THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 1992.44B ( I ) 57-73
Role of Context in Perceptual Learning in Maze Discriminations J. B. Trobalon and V. D. Chamizo University of Barcelona, Spain
N. J. Mackintosh University of Cambridge, U.K.
Three experiments with rats in a maze examined the effects of pre-exposure to the relevant discriminative stimuli (rubber- and sandpaper-covered maze arms) or the extra-maze context (the maze was surrounded either by black curtains or by variety of extra-maze landmarks) on the learning of a discrimination between rubber and sandpaper arms. In Experiment 1, pre-exposure to the extra-maze context facilitated subsequent discrimination learning. Experiments 2 and 3 showed that pre-exposure to rubber and sandpaper arms facilitated subsequent discrimination learning only when these cues were presented in the same context during pre-exposure and discriminative training. Taken together, the results are consistent with the hypothesis that a major cause of perceptual learning is the latent inhibition of stimuli or features common to the two discriminative stimuli, and that such latent inhibition may be disrupted by a radical change of context.
Prior unreinforced exposure to a stimulus that later serves as a conditioned stimulus (CS) in Pavlovian conditioning typically retards conditioning. This is the phenomenon of latent inhibition (Lubow, 1989). But unreinforced exposure to the stimuli that later serve as discriminative stimuli in a discrimination task often facilitates the learning of that discrimination. This is the phenomenon of perceptual learning (Hall, 1980). The apparent conflict between these two results has led to a search for the critical variables that determine the outcome of pre-exposure experiments. One widely accepted suggestion is that pre-exposure may produce a latent inhibition effect when pre-exposure and conditioning occur in the same context, but a perceptual Requests for reprints should be sent to J. B. Trobalon or V. D. Chamizo. Universidad de Barcelona, Departament de Psicologia Basica, Adolf Florensa s/n, 08028-Barcelona, Spain. This research was supported by grants from the U.K. British Council and the Spanish i6n y Ciencia to N. J. Mackintosh & V. D. Chamizo.
0 1992 The Experimental Psychology Society
58
TROBALON, CHAMIZO, MACKINTOSH
learning effect when pre-exposure and discriminative training occur in different contexts (Channell & Hall, 1981; Hall & Honey, 1989). There are, however, several reasons to question this particular resolution. The fact of the matter is that at a theoretical level the conflict between the phenomena of latent inhibition and perceptual learning may be more apparent than real. According to one analysis, latent inhibition occurs because of a decline in the associability of a pre-exposed CS with the outcome of conditioning trials, whereas perceptual learning reflects an increase in the discriminability of the two pre-exposed stimuli from one another (Hall & Honey, 1989; McLaren, Kaye, & Mackintosh, 1989). Moreover, one potential mechanism of this increase in discriminability or decline in generalization may be latent inhibition: if generalization between two stimuli A and B occurs because they share elements in common (each stimulus comprising unique elements a and b and common elements x), then 50 trials of preexposure to both A and B will result in 50 trials of exposure to their unique elements but twice as many trials of exposure to their common elements. If pre-exposure reduces associability, then it will reduce the associability of common elements more than that of unique elements, thus ensuring that the outcome of each trial of discriminative training is preferentially associated with the unique rather than the common elements and reducing generalization between the two stimuli. The clearest demonstration that perceptual learning is more likely to occur when pre-exposure takes place in one context and discriminative training in another was provided by Channell and Hall (1981), who found that pre-exposure to horizontal and vertical striations in animals’ home cages enhanced subsequent discrimination learning, whereas pre-exposure to these stimuli in the test apparatus retarded subsequent learning. Their interpretation of this result rested on the assumption that latent inhibition effects are disrupted by a change of context between pre-exposure and conditioning (Lubow, 1989). But this interpretation is mildly suspect, since McLaren (1990) has shown that prior exposure to the context in which unreinforced presentations of a stimulus subsequently occur results in a latent inhibition effect that generalizes across contexts. By this token, exposure to horizontal and vertical striations in the animals’ home cages, where they had lived all their lives, should not have prevented a latent inhibition effect occurring when the stimuli were subsequently presented in another apparatus. Indeed, Hall and Channell (1986) themselves found no evidence of a reduction in latent inhibition when pre-exposure to flavour stimuli occurred in animals’ home cages and conditioning in another cage. Finally, there are good reasons to suppose that the presentation of a preexposed stimulus in a novel context will result in relatively slow conditioning or discriminative learning. There is ample evidence from conditioning experiments that the experimenter’s nominal CS may compete with contex-
