Putting an object in context and acting on it: neural ... - CiteSeerX

DOI 10.1515/revneuro-2012-0073 Rev. Neurosci. 2012; aop

Inah Lee* and Sang-Hun Lee

Putting an object in context and acting on it: neural mechanisms of goal-directed response to contextual object Abstract: Animals including humans experience objects in a certain environment, that is, a context. Same objects may have to be treated differently, or different objects may need to be treated similarly depending on contexts. Flexible behavioral choice in such ambiguous situations involves dynamic interactions among brain regions, but underlying neural mechanisms are poorly understood. In this article, prior studies that have examined (mostly in rodents) some of the brain regions involved in contextual processing of object information using goal-directed tasks are selectively reviewed. The current review identifies the hippocampus, prefrontal cortex (PFC) and perirhinal cortex (PER) as key regions for associating the same objects with different reward values and responses depending on the background visual context. The hippocampus is particularly important for contextual choice behavior when the context must be used as a conditional cue that can disambiguate reward-related ‘meanings’ of objects. The PER appears to play significant roles in such tasks during initial learning (but not so much for retrieval) because perturbations in the PER produce severe deficits in the acquisition of the contextual object memory task. Perturbations in the PFC also affect performance when flexible contextual responses should be made toward otherwise ambiguous objects. Keywords: context; hippocampus; object recognition; paired association; perirhinal cortex; prefrontal cortex; spatial memory.

*Corresponding author: Inah Lee, Department of Brain and Cognitive Sciences, Seoul National University, Gwanak, 151–742 Seoul, Korea, e-mail: [email protected] Sang-Hun Lee: Department of Brain and Cognitive Sciences, Seoul National University, Gwanak, 151–742 Seoul, Korea

Introduction An object seldom exists in isolation, detached from its background context. A pencil sharpener, for example,

may be found on top of a desk in somebody’s office, and an apple can be found in a fruit basket on someone’s kitchen table. In other words, objects are contextually perceived in most cases with other objects and backgrounds in natural settings. Processing object information contextually brings several advantages for adaptation for survival. Examples of such advantages are easily found especially when the significance or value information associated with an object is ambiguous and when additional information is required to disambiguate the meaning of the object. For example, street rats can be extremely cautious in approaching (or never even touch) a small piece of cheese found in a context where foods are not normally found because of the possibility that it may well be bait for a rodent trap (Figure 1). If individual animals or people are also considered as objects, more powerful and numerous examples can be found with respect to the importance of the contextual disambiguation of objects. Theories have been put forth for studying how the brain processes object information and its associated contextual information. As the hippocampus is known anatomically as one of the regions that receive highly processed and associative sensory information (Witter and Amaral, 2004; Amaral and Lavenex, 2007), it is generally hypothesized that the hippocampus functions as a ‛hub’, where associative representations of objects and contexts are formed (Hargreaves et al., 2005; Knierim et al., 2006; Kerr et al., 2007). That is, with respect to the nature of hippocampal inputs, a dominant working hypothesis has been that a group of regions represent spatial information and the other group of regions represent nonspatial information and these two separate spatial-nonspatial information processing streams remain relatively segregated until they finally become associated within the hippocampus (Burwell, 2000; Lavenex and Amaral, 2000; Witter et al., 2000). That is, in the medial temporal lobe (MTL), people have highlighted a group of brain regions, including the perirhinal cortex (PER) and the lateral entorhinal cortex (LEC) as important functional units for processing the identities of discrete and individual stimuli such as objects. However, the POR, the rodent homolog of the parahippocampal cortex of primates, and

Brought to you by | Seoul National University Authenticated | [email protected] author's copy Download Date | 11/25/12 1:06 AM

2 I. Lee and S.-H. Lee: Contextual object memory

Figure 1 A cartoon illustrating the importance of contextual object information processing. In a natural environment, rodents usually avoid an object (e.g., a piece of cheese) that has not been encountered in a familiar context (e.g., back street in a city).

the medial entorhinal cortex (MEC) gathered attention as functional groups where spatial information (presumably including contextual information) is mainly processed. Detailed reviews are available in other articles on this topic (Hargreaves et al., 2005; Knierim et al., 2006; Manns and Eichenbaum, 2006; Kerr et al., 2007), and the same contents will not be repeated here. In the current review, we will rather selectively review prior studies that examined contextual object memory in animal models (mostly rodents). Interesting human studies (especially fMRI imaging studies) have been published on this topic (Goh et al., 2004; Staresina et al., 2011), but we will restrict the scope of the review to animal studies. In the current review, we will also present the results from a series of experiments conducted in our laboratory on this issue. Studies that investigated only object memory or contextual memory by itself without studying the interactions between the two will not be reviewed here. It is important to pre-define some of the terms that will be used throughout the current review. First, the term ‘object’ in this review will refer to a thing (or a person in real world) to which a specific goal-directed action is executed. That is, according to this definition, an object should be able to be ‘manipulated’ by a subject. A typical example will be a small toy object that can be displaced by the rat in order to obtain a food reward underneath it. Such an operational definition will thus exclude calling

distal room cues (that are not reachable by the rat, for example, from a maze) objects. This operational definition will also make it difficult to call oversized figures ‘objects’ as used in some studies (Norman and Eacott, 2005; Bartko et al., 2007) because these figures are used mainly for observational or informational purposes (thus, more similar to landmarks that only need to be recognized without involving any manipulating interactions) and because no specific actions toward these figures can be clearly pre-defined. The definition of the term ‘object’, however, will allow a computer-generated 2D image on a touchscreen monitor to be called an object (Buckley et al., 2001; Talpos et al., 2008; Clelland et al., 2009) because the subject could touch (and thus manipulate and interact with) the images according to task demands. In addition, the term ‘object’ in this review mostly relies on visual perception, and it is hypothesized that complex computations are performed in visual cortices and also in higher associative cortical areas, including the inferior temporal cortex (or temporal association cortex in rodents) and rhinal cortical areas (e.g., PER, entorhinal cortex) before the identity of an object is extracted. In that sense, it may be somewhat different from a simple sensory stimulus such as an olfactory cue (Wood et al., 1999; Komorowski et al., 2009), the information of which can be transmitted rather directly in a disynaptic manner to the hippocampus from neurons in


