Extinction of the Context and Latent Inhibition

Transcription

1 LEARNING AND MOTIVATION 13, (1982) Extinction of the Context and Latent Inhibition A. G. BAKER AND PIERRE MERCIER McGill University The hypothesis that latent inhibition could be reduced by extinguishing the experimental context, that is, exposing the rats to the context between exposure to the conditional stimulus (CS) and conditioning, was tested. All experiments used the conditioned emotional response procedure. In Experiment 1, extinction was not effective when the animals were exposed to the clicker 40 times off the baseline of responding for food and when the clicker CS was partially reinforced with shocks during the test phase. In Experiments 2 and 3, latent inhibition could be reduced by extinction if the animals were exposed to the CS 24 or 16 times on-baseline, and if continuous reinforcement was used during the test. In Experiments 4, 5, and 6, we attempted to determine which variable was responsible for the discrepant results. In Experiment 4, extinction was effective with 20 or 40 on-baseline exposures to the CS, using continuous reinforcement during the test. In Experiment 5, extinction was not effective with exposure on- or offbaseline, using 24 exposures and partial reinforcement. Finally, in Experiment 6, extinction reduced latent inhibition using continuous, but not partial, reinforcement with 40 exposures off-baseline. From these results, we concluded that Wagner s model of habituation was not sufficient to account for latent inhibition and that a hybrid model, using both associative and cognitive representational processes, was necessary. If animals are exposed to the conditioning stimuli before subsequent conditioning, learning is often retarded. Although the magnitude of the retardation varies, it may be produced by exposing the animals to the conditional stimulus (CS) (Lubow, 1973), the unconditional stimulus (US) (e.g., Baker, Mercier, Gabel, & Baker, 1981), or uncorrelated presentations of both of these stimuli (e.g., Mackintosh, 1973; Baker & Mackintosh, 1977). The focus of this paper is the first of these effects-the interference caused by exposing the animals to the CS prior to conditioning. This effect, called latent inhibition by Lubow and his colleagues (e.g., Lubow & Moore, 1959), has been extensively studied and a number of explanations have been offered for it. These include a general atten- Please address all correspondence and requests for reprints to Dr. A. G. Baker, Department of Psychology, McGill University, 1205 Docteur Penfield Ave., Montreal, P.Q. H3A IBl, Canada /82/ $02.00/O Copyright by Academic Press, Inc. All rights of reproduction in any form reserved.

2 392 BAKER AND MERCIER tional model proposed by Lubow (e.g., Lubow, 1973). However, the explanation that is most relevant here is a unified account of conditioning offered by Rescorla and Wagner (1972) and later expanded by Wagner (1978). The main tenet of Wagner s recent version of the model is that animals can form associations between stimuli and that these associations will be more easily formed if the stimuli, or more properly representations of them, are not already present in the animals short-term memory. Short-term memory can be primed with these representations in at least two ways: (1) by actually presenting the stimuli, or (2) by calling them out of a long-term store as part of some previously learned association. The explanation of latent inhibition comes quite parsimoniously from the second of these assumptions. Associations are formed between the experimental context and the CS during exposure. Later, during conditioning, the context causes a representation of the CS to be primed in short-term memory and this priming retards the formation of associations between this stimulus and any other stimulus. An implication of this argument is that latent inhibition should obey the same laws as do other associative phenomena and there is some evidence that it does. Like other associations, latent inhibition does not dissipate much with the passing of time (Siegel, 1970). If exposure to the CS occurs in a context other than the test context, associations should form between that other context and the CS, thereby resulting in little retardation. There is at least one experimental result that is consistent with such an interpretation but it used an operant task and confounded changes in test context with the novelty of context of exposure (Lubow, Rifkin, & Alek, 1976). Contrary to this finding, using a conditioned suppression procedure Anderson and his colleagues showed that there was little reduction of latent inhibition if conditioning was carried out in a context different from that used for exposure (Anderson, O Farrell, Formica, & Caponigri, 1969; and Anderson, Wolf, & Sullivan, 1969). They did show that the context in which conditioning was assessed might be an important factor. Other procedures which disrupt or weaken the association between context and CS should also reduce latent inhibition. For instance if, during exposure to the CS, the CS is involved in an association with a more salient stimulus than the context, then this association should block or retard formation of the context-cs association and thereby reduce interference. Again there is some evidence for such an effect. Lubow and his colleagues (Lubow, Alek, & Arzy, 1975; Lubow, Schnur, & Rifkin, 1976), Szakmary (1977), and Dot-C (1980) have discovered that if the exposed CS predicts a second CS during exposure, latent inhibition is reduced. There is less evidence on the inverse situation, in which the CS is preceded by another stimulus.

