulus. The ratio has limits of .OO and 1.00 where .OO represents complete response suppression; JO, no effect of stimulus on response rate; and 1.OO, the case where no responses are made prior to the stimulus interval, but some are made during it. Inhibition

Training

After the second pretest day, 3 days of inhibition training were given. There were, as previously mentioned, two types of inhibitory operation employed: experimental extinction and food satiation. In the extinction procedure animals were maintained at 75% of their free-feeding weight, but during each experimental session no food reinforcement was programmed. In the food-satiation condition, animals were allowed free access to food in their home cages for 1 hr prior to their introduction to the Skinner box. During the experimental session, however, the VI food reinforcement schedule remained in effect. Test for Disinhibition After the third day of inhibiticn training, testing for disinhibition began. During the test phase of the experiment, the inhibitory operation of extinction remained in effect for the extinction animals as it had during the three prior days. Satiation animals, however, were switched from a 1-hr to a 23-hr ad libitum feeding schedule in the home cage. This change was necessitated by the fact that for some animals the 1-hr free feeding prior to Skinner box experience had not suppressed lever-pressing to the requisite low level. Throughout the test, satiation animals continued to receive food reinforcement for lever pressing on the 2.5-min VI schedule. In order to ensure that the baseline response rate was equally inhibited in all animals at the time of testing, the presentation of the extraneous stimulus was made contingent on a criterion of three consecutive minutes without a response.The number of responses that occurred in the 3-min interval after stimulus onset was then recorded. In order to assess the effect of stimulus presentation, each animal acted as its own control. This was accomplished as follows. On each test day, each animal received one dummy and one stimulus sequence, the order of presentation being coun-

350

C.

J. BRIMER

terbalanced both between Ss in each experimental group and from day to day within each S. The dummy presentation simply consisted of counting the number of responses that occurred in the 3-min interval after the 3-min no-response criterion. Thus, disinhibition could be calculated by comparing an animal’s response rate during the 3-min stimulus interval with its rate during the comparable 3-min dummy of the same test day. A disinhibition test trial, therefore, refers to both the stimulus and dummy presentation which occurred on the same test day. The programming of test presentations was contingent on the subject’s behavior, but the contingency did not come into effect until 9 min after the beginning of the daily session. The contingency was again suspended for a 6-min interval beginning with the onset of the first test presentation. There were never more than two test presentations (one stimulus and one dummy) for any S on any test day. The experimental plan required that the two presentations be given within the 1-hr session, as the boxes had to be utilized by other Ss. The original test schedule called for one stimulus and one dummy presentation to be administered to each animal on each of four consecutive test days. This program was successfully carried out with all extinction Ss. However, the scheduled total of four test trials was not obtained with some of the satiation Ss, due to their relatively high baseline response rates. Complete data were available, however, for at least two test days for all satiation Ss. Results For the statistical analyses associated with this and all subsequent experiments, a two-tailed .05 level of rejection was employed. Consequently, throughout the report no further reference is given to specific p values. Pretest Ratios Figure 1 presents the median inflection ratios for each of the four stimulus conditions on each of the four pretest trials. On the first pretest trial a significant proportion of animals in the 1.5set noise (14116) and 3-min light (16/ 16) groups have ratios below 50, whereas the proportion of Ss in the 3-min noise condition falls short of significance (12/16).-However, an additional five Ss were subsequently run with 3-min noise for comparison with groups in Expts. 2 and 4. When these Ss are added to the present 16, the proportion of animals with ratios below .50 (16/21) is significant. Thus, on initial presentation, all stimuli but the 1.5 set light tend to produce response suppression.

351

DISINHIBITION .---.

IfNOISE 3’ NOISE I.

.

2i,-:

.60

.+.I$?

r” E I

A

.20

>

.oo

’ 1

I 2

PRETEST FIG.

1 3

I 4

TRIAL

1. Median inflection ratios as a function of pretest trial.

In order to index the overall magnitude of pretest suppression a mean inflection ratio was calculated for each subject based on the average of the four pretest ratios The Kruskal-Wallis ranked analysis of variance demonstrated that the differences in mean ratios between groups was statistically significant, H(3) = 32. A series of Mann-Whitney Cl tests revealed that the 3-min-light animals had lower ratios than the Ss in any of the three other stimulus conditions, but that none of the other groups differed reliably. It is clear from Fig. 1 that the suppressant effect of stimulus presentation dissipates with repeated trials. For each of the three groups which initially showed response suppression, the inflection ratios were significantly higher on trial 4 than on trial 1 of the pretest. Baseline Response Rates

In terms of the total responses emitted during each 1-hr experimental session, the extinction and satiation groups did not differ on either of the two pretest days. With the introduction of inhibition training, all animals decreased their response rates from the previous day’s level with the satiation Ss dropping to a significantly lower rate than the extinction animals, U(32,32) = 137. By the third inhibition day, however, this situation had been reversed, so that the response rate of the extinction group was significantly below that of the satiation Ss, 1/(32,32) = 150. This latter relationship was maintained over the four test days. Thus, although