I. Lee and S.-H. Lee: Contextual object memory 3

the olfactory tract via LEC (Schwerdtfeger et al., 1990; Biella and de Curtis, 2000). Second, the term ‘context’ will also be used with a specific definition to allow a more focused review. Context is an umbrella term used in a very broad sense in cognitive science (Hirsh, 1974; Nadel et al., 1985; Kim and Fanselow, 1992; Maren et al., 1997; Holland and Bouton, 1999; Burwell et al., 2004; Kennedy and Shapiro, 2004; Rajji et al., 2006; Rudy, 2009; Smith et al., 2012). The term context in this review, however, will only refer to a ‘visual background’ that conveys useful information for removing critical ambiguity in target stimuli during choice behavior. It does not necessarily coincide exactly with the term ‘spatial context’ because the term spatial context may include egocentric information (e.g., path integrative and head direction information) and complex combinations of distal and local cues. By the current operational definition in this review, a context should not be a direct target of a response but should remain in the background while actions are directed toward, for example, some objects in the foreground. Such operational definition of context in this review may disqualify calling a cue card in a recording chamber or cylinder (that has been typically used for recording place cells in the hippocampus; Muller et al., 1987; Hetherington and Shapiro, 1997; Hargreaves et al., 2005) a contextual cue because the cue card itself can be considered as both a background and a foreground cue (depending on the rat’s position relative to the cue card) interchangeably during a behavioral session. Furthermore, by this definition, when the rat swims toward a patterned large visual cue in a modified water maze to reach the platform placed directly underneath the cue (Prusky et al., 2004), such visual cues will not be considered as a context in this review because the visual cue in that case is a target stimulus itself and does not serve as a background. The operational definition of context in this review emphasizes that a context should evoke a sense of background environment to the subject. This requires the contextual stimuli to be relatively large in size and complex in visual property. That is, a context should not be easily mistaken as an object in terms of size and complexity. Complexity is important because a truly contextual stimulus would make it counterproductive to focus on a particular element or feature in it. A good contextual stimulus should instead bias the subject to perceive it as a whole if the goals are to achieve rapid and efficient information processing. Referring to our recent studies using LCD monitor-based visual contexts (Kim et al., 2012; Lee and Shin, 2012) may help readers understand the operation definition of context better. Other researchers also categorize internal states, such as motivation or

task demand, into context (Kennedy and Shapiro, 2004; Smith and Mizumori, 2006), but the current review will not discuss those studies. Finally, with respect to the scope of the studies discussed, the current review will only review prior studies that used goal-directed tasks. Studies that used other behavioral paradigms in which animals were allowed to freely forage for food (Muller and Kubie, 1987; Lee et al., 2004; Leutgeb et al., 2005) or spontaneously explore objects (Save et al., 1992; Aggleton et al., 1997; Winters et al., 2004; Lee et al., 2005; Norman and Eacott, 2005; Knierim et al., 2006) will not be covered in this review. Specifically, prior studies used several behavioral paradigms to examine the neural mechanisms of contextual information processing for objects. As objects are almost always experienced simultaneously with a certain context, the goals of these paradigms were to dissociate the two information contents (i.e., object and context) experimentally. Among those, a spontaneous object exploration (or spontaneous object recognition) paradigm has been used most frequently in rodent studies for its simplicity. This paradigm requires animals to explore an arena in which small junk objects are placed in a certain spatial configuration. The paradigm often involves several successive behavioral sessions, each of which includes a set period of spontaneous object exploration. The object exploration behavior is believed to be spontaneous and innate because no apparent goals (e.g., seeking of food reward, avoidance of obnoxious stimulus, etc.) are identified in this paradigm. Typically, the same objects are moved to different places in some sessions, and the differential amounts of exploration observed across such manipulations are measured and used to dissociate object information from contextual information. As the object exploration paradigm requires minimal, if any, pre-training and relies only on observational measurements, its simplicity has been considered as its biggest advantage and has been adopted by many studies using rodents. If the animal explores a previously habituated object more in a later session, it is interpreted as a sign that something novel has been detected with the object. For example, if an object is displaced to a novel location in the arena, and the rat explores the object disproportionately more compared to other objects that have not been moved, it is interpreted that the novel place information associated with the object (and the rat’s innate tendency to explore novelty in its environment more) is the main underlying cognitive factor for the increased exploration activity. By the same logic, when a new object is placed where a familiar object was located previously and the rat explores the novel object more than the other objects



in the arena, this is considered as a sign of object-based, but not place-based, behavior. Further investigation is needed, however, to determine whether the information processing in the brain for the object and its associated context can be dissociated by this seemingly simplistic approach. For example, when the rat explores an object moved to a new location, it is likely that the identity of the object, its previous location information and its current location information are all processed at the same time in the brain in order to compute novelty information. If the new location for the object had been occupied by a different object previously, the complexity of cognitive processes could increase more. What are the mental processes ongoing in the brain when the rat explores the displaced object in its new location? In other words, exactly why does the rat explore the object? In addition to these theoretical issues that need to be clarified, what constitutes object exploration behavior operationally has never been seriously examined. In rodents, most studies have used the frequency of orienting response toward an object within a (arbitrarily defined) close distance as a dependent measure. It could be subjective, however, for measuring such behaviors by just watching videos. For example, when a rat stands right beside an object with its head orienting toward somewhere else, does it indicate that the rat has recognized the object or not? How long should the animal contact the object for that behavior to be counted as an exploration? When a mouse rears and leans against an object with its front paws touching the top of the object, but with its head looking up for room cues, is this an object exploration or not? These ambiguities may undoubtedly exist in coding different behaviors near objects and future studies need to make efforts to improve measurement techniques for the spontaneous exploration paradigm and operational definitions (both inclusive and exclusive criteria) should be agreed upon regarding what constitutes object exploration in rodents among researchers. Mainly for the reasons laid out above and other reasons (for example, is maximal exploration guaranteed in all sessions due to the lack of explicit goals or motivation?) and because there are other excellent reviews available (Mumby, 2001; Winters et al., 2008; Brown et al., 2010), we will not review spontaneous exploration paradigms here. In comparison to the spontaneous exploration paradigm for objects, there is a category of behavioral tasks in which motivated animals work toward achieving goals by learning some key task demands. Such behavioral paradigms are called goal-directed tasks here, as opposed to the spontaneous exploration- or foragingtype of behavioral paradigms, and we will only review