CONTEXT AND PERCEPTUAL LEARNING
59
tual cues for association with the reinforcer (Mackintosh, 1983). A novel context might compete more effectively than one whose associability has also been reduced by prior exposure. Similarly, in discriminative experiments, contextual cues can be regarded as incidental or irrelevant stimuli, attention to which interferes with the acquisition of control by the relevant discriminative stimuli (Sutherland & Mackintosh, 1971). Latent inhibition of such irrelevant stimuli ought only to enhance discrimination learning; thus, learning ought to be particularly slow when the relevant stimuli have suffered a loss of associability through pre-exposure but the irrelevant contextual stimuli have not. We report here a series of experiments that examine some of these questions. All experiments employed a maze and required rats to learn a discrimination between two intra-maze cues, rubber- and sandpaper-covered arms. In Experiment 1, we asked whether pre-exposure to incidental or irrelevant contextual cues would facilitate subsequent discrimination learning. This required the provision of two distinctive contexts. The maze was located in a large, brightly lit room. One context was provided by a variety of extra-maze cues or landmarks. For the other, the maze was surrounded by heavy black curtains, which hung from ceiling to floor. In previous experiments (Chamizo & Mackintosh, 1989; Trobalon, Sansa, Chamizo, & Mackintosh, 1991), we have shown that pre-exposure to the intra-maze cues themselves reliably facilitates subsequent discrimination, i.e. produces a perceptual learning effect. This result is confirmed in Experiments 2 and 3, where we also ask whether changing the incidental, contextual cues between pre-exposure and discrimination training affects the magnitude of this perceptual learning effect.
EXPERIMENT 1 One mechanism proposed by McLaren et al. (1989) to account for perceptual learning is that pre-exposure to two stimuli that share many features or elements in common will result in a greater decline in the associability of their common elements than of their unique elements. Consistent with this account, Chamizo and Mackintosh (1989) and Trobalon et al. (1991) found that pre-exposure was more likely to facilitate maze discriminations when the discriminative stimuli shared more common elements. But the context in which an intra-maze discrimination is learned, here the extra-maze landmarks or the black curtains, constitutes a large set of features or elements effectively common to both positive and negative stimuli. Learning the intramaze discrimination must depend on associating the outcome of each trial with the unique intra-maze cues rather than the common extra-maze stimuli. It follows that one way in which pre-exposure could benefit intra-maze
60
TROBALON, CHAMIZO, MACKINTOSH
discrimination learning is by reducing the associability of these common extra-maze stimuli. In Experiment lA, therefore, we exposed some animals to the intra-maze cues in the presence of the same extra-maze stimuli (landmarks or curtains) as would be present in the second, discrimination phase of the experiment, and others in the presence of the different set. Only in the former group will there be a decline in the associability of the extra-maze stimuli present in discrimination training. Where putative common elements are an intrinsic part of the two discriminative stimuli (for example, the shape and size of the two maze arms), it is difficult, if not impossible, to vary their exposure independently of the unique differentiating features of the two discriminative stimuli. As exposure to contextual stimuli can be manipulated independently of exposure to the relevant discriminative stimuli, the present experimental design should permit a direct test of the proposition that pre-exposure facilitates subsequent discrimination learning by reducing the associability of features or elements common to the two discriminative stimuli. Experiment 1A contained four groups, all pre-exposed to the intra-maze cues and then required to learn the intra-maze discrimination. The design was a 2 x 2 factorial with the presence of extra-made landmarks or of black curtains, in pre-exposure or in discrimination training, as the two factors. The design of Experiment 1B was exactly the same, but in this case all animals were exposed to plain maze arms (without the intra-maze cues) in the first stage of the experiment. In each experiment, the four groups are labelled E-E, C-E, C-C, and E-C, where E=extra-maze landmarks and C=curtains; the letter before the hyphen applies to conditions in pre-exposure, that after the hyphen to the conditions in discrimination training.