352

C. J. BRIMER

the satiation procedure initially produced a dramatic drop in response rate, with continued training, extinction proved to be the more effective response inhibitor. Test for Disinhibition Complete test data were available for only two test trials. To analyze these data, the responses emitted during the first two stimulus and first two dummy presentations were summed separetely for each S. The data are summarized in Fig. 2, which presents the median number of responses for each type of presentation for the eight experimental conditions. The overall rate of responding was significantly higher in the satiation than in the extinction procedure during both the stimulus, U(32,32) = 254, and the dummy, U(32,32) = 3 15, intervals. These differences seem to reflect the previously noted fact that food satiation was not as effective an inhibitory operation as experimental extinction. The more important finding was that disinhibition clearly occurred, as demonstrated by the consistent tendency for more responses to occur during the stimulus than during the dummy interval. In all of the experimental conditions, the response rate was higher during the stimulus than during the dummy interval for a significant proportion of the Ss (5 1.5/64). Applied to each experimental group, Wilcoxon’s test revealed that all but the satiated 1.5set light animals responded significantly more during the stimulus than during the dummy interval. Thus even with only eight Ss

1;

w

NOBE EXTINCTION

LIGHT

3’

3’

rt’ NOISE

’f’ LIGHTa’

SATIATION

FIG. 2. Median number of responses (trials l-2) during stimulus and dummy intervals for the eight experimental conditions.

DISINHIBITION

353

per group, seven out of the eight experimental conditions produced reliable disinhibition. To examine differences between groups in the magnitude of the disinhibition effect, each animal was assigned a difference score calculated by subtracting the number of responses emitted during the first two dummy presentations from the number of responses that occurred during the first two stimulus presentations. The analysis of variance carried out on these difference scores revealed two significant main effects: inhibitory operation and stimulus duration. Greater disinhibition was produced by the satiation procedure than by extinction, F( 156) = 4.86, which may reflect the fact that the overall level of responding was higher with the satiation procedure. The 1.5set stimuli were less effective disinhibitors than the 3-min stimuli, F( 1,X) = 6.52, when response rate was measured for the 3 min after stimulus onset. There were no significant interactions. There is an interesting parallel between the pretest (inhibition) and test (disinhibition) results. In pretest, the 3-min light produced the most response inhibition and the 1.5set light, the least. In test, these two stimuli tended to produce, respectively, the most and least disinhibition. If one rank orders the eight experimental groups in terms of the median inflection ratio on the first pretest trial and the median difference score on the first disinhibition test day, a significant correlation is found (rho = -.67). Those stimuli which, before application of an inhibitory operation, were the most effective response inhibitors were, afterwards, the most potent disinhibitors. To look at the disinhibition effect as a function of repeated test trials, the analysis was restricted to the experimental extinction group for it was only in this group that data existed for all animals for all four test trials. It was clear that the disinhibition effect, calculated as the difference between stimulus and dummy responding, progressively decreased from trial 1 to trial 4, Wilcoxon T(32) = 150. There was also, however, a significant decrease in the dummy response rate from test day 1 to test day 4, Wilcoxon T(32) = 1 lg. The significance of the former finding is consequently ambiguous, for consecutive trials involved increasing familiarity with the stimulus and also a progressively decreasing baseline response rate. After completion of the main portion of Expt. 1, a final group of eight rats was subjected to a different inhibitory operation, as an attempted tour &force. In this case lever pressing was inhibited by a punishment contingency (electric grid shock for each lever press) which remained in effect through the test for disinhibition. The basic procedure was the same as that previously outlined, except that there were no inhibition training days prior to the first testing day. The shock intensity was individually adjusted

354

C.

J. BRIMER

for different animals to a level that effectively inhibited responding. The intensities employed varied from .25 to .50 mA calibrated on a Grason-Standler Model E1064GS shock generator. Four test days were given immediately after pretest day, each animal receiving one dummy and one stimulus presentation on each day. The test presentations were, as before, contingent on a 3-min no response criterion. Throughout the test phase of the experiment, the VI food reinforcement schedule remained in effect. The disinhibiting stimulus consisted of both the light and noise presented concurrently for 3 min. In all other respects, the procedure was identical to that of Expt. 1. All eight animals had a higher response rate during the stimulus than during the dummy interval. Discussion

The major conclusion to be drawn from Expt. 1 is, simply, that the disinhibition effect has considerable generality. The previous observation of response acceleration to an extraneous stimulus (Brimer & Kamin, 1963), clearly, is not specific to that situation. It has now been demonstrated that response acceleration occurs under four types of inhibitory operation (unsignalled shock, extinction, food satiation, and punishment), two stimulus modalities (light and noise) and two stimulus durations (1.5 set and 3 min). This generality of the effect appears to justify identification of the phenomenon with Pavlovian disinhibition. The phenomenon, indeed, appears even more general than Pavlov reported; for it was observed when responding was diminished by satiating the animals with food without any experimental extinction (or, in Pavlovian language, reducing the excitability of the center of the unconditioned reflex). The fact that a response acceleration would occur even when each lever press was punished with shock, dramatically demonstrates the strength of the tendency to disinhibit. As Pavlov and others have reported, the stimuli which disinhibit the suppressed response, inhibit the nonsuppressed response. The relationship between the inhibitory and disinhibitory capacity of stimuli appeared to be monotonic, however, rather than an inverted-U as Pavlov had suggested (cf. e.g., Pavlov, 1928, p. 138). The better inhibitors were the better disinhibitors. However, regardless of the exact shape of this function, the important point seems to be that the same stimulus which reduced the probability of responding when the animal was in one state, increased the probability of responding when the animal was in another state. Therefore, the effects of the extraneous stimulus on lever-pressing

