Brain Research Bulletin 67 (2005) Toronto, Ont., Canada M6A 2E1 b Department of Psychology, University of New Mexico, Albuquerque,

Brain Research Bulletin 67 (2005) 62 76 Differential contributions of hippocampus, amygdala and perirhinal cortex to recognition of novel objects, contextual stimuli and stimulus relationships Sandra N. Moses a,b,, Carrie Cole b, Ira Driscoll c, Jennifer D. Ryan a a Rotman Research Institute, Baycrest Centre for Geriatric Care, 3560 Bathurst St., Toronto, Ont., Canada M6A 2E1 b Department of Psychology, University of New Mexico, Albuquerque, NM 87131, USA c Department of Psychology and Neuroscience, University of Lethbridge, Lethbridge, Alta., Canada Received 24 May 2005; received in revised form 24 May 2005; accepted 25 May 2005 Available online 11 July 2005 Abstract This study examined contributions of the hippocampus, amygdala and perirhinal cortex to memory. Rats performed a cover task, and changes to stimulus identity or relationships were used to test incidental memory. Rats with hippocampal damage showed deficient responses to relationship changes, but demonstrated knowledge of the position and identity of the target object. They over-focused on the most predictive stimuli, and failed to acquire associations including surrounding cues. Rats with amygdala damage responded to changes involving distal stimuli, and showed deficient responses to novel objects and object relationships. These rats may be highly reliant on relational representations, resulting in a reduced salience for individual novel stimuli. Rats with perirhinal damaged responded to novel stimulus relationships and distal cues, but showed deficient responses to novel objects, suggesting that changes in identity had reduced salience. Implications for declarative and conjunctive hippocampal theories are discussed. 2005 Elsevier Inc. All rights reserved. Keywords: Learning; Memory; Novelty; Medial temporal; Incidental; Relational 1. Introduction Different neural structures may be involved in learning about different types of stimuli. The current study examines the effects of hippocampal, amygdala and perirhinal cortex lesions on recognition of changes in proximal objects, distal cues and the relationships among them. Evidence suggests that these structures are differentially involved in learning about stimulus relationships, distal stimulus identity and proximal object identity. Corresponding author. Tel.: +1 416 785 2500x3362; fax: +1 416 785 2862. E-mail address: smoses@rotman-baycrest.on.ca (S.N. Moses). 1.1. Hippocampus Theories regarding the role of hippocampal function have implicated this structure in configural, conjunctive and declarative memory. While these theories have subtle differences regarding the nature of the representations that are maintained by the hippocampus, all theories have a common thread in postulating that the hippocampus is critical for learning about relationships among stimuli [24]. An essential feature of this type of learning is that associations among multiple elements are combined to form a unique long-term memory representation [9,11,37 39,44]. Memory for spatial locations often requires relational memory, specifically when the subject uses cues to navigate to a goal from several different start locations [22,30,42]. 0361-9230/$ see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.brainresbull.2005.05.026

S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 63 However, some types of navigational tasks can be nonrelational in nature, and remain intact after hippocampal lesions. For example, rats with fornix lesions show normal performance on the Morris water task if they are released from the same start location and required to swim to the same goal location each time. They are subsequently impaired relative to controls when starting from a novel location. Compared to controls, their behavior is abnormally dependent on the particular cues directly in front of them [12]. This likely occurs because, without a relational representation, animals can only blend the distal cues, as viewed from one vantage point, into a unified representation. A relational representation is often composed of more than just spatial cues. For example, the simultaneous representation of discrete predictive cues and surrounding non-essential contextual information requires an intact hippocampus. Rats with hippocampal damage can learn a simple association between contextual cues and an aversive event, such as a shock. However, if a discrete stimulus is added that immediately precedes the aversive event, such as a tone, rats with hippocampal damage show fear responses to the tone, and not to the contextual cues [32,33]. Thus, rats with hippocampal damage over-focus on the stimulus most directly predictive of the event, and fail to acquire a relational representation including associations involving the surrounding cues. A similar phenomenon occurs with latent inhibition. Latent inhibition occurs when previous exposure to a stimulus results in slower acquisition of conditioned responses to that stimulus when it is paired with a positive or negative outcome. Hippocampal lesion rats show normal latent inhibition for discrete stimuli paired with reward. For example, pre-exposure to stimulus X leads to slower response acquisition rates for stimulus X compared to stimulus Y. However, hippocampal lesions lead to impaired latent inhibition for the stimuli surrounding the rewarded object. For example, following pre-exposure to stimulus X in context A rats with hippocampal lesions do not show normal reductions in response acquisition rates when trained in context A compared to context B [15]. They also fail to show a normal disruption in responding to a rewarded stimulus when transferred to a different context [15]. Again, these results suggest that the rats with hippocampal damage focus only on the cues that are most predictive of outcome, and fail to form a representation including the less-predictive surrounding cues. Taken together, the previous studies support the idea that rats with hippocampal damage lack a relational representation of the environment. Rather, they form a rigid association comprised of the stimulus/stimuli most immediately predictive of outcome. This representation likely excludes the relationships among the predictive stimulus/stimuli and the surrounding cues, the nature/identity of the surrounding cues and the relationship of the surrounding cues to each other. Thus, we anticipated that in a distraction paradigm rats with hippocampal damage would show impaired incidental learning about all cues surrounding those that immediately control their behavior. They were expected to be disrupted following changes to those essential cues, and not following changes in surrounding cues. 1.2. Amygdala The amygdala is important for associating objects or spatial relationships with positive or aversive outcomes in a passive classical conditioning setting [17,20,21,33,43].However, it is unnecessary for remembering information about objects or spatial relationships [1,16,20,28,35,47]. There is evidence that amygdala lesions may lead to subtle deficits in spatial navigation due to an impaired ability to form a meaningful association between a specific location and its predicted valence [23]. The amygdala also appears to mediate fear responses to novel stimuli or neophobia. Rats with amygdala lesions show attenuated neophobia for novel foods [5,10,36,43].In a familiar open-field setting, amygdala damage leads to abnormal responding to a novel proximal object, but exploration of novel distal cues remains unaffected [25]. This pattern of results may occur because the hippocampus can