3 CONTEXT AND LATENT INHIBITION 393 Another feature of associations is that they are susceptible to extinction. If the cause of latent inhibition is context-c!3 associations, and if the context but not the CS is presented between exposure and conditioning these context-(x associations should be weakened, and so should latent inhibition. Wagner (1978) has reported experiments in which just such a manipulation has caused a habituated response to dishabituate. Furthermore, Randich (1981) and Baker and Mercier (1982) have found that extinguishing the context by presenting it alone between exposure and conditioning reduces the US preexposure effect. The present series of experiments was designed to determine whether a similar extinction effect could be found for latent inhibition and to delineate the conditions under which such an effect could be found. EXPERIMENT 1 This experiment was designed to determine whether extinction of the context would reduce latent inhibition. The experiment included three groups which were exposed to the CS prior to a conditioning test in which the CS (a clicker) was paired with the US (an electric shock). In Group Immediate, which was a standard latent inhibition group, the conditioning phase began in the session immediately following the last exposure session. Group Extinction received the same treatment as Group Immediate except that they received five sessions of extinction of the experimental context between the last exposure session and the test. Because it is possible that any disruption of conditioning in Group Extinction compared to Group Immediate might be due simply to the passage of time between exposure and test rather than to exposure to the context during this period, Group Home was included. Group Home received the same treatment as Group Extinction except that during the five-session extinction phase these animals were left in their home cages and were not exposed to the context. A final control received no exposure to the context. There are three other features of our procedure that deserve comment. First, prior to the experimental treatment we exposed the animals in all groups to the clicker twice because previous experiments had shown that one or two exposures to the CS would eliminate any disruption of bar pressing by the CS on the first trial. This elimination of suppression is often quite longlasting. This unconditioned suppression of lever pressing can confound demonstrations of latent inhibition. In groups preexposed to the CS this unconditioned suppression habituates, but in nonpreexposed control groups the CS is novel on the first presentation and the animals may be quite suppressed. This suppression is in the same direction as the difference expected in latent inhibition, but it occurs on the first test trial before any differences in conditioning could appear. Thus it is imperative to eliminate this suppression if an accurate estimate

4 394 BAKER AND MERCIER of the magnitude of any latent inhibition is desired. Second, in our previous experiments on the US-preexposure effect, we carried out exposure off-baseline in order to avoid spurious response-shock associations. We used the same procedure here in order to keep both series of experiments comparable and to prevent response-clicker associations (i.e., the animals might associate a given rate of responding with the CS and this might reduce acquisition of suppression). Finally we used a 50% reinforcement schedule (i.e., only half of the clickers were followed by shocks) during the test phase because associative manipulations using partial reinforcement seemed to be more effective in reducing the US preexposure effect (compare Baker et al., 1981, and Randich & LoLordo, 1979). Method Subjects. The subjects were 32 male hooded Long-Evans rats at initial weights of g, obtained from Canadian Breeding Farms Ltd. The animals were individually housed and, after 5 days adaptation to the laboratory, were reduced to 80% of their free-feeding weights. Apparatus. Four Campden Instruments Inc. lever boxes with only the left-hand lever inserted were used. Lever-press responses were reinforced with 45mg Noyes pellets. Scrambled electric shocks (1 ma, 0.5 set) from Grason Stadler E1064GS shockers could be delivered through the grid floors. During the experimental sessions, the overhead houselight was dimly illuminated (General Electric No V bulb in series with a 470-ohm resistance). The 90-set clicker CS was a 70-dB SPL, 40- Hz string of clicks produced by a Campden Instruments Audio generator. The CS was presented through an overhead speaker. Procedure. The rats were magazine trained and trained to press the lever for food on a continuous reinforcement schedule over 3 days (see Baker et al., 1981 for specific details). All experimental sessions except these first three training sessions lasted 50 min. During the next eleven sessions, all animals were reinforced according to a variable-interval (VI) I-min schedule of reinforcement. On the last of these VI training days, the animals were exposed to the clicker twice. These exposures commenced on the 21st and 43rd minutes of the session. The next 10 days were the treatment phase during which the levers were removed from the boxes and the VI food reinforcement schedule was discontinued. The animals were divided into four groups of eight. During the first 5 days of the treatment phase the animals in Groups Extinction and Home received 40 exposures to the clicker (8 per day). Each day these clickers were programmed according to a different irregular schedule. The only temporal constraint in the schedule was that at least 1 min elapsed between successive clicker presentations. The other two groups (Groups Immediate and Control) were placed in the