Method Subjects and Apparatus. The subjects were 34 male and 36 female Hooded Lister rats approximately three months old at the beginning of the experiment. They were housed in cages of two or three, put on food deprivation until they had been reduced to 85% of their ad lib. weights, and kept at this weight by being fed a fixed amount of food one hour after each day’s trials. Forty-two (22 male and 20 female) rats were used in Experiment 1A and 28 (12 male and 16 female) rats in Experiment 1B. In Experiment IA, they were divided into three groups of 1 1 and one of 9 (group E-E); in Experiment 1B, groups E-E and E-C had 8 subjects each, C-C and C-E 6. The apparatus was an elevated eight-arm radial maze made out of wood, Perspex, and aluminium (see Chamizo & Mackintosh, 1989). Only three of the arms were used on any one trial. They were connected to an octagonal platform 37 cm in diameter, with eight 20-cm-high Perspex guillotine doors, which were
CONTEXT AND PERCEPTUAL LEARNING
61
controlled by a hand-operated device located in the middle of one wall of the experimental room. Each arm was 76 cm long x 9 cm wide and had 2.5-cmhigh hardboard walls along the two sides and far end. One arm, used as the start-arm, was covered with green plastic. The other two were used as goalarms; one was covered with black rubber, the other with dark red sandpaper (aluminium oxide paper, 100-grit). Each goal-arm had a recessed 4.5-cmdiameter food-cup located 4 cm from the far end and triangular baffles 12 cm high at the central platform extending 12 cm along the arm from the platform to prevent rats jumping from one goal-arm to another. The maze was placed in the centre of a room, 3.45 x 3.20 x 2.85 m high, illuminated by an overhead light. The plan of the room and the various extra-maze landmarks are shown in Figure 1. Where called for, the maze could be surrounded by thick black curtains, which hung from ceiling to floor, forming an enclosure 2.5 m square and excluding as far as possible all extra-maze landmarks. N
ar
1 0 -
S
FIG. 1. Plan (to scale) of the experimental room and maze, showing all possible eight arms. 1 =controls for maze doors; 2= coat hanger; 3 =chair and experimenter; 4= table; 5 =desk lamp; 6 = toy animals hanging on wall; 7 = rat waiting cages; 8 =black curtains pulled back and tied up; 9 = wastepaper bin; 10 =bright green blanket hanging on wall, with a small toy animal in front; I I = wastepaper bin; 12 = poster on wall; 13=ashtray; 14 =poster on wall; 15 =lamp with blue light; 16=rack with empty rat cages; 17=poster on door; 18=chair. When the maze was surrounded by curtains, the four curtains at 8 were drawn to form a square enclosure.
62
TROBALON, CHAMIZO, MACKINTOSH
Procedure. On the first day of the experiment, all subjects received two unreinforced pre-test trials. They were placed on the start-arm, pointing NW on one trial and SE on the other, and allowed to run through to the central platform from which the rubber and sandpaper arms ran off NE and SW. Rats were assigned to the four groups in each experiment in such a way as to match their sex, experience, and preferences on these two trials. During pre-exposure all rats received 16 non-reinforced trials daily for eight days, being run in squads of three or four animals. Only one arm (alternately rubber or sandpaper in Experiment 1A, a special goal-arm covered with green plastic in Experiment 1 B) was used on each pre-exposure trial. A trial started by placing a rat at the end of the arm, with all doors closed, facing the central platform; 30 sec later, the rat was picked up from the arm by the experimenter and placed into a waiting cage. When all rats in a squad had visited that arm, it was rotated 45" clockwise, and the same procedure was repeated until all eight possible locations with this arm had been visited by all rats. Exactly the same procedure was then run through again, either with a different arm in Experiment 1A or the same arm in Experiment 1B. Each day started with a different location and with a different arm. At the end of the pre-exposure phase, all subjects received four more nonreinforced trials to re-measure their preferences for the intra-maze cues. The same configurations as in Trials 1-4 on Day 1 of discrimination training were used. Trials 1 and 2 were free, Trial 3 could be free or forced, and Trial 4 was always forced, to equate animals' experience with the two intra-maze cues. In the discrimination training phase, rats were trained against their preference, defined either as the arm chosen on both Trials 1 and 2 of these preference