355

DISINHIBITION

rate cannot be attributed to any direct property of the stimulus that simply interferes with, or facilitates, lever-pressing behavior (e.g., Skinner, 1936). The state of the animal at the time of presentation of the stimulus is crucial. The fact that disinhibition diminishes with repeated trials, at least when the inhibitory operation is experimental extinction, raises a question. Is this diminution to be attributed to increasing familiarity with (adaption to) the stimulus, or to the progressive weakening of the basic tendency to perform the lever-pressing response? The second experiment was designed to investigate this question by assessing the effect of differential degrees of familiarity with the stimulus on the magnitude of disinhibition. EXPERIMENT

2

Method Within Expt. 2, and in all subsequent experiments, the inhibiting operation was extinction, and the disinhibiting simulus was of 3-min duration. Three new groups of rats (32 Ss) were trained for Expt. 2. Two of these groups were trained exactly as were the 3-min noise and 3-min light (extinction) groups of Expt. 1, except that no stimulus was presented on the pretest days. The third new group was pretested with noise for 8 days (i.e., 16 pretest trials) prior to the introduction of the inhibitory operation. The training procedure for this group was identical to that of the other groups with the exception that the extra pretests necessarily involved six extra VI training days prior to the introduction of experimental extinction. The 3-min noise and 3-min light groups which received extinction training in Expt. 1 were included for comparison in Expt. 2. Five additional rats were trained with four noise pretest trials to increase the size of this group for purposes of comparison in a subsequent experiment. The total number of animals in each of the five groups considered in Expt. 2 was as follows: Group

N

Noise, 0 pretest trials Noise, 4 pretest trials Noise, 16 pretest trials Light. 0 pretest trials Light, 4 pretest trials

12 13 8 12 8

Thus, within the noise condition, three degrees of familiarity with the stimulus could be compared. Within the light condition, two degrees of familiarity could be assessed.

356

C. J. BRIMER

I

NOISE NUMBER

OF PRETEST

LIGHT TRIALS

FIG. 3. Median number of responses (trials 1-4) during stimulus and dummy intervals for noise and light subjects as a function of number of prior pretest trials.

Results Pretest The pretest data for groups given four pretest trials have already been analyzed in Expt. 1 (cf. Fig. I). The pattern of responding for Ss given 16 pretest noise presentations was similar to that previously observed. On the initial stimulus presentation, there was a significant drop in the response rate from the prestimulus level, Wilcoxon T(8) = 1.O. However, there was no consistent effect of the noise over trials 3-16. Test for Disinhibition Figure 3 presents the median number of responses emitted during the stimulus and dummy intervals for all groups. In all cases, the scores are based on disinhibition test days 1 through 4. As in Expt. 1 the amount of disinhibition was measured for each S by the difference in number of responses made during the stimulus and dummy intervals. The figure makes it obvious that there was no difference between the light Ss that did, and did not, undergo pretest. Pooling both light groups, however, revealed that a significant proportion of the animals (19.5/20) had higher response rates during the stimulus than during the dummy intervals. Within the three noise groups, there was again no significant effect of number of pretest trials on the magnitude of the disinhibition effect. Again, as with light, a significant proportion of the noise Ss (27.5/33) showed the disinhibition effect.

351

DISINHIBlTION

Discussion The results of Expt. 2 seem clear. The amount of previous experience with the stimulus does not differentially affect the magnitude of disinhibition-at least not within the limits tested. Thus, it seems safe to conclude that familiarity with the stimulus is not a very important variable in the disinhibition phenomenon. This is in contrast to Pavlov’s suggestion that novel stimuli make the best disinhibitors (cf. e.g., Pavlov, 1927, p. 84). It would appear that the diminishing amount of disinhibition observed in Expt. 1 with repeated test trials cannot be attributed solely to stimulus familiarity. This, in turn, encourages the speculation that the lessening amount of disinhibition with repeated trials is attributable to the progressive weakening of the basic tendency to lever press. Possibly, then, disinhibition can only be demonstrated when the basic tendency to respond is not too inhibited. This problem is considered again in the context of Expt. 5. EXPERIMENT

3

In Expt. 3, the characteristic of the stimulus that was investigated was onset versus termination of a physical energy. The early Pavlovian literature seemed to imply that conditioning, at least, was a direct function of the total amount of physical energy impinging on the “cortical analyzers” (cf. e.g., Kupalov & Gantt, 1927). This idea is, in many essential respects, similar to Hull’s concept of “stimulus intensity dynamism” (Hull, 19.5 I). Taken literally, the notion suggests that conditioning should be more effective when the conditioned stimulus consists of the onset rather than the termination of a physical stimulus. In the few conditioning studies that have attempted to investigate this variable there is the suggestion that stimulus onset may be superior to termination when an auditory stimulus is employed (Kamin, 1965; Kish, 1955), but not different with visual stimuli (Schwartz & Goodson, 1958; Hansche & Grant, 1960; Logan & Wagner, 1962). Experiment 3 sought to study the effect of employing as the extraneous stimulus the onset and termination of light and noise. This is a stimulus variable which has not previously been investigated in the disinhibition situation. Method For Expt. 3, two new experimental groups (20 Ss) were trained. These groups received exactly the same training as the 3-min noise and 3-min light groups of Expt. 2, with one exception, From the second day of preliminary training either the 80 db noise or the light was continually on