5 CONTEXT AND LATENT INHIBITION 395 chambers on each of these 5 days but received no programmed stimuli. During the second 5 days, Groups Extinction and Control were placed in the conditioning chambers but received no programmed stimuli. The animals in Group Home were handled and weighed on each of these days but were left in their home cages. The animals in Group Immediate received 40 exposures to the clicker during these 5 days. All animals which received clickers experienced the same schedule of exposure on corresponding exposure days. The next 5 days constituted the test phase of the experiment. During these days the levers were again present and the VI reinforcement schedule was reinstated. Each day, the clicker was presented four times and followed with a shock twice. The sequence of reinforced clickers was or on alternate days. The interval between clickers was 10 min. Response measures and statistical treatment. On all conditioning days, the number of responses during the clicker and the 90 set immediately preceding it (PreCS period) was recorded. The measure of conditioning used was the standard suppression ratio (cf. Annau & Kamin, 1961), which is the proportion of total responses during the PreCS and CS periods which occur during the CS [CS/(CS + PreCS)]. A suppression ratio of 0 represents complete suppression during the CS whereas a ratio of.5 represents no relative suppression. Whenever an animal produced a PreCS score less than 6 this score was not used in the analyses of conditioning. However, these scores were included in all analyses of baseline suppression. To assess general suppression of baseline responding produced by exposure to the shocks, we calculated a measure we call a PreCS suppression ratio (cf. Baker et al., 1981). This measure is similar to Kamin s suppression ratio except that it compares the rate of responding on the last VI training day with the response rate during the PreCS periods of the test day in question. Again, a PreCS suppression ratio of 0 represents complete suppression and a ratio of.5 represents no relative suppression of baseline responding. All comparisons in this and subsequent experiments were carried out using Rodgers (1967) a posteriori method. This method involves calculating a standard analysis of variance but comparing the resultant F ratio with Rodgers (1975) test statistic rather than Fishers. If the F ratio is reliable, then a set of orthogonal comparisons using Scheffe s (1953) method is carried out; these contrasts are then compared with Rodgers (1975) statistic. An advantage of this method is that it uses a decision-based rather than an experiment-based error rate and hence has a relatively constant power to detect a given difference between two means in experiments with varying numbers of groups. A rejection level of.05 was used throughout the experiments.

6 396 BAKER AND MERCIER Results All animals learned to respond for food. On the last VI training day the groups mean response rates varied from 34 to 39 responses per minute (rsp/min) and did not differ reliably [F(3, 28) =.15]. The animals in all groups showed moderate suppression to the clicker on the first preexposure trial (mean suppression =.26). This suppression was largely dissipated by the second trial [mean suppression = 52; F(1, 31) = , pooled across groups]. There were also no between-group differences on the first test trial to the clicker [F(3, 28) = The groups did not differ on suppression to the clicker during this exposure treatment [F(3, 28) = 1.20, pooled across trials]. Figure 1 shows the groups daily mean suppression ratios to the partially reinforced clicker presentations of the test phase of the experiment. The general impression given by this figure is that the control group learned more rapidly than the other three groups, which did not themselves differ. Statistical analyses support this impression. The groups differed reliably on the first three test days [Day 1 F(3, 28) = 4.99, Day 2 F(3, 28) = 50.74, Day 3 F(3, 23) = 3.38, Day 4 F(3, 20) = 1.98, Day 5 F(3, 17) = A posteriori tests indicated that on each of the first three test days, the control group was more suppressed than the other three groups [Day 1 F(3, 28) = 4.94, Day 2 F(3, 28) = 40.31, Day 3 F(3, 23) = Analyses were also carried out on the raw response rates during the CS periods. These rates differed reliably on the second and fourth test days [Day 1 F =.23, Day 2 F = 9.21, Day 3 F = 2.13, Day 4 F = 2.41, Day 5 F = 1.57; all df = 3, 281. On the second test 0 Clicker-Home 0 Clickarl - Clocker FIG. 1. Acquisition of suppression to the clicker for the four groups in Experiment I. Exposure was carried out off-baseline and the test used a partial reinforcement schedule. Clicker-Home = Group Home; Clicker - = Group Extinction; - Clicker = Group Immediate.

7 CONTEXT AND LATENT INHIBITION 397 day the control group (mean = 6 rsp/min) responded more slowly than the other groups [range of means rsp/min; F(3, 28) = On the fourth test day, the control group (mean = 1 rsp/min) was more suppressed than the extinction and the immediate groups [both means = 7 rsp/min; F(3,28) = Although there was a considerable amount of baseline suppression on the later test days (mean baseline suppression ratio =.14 on Test Day 5), there were no reliable differences in baseline suppression on any test day [maximum F(3, 28) = Discussion This experiment speaks to two related predictions arising from Wagner s (1976) analyses of latent inhibition and habituation. First and more central to present concerns is the observation that an extinction treatment in the experimental context did not reduce latent inhibition. This is clearly inconsistent with the associative explanation that claims that latent inhibition occurs because associations form between the CS and context during preexposure and that these associations block later conditioning. But although it is inconsistent with the model, our finding does not constitute disconfirmation. A major problem with manipulations intended to alter the strength of a postulated context-cs association is that, because it is difficult to assess the strength of context-cs associations, it is difficult to independently assess whether a manipulation which is claimed to have reduced such associations was actually effective. Without such independent assessments, associative theory can become a tautology. An associative manipulation such as extinction is carried out and its effect on latent inhibition is assessed; but if the only way to assess the effectiveness of this manipulation is by its effect on latent inhibition, then it becomes impossible to disconfirm the model. If the associative manipulation reduces latent inhibition then this supports the model; however, if it does not reduce latent inhibition then we do not know whether the model was wrong or whether the associative manipulation failed to reduce context-cs associations. Experiment 1 is open to the same criticism, but we have at least some independent evidence to counter this argument. We have recently completed a series of experiments on the US-exposure effect in which we attempted to extinguish context-us associations by extinguishing the context (cf. Baker & Mercier, 1982). In these experiments, the US was an electric shock and associations between it and the context could be measured by assessing fear of the context produced by this context-us association. In these experiments, the extinction procedure was very effective in reducing fear of the context (as measured by suppression of a baseline of lever pressing for food). In the present experiment, we exposed the animals to the CS the same number of times as these other