tests, or the arm chosen on Trial 3 if an animal had alternated choices on Trials 1 and 2. This meant that in Experiment IA, groups C-E, C-C, and EC had 6 animals trained with rubber positive and 5 with sandpaper, and for group E-E the numbers were 4 and 5, respectively. In Experiment 1 B, groups E-E, C-E, and C-C had 4 animals trained with rubber positive, and group E-C had 3. During this final, discriminative training phase of the experiment, one goal-arm on each trial was always covered with rubber, the other with sandpaper. The start-arm was green plastic. One goal-arm always pointed to the east comer of the room, whereas the other could point south, south-west, west, or north-west, i.e. making an angle of at least 90" with the first. The start-arm also always made an angle of at least 90" with both goal-arms. Two additional constraints were that the goal-arm pointing to the east was equally often rubber and sandpaper, and a right turn from the start-arm led equally often either to rubber or to sandpaper and either to the east comer or to one of the other directions. A total of 16 different configurations of the maze were used. Rats received 64 trials of discrimination training at the rate of 8 trials
CONTEXT AND PERCEPTUAL LEARNING
63
daily, with the exception of the first two days, when they received only 4 trials. The correct arm was baited with a piece of chocolate, approximately 240mg in weight (a quarter of a coloured candy-coated chocolate); the incorrect arm was unbaited. After a choice, defined as the rat placing all four feet onto a goal-arm, the door to the central platform was closed behind the rat, and no retracing or correction was allowed. The rat was removed from the correct goal-arm after eating the food, and from the incorrect arm after 30 sec.
ResuI ts Figure 2A shows the mean percentage of correct choices throughout discrimination training for the four groups of Experiment 1 A, and Figure 2B for Experiment 1B. It is clear that in both experiments the expected interaction occurred. When animak learned the intra-maze discrimination in the presence of extra-maze landmarks, pre-exposure to those landmarks facilitated learning; when they learned in the presence of the curtains, pre-exposure to the curtains facilitated learning. An analysis of variance of Experiment 1A revealed that although the only significant main effect was days, F(7,266)=30.50, p
Role of Context in Perceptual Learning in Maze Discriminations J. B. Trobalon and V. D. Chamizo University of Barcelona, Spain
N. J. Mackintosh University of Cambridge, U.K.
Three experiments with rats in a maze examined the effects of pre-exposure to the relevant discriminative stimuli (rubber- and sandpaper-covered maze arms) or the extra-maze context (the maze was surrounded either by black curtains or by variety of extra-maze landmarks) on the learning of a discrimination between rubber and sandpaper arms. In Experiment 1, pre-exposure to the extra-maze context facilitated subsequent discrimination learning. Experiments 2 and 3 showed that pre-exposure to rubber and sandpaper arms facilitated subsequent discrimination learning only when these cues were presented in the same context during pre-exposure and discriminative training. Taken together, the results are consistent with the hypothesis that a major cause of perceptual learning is the latent inhibition of stimuli or features common to the two discriminative stimuli, and that such latent inhibition may be disrupted by a radical change of context.
Prior unreinforced exposure to a stimulus that later serves as a conditioned stimulus (CS) in Pavlovian conditioning typically retards conditioning. This is the phenomenon of latent inhibition (Lubow, 1989). But unreinforced exposure to the stimuli that later serve as discriminative stimuli in a discrimination task often facilitates the learning of that discrimination. This is the phenomenon of perceptual learning (Hall, 1980). The apparent conflict between these two results has led to a search for the critical variables that determine the outcome of pre-exposure experiments. One widely accepted suggestion is that pre-exposure may produce a latent inhibition effect when pre-exposure and conditioning occur in the same context, but a perceptual Requests for reprints should be sent to J. B. Trobalon or V. D. Chamizo. Universidad de Barcelona, Departament de Psicologia Basica, Adolf Florensa s/n, 08028-Barcelona, Spain. This research was supported by grants from the U.K. British Council and the Spanish i6n y Ciencia to N. J. Mackintosh & V. D. Chamizo.