mediate responses to novel contextual information in the absence of an intact amygdala. In fact, the hippocampus and amygdala have competitive interactions for control of behavior. McDonald and White [21] demonstrated that additional relational knowledge of the cues in the test room, acquired in a pre-exposure session, led to slower acquisition of a cue preference task in normal rats. Disruption of the hippocampal system eliminated competition between amygdalaand hippocampal-mediated strategies, and hence rate of task acquisition was enhanced in fornix lesion rats. Conversely, it is possible that animals with amygdala damage may focus more on context and stimulus relationships, rather than on individual proximal items. We expected that rats with amygdala damage would show normal incidental learning about relationships among cues, but may show impaired responding to individual novel proximal objects. We expected them to show normal disruption by the presence of novel stimulus relationships and novel contextual stimuli, as these behaviors may be mediated by the hippocampus. However, abnormal responses were anticipated in response to individual novel objects. 1.3. Perirhinal cortex The perirhinal cortex is important for perception and memory of object features and identity but not spatial arrangements [3 5,7,8,13,18,19,23,27,29,34,46]. However, perirhinal cortex lesions can lead to subtle deficits in memory for spatial locations. This may occur because perirhinal lesions impair processing of cue identity information, which normally makes a non-essential contribution to solving a task that requires the formation of relationships among these cues [6,18,23,27]. It appears that when object identity becomes more important for solving a spatial task the perirhinal cortex becomes more essential to performance (for review,

64 S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 see [2]). The perirhinal cortex does not appear to be essential for relational memory [4], unless the only solution to a relational task requires simultaneous discrimination between several objects [23]. We expected that rats with perirhinal cortex lesions would show normal incidental learning about relationships among cues. On the other hand, we expected them to show impaired disruption following changes to identity or form of stimuli. 1.4. Behavioral predictions In the present study, rats were trained on a simple cover task. They were required to displace one object among five to obtain food reward. This is a goal-correlated task, as opposed to incidental, due to the explicit pairing of a specific behavioral response with a desired outcome. However, incidental learning about the non-rewarded cues can be measured using this paradigm. Once the rats acquired this task, different distal and proximal cues and relationships were manipulated as follows. The testing apparatus was rotated 90 relative to the distal cues in the room, and it was placed into a novel room with novel distal cues. These manipulations led to a change in the relationships between the proximal objects and the distal stimuli in the room. A familiar non-rewarded proximal object was replaced with a novel object, leading to a change in proximal stimulus relationships and non-essential object identity information. All of the objects were rearranged, including the target, which led to changes in relationships between the objects and with other proximal and distal stimuli. Object identity remained constant. All of the familiar objects, including the target, were replaced with novel ones, which led to a change in relationships between the objects and other proximal and distal stimuli, as well as changes in object identity information. The degree of disruption following each manipulation was recorded in terms of latency to displace the object, number of contacts with proximal objects prior to displacing the object, number of rears and number of errors. Disruption of task performance was taken as evidence that the animals detected novel stimuli or relationships. Increases in latency to displace the object were interpreted as the animals becoming distracted from the target object by salient changes in stimulus relationships or identity. Contacts with specific objects were interpreted as exploration of the proximal stimuli and rearing as exploration of distal stimuli, as in ref. [25]. Increases in errors were interpreted as evidence that behavior was controlled by several sets of stimuli, such as the stimuli comprising the target object and stimuli surrounding the target object (contextual cues). This could only become apparent when the manipulations led to a novel conflict between previously compatible stimulus arrangements, such as rotation of the apparatus relative to the distal stimuli. Our hypotheses were as follows: (1) animals with hippocampal damage would show abnormal disruption following changes to all contextual stimuli surrounding those cues immediately controlling their behavior, and would be only disrupted following changes to essential stimuli. This means that they would show no change in task performance in response to the rotation, novel room or non-rewarded novel object. They would only be disrupted when all the objects, including the target, are rearranged or replaced. (2) Animals with amygdala damage would show normal disruptions in the presence of novel distal stimuli and distal relationships, but would show abnormal disruption in the presence of novel proximal objects. This means that they would show normal disruption in response to the rotation and novel room, and would show impaired disruption for a novel non-rewarded object, rearrangement of all objects and all novel objects. (3) Animals with perirhinal cortex damage would show normal disruption in the presence of novel distal stimuli and distal relationships, but would show abnormal disruption in the presence of novel proximal objects. This means that they would show normal disruptions following the rotation, and the novel room, and would show impaired disruption for a novel non-rewarded object, rearrangement of all objects and all novel objects. 2. Methods 2.1. Subjects Thirty-two male Long Evans hooded rats were housed individually on a 12-h light:12-h dark cycle. Eight rats were given NMDA neurotoxic lesion of the amygdala, eight were given NMDA neurotoxic lesions of the hippocampus, eight were given NMDA neurotoxic lesions of the perirhinal cortex and eight were given sham surgery. Throughout the duration of testing, rats were maintained at 85% of their original bodyweight. 2.2. Surgery Rats were induced into anesthesia with 4% halothane, 2% oxygen and maintained at 2 2.5% halothane, 2% oxygen. They were placed into a stereotaxic apparatus. The scalp was opened using a scalpel and holes were drilled into the skull. Cannuli were then placed into the holes and lowered into the brain. A 10 mg/ml NMDA in buffered saline solution was injected through the cannuli using a micro infusion pump. For hippocampal lesions, the injection sites in relation to bregma were as follows: (1) anterior 3.1, lateral ±1.5, ventral 3.6; (2) anterior 4.1, lateral ±3.0, ventral 3.0; (3) anterior 5.0, lateral ±3.0, ventral 4.0; (4) anterior 5.3, lateral ±5.2, ventral 7.3; (5) anterior 6.0, lateral ±5.0, ventral 7.3. NMDA was injected at a rate of 0.15 l/min for 4 min. For amygdala lesions, the injection sites relative to bregma were as follows: anterior 2.5, lateral ±4.7, ventral 9.4. NMDA was injected at a rate of 0.125 l/min for 4 min. For perirhinal cortex lesions, the injection sites relative to bregma were as follows: (1) anterior 5.0, lateral ±5.4, ventral 8.2; (2) anterior 6.8, lateral ±5.4, ventral 7.5.