8 398 BAKER AND MERCIER animals were exposed to the US, we gave the animals the same amount of time to extinguish associations with the context, and we used the same apparatus and other training and test parameters as in the previous experiments. It would seem reasonable to assume that the US would be at least as effective as the CS in reinforcing associations with the context. And, once formed, context-us associations should be at least as resistant to extinction as context-cs associations. If these assumptions are accepted, then it would seem reasonable to assume that the extinction procedure is effective in reducing context-cs associations regardless of whether it reduces latent inhibition or not. A second point arising from the present results concerns Wagner s model of habituation. Wagner claimed that animals habituate to a stimulus because it becomes primed in short-term memory because of the associations which form between it and the context. As a consequence of this analysis, associative manipulations of the context should influence habituation (cf. Marlin & Miller, 1981). The present series of experiments, whose main concern is latent inhibition, may be considered as also providing a series of tests of this habituation model. It will be recalled that there is a strong unconditioned suppression of lever-press responding which occurs when a novel CS is first presented and that this suppression rapidly habituates. In our experiments, this initial presentation of the CS is followed by some associative manipulation of any CS-context associations. The first trial of the conditioning phase, before the first shock has occurred, may then be used to assess whether the associative manipulation has influenced this habituated response. The present experiment is a case in point. Animals which had received 2 trials of clicker presentations either received 40 more trials or, in the case of the control group, received 10 days for the context to extinguish. Yet on the first test trial, there was no evidence that even the animals which had received only two exposures to the clicker followed by 10 days of extinction of the context had dishabituated. This is consistent with the results obtained by Marlin and Miller, who found no effect of extinction on habituation of a startle response (1981, Experiment 4). EXPERIMENT 2 Experiment 1 produced no evidence that extinction of the context could reduce latent inhibition. In spite of our earlier argument that the parameters used in Experiment 1 should have reduced context-clicker associations it is possible that our manipulation was too weak or the associations were too strong for extinction to be effective. In order to determine whether it was possible to extinguish latent inhibition under at least some circumstances we carried out an experiment similar to Experiment 1 but, because the associations in the first

9 CONTEXT AND LATENT INHIBITION 399 experiment may have been too strong, we reduced the number of exposures to the CS in this experiment. In addition, although in our earlier experiments we used partial reinforcement during the test phase (e.g., Baker & Mackintosh, 1979; Baker et al., 1981), most other studies have used continuous reinforcement (e.g., Rudy, Krauter, & Gaffuri, 1976). So we used a continuous reinforcement procedure in this experiment. In our US-preexposure experiments we found that exposing the animals to the CS on the baseline of lever-press responding produced results that were more compatible with the associative account of that phenomenon (cf. Baker & Mercier, 1982); so in this experiment we carried out exposure on baseline. Method Subjects and apparatus. The subjects were 24 male hooded rats obtained and maintained in the same manner as in previous experiments. However, one animal died from a sialo dacryo adenitis (SDA) virus immediately prior to the treatment phase so only seven animals were assigned to the control group. The same four conditioning chambers were used. The clicker and the shock were of the same intensity and duration as in Experiment 1. Procedure. The animals were trained to press the lever in the usual manner and then received 5 days of training on the VI l-min food reinforcement schedule before the exposure phase began. During the exposure phase, which lasted for 3 days, two groups of rats received 24 exposures to the clicker CS (8 per day). The clickers were presented according to three irregular schedules similar to those used in the previous experiment except that the minimum time interval between clicker presentations was 180 sec. The final, control group, which consisted of seven rats, received no programmed stimuli on the first two exposure days but on the third day was exposed to the clicker twice. Following the exposure phase, the control group and one of the groups exposed to the clicker (Group Clicker) received the test phase which consisted of 5 days during each of which the animals received three reinforced clicker presentations. The first clicker presentation began in the 15th minute of the session and the subsequent two occurred at 16min intervals after it. The final group (Group Extinction) received 5 days of VI training following exposure to the clicker and prior to the test phase. Results On the day immediately prior to the test for all groups, the groups differed reliably on rate of lever pressing [F(2, 20) = A posteriori tests indicated that Group Extinction responded at a higher rate (61 rsp/ min). However, it should be noted that because of the extinction phase,