experimental results from goal-directed behavioral tasks. The goal-directed behavioral paradigm first motivates animals. For example, the amount of food is controlled for motivating the animals to make them actively search for food reward or a noxious stimulus is introduced for motivating the animals to avoid it. In contrast to the spontaneous exploration paradigm, therefore, it is relatively clear to interpret the behaviors of animals because the motivation underlying the behavior is well defined, and the behavioral responses leading to the goals are clearly pre-defined via training. In nature, animals arguably spend more time in goal-directed behaviors, such as searching for food, avoiding predators, interacting with other animals and so on, for survival and adaptation (as opposed to spontaneously exploring objects in open space).

Goal-directed choice behavior based on contextual object memory Primate studies A surprisingly small number of studies have been conducted if only goal-directed memory tasks (but not spontaneous exploration paradigms) are reviewed with the topic of contextual object information processing in animal models. Constraining the scope of review using the operational definition of context explained above further reduces the number of studies available for review because this type of context is difficult to implement in primate studies. Specifically, monkeys are often tested using a Wisconsin General Testing Apparatus (Parkinson et al., 1988; Angeli et al., 1993; Malkova and Mishkin, 2003; Belcher et al., 2006) or computer monitors (Cahusac et al., 1989; Rolls et al., 1989; Rao et al., 1997; Suzuki et al., 1997; Dore et al., 1998; Rainer et al., 1998; Charles et al., 2004; Baxter et al., 2007; Bachevalier and Nemanic, 2008; Wirth et al., 2009). Although object-place associations were tested in those settings, the visual context in the background was often made irrelevant to the task (by darkening the background area). Place information associated with an object was in this case conveyed by foodwell locations in a tray in front of the monkey’s cage or by locations within a small computer monitor. In such experimental conditions, it is reasonable to assume that the locations of objects are mostly identified using an egocentric frame of reference (i.e., with respect to the animal’s body).



The nature of stimuli used in the Gaffan group’s studies (i.e., object-in-place scene memory task) may qualify those as proper studies to be reviewed here. That is, an object in the Gaffan’s paradigm was a small typographic character that can be touched by the monkey in a computer monitor. A context was a visual image covering the entire computer screen and served as the background of the typographic objects. The background image (i.e., ‘scene’ according to the Gaffan group) was composed of a random number of ellipses and segments of ellipses. The colors, positions and sizes of elliptical shapes were randomly determined by a computer algorithm. The monkey was required to learn where in the monitor a certain typographic object was rewarded when a particular scene was presented as a contextual background. The Gaffan group showed that lesions produced in the fimbria-fornix (i.e., a fiber bundle that connects the hippocampus to subcortical regions such as mammillary bodies and thalamus), entorhinal cortex and orbital prefrontal cortex resulted in learning deficits in the task (Gaffan, 1994; Charles et al., 2004; Baxter et al., 2007). Disconnecting the interactions between the fimbria-fornix and the PER also produced impairment (Gaffan and Parker, 1996). The operational definition of the term ‘context’ in this review emphasized the role of context in removing critical ambiguity in target stimuli (e.g., objects) during choice behavior. Judging from that perspective, the background image used in the object-inplace scene memory task by the Gaffan group might have been used as a disambiguating cue. That is, by providing configural spatial information between the target object and other background elliptical shapes, the background image was also used as a spatial cue for helping the rats to determine whether the target character was in the right place or not for the purpose of obtaining a reward. For the purpose of the current review, it is arguably difficult to differentiate foreground objects from the background stimuli in this type of setting because the elliptical shapes contained in a scene may well be perceived as objects also (see also Suzuki et al., 1997). A more proper example of study, for the purpose of the current review, in which visual background context was used as a critical disambiguating cue in primates was conducted by the Rolls group (Rolls et al., 2005). In that experiment, monkeys were required to associate object-displaying monitors with local room contexts. The positions of both the monkey and rhe monitors were varied across trials, but arrangements were made in such a way that the monkey could view certain room cues constantly in association with a particular monitor. There were two monitors in the room, and the monkey

should lick a sipping tube to obtain a reward when an object, image A, for example, appeared in one monitor, but should avoid licking the tube (to avoid saline solution, SAL) when the same object image appeared in the other monitor. Another object, image B, appeared always with object A but with opposite reward contingencies (thus making the task a biconditional objectplace paired-associated task). Object-associated ambiguity was maximal in the task because the same objects were used as both positive and negative stimuli depending on the background room context in which they were presented. It is important to note that, unlike the Gaffan group’s scene memory task, the monitor itself did not contain any background image (only two objects were shown), thus allowing more straightforward definitions for objects and contexts in the task. The Rolls et al. study is exceptional compared to other primate studies because object images in computer monitors were associated with actual room cues in the background, as in rodent behavioral studies. Rolls et al. verified the dominant usage of the room context (but not the egocentric reference frame) in the task by recording ‛spatial view cells’ that responded equally to the monitor positions, even when the monkey ’s egocentric relationships with the monitors changed. Neurons in the hippocampus and PER responded to object identity, the background context in which they appear (i.e., place), and the combination of those two factors, according to the study. It is reported in the study that firing patterns of only approximately 10% of neurons were significantly modulated by object information both in the hippocampus and the PER, which is somewhat surprising because the PER is considered to be the major area for representing object identity information in the medial temporal lobe. What is more interesting in this study, given that the PER is considered to process nonspatial information by many researchers, is that approximately equal proportions of neurons in the PER responded significantly to place information as well as to object-place paired associative information. Rolls et al. emphasized that the results might be attributable to the recording locations in the brain being at the boundaries between the posterior PER and the parahippocampal cortex. However, it is also possible that the results might reflect the actual functional firing properties of perirhinal cortical neurons in contextual object information processing. Except for the Rolls et al. study, almost all studies using primates used local positions within a computer monitor or within a testing tray when investigating object-place associations.