S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 65 NMDA was injected at a rate of 0.1 l/min for 4 min into site 1 and 3 min into site 2. Cannuli were left in place for 3 4 min following each injection to allow the NMDA to diffuse into the tissue. The cannuli were then removed. The incision was closed using wound clips. Rats were given 0.2 cm 3 of benzodiazepine upon waking and then 0.1 cm 3 as needed to control seizures. 0.1 cm 3 morphine was given every 4 h to provide analgesia. 2.3. Histology All the rats were anesthetized with halothane, and were perfused intracardially with saline followed by a 10% formalin solution. The brains were removed and stored in a 10% formalin solution for 1 week. They were then placed in a formalin, 20% sucrose solution for 48 h. Using a cryostat they were sectioned coronally at 40 m and stained with cresyl violet. A microscopic examination was performed in order to assess the location and size of the lesions, in a similar manner to previous work [45]. 2.4. Behavioral testing 2.4.1. Apparatus Rats were placed on a plastic cart (71.5 cm 64 cm 80 cm) which had outer edges extending 7 cm high. The cart was placed in a room in which all the distal cues and the relationships among them were fixed. The cart remained in a fixed position relative to the cues in the room throughout training. Five different plastic objects (about 2.5 5 cm across and 4 5.5 cm high) were placed along one of the long edges of the cart. Objects were placed about 8 cm apart. All objects contained a hole at the bottom, but were solid on top and one all sides (see Fig. 1). Duct tape was placed on one side of each Fig. 1. Schematic of room in which incidental learning experiment was conducted. Black shapes on the cart represent the objects used. object, on the outside facing away from the rat, connecting it to the cart. This was done in order to secure one side of the objects onto the cart, so that the objects could be tipped over but not moved. During the first week of training, an additional piece of duct tape was placed on the inside of all but one of the objects. This piece was placed across from the first piece of tape and secured to the cart, so that the objects could not be knocked over or moved, and the tape could not be seen. A small piece of Froot Loops cereal was attached to the top of the inside of each object where the rats could not reach it. This ensured that all of the objects smelled like Froot Loops. A radio was played in the room throughout the experiment to provide background noise. 2.5. Procedure For each rat, one of the five objects always had one accessible piece of Froot Loops cereal underneath it, and the other four never did (although all objects contained an inaccessible piece of cereal as described earlier). Rats were randomly assigned one of two objects as their correct object. For half of the rats, food was always under the second object from the left (object 2) and for the other half, food was always under the second objects from the right (object 4). Objects remained in a constant position throughout training. Each rat was placed on the cart at the center of the long side of the cart, opposite to the row of objects. They were placed on the cart with their back to the objects, and given 120 s to knock over the correct object in order to obtain food. Latency to knock over the object was recorded, as well as which objects the animals contacted, number of contacts and number of rears. Contacts with the object included touching of the objects with the forepaws or snout. Rearing consisted of the standing on the hind legs with forepaws in the air, and head tilted upwards. If the rat failed to knock over the correct object in 120 s, the object was knocked over by the experimenter. The rat was then given 60 s to eat the food, and then returned to their cage. Each rat was given 10 such trials per day. In between each trial for each rat, the objects and the cart were cleaned with a 0.1% alcohol solution. During the first week of training, only the correct object could be knocked over. Following this period, all objects could be knocked over, but only one contained accessible food. If the rat knocked over an incorrect object, it was immediately removed from the cart and placed in its cage. Number of correct responses for each rat was recorded. Rats were run in squads of eight animals. Each squad contained two rats with amygdala lesions, two with hippocampal lesions, two with perirhinal cortex lesions and two with sham lesions. Once each squad obtained an average correct responses/day of over 99%, and an average latency to knock over the object of below 10 s, different cue manipulations were carried out. (1) The cart was rotated 90 relative to the distal cues in the room, but the relationships among the proximal objects relative to each other and the cart remained constant. (2) The cart was placed in a new room that contained

66 S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 entirely new distal cues. The relationships among the objects relative to each other and the cart remained constant. (3) The center object (object 3), which was a red triangle, was replaced with a novel object, a slightly larger green cylinder. (4) The positions of the objects were rearranged pseudorandomly on each trial. (5) All of the objects were replaced with novel objects whose positions were pseudo-randomly varied on every trial. 2.5.1. Rotation, novel room and novel object For the first three manipulations, each change was maintained for all 10 trials on that day, and then on the following day all of the cues were returned to their original state, and each rat was given only 5 trials. If the average latency per squad was above 10 s on this day, an additional day of five trials with the cues in their original state was added. parietal cortex (Fig. 2). One rat suffered an infarct, which caused right visual cortex, right subicular cortex and right retrosplenial damage, as well as thinning of overlying cortex on the right. This animal, however, had 100% bilateral anterior, 100% left posterior and 50% right posterior hippocampal damage, and was not removed from the data analysis. 