10 400 BAKER AND MERCIER Group Extinction had 5 more VI training days before the test. If the groups response rates are compared on the last day of exposure for all groups, then they do not differ reliably [F(2, 20) = 1.221, although Group Extinction (43 rsp/min) did respond quite rapidly. Suppression to the clicker was assessed on the first trial of the test for all groups. The groups did not differ reliably [F(2, 20) = 1.631, although Group Clicker (suppression ratio = 58) had suppression ratios higher than those of the other groups (means = 51 and 52). The groups daily mean suppression ratios to the clicker are shown in Fig. 2. The extinction procedure was effective in reducing latent inhibition. Statistical analysis supports this contention. The groups differed reliably on the second and fourth test days [Day 1 F(2, 19) =.89, Day 2 F(2, 19) = 9.79, Day 3 F(2, 16) = 2.66, Day 4 F(2, 14) = 7.88, Day 5 F(2, 16) = On the second test day, the control group was more suppressed than the other groups [F(2, 19) = 9.161, while on the fourth test day the control group and the extinction group were more suppressed than Group Clicker [F(2, 14) = As the degrees of freedom indicate, there was a fair amount of baseline suppression in this experiment; so a number of animals did not reach our PreCS period criterion and hence were eliminated from some of the analyses. This suppression was fairly evenly distributed except on Test Day 4 when three of the six totally suppressed animals were in Group Extinction. We analysed both PreCS baseline suppression ratios and raw response rates. The only reliable difference in these measures was in the groups raw response rates on the first test day [F(2, 20) = Not surprisingly, this difference reflected the higher response rates of Group Extinction (56 rsp/min) as A Control 0 Clicker 0 Extinction FIG. 2. Acquisition of suppression to the clicker for the three groups in Experiment 2. Exposure was carried out on-baseline and the test used a continuous reinforcement schedule.

11 CONTEXT AND LATENT INHIBITION 401 compared to the other groups (range of means = 25 to 32 rsp/min) on the last exposure day. Discussion Contrary to Experiment 1, extinguishing the context did reliably reduce latent inhibition. This finding suggests that, at least with some parameters, extinction of the context can reduce latent inhibition. In addition to this finding, there was a nonreliable trend in support of the associative explanation of habituation of unconditioned suppression to the CS. In this experiment, the animals in Group Clicker had suppression ratios higher than those of the other groups, which suggests that the extinction of the context may have reduced habituation to the clicker. EXPERIMENT 3 Experiment 2 showed that extinguishing the CS can reduce latent inhibition. However, there were several problems with interpreting the data. The extinction group received 5 extra days of VI training and had higher response rates prior to, and on, the first test day. The animals in the extinction group received the test phase on the 5 days following the test phase of the other groups. Finally, some of the animals contracted SDA virus during the experiment. This virus produced substantial temporary perturbations of individual response rates. Experiment 3 included three groups. An extinction group was exposed to the CS and then received 5 days of extinction of context while responding for food prior to the test (Group Extinction). A latent inhibition control group was exposed to the clicker and then spent 5 days in home cages prior to the test (Group Home). The final, control, group received two exposures to the CS prior to the test phase. In this experiment the animals received only 16 trials of exposure to the CS in the hope of optimizing the disruptive effects of the extinction procedure. Method Subjects and apparatus. Twenty-four hooded rats similar to those used in the other experiments were used. The apparatus and stimuli described earlier were used. Procedure. All animals were trained to lever press and received seven sessions of VI training using our standard procedure. On the last day of VI training the animals were exposed to the!wsec clicker stimulus twice. The next 2 days were the exposure phase during which two groups received 16 clicker The control group received no programmed stimuli during this phase. For the next 5 days, the control group and one of the exposed groups (Group Extinction) received 5 more days of VI reinforced lever-press responding while the other exposed group remained in the home cages (Group Home). Following this phase

12 402 BAKER AND MERCIER were the 4 test days during each of which all animals received three shock-reinforced clicker presentations, programmed with the same temporal parameters as in Experiment 2. Results Although the control group (39 rsp/min) responded somewhat more rapidly than the other groups (29 and 33 rsp/min) on the last day of VI responding before the test, this difference was not significant [F(2, 21) =.63]. On the first test trial, the groups mean suppression to the clicker varied from.45 to 52 and did not differ reliably [F(2, 21) = The results of the test phase are shown in Fig. 3. The extinction procedure reduced latent inhibition. The mean suppression ratios differed reliably on the second and third test days [Day 1 F(2, 21) = 1.95, Day 2 F(2, 19) = 8.96, Day 3 F(2, 15) = 7.15, Day 4 F(2, 15) =.58]. A posteriori tests indicated that on the second test day Group Control was more suppressed than Group Home [F(2, 19) = 8.711, whereas on the third test day Group Home was less suppressed than the other groups [F(2, 15) = Six animals made less than five responses during the precs on the third test day. Three of them were in Group Home and two were in Group Extinction. Again in this experiment, there was fairly marked baseline suppression during the test phase. On the last 3 test days the control group was less suppressed than the other groups. This difference was only reliable on the last test day [F(2, 21) = I Doys FIG. 3. Acquisition of suppression to the clicker for the three groups in Experiment 3. Exposure was carried out on-baseline and the test used a continuous reinforcement