0 1992 The Experimental Psychology Society
58
TROBALON, CHAMIZO, MACKINTOSH
learning effect when pre-exposure and discriminative training occur in different contexts (Channell & Hall, 1981; Hall & Honey, 1989). There are, however, several reasons to question this particular resolution. The fact of the matter is that at a theoretical level the conflict between the phenomena of latent inhibition and perceptual learning may be more apparent than real. According to one analysis, latent inhibition occurs because of a decline in the associability of a pre-exposed CS with the outcome of conditioning trials, whereas perceptual learning reflects an increase in the discriminability of the two pre-exposed stimuli from one another (Hall & Honey, 1989; McLaren, Kaye, & Mackintosh, 1989). Moreover, one potential mechanism of this increase in discriminability or decline in generalization may be latent inhibition: if generalization between two stimuli A and B occurs because they share elements in common (each stimulus comprising unique elements a and b and common elements x), then 50 trials of preexposure to both A and B will result in 50 trials of exposure to their unique elements but twice as many trials of exposure to their common elements. If pre-exposure reduces associability, then it will reduce the associability of common elements more than that of unique elements, thus ensuring that the outcome of each trial of discriminative training is preferentially associated with the unique rather than the common elements and reducing generalization between the two stimuli. The clearest demonstration that perceptual learning is more likely to occur when pre-exposure takes place in one context and discriminative training in another was provided by Channell and Hall (1981), who found that pre-exposure to horizontal and vertical striations in animals’ home cages enhanced subsequent discrimination learning, whereas pre-exposure to these stimuli in the test apparatus retarded subsequent learning. Their interpretation of this result rested on the assumption that latent inhibition effects are disrupted by a change of context between pre-exposure and conditioning (Lubow, 1989). But this interpretation is mildly suspect, since McLaren (1990) has shown that prior exposure to the context in which unreinforced presentations of a stimulus subsequently occur results in a latent inhibition effect that generalizes across contexts. By this token, exposure to horizontal and vertical striations in the animals’ home cages, where they had lived all their lives, should not have prevented a latent inhibition effect occurring when the stimuli were subsequently presented in another apparatus. Indeed, Hall and Channell (1986) themselves found no evidence of a reduction in latent inhibition when pre-exposure to flavour stimuli occurred in animals’ home cages and conditioning in another cage. Finally, there are good reasons to suppose that the presentation of a preexposed stimulus in a novel context will result in relatively slow conditioning or discriminative learning. There is ample evidence from conditioning experiments that the experimenter’s nominal CS may compete with contex-
CONTEXT AND PERCEPTUAL LEARNING
59
tual cues for association with the reinforcer (Mackintosh, 1983). A novel context might compete more effectively than one whose associability has also been reduced by prior exposure. Similarly, in discriminative experiments, contextual cues can be regarded as incidental or irrelevant stimuli, attention to which interferes with the acquisition of control by the relevant discriminative stimuli (Sutherland & Mackintosh, 1971). Latent inhibition of such irrelevant stimuli ought only to enhance discrimination learning; thus, learning ought to be particularly slow when the relevant stimuli have suffered a loss of associability through pre-exposure but the irrelevant contextual stimuli have not. We report here a series of experiments that examine some of these questions. All experiments employed a maze and required rats to learn a discrimination between two intra-maze cues, rubber- and sandpaper-covered arms. In Experiment 1, we asked whether pre-exposure to incidental or irrelevant contextual cues would facilitate subsequent discrimination learning. This required the provision of two distinctive contexts. The maze was located in a large, brightly lit room. One context was provided by a variety of extra-maze cues or landmarks. For the other, the maze was surrounded by heavy black curtains, which hung from ceiling to floor. In previous experiments (Chamizo & Mackintosh, 1989; Trobalon, Sansa, Chamizo, & Mackintosh, 1991), we have shown that pre-exposure to the intra-maze cues themselves reliably facilitates subsequent discrimination, i.e. produces a perceptual learning effect. This result is confirmed in Experiments 2 and 3, where we also ask whether changing the incidental, contextual cues between pre-exposure and discrimination training affects the magnitude of this perceptual learning effect.