3.1.2. Amygdala The average lesion involved 65% of the anterior portion, 62% of the posterior portion and 64% of the entire amygdala. Some lesions included one of the following additional sights of damage: slight ventral caudate damage, slight disruption 2.5.2. All objects rearranged For this manipulation the object that had been correct throughout training (either two or four, depending on assignment) remained the correct choice regardless of its location. Thus, reward was predicted by object identity and not position. For the first trial on this day, rats were able to knock down all of the objects. However, for the remaining 9 trials, and for all 10 trials on the next day, only the correct object could be knocked over. On the following day, all objects could be knocked over, and correct choice was recorded. On the next day all objects were returned to their original position and remained there for all five trials. 2.5.3. All new objects In this case, the position where the correct object had been throughout training was the correct choice. Thus, object position predicted reward, regardless of object form. For the first trial on this day, rats were able to knock down all of the objects. However, for the remaining 9 trials, and for all 10 trials on the next 3 days, only the correct object could be knocked over. The rats were then given 5 days of 10 trials on which all objects could be knocked over. 3. Results 3.1. Histology 3.1.1. Hippocampus One of the eight animals that received hippocampal lesions was removed from the data analysis. This animal had an asymmetrical lesion encompassing lateral temporal and perirhinal cortices, right lateral occipital cortex and 20% of CA1 cells on the right side. The majority of the hippocampus was spared bilaterally. For the seven remaining animals, the average lesion involved 90% of the anterior portion, 78% of the posterior portion and 84% of the entire hippocampus. The largest lesions included damage to the entorhinal cortex, ventral subicular damage and thinning of the posterior Fig. 2. Examples of: (a) hippocampus, (b) amygdala and (c) perirhinal cortex lesions.

of left stria terminalis, right pyriform damage and very slight thinning of the ventral bank of the perirhinal cortex. Some lesions included sparing of the lateral nucleus, left endopyriform nucleus and right medial nucleus (Fig. 2). S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 67 3.1.3. Perirhinal cortex The average lesion included 61% of the anterior portion, 68% of the posterior portion and 61% of the entire perirhinal cortex. Larger lesions included slight CA1 damage, damage to the corner of left posterior caudate, pyriform damage, TE3 and partial TE1 damage, ventral subiculum damage, entorhinal damage, ventral hippocampal damage and overlying cortical damage (Fig. 2). 3.2. Behavioral results 3.2.1. Interpretation of the different measures Increases in latency to displace the object were interpreted to suggest that the rats were distracted from the task by a particular manipulation. Contacts with the objects were interpreted as exploration directed at specific objects. Rearing was interpreted as exploration directed at distal cues. Decreases in the number of correct responses on the task following a manipulation were interpreted as a conflict between different sets of stimuli controlling behavior. For example, if behavior was controlled by an object and a set of distal cues, and when those distal cues changed position relative to the object the rats failed to correctly select the rewarded object, this suggests that the rats continued to respond to the reward contingency of the distal cues rather than to the object. The rats showed very little rearing in the current experiment and few manipulations elicited rearing. Therefore, rearing results are presented merely for descriptive purposes, and are not statistically quantified. Similarly, rats usually showed performance accuracies of 100% correct responses, and few manipulations lead to errors. Thus, error results are also presented for descriptive purposes only and not statistically quantified. The measures of latency and contacts with the objects were sensitive to all five manipulations. This leaves two variables and five different manipulations, which yields 10 overall ANOVAs. The conservative Bonferroni corrected p value thus becomes 0.005. Most of the comparisons yielded Fig. 3. Correct responses on the first trial of the incidental learning task. All groups acquired this task with nearly 100% accuracy. Following the rotation of the cart relative to the distal cues the rats with sham, amygdala and perirhinal cortex lesions showed a decrease in number of correct responses. The rats with hippocampal lesions did not make any errors. Following the rearrangement of all the familiar objects, including the goal object, all groups showed a decrease in number of correct responses. Following the replacement of all the familiar objects with novel ones, including the goal object, all groups showed a decrease in the number of correct responses. p values of below 0.005. However, since this is a rather stringent test for significance, we present all p values below 0.05 as well, but point out that these comparisons exceeded the allowable value according to the Bonferonni correction. 3.2.2. Baseline All groups acquired the task similarly, and performed with nearly 100% accuracy (see Fig. 3). Decreases in latency to displace the object and in the number of contacts with the objects occurred from the first 5 days to the last 5 days, and from the first trial to the last trial on each day (see Tables 1 and 2 for overall ANOVA). None of the groups showed any rearing on the last 5 days before the rotation, suggesting that they were not exploring the distal cues. 3.2.3. Rotation The rotation of the apparatus relative to the distal cues entailed a change in the relationships among the proximal objects on the apparatus and the distal cues in the room. This change led to a conflict of distal and proximal information for solving the task. That is, the location of the rewarded object Table 1 Overall ANOVA for contacts with the objects following each manipulation compared to the previous day over the first five trials Latency Habituation Rotation Novel room Novel object Rearrangement All new objects Group d.f. = 3,27 N.S. 142.163 ** 106.871 ** N.S. N.S. N.S. Day d.f. = 1,27 261.776 ** 88.574 ** 58.783 ** N.S. 48.710 ** 86.939 ** Trial d.f. = 4,108 44.693 ** 77.154 ** 19.590 ** N.S. 7.932 ** 4.297 ** Group day d.f. = 3,27 N.S. 5.595 ** 6.258 ** N.S. N.S. N.S. Group trial d.f. = 12,108 N.S. 6.761 ** 4.871 ** N.S. N.S. N.S. Day trial d.f. = 4,108 25.261 ** 71.255 ** 43.676 ** N.S. 10.655 ** 4.038 ** Group trial day d.f. = 4,108 N.S. 6.469 ** 4.941 ** N.S. N.S. N.S. ** Denotes significance at p < 0.005 in accordance with Bonferroni correction.