13 Discussion CONTEXT AND LATENT INHIBITION 403 The results of this experiment are similar to those of Experiment 2. Again, extinguishing the context reduced latent inhibition. Although there were differences in baseline suppression, these differences were not reliable on the crucial days in which there were reliable differences between the groups. In any case, the baseline suppression of the two crucial groups, Group Extinction and Group Home, did not differ reliably on any test day. This experiment was also an appropriate test of the notion that associative manipulations might reduce habituation of the unconditioned suppression of lever pressing caused by presentation of the clicker. Groups Extinction and Home received 18 habituation trials to this CS and then had either 5 days of exposure to the context in which context-cs associations could extinguish or 5 days spent in their home cages. The analysis of suppression to the clicker on the first test trial indicates that there was little evidence of disruption of habituation. EXPERIMENT 4 Extinction of context did little to reduce latent inhibition in Experiment 1, in which we exposed animals to 40 CS presentations off-baseline and used partial reinforcement during the test. In Experiments 2 and 3 we exposed the animals on-baseline to fewer stimulus presentations (16 or 24) and tested with continuous reinforcement. In these experiments extinction was effective in reducing interference. We carried out Experiments 4, 5, and 6 in an attempt to determine which of these three confounded variables might be responsible for the contradictory findings. The simplest associative explanation of the difference in results between these experiments would note that any associations formed between the CS and the context during exposure would be stronger following 40 rather than 16 or 24 exposures. These stronger associations would be more resistant to the 5 days of extinction, thereby producing the contradictory results between the experiments. Experiment 4, which employed a simple factorial design, was carried out to test this notion. If the important difference between the two procedures was the amount of exposure, then it should be possible to reduce the effects of extinction by exposing the animals to more CS presentations. In this experiment, two groups of animals were exposed to the CS 40 times and two groups were exposed to it 20 times. One group of animals which received 40 trials of exposure and one group which received 20 trials received 5 days of context extinction before the test (Groups 40E and 20E). Groups 20 and 40 underwent the test on the day following their 20th and 40th exposure trial, respectively. In this experiment exposure was carried out on-baseline and continuous reinforcement was used during the test. These

14 404 BAKER AND MERCIER are the parameters which produced reliable extinction in Experiments 2 and 3. Method Subjects and apparatus. We used 31 hooded rats which were obtained and maintained in the same manner as in our earlier experiments. The apparatus and stimuli described earlier were again used. Procedure. The animals were trained to lever press for food and received 5 days of VI training following the regimen described earlier. On the last day of VI training all animals were exposed twice to the clicker. Following this came the exposure phase which lasted for 10 days and was carried out on-baseline. During the first 5 days of this phase, Groups 20 and 40 (Group 40 consisted of 7 rats, all other groups had 8) received no programmed stimuli. On each of these 5 days Group 40E received eight clicker presentations programmed with the same constraints as described in our earlier experiments. On the first two days Group 20E received no clicker presentations, on the third day they received four clicker presentations, and on the fourth and fifth days they received eight clicker presentations. The second 5 days of this exposure period were the extinction of context days for Groups 20E and 40E. On these days, these groups received no exposure to the clicker. During each of the second 5 days, Group 40 received eight clicker presentations. Group 20 received no clickers on the first 2 of these days, four clickers on the next day, and eight clickers on the final 2 days. We did not use homecage control groups in this experiment because we chose to equate the groups for total time in the apparatus, total time and experience with the VI schedule, and period in which the conditioning trials occurred. Immediately following the exposure phase came the test phase which lasted for 4 days. On each of these 4 days, each animal received three clicker presentations, each of which was followed by a shock. This phase was identical to the test phases of Experiments 2 and 3. Statistical analysis. Because the design of this experiment was factorial and the hypotheses of interest were consistent with those tested in a factorial design, we carried out daily two-way independent factorial analyses of variance on the conditioning scores. The two factors were number of exposure trials and extinction vs no extinction. All other analyses were the same as discussed earlier. Results All animals learned to respond in the usual manner. On the last VI day, the groups mean response rates varied from 14 to 24 rsp/min but did not differ reliably [F(3, 27) = On the final day of exposure, the groups mean response rates varied from 36 to 41 rsp/min and did not differ reliably [F(3, 27) = A factorial analysis of suppression

15 CONTEXT AND LATENT INHIBITION 405 to the clicker on the first test trial showed no significant differences: main effect for exposure trials [F( 1,26) = 0.231, main effect for extinction [F(l, 26) = 4.211, interaction [F(l, 26) = The groups mean suppression ratios to the clicker during the 4 test days are shown in Fig. 4. Daily analyses of variance indicated that the groups differed reliably on only the third test day [Day 1 F(3, 27) = 1.89, Day 2 F(3, 21) = 1.16, Day 3 F(3, 19) = 5.97, Day 4 F(3, 17) = On Day 3 the comparison representing the main effect for trials (40 vs 20 trials) was reliable [F(l, 19) = , and the main effect for treatment (extinction vs no extinction) was also reliable [F(l, 19) = The interaction was not reliable on that test day [F(l, 19) = As in our other experiments there was considerable baseline suppression on the test days but again this suppression was evenly distributed among the groups and the groups did not differ reliably on our baseline suppression ratio measure on any of the test days [maximum F(3, 27) = 1.54 on Test Day 21. Discussion The results of this experiment were quite clear-cut; 40 trials of exposure produced a stronger latent inhibition effect than 20 trials. Furthermore, and of more concern here, the extinction procedure reduced interference following both 20 and 40 exposure trials. It should be noted that Groups 20 and 40 were both tested immediately following exposure, whereas Groups 20E and 40E were tested 5 days later. If this experiment was considered in isolation then it would be possible to argue that it was not extinction of the context but passage of time which reduced interference..5-0 g.4 l 40 MM 0 40EXT V 201MM V POEXT s a z h.3- z.2-5 I.I - FIG. 4. Acquisition and extinction of suppression to the clicker for the four groups of Experiment 4. Exposure was carried out on-baseline and the test used a continuous reinforcement schedule.