358

C. J. BRIMER

inside the Skinner box. For both pretest and disinhibition testing, the experimental stimulus consisted of a 3-min interruption in either the noise or light. The inhibition training procedure was, as usual, experimental extinction. For comparison, the appropriate noise-on and light-on groups from Expt. 1 were incorporated in the analyses. The number of animals in each of the four experimental groups were: noise-on, 13; noise-off, 12; light-on, 8; light-off, 8. Results Pretest In each of the four experimental groups there was a significant tendency for all but the noise-off animals to suppress responding on the first pretest trial. Within either the noise or light treatment the difference in suppression ratios between the on and off groups was not significant. However, the light groups pooled had significantly lower ratios on the initial pretest trial than the noise groups polled, U(16,25) = 36. Test for Disinhibition Figure 4 presents the median number of responses emitted during the stimulus and dummy intervals for each of the four experimental groups. Our usual difference score was the measure of disinhibition and the light-on and light-off groups did not differ. The light subjects overall showed a significant disinhibition effect as 15 of the 16 subjects had higher response rates in the stimulus than in the dummy intervals. The difference between the noise-on and noise-off groups was not quite significant. However, when the groups were considered separately, only

3 %

@2 5TlM”L”S 0 DUMMY

15

b 8L * 8.3 nk 1

‘0 5

% e 1

ON LIGHT

OFF

ON

OFF NOISE

FIG, 4. Median number of responses (trials 1-4) during stimulus and dummy intervals for the four experimental conditions.

359

DISINHIBITION

the noise-on Ss had significantly higher response rates during the stimulus than during the dummy intervals, Wilcoxon T(l3) = 5.5. Discussion The results of Expt. 3 suggest that whether the extraneous stimulus is the onset or termination of light appears to make no difference. This was true both for the inhibition of nonsuppressed responding (pretest) and for the disinhibition of inhibited responding. With noise, however, stimulus termination was not sufficient to produce either significant inhibition or significant disinhibition: conversely, onset of noise was a sufficient stimulus to produce both effects. These findings correspond with the previously noted conditioning studies where both light-on and light-off were found to be equally effective conditioned stimuli, but noise-on was superior to noise-off. Thus, to at least some degree, the dimensions of the stimulus which control amount of conditioning appear also to control amount of disinhibition. This correspondence, however, need not argue for any “central” similarity between processes involved in conditioning and disinhibition. Since both phenomena depend upon stimulus reception by the subject the common factor may lie in a relatively peripheral sensory mechanism. Put simply, the rat may not sense the change involved in noise termination as well as it senses the change involved in noise onset. Thus, noise-off would for any purpose be a relatively ineffective stimulus for the rat. EXPERIMENT

4

The preceding experiment indicated that at least one stimulus dimension affects both conditioning and disinhibition similarly. In Expt. 4, another stimulus dimension-intensity-is examined. Previous work (Kamin & Schaub, 1963; Kamin & Brimer, 1963) has shown a direct monotonic relation between magnitude of the conditioned emotional response in the rat and the intensity of a white noise CS over the range 45-82 db. The present study explores the effects on inhibition and disinhibition of varying the intensity of a white noise extraneous stimulus. Pavlov at one time clearly suggested that stimuli of “moderate” intensity produced the most disinhibition; “too weak,” or “too strong” stimuli producing either less disinhibition or none (Pavlov, 1928, pp. 138, 2 1 1). This corresponded to Pavlov’s views on the effects of conditioned-stimulus intensity, since he believed that “too strong” a CS produced the “paradoxical” effect of poorer conditioning (Pavlov, 1927, pp. 271-272). However, in the only available systematic study of the intensity of a disinhibition stimulus, the results were sufficiently ambiguous to lead the authors to conclude “that the disinhibition effect is a fact

360 but not an easily reproducible 576).

C. J. BRIMER

one” (Yamaguchi

& Ladioray,

1962, p.

Method Three new groups (13 Ss per group) were trained for this experiment. The 13 Ss in Expt. 3, trained with an 80-db noise-on stimulus, were included in the design. The training procedure for the new groups was identical to that used with the 80-db noise-on group of Expt. 3 with the exception of the intensity of the noise which was employed. The three new groups received white noise stimuli set at 45, 65, and 100 db respectively. Results Pretest Although there was a tendency for the 100-db noise to produce the most suppression in pretest, due to the considerable variability that existed within each of the four groups no significant differences could be detected between the different intensities on either trial 1 or over trials 1 to 4. The noise intensity groups were therefore combined, and a significant proportion of the Ss on trial 1 had suppression ratios below 50 (41/52). Test for Disinhibition Figure 5 presents the median number of stimulus and dummy responses on test trials 1 through 4 for the four intensity groups. Although the data may suggest some slight tendency for the difference scores to increase

NOISE

INTENSITY

( D B 1

FIG. 5. Median number of responses (trials 1-4) during stimulus and dummy intervals as a function of noise intensity.

361

DISINHIBITION

with stimulus intensity, the differences between the groups did not approach significance. A significant proportion of animals responded more during the stimulus than during the dummy intervals (44.5/52), and a significant proportion of the Ss in each group, with the exception of the 45db group (IO/l 3) displayed disinhibition. This distinction of the 45db group may suggest some small effect of stimulus intensity, since it represents the weakest intensity studied. Discussion