68 S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 Table 2 Overall ANOVA for latency to displace the object following each manipulation compared to the previous day over the first five trials Contacts Habituation Rotation Novel room Novel object Rearrangement All new objects Group d.f. = 3,27 N.S. 32.123 ** N.S. N.S. N.S. N.S. Day d.f. = 1,27 88.4 ** N.S. 35.688 ** 13.700 ** 213.913 ** 131.545 ** Trial d.f. = 4,108 7.103 ** 9.563 ** 59.917 ** 2.680 * 5.063 ** 6.166 ** Group day d.f. = 3,27 N.S. N.S. N.S N.S. N.S. N.S. Group day d.f. = 12,108 N.S. N.S. N.S. N.S. N.S. N.S. Day trial d.f. = 4,108 5.045 ** 9.414 ** N.S. N.S. 6.124 ** 5.623 ** Group trial day d.f. = 4,108 N.S. N.S. N.S. N.S. N.S. N.S. * Denotes significance at p < 0.05. ** Denotes significance at p < 0.005 in accordance with Bonferroni correction. was different relative to the distal cues, but not relative to the proximal cues. This manipulation permitted an investigation of the extent to which the animal s behavior was controlled by the distal proximal cue relationships, the distal cues alone or the proximal cues. Following the rotation of the cart relative to the cues in the room, performance of the sham, amygdala and perirhinal cortex lesion groups was disrupted, suggested that these rats detected the change in stimulus relationships. These groups showed increases in latency to displace the object, contacts with the objects (Fig. 4) and incorrect responses (Fig. 3). In contrast, the hippocampal damaged group showed very little or no disruption on these measures (Table 3; see Tables 1 and 2 for overall ANOVA). Fig. 4. Behavioral changes in response to the rotation of the apparatus relative to the distal cue. Differences between rotation and baseline 1 are represented in (a and b). Rats with sham, amygdala and perirhinal cortex lesions showed significant increases in (a) latency to displace the object and (b) contacts with the objects on the first trial, but rats with hippocampal lesions did not show the same increase as the other groups. Follow-up ANOVA comparing performance on only the first trial following the manipulations revealed increased contacts with the objects (F(1,27) = 56.479, p = 0.0001) and increased latencies to knock over the target object (F(1,27) = 102.209, p = 0.0001). However, all groups were not disrupted to the same extent, as illustrated by significant group by time interactions (contacts: F(3,27) = 4.731, p = 0.009; latency: F(3,27) = 8.015, p = 0.001). Rats with sham, amygdala and perirhinal cortex lesions were disrupted to a greater extent than the rats with hippocampal lesions on the first trial following the rotation compared to the first trial on the previous day. Contrasts revealed that rats with hippocampal lesions showed greatly attenuated changes in contacts, latency and correct responses, suggesting a failure to react to the novel stimulus relationships. All groups, except for the hippocampal group, showed a significant increase in contacts on the first trial following the rotation (sham: t(1,7) = 3.366, p = 0.012; amygdala: t(1,7) = 5.426, p = 0.001; perirhinal cortex: t(1,7) = 5.346; p = 0.001; hippocampus: t(1,6) = 1.648, p = 0.150). All groups had significantly higher latencies following rotation (all t s < 2.0). However, contrasts comparing latency of the rats with hippocampal damage to that of the shams revealed a significantly lower latency in the hippocampal group following (t(1,27) = 3.750, p = 0.005), but not prior to (t(1,27) = 1.155, p = 0.275) the rotation. Only sham and amygdala lesion rats reared following the rotation. Following the rotation the sham, amygdala and perirhinal cortex lesion groups showed a reduction in correct responses during the first trial following the rotation. In contrast, no rat with hippocampal damage made a single incorrect response (Fig. 3). Thus, the performance of rats with hippocampal lesions was significantly less disrupted by the rotation than that of the other groups. In summary, the rats with amygdala and perirhinal cortex lesions showed increases in latency and contacts on measures of memory for the distal cues and distal cue relationships. These rats were affected by the conflicting information provided by the distal and proximal cues, demonstrated by increased number of errors. In contrast, the rats with hippocampal damage showed impairments on all measures of memory for the distal cues and distal proximal cue relationships, and were not affected by conflicting

S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 69 Table 3 Summary of behavior of each lesion group compared to sham group following each manipulation Group Measure Rotation Novel room Novel object Rearrange all objects All novel objects Hippocampus Latency Faster Faster Normal trial 1; Normal faster trial 1 5 Contacts Fewer Fewer Fewer Normal Normal trial 1; fewer all 10 Errors Fewer (0) Normal Normal Amygdala Latency Normal Normal Normal Faster Contacts Normal Normal Fewer trial 1; Fewer More more trial 1 5 Errors Normal Normal Normal Perirhinal Latency Normal Normal Normal Faster Contacts Normal Normal Normal trial 1; Normal Normal more trial 1 5 Errors Normal Normal Normal Rats with hippocampal damage over-focused on the target object, and showed normal disruptions only following manipulations that involved that specific object. Rats with amygdala damage may have been biased to use a more relational strategy than sham rats, as they were disrupted following manipulations involving the distal cues and distal relationships, but not following changes to proximal objects. Rat with perirhinal damage showed mild abnormalities for object memory, as they were disrupted following all distal and proximal relational manipulations, but showed abnormalities in the presence of one novel object or all novel objects. Normal = same as shams; Faster = shorter latency than shams; Fewer = fewer contacts/errors than shams; More = more contacts than shams. information provided by proximal objects and the distal stimuli. Interestingly, these rats did show a small, extremely attenuated, yet significant increase in latency to displace the object. This finding suggests that they did detect the rotation, although they did not respond to it in a manner that resembled the response of the other groups. Thus, these rats were focusing on the objects rather than on the surrounding contextual stimuli and their relationships to the objects. 