A

Rodent studies Hippocampus and contextual object memory – behavioral studies

Object-in-context Response-in-context

100

Percent correct

B

Object-in-context arms Control Hippocampal lesion

90 80 70

Response-in-context arms Control Hippocampal lesion

60 50 40

1

2

3

4 5 Day

6

7

8

9 10

C 100 90 Percent correct

More experiments have been conducted using rodents for studying brain regions involved in contextual object processing. For example, Gilbert and Kesner did a series of experiments on this topic (Gilbert and Kesner, 2002, 2003, 2004). In their typical behavioral paradigm, a rat exited a start box located in one corner of a circular open field (i.e., cheeseboard maze) and had to choose whether to approach and displace an object located in a certain place in the arena or not (go/no-go responses). Two different objects were used in the experiment and only one of the objects appeared in a given trial. The object appeared in one of two fixed locations in the circular field, and each object contained a reward underneath it only when it appeared in a particular location. In this task, as the start box door was opened, the rat saw an object against a certain visual background in the room because the open field was not walled off from the environment. Normal rats learned to approach the rewarding object but to inhibit such responses to the unrewarded object, whereas electrolytic lesions in the hippocampus completely abolished such learning capability (Gilbert and Kesner, 2002) as well as performance when lesions were produced after learning (Gilbert and Kesner, 2004). The same results were reported when different odors must be discriminated contextually (but not when object-odor associations were required). However, the hippocampal-lesioned rats were normal in discriminating objects, odors and places when tested separately without any associative demand. The Kesner group also showed that CA3 in the dorsal hippocampus was critical in learning the task (Gilbert and Kesner, 2003), presumably attributable to CA3’s autoassociative network properties (O’Reilly and McClelland, 1994; McClelland and Goddard, 1996; Rolls and Kesner, 2006). This was also verified in a similar study using transgenic mice in which the gene for the NR1 subunit of NMDA receptor could be temporally manipulated. The CA3 NR1-deleted mice were impaired in learning associations between new odors (scented playground sand in a plastic cup) and the new surrounding contexts (the cage where the cup was placed and the room where the cage was located) in which the odor cups appeared (Rajji et al., 2006), but not when paired associatve learning was required using either familiar odors or familiar contexts. In our laboratory, we developed a behavioral paradigm (Figure 2A) in which different arms of a radial-arm

80 Response-in-context Object-in-context

70 60 50 SAL MUS MUS

SAL

Figure 2 Disambiguation of objects using contexts and associated task demands. (A) Contextual object memory task. Illustration of the radial maze (overview) shows four arms (arms 1, 3, 5 and 7, clockwise from the left) used in the task (gray color) in addition to the other three arms not used in the task (light gray). An objectin-context rule was applied to arms 1 and 5 (lighter gray) and a response-in-context rule was valid in arms 3 and 7 (darker gray). Possible configurations of objects and correct choices (dotted circles) are shown. In a given trial, only one arm was accessible, and objects were available only in the open arm. (B) Post-surgical performances of hippocampal-lesioned rats and controls in object-in-context and response-in-context arms. The hippocampal-lesioned rats showed severe and sustained deficits in performance in object-in-context arms, whereas gradual improvement was observed in response-in-context arms after the initial severe impairment. The control group showed good performance in both task demands, although the performance in the response-in-context condition was higher than the performance in the object-in-context condition. Dotted line: chance performance level. (C) Post-surgical performances in object-in-context versus response-in-context conditions with or without mPFC inactivations in the same animals. Inactivation of MUS produced severe deficits in performance as compared to saline conditions in object-in-context arms, whereas mPFC inactivations did not result in such deficits in response-in-context arms. Dotted line: chance performance level.



maze were associated with either an object-in-context rule or a response-in-context rule in order to test strategic goaldirected behavior maximally (Lee and Solivan, 2008). Rats were required to make an explicit choice of an object associated with reward in a given arm context. Specifically, a pair of different toy objects (e.g., spider man and LEGO block in Figure 2A) was presented in one of the four arms in a large radial-arm maze with seven arms. The two objects were placed on top of food wells in a small rectangular platform at the end of a given arm, and the rat was given only a single opportunity for exposing one of the food wells covered by the objects (by displacing the object with its front paw or snout). The configuration of placing the two objects in a given arm varied from trial to trial (e.g., object A can be on the left or right side of object B). Two of the arms were associated with the object-in-context rule and, when the rat entered the object-in-context arms, object identity information was a critical factor because the rat was required to ignore the food-well locations associated with the objects and should displace a particular object in that context regardless of whether the object appeared on its left or right side. If the rat entered one of the two arms associated with the response-in-context rule, the rat should ignore object identity because displacing any object occupying the food well on a particular side of the rat (or over a particular food well) was rewarded in a given context. The design of the task thus required the rat to pay attention to (a) the arm-associated background context for a given trial, (b) the strategy or rule that was relevant in the context and (c) specific object or response information associated with reward in a given context for obtaining reward. As the maze was surrounded by rich visual cues in the background, presumably rats paid attention to the visual context in the above task to determine a contextually correct object or response. Despite the seemingly complex task demands, normal rats pre-trained to displace an object to retrieve a food reward learned the task in approximately 1 week on average. The learning curve itself is interesting because of its nonlinear property. That is, rats typically showed almost 50% chance-level performances for several days (approx. 6–7 days) but suddenly exhibited a jump in performance at around day 7 or so during the acquisition. Such a learning curve suggests that the rat experienced an ‘a-ha’ moment suddenly after several days of trying to figure out what the task demands were. This is likely because a typical source of poor performance during acquisition was the response bias erroneously associated with object-in-context arms. For example, when a rat happened to displace a correct object (e.g., object A) on the left food well in arm 3 (one of the