The obvious conclusion to be drawn from this experiment is that both inhibition and disinhibition are relatively insensitive to noise intensity. There is some vey siight suggestion that both inhibition and disinhibition may increase with noise intensity, but the striking fact is that over a range as vast as 45- 100 db, the differences in the response rate changes are so minimal. These minimal differences stand in marked contrast to the CER conditioning studies where far smaller variations in CS intensity produce significantly different acquisition curves. This discrepancy between the effects of stimulus intensity on conditioning and on inhibition and disinhibition encourages speculation that the stimulus plays very different roles in the two phenomena. Perhaps, since conditioning requires that the S associate the onset of the CS with a subsequent event in time (vis., the US), the role of CS intensity in conditioning is to provide a long-lasting neural “stimulus trace” to be contiguous in time with the US. This would be consonant with the observation by Kamin & Schaub (1963) that CS intensity is a more important variable in trace, as opposed to delayed conditioning. This “time-bridging” function, of course, would not be present in disinhibition where only a single stimulus is involved. EXPERIMENT

5

Experiment 5 was designed to investigate the relation between the amount of inhibition of the baseline response rate and the magnitude of disinhibition by “mapping the reaction to white noise at consecutive stages of extinction training, as the baseline response became progressively more inhibited. This involved a considerably different experimental procedure from that previously employed. Method Preliminary Training

Forty-eight additional animals received preliminary training which was basically the same as that adopted in the previous investigations. After

362

C.

J. BRIMER

initial train& to lever press for food, there were six VI practice days. The last LWO of these days included four pretest stimulus presentations for all groups. The stimulus employed was an 80-db white noise. In this experiment, the house light in the Skinner box was on throughout all stages of training. Inhibition

Training

and Disinhibition

Testing

On the day after the second pretest session, the inhibitory operation of extinction was introduced, and testing for disinhibition was carried out. The first response made by each animal on this day produced a food pellet, but no subsequent response was reinforced. All testing was done during a single 2-hr. experimental session. There were no inhibition training days prior to the testing day, and presentation of the test stimulus was not contingent on any response criterion. The test stimulus was presented, instead, at one of five fixed times after the start of the experimental session. Independent experimental groups of 10,9, 10, 9, and 10 Ss each received the stimulus presentation after 3, 30,60,90, or 120 min of extinction, respectively. Responses were recorded at each test point for the 3-min stimulus interval and for the 3-min interval immediately preceding stimulus onset. Because the response rate might be expected to decline steadily through a large part of the testing session, dummy scores were necessary as a control measure against which to evaluate stimulus-induced responding. The 120-min experimental group served as the control Ss to provide such dummy measures at 3, 30, 60, and 90 min. There were no control scores available at 120 min. Unlike the previous experiments, disinhibition was indexed in this study by a between-, rather than by a within-subject comparison. Results Pretest The pattern observed with 80-db noise in previous experiments was again evident in pretest. On the initial noise presentation, a significant proportion of Ss had inflection ratios below .50 (35/48), but this was no longer true by the second trial (25.5/48). Test for Disinhibition Figure 6 presents the median response rates for the experimental and control Ss tested at different stages of extinction. It is obvious from the control measures that the baseline response rate progressively decreased

363

DISINHIBITION

EXTINCTION

pg

EXPERlMENTAL

a

CONTROL

(in min.)

FIG. 6. Median rate (responses per minute) for experimental function of degree of extinction training.

and control animals as a

throughout the extinction session, reaching a level close to zero by 90 min. However, the only significant difference between the response rates of the experimental and control Ss occurred after 60 min of extinction training U( 10,lO) = 22. In order to analyze the data further, each animal was classified as displaying disinhibition if it made more responses during the stimulus than during the prestimulus interval. Fisher’s Exact Test was applied to these data and showed that the proportions of “disinhibitors” in the experimental and control groups differed significantly after 60 and after 90 min of extinction training. In the case of both of the above analyses there are no dummy scores for 120 min so that no test can be made. However, it is unlikely that significant disinhibition could have been demonstrated at this stage as 4 of the 10 experimental animals failed to respond during either the stimulus or the prestimulus intervals. Thus, it is only after 60 min of extinction (when the baseline rate is about two responses per minute) that disinhibition was first demonstrated. Although the baseline response rate has shown a significant drop after 30 min of extinction, Wilcoxon T(9) = 3.0, there is no suggestion of a tendency for the stimulus to produce an increase in responding.

364

C. J. BRIMER

Discussion The results of Expt. 5 suggest that disinhibition will occur only when the baseline response rate lies within some low critical range of values, being neither “too high” nor “too low.” The idea that baseline responding may be too inhibited for maximal disinhibition to occur receives support from a number of observations. On the one hand, in Expt. 5 there is an obvious suggestion that when the baseline rate is literally zero, many animals do not show disinhibition. On the other hand, in Expt. 1 the greater disinhibition seen in satiation, as compared to extinction, could be interpreted in terms of the different levels of inhibition produced by the two procedures. Finally, in Expts. 1,2, and 3 it was repeatedly observed that the amount of disinhibition declined as the baseline response rate became progressively more inhibited with repeated testing. Thus, we might now postulate that disinhibition occurs when, and only when, the baseline response rate lies within a critical range of values. Although this notion is theoretically very simple, it has some interesting consequences, to which the final experiment was addressed. EXPERIMENT

6

The experiments already discussed contain a fundamental consistency. The same extraneous stimulus which, before inhibition training, produces a decrement in operant responding, later produces an increment in the same response. One must now ask, what do the various inhibitory operations (experimental extinction, satiation, punishment, unsignalled shock), each of which permits disinhibition, have in common? Without any reference to Pavlovian cortical dynamics it is possible to describe the outcomes of the preceding experiments in two very different ways. The first is to state that disinhibition occurred when a response which once had a much higher probability of occurrence had been reduced to a much lower probability. Perhaps disinhibition will be observed whenever any operation (excluding injury, death, etc.) accomplishes this kind of change in response probability over time (cf. Valenstein, 1959). Alternatively, one might suggest that disinhibition occurred when the probability of the baseline response was low, but greater than zero (cf. Harrison & Abelson, 1959, p. 20). This latter statement makes no reference to the fact that the probability of response was once at a higher level, although, as was the case with Pavlov’s demonstrations, this was true of the current experiments. In Expt. 6, an attempt was made to test animals for “disinhibition” during early acquisition of the lever-pressing response, when the probability of responding had risen from the operant level to the requisite low level.