3.2.4. Novel room The placement of the apparatus in a novel room with novel distal cues entailed a change in the relationships among the proximal objects on the apparatus and the distal stimuli in the room, but did not lead to conflicting information between distal and proximal stimuli. This manipulation permitted an investigation of the extent to which the animal noticed changes in distal proximal relationships and changes to the distal stimulus identity. Performance of the sham, amygdala and perirhinal cortex lesion rats was disrupted following the placement of the cart in the novel room, suggesting that these rats detected the change in distal stimulus identity and distal proximal stimulus relationships. These groups showed increased latency to displace the object and an increase in number of contacts with the objects (Fig. 5). In contrast, the hippocampal lesion group showed an attenuated disruption on all measures (Table 3; see Tables 1 and 2 for overall ANOVAs). Additionally, the amygdala damaged rats showed some rearing. Rats with sham, amygdala and perirhinal cortex lesions were disrupted to a greater extent than the rats with hippocampal lesions on the first trial in the novel room compared to the first trial on the previous day, suggesting that the novel stimuli and relationships had reduced salience for the rats with hippocampal damage. ANOVA revealed that all groups showed increased contacts with the objects (F(1,27) = 25.297, p = 0.0001), and increased latencies to displace the target object F(1,27) = 60.745, p = 0.0001). Fig. 5. Behavioral changes in response to the placement of the cart into a novel room. Differences between novel room and baseline 2 are represented in (a and b). Rats with sham, amygdala and perirhinal cortex lesions showed a significant increase in latencies to complete the task on the first trial, but rats with hippocampal lesions did not show the same increase as the other groups. Rats with sham, amygdala and perirhinal cortex lesions showed increases in contacts with the objects on the first trial and over the first five trials, but rats with hippocampal lesions did not.

70 S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 However, the different groups were not affected in the same way, illustrated by significant latency by group (F(3,27) = 6.522, p = 0.002). Contrasts revealed that all groups, except for the hippocampal group, had significantly higher latencies in the novel room (sham: t(1,7) = 8.725, p = 0.0001; amygdala: t(1,7) = 4.752, p = 0.002; perirhinal cortex: t(1,7) = 4.537, p = 0.003; hippocampus: t(1,6) = 1.784, p = 0.125). The hippocampal group also did not rear at all, in contrast to the other groups showing some rearing. Correct responses were unaffected by the novel room (Fig. 3). Thus, the performance of the hippocampal lesion rats was less disrupted in the novel room compared to the other groups. In summary, the rats with sham, amygdala and perirhinal cortex lesions showed increases in latency and number of contacts with the objects on measures of memory for the distal stimulus identity and distal proximal relationships. In contrast, the rats with hippocampal damage showed impairments on all measures of memory for the distal stimulus identity and distal proximal relationships. 3.2.5. Novel object The replacement of a familiar (non-essential) object with a novel one caused a subtle change in the distal proximal cue relationships, a change in the proximal proximal cue relationships and a change in object identity information. This manipulation may have been less salient than the others, since it did not elicit increases in latency to displace the object. Rather, the sham lesion rats showed an increased number of contacts with all of the objects, and specifically with the novel object. Rats with amygdala and perirhinal cortex damage showed abnormal contacts, and rats with hippocampal damage showed the most severely impaired contacts (Fig. 6; Table 3). Latency to displace the object, rearing and correct responses were unaffected (see Tables 1 and 2 for overall ANOVAs). Overall, the groups showed increased contacts with the novel object during only the first trial, compared to the previous day, illustrated by a significant main effect of the manipulation on contacts (F(1,27) = 10.285, p = 0.003). However, rats with hippocampal and amygdala lesions showed a different pattern of contacts than the sham rats. Planned comparisons revealed that rats with hippocampal damage contacted all the objects fewer times than all the other groups combined or than the shams alone (t s>2, p s < 0.05), but this was not the case on the previous day (t s > 1). In addition, unlike the rats with sham and perirhinal cortex lesions ( t s < 2), rats with amygdala and hippocampus lesions did not show increased contacts with the novel object compared to familiar object on the day before (t s < 1). We also investigated the behavior of the rats during the first five trials with the novel object compared to the previous day. All lesion groups showed different changes in behavior compared to the sham group. Sham (F(1,7) = 8.758, p = 0.021) and amygdala (F(1,7) = 12.916, p = 0.009) lesion rats showed Fig. 6. Behavioral changes in response to the replacement of a familiar, nonessential object with a novel object. Differences between novel object and baseline 3 are represented in (a c). No groups showed increases in latency to displace the object (a). Contacts with all the objects are represented in (b; note smaller scale). Rats with sham and amygdala lesions showed significant increases in contact with all the objects during the first trial and over all five trials. Rats with perirhinal cortex lesions did not show increased contacts over the first five trials. Rat with hippocampal lesions did not show increased contacts during the first trial or over the first five trials. Contacts with only the novel objects are represented in (c). Rats with sham and perirhinal cortex lesions showed significant increases