16 406 BAKER AND MERCIER However Experiment 3, which included both extinction and home-cage groups, used parameters similar to those for Groups 20 and 20E of this experiment and found a reliable extinction effect. As well, in Experiment 1 we included extinction, home-cage, and immediate groups and found little evidence of fading of latent inhibition with time. Siegel (1970) also found that latent inhibition did not dissipate with time. As in experiment 2, where extinction was also effective in reducing latent inhibition, there was nothing more than an unreliable trend suggesting possible return of unconditioned suppression on the first test trial (means: Group 40 = 0.52, group 40E = 0.45, group 20 = 0.53, group 20E = 0.47). EXPERIMENT 5 Experiment 4 rules out the simple associative explanation that our differences in the efficacy of extinction procedures reflected the differing numbers of exposure trials and, hence, the strength of context-cs associations. The two remaining variables that could singly or jointly be responsible for the differences are on- vs off-baseline exposure and partial vs continuous reinforcement. In Experiment 5 we decided to investigate the effect of the former variable because in certain of our US-preexposure experiments it appeared that the on- to off-baseline manipulation was an important determinant of the effectiveness of associative manipulations. Experiment 5 was a simple factorial design in which two groups of animals were exposed to 40 clicker presentations on-baseline and two groups were exposed off-baseline. One group which had been exposed on-baseline (Group On-Ext) and one group which had been exposed offbaseline (Group Off-Ext) received five sessions of context extinction in the exposure context following exposure. The other two groups (Groups On-HC and Off-HC) were simply left in their home cages for these 5 days. Following these days, all animals were conditioned to the clicker using our 50% partial reinforcement procedure. We chose to use partial reinforcement and 40 trials of exposure because we wished to determine whether the on- vs off-baseline variable was sufficient to determine whether or not extinction was effective. And the impression given by the previous experiments was that this would be the best combination of parameters to do this. Method Subjects and apparatus. The 32 hooded rats used in this experiment were obtained and maintained in the same manner as in our previous experiments. We used the same apparatus and stimuli described earlier for this experiment. Procedure. The animals were trained to respond for food and received 10 days of VI training prior to the exposure phase. During the last of

17 CONTEXT AND LATENT INHIBITION 407 these VI days all animals received one exposure to the clicker in order to assess whether the groups differed on unconditioned suppression of lever pressing to the clicker. The animals received one preexposure rather than the usual two because all animals were to be subsequently exposed to the clicker so they ail would be expected to habituate to this stimulus. The next 5 days were the exposure phase of the experiment. On each of these days all groups of animals received eight exposures to the clicker, programmed to occur at irregular intervals in the manner described earlier. Groups On-Ext and On-HC received these exposure trials while responding for food whereas Groups Off-Ext and Off-HC received their exposure off-baseline (for these animals the levers were removed and food reinforcement was discontinued). The next 5 days constituted the extinction phase of the experiment. During these 5 days, Groups On-Ext and Off-Ext were placed in the experimental context. For Group On-Ext the lever was present and the animals were allowed to respond for food. Group Off-Ext was also placed in the experimental context but the levers remained out and food reinforcement was not in effect for these 5 days. Groups On-HC and Off- HC were weighed and fed in the normal manner but otherwise were left in their home cages in the colony room for this period. The final 5 days of the experiment were the test days. On each of these test days the animals received two shock-reinforced and two nonreinforced clicker presentations. The sequence of reinforced trials was and on alternate days using the temporal parameters described for Experiment 1. These 5 days were carried out on-baseline for all groups. Results On the final day of VI training, the groups mean response rates varied from 30 to 33 rsp/min and did not differ reliably [F(3, 28) = The acquisition of suppression to the clicker for the four groups is shown in Fig. 5. The computer failed to collect the data for four animals in Group Off-HC on the first test day so the mean for this group on this day includes only four animals. These animals did receive the appropriate experimental events on this day. There was no difference between the groups on the first test trial [F(3, 24) = It is clear from Fig. 5 that our four groups did not differ much from one another and that, compared to the control groups of our earlier experiments, they all learned quite slowly. In general, our analysis of the data from the four groups supports this contention; however, there was a marginally significant difference among the groups on Test Day 2 [Day 1 F(3, 24) = 1.29, Day 2 F(3, 26) = 2.07, Day 3 F(3, 24) =.32, Day 4 F(3, 22) =.47, Day 5 F(3, 22) = No meaningful singificant contrasts could be found on the second test day, but 90% of the variance