DISINHIBITION

365

Initially it was necessary to determine the requisite low level of response probability which permits disinhibition. In Expts. 1 to 4, which shared the same basic experimental procedure, the baseline (dummy) rates during disinhibition testing were very low. These experiments, however, are not the best possible guide, since their design guaranteed that the baseline for the 3 min preceding the disinhibitory stimulus would be zero. Experiment 5, shows clearly that disinhibition can occur when the response rates, over a 3-min interval, are higher than the very low rates in Expts. l-4. In Expt. 5, with 3-min 80-db white noise, significant disinhibition was observed after 60 and 90 min of extinction training. The median response rates for the control group at these two stages of training were 1.7 and 0.3; the mean rates, 2.1 and 0.7. However, after 3 and 30 min of extinction, with median rates of 10.7 and 8.7 (means of 12.6 and 7.8), there was no sign of the disinhibition effect. It thus seems clear that 8 responses per minute represent a rate too high to demonstrate disinhibition, but the effect can be shown with response rates up to at least 2 per minute. Prior to Expt. 6, a pilot study was run to obtain information on how rate changes develop during the early acquisition of the lever-pressing response. Animals were first given 60 food pellets in the Skinner box with the response lever removed and then tested on the following day with the lever returned. Testing simply consisted of counting the number of continuously reinforced responses emitted in a 3-min interval after some particular response criterion. There were 43 Ss in four groups tested after having made 1, 3, 6, or 10 responses. The median response rates after these criteria were 3.0,5.7, 8.7, and 8.5, respectively. Thus, although the response rate after one reinforcement was slightly higher than what would have been ideal, the level of responding was very close to a value in which disinhibition occurred in a between-subject comparison in Expt. 5. At the other end of the scale, the rate after 10 reinforcements was unquestionably outside the limits of previous observations of disinhibition. Therefore, in Expt. 6, experimental groups were added which received the 3-min noise stimulus immediately on making the first or the tenth reinforced response. In addition, experimental and control groups were tested during extinction after having made one reinforced response. Method

In a manner similar to that adopted for the pilot study, animals were first put on a 24-hr feeding rhythm and reduced to 75% of their free-feeding body weight. The Ss were then introduced to the Skinner box, in the absence of the response lever, and given 60 reinforcements on a

366

C. J. BRIMER

I-min VI schedule. On the next day, Ss were reintroduced to the Skinner box, where the response lever was now present, and testing was carried out. Two continuously reinforced experimental groups of 13 and 14 Ss, respectively, were tested with the 80-db noise after either 1 (CRF-I) or 10 (CRF- 10) responses. The 1- and 1O-criterion groups of the pilot study, containing 13 and 12 Ss each, were utilized as controls. Additional Ss were tested after a response criterion of one, but only this first response was reinforced; subsequent lever presses produced neither the magazine click nor a food pellet. There were 18 Ss in the EXT- 1 condition, divided into experimental and control groups of 9 Ss each. Results Figure 7 depicts the median response rates for the various experimental and control groups. It is clear, that in the CRF condition there were no differences between the experimental and control groups tested after one reinforced response. The Ss receiving the test stimulus after the tenth response, however, display a significantly lower response rate than the corresponding control animals, U( 14,12) = 36. Similarly, the extraneous stimulus suppressed responding when it was presented during extinction after one reinforced response, U(9,9) = 12.5. Suprisingly, the control Ss tested during extinction, after one reinforced response, did not differ significantly from animals that continued to receive reinforcement.

TESTED UNDER EXTINCTION RESPONSE

TESTED UNDER CRF ACQUISITION

CRITERION

Fro. 7. Median rate (responses per minute) for experimental and control subjects tested under extinction and continuous reinforcement following response criteria of 1 and 10.

DISINHIBITION

367

Discussion The noise stimulus produced neither inhibition nor disinhibition when the response rate was low, but did produce inhibition when the response rate was high. Discontinuing reinforcement during the test interval also led to an inhibitory effect of noise. This latter outcome presents a new problem. The response rates after one reinforced response are at essentially the same low level, whether or not extinction is introduced. Why, then, does the stimulus have an inhibitory effect only under the extinction procedure? One possible answer is that continuing reinforcement during the test might well offset any basic inhibitory effect of the stimulus. Perhaps the basic effect of an extraneous stimulus is always inhibitory, regardless of the baseline probability of response. The only exception to this possibility in the present data is the case where response probability is low, having once been at a higher level. This exception, of course, constitutes the phenomenon of disinhibition. Although disinhibition was not observed when the response rate was low, the argument could be made that the rates in Expt. 6 were not low enough. However, attempts to lower the response rate further by delaying reinforcement or satiating the animals prior to testing were unsuccessful. Consequently, in a final gesture, seven rats were simply placed in the Skinner box (with no prior magazine training), and noise was presented for 3 min after the first response. The response rates observed during the 3-min test interval were not significiantly greater than zero, and, in fact, were lower (although not significantly) that those of a control group of nine Ss that were similarly treated, but received no noise presentation. Thus, once again, a very low probability of response is not, by itself, a sufficient condition to produce disinhibition. GENERAL