in contacts with the novel object on the first trial. Rats with amygdala and hippocampal damage did not. Rats with sham and hippocampal lesions did not show signifcant increases in contact with the novel object over the first five trials. Rats with amygdala and perirhinal cortex lesions showed increases in contacts with the novel object over the first five trials. increased contacts with all the objects, but perirhinal cortex and hippocampal lesion rats did not (note that these p values are above the 0.005 Bonferroni correction). In addition, contacts with only the novel object were compared to contacts with only the familiar object in the same position on the previous day. The rats with amygdala (F(1,7) = 13.432, p = 0.008) and perirhinal cortex (F(1,7) = 24.652, p = 0.002) lesions showed increased contacts with the novel object over the first five trials, but the sham and hippocampal groups did not (note that the p value for the amygdala group is above the 0.005 Bonferonni correction).

S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 71 In summary, the rats with amygdala damage did not show increased contacts with the novel object comparable to the shams during the first trial, but showed more contacts than the shams during the first five trials. The perirhinal cortex lesion group showed increased contacts with the novel object over the first five trials, and did not show an increase in contacts with all the objects, in contrast to the sham group. The rats with hippocampal damage failed to show increased contacts with both the novel object and with all the objects during the first trial or the first five trials. Thus, rat with amygdala and perirhinal cortex lesions show abnormal exploration following the novel object, and the rats with hippocampal damage show the most severely impaired exploration. The performance of the rats with hippocampal damage appears to be abnormally reliant on the specific rewarded object, as the novel object did not lead to any disruption. Rats with amygdala damage may be less sensitive compared to the shams to the presence of a novel objects, as they respond to it later than the sham rats. Rats with perirhinal cortex initially respond to the novel object normally, however likely due to deficits in object perception and memory, they continue to explore it for longer than the shams. 3.3. Rearrangement of objects The rearrangement of all the familiar objects caused changes in the proximal distal and proximal proximal cue relationships. It also entailed a change in the relationship of the rewarded target object to all the other objects and to the distal cues. Identity of the object remained constant. Following the rearrangement of all familiar objects all groups showed increased latency to displace the object and contacts with the objects (Fig. 7; see Tables 1 and 2 for overall ANOVAs), suggesting that they were distracted by the novel relationships and re-explored the objects in their novel positions. However, rats with amygdala damage showed an abnormal change in contacts with the objects, suggesting abnormal exploration, and rats with hippocampal damage did not show a comparable change in latency to the other groups, suggesting they were less distracted by the novel relationships (Table 3). All groups showed comparable increases in number of incorrect responses (F(1,27) = 291.587, p < 0.0001) (Fig. 3). Rearing was unaffected. On the first trial with the object rearranged compared to the previous day, the sham, perirhinal cortex and hippocampal groups showed an increased number of contacts with the objects (all t s < 2.5, p s = 0.02 0.03; note these p values are above the Bonferroni correction of 0.005). In contrast, the amygdala group damage failed to show an increased number of contacts (t(1,7) = 1.758, p = 0.122). We also examined behavior during the first five trials with the objects rearranged compared to the previous day. The hippocampal group did not show a normal increase in latency over the first five trials. This was illustrated by repeated measures ANOVAs conducted separately for each group that found significant time by trial interactions for Fig. 7. Behavioral changes in response to rearrangement of all the familiar objects, including the goal object. Differences between rearrangement of objects and baseline 4 are represented in (a and b). (a) All groups showed significant increases in latency to displace the object on the first trial and on the first five trials. (b) Rats with sham, perirhinal cortex and hippocampal lesions showed significant increases in contacts with the objects on the first trial and on all five trials. Rats with amygdala lesions did not show increased contacts on the first trial, but did show significantly increased contacts on the first five trials. the sham, amygdala and perirhinal cortex group (all F s > 3, p s = 0.02 0.05; note that these values are above the Bonferroni correction of 0.005), but not the hippocampal group. Rats with hippocampal lesions may not have been as distracted as the other groups, since they performed the task faster. All groups acquired this new version of the task, with object identity regardless of position predicting reward, and performed it with the same degree of accuracy. On the third day of training with object form predicting reward and object location varied pseudo-randomly from trial to trial, one-way ANOVAs comparing mean performance among all the groups revealed that the groups did not differ on number of contacts, latency, rearing or correct responses (all F s < 1). In summary, all the groups were disrupted, suggesting they had some knowledge of the position of the rewarded object relative to the proximal and/or distal cues in the room. The group with hippocampal damage was less distracted by the novel proximal relationships than the sham group. The group with amygdala lesions showed less exploration of the novel proximal relationships than the sham group, suggesting that the salience of the proximal objects may be reduced. The group with perirhinal cortex lesions did not differ from the sham group. All groups demonstrated knowledge of the identity of the rewarded target object, by learning to solve the task correctly.