object-in-context arms) at an early learning stage, the rat appeared to become strongly inclined to adopt a response strategy almost immediately. That is, the rat seemed to make an erroneous association and ‘think’ that it obtained a reward because it turned to the object on its left side in the context (instead of associating the object identity with the context and reward). It appears that this dominance of response bias over object-in-context strategy is an innate bias that all rats show without exception, albeit to different degrees among individual rats. During the earlier phase of learning, rats acted as if they were only governed by the response-in-context strategy, but they started to show inhibitory behavior immediately in front of a wrong object after a few days of training, which was usually a sure sign that foretold an upcoming performance surge in the task in 2–3 days. Rats with axon-sparing excitotoxic lesions in the hippocampus were severely impaired in the task described above, and their performance stayed almost at chance level throughout 10 days of post-surgical testing (Lee and Solivan, 2008), which suggests that normal hippocampal function was required for the task. Interestingly, the performance deficits were more severe in object-in-context trials than in response-in-context trials with hippocampal lesions (Figure 2B). That is, the lesioned rats showed gradual improvement in performance in response-in-context trials, whereas the same rats exhibited chance-level performances throughout 10 days of testing in objectin-context trials. The severe performance deficits in the object-in-context arms was not attributable to impairment in object recognition because the hippocampal-lesioned rats were normal in discriminating objects when only object recognition was required without any demand for contextual information processing. Overall, it is important to note that the hippocampal-lesioned rats were normal in simple object discrimination and also gradually improved when no contextual object information processing was required (i.e., in response-in-context arms) but showed severe and irreversible impairment only in objectin-context arms where processing object information in the specific arm context was critical. It is unlikely that the performance deficits were due to the strategic complexity associated with the task (i.e., intermixing the object-incontext trials with response-in-context trials) because we confirmed that hippocampal inactivations with a GABA-A receptor agonist muscimol in the absence of response-incontext trials (thus only with the object-in-context rule in effect for processing objects) still produced severe deficits in the same task ( Jo and Lee, 2010b). It is important to note, however, that the object-in-context trial type was apparently harder for rats than the response-in-context



the task was more simplified and only two arms (instead of four arms in the original task) associated with the object-in-context rule were used in the study without imposing the response-in-context task demand (Figure 3A). The two arms were spaced relatively close to each other to create a situation where the DG’s pattern separation became essential (but not too close to establish good baseline performance). As computational models have suggested, DG-lesioned rats showed no sign of improvement in performance from chance level throughout 6 days of post-surgical testing (Figure 3C), resulting in virtually the same results as those observed in total hippocampal-lesioned rats. When the rats were tested with

A

B

C

100

Percent correct

trial type, as controls performed significantly better in response-in-context arms (Figure 2B). This means that it would be premature to completely rule out the possibility that the differential performance deficits between object-in-context and response-in-context trial types were related to general task difficulty. What functions of the hippocampus would make the structure so critical for processing contextual object information? The contextual memory task described above required the rats first to identify the visual context of the environment associated with a visiting arm because the contextual information was used for determining the correct strategy and/or correct object information in that context. The literature suggests that the hippocampus is essential for either dissociating or generalizing similar contextual representations (Marr, 1971; McNaughton and Morris, 1987; O’Reilly and McClelland, 1994; Guzowski et al., 2004; Lee et al., 2004; Leutgeb et al., 2004; Vazdarjanova and Guzowski, 2004). Dissociating similar neural representations is often called ‘pattern separation’ or ‘orthogonalization’, and the hippocampus appears to possess anatomically well suited networks for decreasing overlap among memory representations. This function of the hippocampus may be critical in representing similar events into unique event memories. In contrast, it is believed that the hippocampal networks also perform a seemingly opposite computational function, and it is called ‘pattern completion’ or ‘generalization’ of neural representations. It is highly likely that both pattern separation and completion processes were needed for identifying arm-associated contexts properly in our contextual object memory task. Computational models have emphasized the functions of the dentate gyrus (DG) and CA3 subfields in the hippocampus for pattern separation and completion processes. Detailed reviews on this topic can be found in other articles (Marr, 1971; McNaughton and Morris, 1987; O’Reilly and McClelland, 1994; Rolls and Treves, 1998; Guzowski et al., 2004). Briefly, it has been suggested that both sparse connectivity between the DG and CA3 via the mossy fiber pathway and the autoassociative network using massive recurrent collaterals within CA3 provide mechanisms with which both computational processes can be performed dynamically. If the suggested computational processes are required in our contextual memory task described above, it is predicted that lesions produced in the DG should disrupt the performance in the task because contextual discrimination should suffer accordingly. So, we tested whether the DG was necessary in the contextual object memory task mentioned above. As we learned that the hippocampus was critically involved more in the object-in-context version of the original task,

Small contextual separation Control DG lesion

90 80 70

Large contextual separation

60

Control DG lesion

50 1

2 3 4 5 Post-surgical day

6

Figure 3 The DG in the hippocampus and contextual disambiguation of objects. (A) Contextual object memory task with only object-in-context arms (without response-in-context arms). The arms that were not used in the task are shown in light gray. Possible configurations of objects and correct choices are shown as in Figure 2A. (B) The same task with different arms that were more widely separated from each other for potentially easier contextual disambiguation. (C) Postsurgical performance graphs for the two versions of the contextual object memory task shown in A and B. DG lesions produced deficits in performance in both versions of the task, but the impairments remained throughout the testing period (6 days) only when the contextual object disambiguation was more difficult (A).