DISCUSSION

Although the preceding experiments were generally conceived of as empirical studies concerned with the effect of extraneous stimulation on operant behavior, it was hoped that they also would cast some light on the Pavlovian concepts of external inhibition and disinhibition. Of course, the data as such, do not constitute a critical test of Pavlovian theory. In fact, it is doubtful if any empirical study might be taken as proof of such a theory. As Konorski (1948) has clearly pointed out, whatever seductive features the Pavlovian neurophysiological model might contain, “there are many mutually contradictory statements which, moreover, do not accord with the general principles of functioning of the nervous system, generalizations which do not fit the facts, and arbitrary adaptation of experimental data to the requirements of the theory (p. 5 I).” However, it

368

C.

J.

BRIMER

is possible to note some similarities between the present findings and Pavlov’s observations. First and foremost, it was repeatedly found that the same extraneous stimulus which decreased response rate in pretest, reliably increased response rate during the test. Second, a relation was found in Expt. 1 between the magnitude of these inhibitory and disinhibitory effects. However, the relation appeared to be monotonic, rather than an inverted-U as Pavlov had suggested. The first experiment also showed that a reliable disinhibition effect could occur when inhibitory operations other than experimental extinction were employed to suppress the baseline rate of responding. It seems clear that the disinhibition observed with a satiation procedure would not have been predicted by Pavlov. Again in contrast to Pavlov, the results of Expt. 2 suggested that familiarity with the stimulus did not attenuate its disinhibition effect. Thus, the decrement in disinhibition that was observed with repeated testing appears to be due to the growing level of inhibition of the response, rather than to adaptation to the testing stimulus. On the other hand, throughout the present studies suppression of responding attenuated rapidly with repeated pretest trials. This suggests that stimulus novelty may play an important role in inhibition but not in disinhibition. Such an asymmetry would be in opposition to Pavlovian theory. In Expt. 3, it was noted that the relative effectiveness of turning a physical stimulus on or off varied with the stimulus modality. This is in distinction to Pavlovian theory, which would predict that the onset of stimuli should consistently be superior to their termination. The similarities of the disinhibition findings with other conditioning studies would have been predicted by Pavlov. However, such parallel results need not suggest any central similarity between conditioning and disinhibition, but might simply reflect the rat’s peripheral sensitivity to different sensory inputs. Such an argument is supported by the results of Expt. 4 where it was found that there was, at best, only a minimal tendency for both inhibition and disinhibition to increase with noise intensity. This finding is in marked contrast to CER conditioning, where far smaller differences in the intensity of a noise CS give rise to conspicuous differences in conditioning. Thus, the results of Expt. 4 seem to suggest, in contrast to Pavlov, that conditioning and disinhibition are not identical in terms of their underlying processes. Throughout this report inhibition has been defined simply as an observed decrease in response rate. It is clear that in each of the first four experiments a substantial degree of inhibition, so defined, was generated prior to testing for disinhibition. Consequently, Expt. 5 attempted to investigate how disinhibition varied with the accumulation of inhibition

DISINHIBITION

369

produced by continued extinction training. This is a question to which Pavlov never specifically addressed himself. However, his early formulation of the disinhibition phenomenon as a case of “inhibition of inhibition” implies that, given a fixed external stimulus as the disinhibitor, the greater the inhibition of the CR, the more difficult it should be to disinhibit it. Indeed, it was observed in Expt. 5 that with the continuation of extinction “below zero,” animals failed to respond to the disinhibiting stimulus. However, it was also observed that the response rate could be substantially reduced without any tendency for the rate to increase in the presence of the stimulus. Thus, the results of Expt. 5 suggested that disinhibition occurred only when responding was reduced to some critically low level. The final series of experiments sought to determine whether this critically low response rate is the sufficient condition for disinhibition. The results suggested that it is not. Rather, it appeared that the response rate must be reduced to the requisite low level from a previously higher value. Presenting an extraneous stimulus to animals with the requisite low response rate when they had never been at a higher level produced either no effect or a decrement in the rate of responding. Thus, without endowing the term ‘* inhibition” with the quasi-neurophysiological properties assigned to it by Pavlov, it is possible to aggree with ‘him that the acceleration in responding produced by extraneous stimuli during extinction is, indeed, a case of disinhibition. Although there are a number of discrepancies between Pavlovian theory and the results of the current studies, it is not obvious how any available psychological theory could incorporate all of the empirical facts contained in the preceding six experiments. The major stumbling block for all theoretical efforts is the paradox to which Pavlov initially drew attention, viz., the fact that the same stimulus will have exactly opposite effects at different stages of training. Since the preceding experiments were completed, a particularly powerful demonstration of these opposing effects has been reported by a number of investigators (Flanagan & Webb, 1964; Hinrichs, 1968; Singh & Wickens, 1969). In all of these studies, it was observed that an extraneous stimulus, introduced when the response rate is low during the early segment of a fixed interval schedule, leads to an increase in responding. On the other hand, if the stimulus is presented later in the interval, when the response rate is high, it produces a decrement in responding. Thus, depending on the location of the test stimulus within a fixed interval cycle, either inhibition or disinhibition is manifested. This type of observation draws clear attention to the possibility that inhibitory and disinhibitory effects may often contaminate studies of operant conditioning. In fact, it would appear prudent to include a