72 S.N. Moses et al. / Brain Research Bulletin 67 (2005) 62 76 3.3.1. All new objects The replacement of all the familiar objects with novel ones entailed a change in the proximal distal and proximal proximal cue relationships, as well as a change in the identity of the objects. Position at which reward was located relative to the apparatus and the room remained constant. On the first trial, following the replacement of all the familiar objects with novel ones compared to the previous day, all groups showed increased latencies to complete the task, and number of contacts with the objects (Fig. 8; see Tables 1 and 2 for overall ANOVAs), suggesting that they were distracted by, and explored, the novel objects. All groups showed increases in number of errors (Fig. 3). However, the groups with amygdala, perirhinal cortex and hippocampal lesions differed from the sham rats on some of these measures (Table 3). The amygdala group showed more exploration of the novel items, demonstrated by a greater number of contacts with the novel objects averaged over all 10 trials than all other groups combined (F(1,27) = 3.228, p = 0.003); although this effect was not significant when compared to the sham group alone. This effect is also illustrated by a repeated measures ANOVA comparing the groups mean number of contacts over all 10 trials with the novel objects and with the familiar objects on the previous day. This analysis yielded a significant Fig. 8. Behavioral changes in response to the replacement of all the familiar objects, including the goal object, with novel objects. Differences between replacement of objects and baseline 5 are represented in (a and b). (a) Rats with sham, amygdala and perirhinal cortex lesions showed significant increases in latency to displace the object on the first trial and on all five trials. Rats with perirhinal cortex lesions did not show increased latency on the first trial, but did show significantly increased latency on the first five trials. (b) All groups showed significant increases in contacts with the objects on all trials. Rats with hippocampal lesions showed significantly fewer mean contacts than the sham operated rats and rats with amygdala lesions showed a significantly greater number of mean contacts than the sham operated rats. interaction of group by day (F(3,27) = 6.306, p = 0.002). In addition, the rats with amygdala lesions were less distracted by the novel stimuli, as they did not show a normal increase in latency during the first trial. This effect is illustrated by repeated measures ANOVAs, conducted separately for each group, comparing the latency on first five trials with the new objects to the previous day. These analyses yielded a significant main effect of trial (F(4,28) = 3.559, p = 0.018), and a significant time by trial interaction (F(4,28) = 3.962, p = 0.011), for the amygdala lesion group (note that these p values are above the Bonferonni correction of 0.005). These effects were absent in all of the other groups. Rats with perirhinal cortex lesions failed to show an increased latency during the first trial (t(1,7) = 1.648, p = 0.121), suggesting that they may have been less distracted by the presence of the novel objects. All the other groups showed this increase in latency (sham: t(7) = 0.005; amygdala: t(7) = 0.026; hippocampus t(6) = 0.022; note that p values for the amygdala and hippocampus groups are above the Bonferonni correction of 0.005). Rats with hippocampal lesions showed less exploration of the novel objects, demonstrated by fewer mean contacts with the objects over all 10 trials than the sham group (t(1,27) = 2.957, p = 0.018) and than all other groups combined (t(1,27) = 3.240, p = 0.016; note these p values are above the Bonferroni correction of 0.005). This effect is also illustrated by the repeated measures ANOVA comparing the groups mean number of contacts over all 10 trials with the novel objects and with the familiar objects on the previous day. As described above, this analysis yielded a significant interaction of group by day (F(3,27) = 6.306, p = 0.002). Sham and perirhinal cortex lesion rats did not rear, but amygdala and hippocampal lesion rats showed some rearing, suggesting that these groups were exploring the distal contextual stimuli. Thus, all groups were disrupted by this manipulation. However, rats with hippocampal damage explored the objects less, as they showed fewer than normal mean contacts over all 10 trials, but did show some rearing. Rats with amygdala damage showed a greater than normal number of mean contacts over all 10 trials, and rearing, suggesting that they were exploring the relationships between the novel objects and the distal contextual cues more than the other groups. Rats with perirhinal cortex damage were less distracted by the novel objects, as they did not show an increased latency during the first trial following the replacement of all the objects. Thus, all of the lesion groups showed abnormalities in responding to the replacement of all the familiar objects with novel ones. All the groups were able to acquire and to perform the new version of the task, with position regardless of object identity predicting reward, with the same degree of accuracy. On the final day of training with object location predicting reward regardless of form, one-way ANOVAs comparing mean performance among all the groups revealed that the groups did not differ on number of contacts, latency, rearing or correct responses (all F s < 1).