a wider separation between the contexts associated with the objects (Figure 3B), DG-lesioned rats improved performance across days as opposed to the sustained deficits observed in the original condition (Figure 3C), presumably because less pattern separation was required with the wider arm separation. The results strongly suggest that the deficits in the hippocampal-lesioned rats in the original contextual task were largely attributable to the loss of DG-CA3 networks, although this should be verified with studies involving other subfield (e.g., CA3, CA1) lesions in the hippocampus. To our knowledge, this is the first experimental evidence showing that the DG is necessary for contextual object information processing in a goaldirected task. Interestingly, the results contradict the results from the Gilbert and Kesner study (2003) in which DG lesions produced in a similar way failed to produce any significant deficit in the acquisition of the objectplace paired-associate task. In addition, CA3-lesioned rats eventually relearned the task in that study. However, it is difficult to directly compare the results from the two studies because the Gilbert and Kesner study tested acquisition of the task, and our study tested performance after learning. More important, task demands were different between the studies. For example, only a single object was presented in a given trial in the Gilbert and Kesner study (thus no reward in no-go trials), whereas two objects were always presented together in the Lee and Solivan study (thus making a reward available for all trials). In addition, the two locations in the open platform in the Gilbert and Kesner study were farther apart from each other (approx. 80°–90°) as compared to the two arms used in our study (51.4° apart from each other). Given the above differences, the discrepancies in results between the two studies with respect to the involvement of the DG in contextual memory may be attributable to less computational demands in the Gilbert and Kesner study, but this should be tested in a more controlled fashion in the same experimental setting in the future. The perturbation studies mentioned so far in this review mostly tested recognition of contextual object representations because animals were presented with objects in certain contexts and were required to respond to the object-context conjunctions according to task demands. Some goal-directed tasks, however, required cued recall in similar settings (Day et al., 2003; Tse et al., 2007; Kesner et al., 2008). For example, in the Day et al. (2003) study, rats were required to learn flavor-place paired associations during a sample phase in a large open field. In the following test phase, the animal was cued with the flavor sampled previously, yet in a remote location (i.e., start box) and was required to revisit the original paired-associate

place on the basis of the olfactory cue sampled. It was shown that the NMDA receptor-mediated plasticity mechanisms in the hippocampus were critical for normal learning to occur in the task. Kesner et al. (2008) also used a similar paradigm (but with objects instead of odors) and reported the importance of CA3 in the hippocampus. The above paradigms, however, test more than only contextual object information processing because the tasks require the animals to ‘hold’ the cue information (e.g., flavor or object) sampled in the start box in working memory during the navigational search for the paired-associate location after they exit the start box. Furthermore, the tasks rely on the intact navigational capability of rats in open space. In other words, even if the rat recalled the paired-associate location at the time of being cued by the cueing odor in the start box, it could still produce performance deficits if normal spatial navigation that should guide the rat to the paired-associate location was impaired as a result of hippocampal perturbations. We tested recently whether those additional cognitive processes mentioned above were critical in testing hippocampal functions for contextual object processing by using a simpler paradigm (Yoon et al., 2012). In our task, the rat ran along a straight track (thus with almost no requirement for spatial navigation) upon exiting a start box and encountered a toy object at the end of the track (Figure 4). The rat was required to simply choose between the two food wells (covered with identical metal discs) on both sides of the cueing object. Two objects were repeatedly used across trials with one of the objects cued the presence of a reward in the left food well, and the other object conveyed the opposite information. The track was located in a circular curtained area, and the curtains were decorated with distinct visual cues to maximally encourage the rats to use visual contexts in the background. The experimental design thus did not require spatial working memory because the paired reward location was found adjacent to the cueing object, and the track guided the rat to the object. Nonetheless, inactivations of the dorsal hippocampus with muscimol severely disrupted normal performance for 2 consecutive days (Figure 4). We confirmed that the rats used visual cues in the background by showing that rats were not able to do this task in the dark, either with or without hippocampal inactivations (Yoon et al., 2012). The rats had no problem in discriminating the cueing objects as well as the food-well location by themselves with hippocampal inactivations. It is important to note that the rats rarely touched the cueing object and predominantly identified the objects using visual information in the task. The results strongly suggest that retrieving object-context conjunctive representations



Percent correct

100 90 80 70 60

SA L

S M

U

S U M

SA L

50

Figure 4 Object-cued retrieval of spatial memory. Left: behavioral testing setup. In a circular curtained room with several visually distinctive cues presented along the curtains, a linear track with a choice platform at its end was placed in the center of the room. A start box was located at the other end of the track. Upper right: reward contingencies for different conditions. In the choice platform, one of two different objects was presented in between two possible reward-containing places (denoted by ‘+’) that were covered by identical metal discs. The availability of a food reward in the food well was determined by the identity of the cueing object. Lower right: performance with or without inactivations in the hippocampus with mPFC (MUS). The MUS inactivations of the dorsal hippocampus severely disrupted performances, whereas no impairment was observed with SAL conditions.

requires the hippocampus, even when additional cognitive demands were not imposed in a goal-directed task.

Hippocampus and contextual object memory – single-unit recording studies Firing patterns of hippocampal neurons indeed suggest that objects are contextually represented in the hippocampus. For example, O’Keefe reported that some hippocampal cells responded by changing firing rates when an unexpected object was found in a certain location or when an object previously found in a particular place disappeared when the place was revisited. These cells were called ‘misplace cells’ (O’Keefe, 1976; O’Keefe and Nadel, 1978). The original experiment by O’Keefe was not conducted using a goal-directed task, although animals were motivated to move around on the three-arm maze to obtain food and water rewards, and some of the objects (e.g., food cups) manipulated for recording misplaced cells were related to reward. As the original report of the misplaced cells was published in 1976, surprisingly few experiments were conducted for specifically examining the firing correlates of contextual object processing in

the hippocampus explicitly in a goal-directed memory task. Some studies were conducted using olfactory stimulus (i.e., scented play sand). For example, the Wood et al. study showed some indication of interaction between odor and location variables (Wood et al., 2000), although the study emphasized the nonspatial coding scheme of hippocampal neurons more. The same research group also recently reported contextual odor information processing in the hippocampus (Komorowski et al., 2009), but it is difficult to find a unit-recording study in which visual object information was critically used in defining object identity in rodents. In order to find functional correlates of hippocampal neurons in the contextual object memory task described previously, we recorded multiple single units simultaneously using the same task illustrated in Figure 3A (Lee and Kim, 2010; Kim et al., 2011). In the unit recording study, neural firing was recorded throughout the acquisition period until the rats showed asymptotic performances in order to find neural correlates of task acquisition. Because of the clearly observable jump in performance during the acquisition (Figure 5A), it was not difficult to draw a boundary between pre-learning and post-learning stages in the task. Hippocampal