18 408 BAKER AND MERCIER 0 ON EXT 0 ON HC 0 OFF EXT l OFF HC I Days FIG. 5. Acquisition of suppression to the clicker for the four groups of Experiment 5. Exposure was on-baseline and the test used a partial reinforcement schedule. among the means on that day was attributable to the greater suppression of Group Off-HC [F(3) 26) = Again, moderate baseline suppression developed in all groups over the test phase. This suppression was fairly evenly distributed; however, Group Off-HC showed somewhat more baseline suppression than the other groups. This difference was only reliable on the first 2 test days [Day 1 F(3, 24) = 2.11, Day 2 F(3, 28) = 2.08, Day 3 F(3, 28) =.79, Day 4 F(3, 28) =.44, Day 5 F(3, 28) =.59]. No meaningful a posteriori contrasts could be found on Test Days 1 and 2 because of the marginal level of these significant F ratios. Discussion This experiment demonstrates that, with 40 exposures to the clicker and a partial reinforcement schedule, extinction of the context does not reduce latent inhibition regardless of whether it is carried out on- or offbaseline. The extinguished and home-cage groups behaved very similarly to one another and to the various clicker groups from the earlier experiments. The only reliable difference in conditioning was between Group Off-HC and the three other groups on the second test day and this difference was in a direction opposite to that predicted by the extinction mechanism. Group Off-HC was slightly more suppressed than the other groups. There was also no evidence of recovery of unconditioned suppression to the clicker whether extinction had been carried out on- or off-baseline. EXPERIMENT 6 Our various experiments on the extinction procedure implicate the schedule of reinforcement during test as the most important variable in

19 CONTEXT AND LATENT INHIBITION 409 the reduction of interference. In each of the three comparisons (Experiments 1 and 5) in which we used partial reinforcement, there was no reduction in interference, whereas in each of the four comparisons in which we used continuous reinforcement, there was a reliable reduction of interference regardless of whether the animals were exposed to the CS 16, 20, 24, or 40 times (Experiments 2-4). The on- vs off-baseline variable does not seem to be important because in Experiment 5 in which on- vs off-baseline exposure was directly compared, there was no evidence of any differential reduction of interference. Although the results of our previous experiments have implicated the schedule of reinforcement during conditioning as the main determinant of the effectiveness of extinction in removing latent inhibition, none of these experiments have directly compared continuous and partial reinforcement. Experiment 6 was carried out to directly compare the two procedures. Accordingly, four groups of rats were exposed to the clicker 24 times. Following exposure, two of these groups (Groups Ext-CRF and Ext-PRF) were given 5 extinction days in the conditioning chambers prior to the conditioning test. The other two groups (Home-CRF and Home-PRF) spent these 5 days in their home cages. Following this extinction phase, one extinction and one home-cage group (Groups Ext- CRF and Home-CRF) received continuously reinforced clicker presentations in the conditioning test, whereas the other two groups received partially reinforced clicker presentations. Exposure and extinction were carried out off-baseline to demonstrate that extinction could be effective with off-baseline exposure and so that the groups would be equated for experience with lever pressing. Method Subjects and apparatus. We used 32 hooded rats that were obtained from the same supplier and maintained in the same manner as in the other experiments. The same apparatus and stimuli were again used. Procedure. The subjects were shaped to respond in the usual manner and, following shaping, given 10 days of training on the VI I-min food reinforcement schedule. These and all subsequent sessions lasted 50 min. Following VI training, the levers were removed and the food reinforcement schedule was discontinued for the exposure and extinction phases of the experiment. The exposure phase lasted for three sessions during which all animals received 24 presentations (S/day) of the 90-set clicker stimulus. In the subsequent 5-day-long extinction phase of the experiment, two groups of eight rats (Groups Home-CRF and Home- PRF) were left in their home cages except for their daily weighing. On each extinction day, the other two groups of eight rats (Groups Ext-CRF and Ext-PRF) received an extinction session during which no stimuli were programmed to occur. Following the extinction phase, the levers were again inserted into the

20 410 BAKER AND MERCIER chambers and the VI food reinforcement schedule was reinstated for the conditioning phase of the experiment. This phase was 5 days long. Each day, the two continuous reinforcement groups (Groups Ext-CRF and Home-CRF) received two shock-reinforced clicker presentations which commenced on the 21st and 43rd minutes of each session. On each conditioning day, the other two groups (Groups Ext-PRF and Home- PRF) received the partial reinforcement schedule. Under this schedule, each day, the animals received four clicker presentations programmed to occur at lo-min intertrial intervals. Each day, two of these presentations were reinforced with shock. The sequence of shock-reinforced clickers was and Results On the last day of VI training, the groups mean response rates varied from 34 to 36 rsp/min and did not differ reliably [F(3, 28) = There was also no difference in suppression to the clicker on the first test trial [F(3, 28) = The results of the conditioning trials to the clicker are shown in Fig. 6. These suppression ratios differed reliably on the fourth and fifth test days [Day 1 F(3, 28) = 1.52; Day 2 F(3, 28) =.14; Day 3 F(3, 25) = 1.31; Day 4 F(3, 27) = 2.92; Day 5 F(3, 26) = A posteriori tests indicated that on Day 4, Group Ext-CRF was more suppressed than the other groups [F(3, 27) = On Day 5, the two continuous reinforcement groups were more suppressed than the two partial reinforcement groups [F(3, 25) = In this experiment, moderate baseline suppression developed during conditioning but again it was fairly evenly distributed among the groups. 0 Ext -Crf I 0 Ext - Prf 0 Home - Crf 9 Home - Prf FIG. 6. Acquisition of suppression to the clicker for the four groups of Experiment 6. Exposure was carried out off-baseline.