EXPERIMENT 1 One mechanism proposed by McLaren et al. (1989) to account for perceptual learning is that pre-exposure to two stimuli that share many features or elements in common will result in a greater decline in the associability of their common elements than of their unique elements. Consistent with this account, Chamizo and Mackintosh (1989) and Trobalon et al. (1991) found that pre-exposure was more likely to facilitate maze discriminations when the discriminative stimuli shared more common elements. But the context in which an intra-maze discrimination is learned, here the extra-maze landmarks or the black curtains, constitutes a large set of features or elements effectively common to both positive and negative stimuli. Learning the intramaze discrimination must depend on associating the outcome of each trial with the unique intra-maze cues rather than the common extra-maze stimuli. It follows that one way in which pre-exposure could benefit intra-maze
60
TROBALON, CHAMIZO, MACKINTOSH
discrimination learning is by reducing the associability of these common extra-maze stimuli. In Experiment lA, therefore, we exposed some animals to the intra-maze cues in the presence of the same extra-maze stimuli (landmarks or curtains) as would be present in the second, discrimination phase of the experiment, and others in the presence of the different set. Only in the former group will there be a decline in the associability of the extra-maze stimuli present in discrimination training. Where putative common elements are an intrinsic part of the two discriminative stimuli (for example, the shape and size of the two maze arms), it is difficult, if not impossible, to vary their exposure independently of the unique differentiating features of the two discriminative stimuli. As exposure to contextual stimuli can be manipulated independently of exposure to the relevant discriminative stimuli, the present experimental design should permit a direct test of the proposition that pre-exposure facilitates subsequent discrimination learning by reducing the associability of features or elements common to the two discriminative stimuli. Experiment 1A contained four groups, all pre-exposed to the intra-maze cues and then required to learn the intra-maze discrimination. The design was a 2 x 2 factorial with the presence of extra-made landmarks or of black curtains, in pre-exposure or in discrimination training, as the two factors. The design of Experiment 1B was exactly the same, but in this case all animals were exposed to plain maze arms (without the intra-maze cues) in the first stage of the experiment. In each experiment, the four groups are labelled E-E, C-E, C-C, and E-C, where E=extra-maze landmarks and C=curtains; the letter before the hyphen applies to conditions in pre-exposure, that after the hyphen to the conditions in discrimination training.
Method Subjects and Apparatus. The subjects were 34 male and 36 female Hooded Lister rats approximately three months old at the beginning of the experiment. They were housed in cages of two or three, put on food deprivation until they had been reduced to 85% of their ad lib. weights, and kept at this weight by being fed a fixed amount of food one hour after each day’s trials. Forty-two (22 male and 20 female) rats were used in Experiment 1A and 28 (12 male and 16 female) rats in Experiment 1B. In Experiment IA, they were divided into three groups of 1 1 and one of 9 (group E-E); in Experiment 1B, groups E-E and E-C had 8 subjects each, C-C and C-E 6. The apparatus was an elevated eight-arm radial maze made out of wood, Perspex, and aluminium (see Chamizo & Mackintosh, 1989). Only three of the arms were used on any one trial. They were connected to an octagonal platform 37 cm in diameter, with eight 20-cm-high Perspex guillotine doors, which were
CONTEXT AND PERCEPTUAL LEARNING
61
controlled by a hand-operated device located in the middle of one wall of the experimental room. Each arm was 76 cm long x 9 cm wide and had 2.5-cmhigh hardboard walls along the two sides and far end. One arm, used as the start-arm, was covered with green plastic. The other two were used as goalarms; one was covered with black rubber, the other with dark red sandpaper (aluminium oxide paper, 100-grit). Each goal-arm had a recessed 4.5-cmdiameter food-cup located 4 cm from the far end and triangular baffles 12 cm high at the central platform extending 12 cm along the arm from the platform to prevent rats jumping from one goal-arm to another. The maze was placed in the centre of a room, 3.45 x 3.20 x 2.85 m high, illuminated by an overhead light. The plan of the room and the various extra-maze landmarks are shown in Figure 1. Where called for, the maze could be surrounded by thick black curtains, which hung from ceiling to floor, forming an enclosure 2.5 m square and excluding as far as possible all extra-maze landmarks. N
ar
1 0 -
S
FIG. 1. Plan (to scale) of the experimental room and maze, showing all possible eight arms. 1 =controls for maze doors; 2= coat hanger; 3 =chair and experimenter; 4= table; 5 =desk lamp; 6 = toy animals hanging on wall; 7 = rat waiting cages; 8 =black curtains pulled back and tied up; 9 = wastepaper bin; 10 =bright green blanket hanging on wall, with a small toy animal in front; I I = wastepaper bin; 12 = poster on wall; 13=ashtray; 14 =poster on wall; 15 =lamp with blue light; 16=rack with empty rat cages; 17=poster on door; 18=chair. When the maze was surrounded by curtains, the four curtains at 8 were drawn to form a square enclosure.