Percent correct

A100

Pre-learning

Post-learning

80

60

40

D-4

D-2 D0 D+2 D+4 Acquisition day

Pre-learning

D+6

Post-learning

Bias index

B1

D-6

0

Contextual object bias Contextual response bias

D-6

D-4

D-2

D0

D+2

D+4

D+6

Figure 5 Acquisition of the contextual object memory task. (A) Average performance of the three rats across 12 days of the acquisition period in the same task shown in Figure 3A. The day on which rats performed better than 75% in both arms for the first time was marked as D0 and was used as a temporal reference point (vertical dotted line) for dividing the learning period into pre-learning and post-learning stages. Horizontal dotted line: chance performance level. (B) A crossover of the contextual response bias and the contextual object bias at the transition point for performance from pre-learning to post-learning.

neurons mostly fired in a spatially localized fashion (e.g., arm-specific place fields) as reported in prior studies (O’Keefe, 1976; O’Keefe and Nadel, 1978; Muller et al., 1987). As has also been reported previously (Breese et al., 1989; Markus et al., 1995; Kobayashi et al., 1997; Lee et al., 2006), cells heavily represented the regions where reward-associated events took place (more so during the pre-learning period than after learning). We were not able to find, however, a significant proportion of neurons that fired only in response to object information, presumably because objects were contextually represented in the hippocampus. Interestingly, there was an interesting transition in the firing properties of neurons in the hippocampus across learning stages. As noted in the perturbation studies (Lee and Solivan, 2008, 2010; Jo and Lee, 2010b), rats first learn to choose objects using the contextual response strategy (e.g., choose any object on the left side in arm 3) at the earlier stage of acquisition in this task (Figure 5B). It was more likely to observe the neurons that fired at higher rates in the trials in which the same turning responses

were observed in a given context (regardless of object identities) during the pre-learning period than in the post-learning period (Figure 6A). As the rats learned the task, and the performance increased during learning, the proportion of neurons whose firing patterns were modulated by the response-in-context strategy decreased significantly, and more neurons started to fire at higher rates in their preferred firing locations when the rat was about to push the contextually correct object, irrespective of the turning directions associated with those choices (Lee and Kim, 2010; Kim et al., 2011). It is known that complex spike neurons (often characterized as place cells) in the hippocampus modulate firing rates within their preferred firing locations (i.e., place fields) in response to subtle yet significant changes in the animal’s external and/or internal environment (Hetherington and Shapiro, 1997; Wood et al., 2000; Lee et al., 2006; Leutgeb et al., 2006). Most unit recording studies were conducted in spatial navigation tasks or in foraging situations. The results from our study strongly suggest that, in a goal-directed task, the spatially bound firing-rate modulation in the hippocampus is likely to be a physiological manifestation of hippocampal networks interacting with other brain regions, as spatial context is utilized according to other task-related variables, such as objects and cognitive strategies. Similar changes in the firing patterns of neurons in the prefrontal cortex (PFC) across learning (see below) support such conjectures.

Prefrontal cortex and contextual object memory The PFC receives direct inputs from the hippocampus and subiculum in rodents. Furthermore, the PFC is considered to be one of the critical regions for coordinating goal-directed cognitive processes in the brain. Therefore, it is likely that the PFC is also critically involved in contextual information processing for objects in goaldirected tasks. The PFC may play key roles in quickly reducing the ambiguity associated with an object by generating a top-down teaching signal in a contextual fashion (Bar, 2004; Fenske et al., 2006) for achieving goals. However, very few studies directly examined the functions of the PFC in contextual object information processing in a goal-directed task setting. Its functions were studied in spontaneous exploration paradigms (Barker and Warburton, 2008; Weible et al., 2012) in rodents and visual patterns in a computer monitor in primates (Rao et al., 1997; Baxter et al., 2007; Browning and Gaffan, 2008), but not in a goal-directed task using a real



A Correct object Chosen object

CA1 cell 1

6

Pre-learning (response-in-context)

0

CA1 cell 2

8

Post-learning (object-in-context)

0

B Correct object Chosen object

mPFC cell 3

1.5

Pre-learning (response-in-context)

0

mPFC cell 4

3

Post-learning (object-in-context)

0.6

100

0.5 CA1 0.4

80

0.3 0.2

60

Percent correct

Similarity in spatial firing (r-to-z score)

C

0

Object-in-context Response-in-context mPFC Object-in-context

0.1

Response-in-context

0.0

40 Pre-learning

Post-learning

Figure 6 Common changes in firing patterns between the hippocampus (CA1) and mPFC across the learning stages. (A) Representative examples of spatial firing rate maps associated with four different conditions in the contextual object memory task. The circled objects in the ‘correct object’ row shows an object associated with rewards in a certain arm in the task (cylinder in this case), and the objects marked with dotted circles in the ‘chosen object’ row indicate the objects actually chosen by the rat (correct choices for the first two conditions, but wrong choices for the last two conditions). Before learning occurred (i.e., pre-learning), CA1 neurons exhibited similar levels of firing between the conditions that could be categorized into the response-in-context-based choice (red-dotted rectangle). However, during the post-learning stage, neurons in CA1 tended to show similar levels of firing between object-in-context arms (blue-dotted rectangle). (B) mPFC neurons also exhibited the strategy- or rule-dependent firing patterns as did CA1 neurons in the same task. A toy girl object was the correct object in this particular example. (C) When similarities in spatial firing patterns were measured by cross-correlating the spatial firing rate maps between different conditions (i.e., object-in-context arms versus response-in-context arm), significant increases in the proportions of neurons whose firing patterns were associated with correct choices made in object-in-context trials were observed in the post-learning stage, both in CA1 and mPFC (with matching decreases in response-in-context trials). Bar graphs show the performance in the pre- and post-learning stages.