370

C. J. BRIMER

disinhibition control whenever test stimuli are superimposed on a suppressed response baseline (Baum & Gleitman, 1967). Skinner (1936), in taking the concept of inhibition to task, pointed out that inhibition subsumes a number of diverse operations (such as lowering drive level, eliciting incompatible responses, fatigue, etc.) under a single term, the only common property of which is a negative effect on a specified response. But, Skinner (1936) suggested that “the use of a single property of the negativity of the change does not lead to the establishment of a significant class of data . . . [and] it must not be assumed that other properties possessed by one case are common to the class [p. 1281.” However, the present data suggest that there may, indeed, be a common property shared by all these divergent operations- they may all be susceptible to disinhibition. It is clear that disinhibition can be obtained readily in the vast majority of subjects, and it is, therefore, susceptible to continued parametric study, which the present experiments have attempted to initiate. REFERENCES BAUM, M., & GLEITMAN, H. “Conditioned anticipation” with an extinction baseline: The need for a disinhibition control group. Psychonomic Science, 1967, 8, 95-96. BRIMER, C. J., & KAMIN, L. J.Disinhibition, habituation, sensitization, and the conditioned emotional response. Journal of Comparative and Physiological Psychology, 1963, 56, 508-5 16. FLANAGAN, B., & WEBB, W. B. Disinhibition and external inhibition in fixed interval operant conditioning. Psychonomic Science, 1964, 1, 123-124. GAGNE, R. M. External inhibition and disinhibition in a conditioned operant response. Journal of Experimental Psychology, 194 1,29, 104- 116. HANSCHE, W. J., & GRANT, D. A. Onset versus termination of a stimulus as the CS in eyelid conditioning. Journal of Experimental Psychology, 1960, 59, 19-26. HARRISON, J. M., & ABELSON, R. M. The maintenance of behavior by the termination and onset of intense noise. Journal of Experimental Analysis of Behavior, 1959, 2, 23-42. HINRICHS, J. V. Disinhibition of delay in fixed-interval instrumental conditioning. Psychonomic Science, 1968,12, 3 13-3 14. HORNS, H. L., & HERON, W. T. A study of disinhibition in the white rat. Journal of Comparative Psychology, 1940,30, 97-102. HULL, C. L. Essentials of behavior. New Haven: Yale University Press, 195 1. HUNTER, W. S. The disinhibition of experimental extinction in the white rat. Science, 1935, 81, 77-78. KAMIN, L. J. Temporal and intensity characteristics of the conditioned stimulus. In W. F. Prokasy (Ed.), Classical conditioning. New York: Appleton-Century-Crofts, 1965. KAMIN, L. J., & BRIMER, C. J. The effects of intensity of conditioned and unconditioned stimuli on a conditioned emotional response. Canadian Journal ofPsychology, 1963, 17, 194-198. KAMIN, L. J., & SCHAUB, R. E. Effects of conditioned stimulus intensity on the conditioned emotional response. Journal of Comparative and Physiological Psychology, 1963, 56, 502-507. KISH, G. B. Avoidance learning to the onset and cessation of conditioned stimulus energy. Journal of Experimental Psychology, 1955,50, 31-38.

DISINHIBITION

371

KONORSKI, J. Conditioned reflexes and neuron organization. New York: Cambridge University Press, 1948. KUPALOV, P. S., & GANTT, W. The relationship between the strength of the conditioned stimulus and the size of the resulting conditioned reflex. Brain, 1927, 50, 44-52. LOGAN, F. A. & WAGNER, A. R. Direction of change in CS in eyelid conditioning: a supplementary report. Journal of Experimental Psychology, 1962,64, 325-326. PAVLOV, 1. P. Conditioned reJlexes.Translated by G. V. Anrep, 1927, New York: Dover. PAVLOV, 1. P. Lectures on conditionedreflexes. Translated by H. Gantt. New York: Intern. Pubs., 1928. RAZRAN. G. H. Decremental and incremental effects of distracting stimuli upon the salivary CRs of 24 adult human subjects. Journal of Experimentul Pspci~ology, 1939. 24. 647-652. SCHWARTZ, M., & GOODSON, J. E. Direction and rate of conditioned stimulus change in avoidance performance. Psychological Reports, 1958,4, 499-502. SINGH, D., & WICKENS, D. D. Disinhibition in instrumental conditioning. Journal of Comparative and Physiological Psychology, 1968,66, 557-559. SKINNER, B. F. A failure to obtain “disinhibition.“Journa/ ofGenetic Psychology, 1936. 14, 127-135. SWITZER, S. A. Disinhibition of the conditioned galvanic skin response. Journal of‘Gcnetic Psychology, 1933, 9,17-loo. VALENSTEIN, E. S. The effect of reserpine on the conditioned emotional response in the guinea-pig. Journal of Experimental Analysis of Behavior, 1959, 2, 219-225. WENGER, M. A. External inhibition and disinhibition produced by duplicate stimuli. Amcricm Journal WINNICK, W.

response 41, 205-2

YAMAGUCHI, stimuhts 572-577.

of Psychology,

1936,48,446-456.

A., & HUNT, J. McV. The effect of an extra stimulus upon strength of during acquisition and extinction. Journal ofExperimentu/ Psychology, 195 I, 15. H. I., & LADIORAY, G. 1. Disinhibition as a function of extinction trials and intensity. Joltma/ of Comparative and Physiologica/ P.sychology, 1962, 55,

(Received February 11, 1970)