62
TROBALON, CHAMIZO, MACKINTOSH
Procedure. On the first day of the experiment, all subjects received two unreinforced pre-test trials. They were placed on the start-arm, pointing NW on one trial and SE on the other, and allowed to run through to the central platform from which the rubber and sandpaper arms ran off NE and SW. Rats were assigned to the four groups in each experiment in such a way as to match their sex, experience, and preferences on these two trials. During pre-exposure all rats received 16 non-reinforced trials daily for eight days, being run in squads of three or four animals. Only one arm (alternately rubber or sandpaper in Experiment 1A, a special goal-arm covered with green plastic in Experiment 1 B) was used on each pre-exposure trial. A trial started by placing a rat at the end of the arm, with all doors closed, facing the central platform; 30 sec later, the rat was picked up from the arm by the experimenter and placed into a waiting cage. When all rats in a squad had visited that arm, it was rotated 45" clockwise, and the same procedure was repeated until all eight possible locations with this arm had been visited by all rats. Exactly the same procedure was then run through again, either with a different arm in Experiment 1A or the same arm in Experiment 1B. Each day started with a different location and with a different arm. At the end of the pre-exposure phase, all subjects received four more nonreinforced trials to re-measure their preferences for the intra-maze cues. The same configurations as in Trials 1-4 on Day 1 of discrimination training were used. Trials 1 and 2 were free, Trial 3 could be free or forced, and Trial 4 was always forced, to equate animals' experience with the two intra-maze cues. In the discrimination training phase, rats were trained against their preference, defined either as the arm chosen on both Trials 1 and 2 of these preference tests, or the arm chosen on Trial 3 if an animal had alternated choices on Trials 1 and 2. This meant that in Experiment IA, groups C-E, C-C, and EC had 6 animals trained with rubber positive and 5 with sandpaper, and for group E-E the numbers were 4 and 5, respectively. In Experiment 1 B, groups E-E, C-E, and C-C had 4 animals trained with rubber positive, and group E-C had 3. During this final, discriminative training phase of the experiment, one goal-arm on each trial was always covered with rubber, the other with sandpaper. The start-arm was green plastic. One goal-arm always pointed to the east comer of the room, whereas the other could point south, south-west, west, or north-west, i.e. making an angle of at least 90" with the first. The start-arm also always made an angle of at least 90" with both goal-arms. Two additional constraints were that the goal-arm pointing to the east was equally often rubber and sandpaper, and a right turn from the start-arm led equally often either to rubber or to sandpaper and either to the east comer or to one of the other directions. A total of 16 different configurations of the maze were used. Rats received 64 trials of discrimination training at the rate of 8 trials
CONTEXT AND PERCEPTUAL LEARNING
63
daily, with the exception of the first two days, when they received only 4 trials. The correct arm was baited with a piece of chocolate, approximately 240mg in weight (a quarter of a coloured candy-coated chocolate); the incorrect arm was unbaited. After a choice, defined as the rat placing all four feet onto a goal-arm, the door to the central platform was closed behind the rat, and no retracing or correction was allowed. The rat was removed from the correct goal-arm after eating the food, and from the incorrect arm after 30 sec.
ResuI ts Figure 2A shows the mean percentage of correct choices throughout discrimination training for the four groups of Experiment 1 A, and Figure 2B for Experiment 1B. It is clear that in both experiments the expected interaction occurred. When animak learned the intra-maze discrimination in the presence of extra-maze landmarks, pre-exposure to those landmarks facilitated learning; when they learned in the presence of the curtains, pre-exposure to the curtains facilitated learning. An analysis of variance of Experiment 1A revealed that although the only significant main effect was days, F(7,266)=30.50, p