visual context physically present in the background of an object. Kesner and Ragozzino (Kesner and Ragozzino, 2003) tested the roles of the medial PFC (mPFC) in rats in a goal-directed task using a go/no-go task that was similar to the task used for the hippocampal experiments. However, a plastic box made of red Plexiglas with high walls was used instead of an open field to prevent the rats from looking at the background visual context in the testing room. The rats in such a situation presumably coded object positions egocentrically instead of using background contextual cues. Nonetheless, mPFC lesions produced severe deficits in the task. In our contextual object memory task shown in Figure 2A (Lee and Solivan, 2008), the mPFC was also manipulated with muscimol in a separate group of rats. As a reminder, the task was composed of the two arms associated with the object-in-context rule and the other two arms dedicated for the response-in-context rule. Inactivations of the mPFC with muscimol dropped the performance in response-in-context trials down to approximately 80% (as compared to almost perfect performance with SAL injections) on the first day of muscimol injection, but the rats still performed the task reasonably well (Figure 2C). On the next day, however, the same drug did not affect the response-in-context performance any more. In contrast, the muscimol inactivations of the mPFC caused more severe deficits in object-in-context conditions (experienced in the same experimental session with the response-in-context conditions), producing almost chance-level performances throughout 2 days of muscimol conditions with no sign of recovery (Figure 2C). The more pronounced deficits in the object-in-context type of trials than in the response-incontext type of trials suggest that the mPFC is necessary for normally disambiguating object identities using background contextual information in rats. As demonstrated in the response-in-context trials, the mPFC-inactivated rats were relatively normal in strategically showing correct responses (turning to the left or right food well) according to the context associated with a visiting arm, and this suggests strongly that the severe deficits of those rats in object-in-context trials were not merely due to inabilities of those rats to remember a context-associated rule or to inhibit inappropriate responses in the task. During the initial learning of the task, however, it appears that the firing patterns of neurons in the mPFC are critically modulated by the acquisition of the rule. Specifically, the rats used in the Lee and Kim study (Lee and Kim, 2010) were implanted with a multi-electrode recording drive that could record single units (as well as local field potentials) in the hippocampus and mPFC simultaneously (Kim et al., 2011). Although the neurons in the mPFC were

not as spatially selective in firing (i.e., arm-specific firing) as those in the hippocampus, mPFC neurons also showed the strategy-dependent similarity in spatial firing patterns across learning as hippocampal neurons did (Figure 6B). The increases in object-in-context-compatible firing patterns in both the mPFC and hippocampus were significantly correlated with performance (Figure 6C), which suggests that the mPFC-hippocampal areas work as a unified functional network for learning how to disambiguate objects using proper strategy and context information. Also in support of such reasoning, when pre-learning and post-learning periods were compared, more neurons in both the mPFC and hippocampus synchronized their spiking timing with theta rhythms (recorded from the same region as well as from the other region) after the learning occurred only when the rats showed object-incontext-compatible responses, but not when the same rats showed response-in-context-based responses (Figure 7A). The proportions of neurons showing significant phaselocking to theta rhythms during response-in-context trials did not change, however, between the pre-learning and post-learning stages. Interestingly, the mPFC neurons were distinguishable from the hippocampal neurons because more global task demands were coded in neuronal firing patterns in the mPFC but not in the hippocampus. That is, mPFC neurons fired similarly when the rat experienced the same type(s) of events between different contexts. For example, more neurons in the mPFC (than in CA1 of the hippocampus) changed their firing patterns significantly when the rat entered an arm and/or chose an object regardless of where those events took place. Interestingly, such specificity in firing associated with a specific event type in mPFC neurons became more pronounced as learning progressed and were correlated with performance of the animals (Figure 7B), whereas no such changes were found in CA1 neurons. The results suggest that the global task structure (e.g., ‘when the door opens, go out and enter an open arm and push the correct object to get reward’) is represented more prominently in the mPFC than in the hippocampus. Although the electrophysiological correlates of mPFC neurons in the contextual memory task suggest that the neural firing in that region is modulated by multiple cognitive factors, it appears that the most critical task demand for recruiting mPFC is to require animals to choose between discrete responses to an object with ambiguous meanings (with respect to reward value) by using the visual context in the background. We recently confirmed this by testing the rat in a visual contextual response selection (VCRS) task (Lee and Shin, 2012 ) in which the animal was required to either push or dig the sand in a sand-filled jar to obtain


14 I. Lee and S.-H. Lee: Contextual object memory mPFC theta

100

100

80

40

60

20 0

40 Pre-learning

100

100

80 80

60 40

60

20 0

40 Pre-learning

Post-learning CA1 unit spiking

Percent correct

80

60

% Units phase-locked to theta

CA1 theta

Percent correct

% Units phase-locked to theta

A

Post-learning

mPFC unit spiking

Object-in-context

Object-in-context

Response-in-context

Response-in-context

% Event type-specific units

B CA1 20

D-6~D-4 D-3~D-1 D0~D+2 D+3~D+6

10

10

0

mPFC

20

-3

-2 -1 0 1 2 Event-type specificity

0 -3

-2 -1 0 1 2 Event-type specificity

Figure 7 Phase relationships between neural spiking and simultaneously recorded theta rhythms during the acquisition of the contextual object memory task. The proportion of single units whose spiking activities were significantly phase-locked to the theta rhythms recorded in CA1 and mPFC increased as learning occurred between pre-learning and post-learning stages both in CA1 and mPFC. Bar graphs show the performance in the pre- and post-learning stages. (B) Changes in the proportion of neurons showing increased levels of firing for common types of events regardless of context (higher event-type specificity) across acquisition days. CA1 neuronal firing was tied to specific contexts (i.e., arms) consistently across learning, whereas mPFC neurons gradually increased the neuronal proportion associated with event type-specific firing across contexts as learning progressed (dotted arrow).

a piece of cereal reward according to the visual context presented in the animal’s background through an array of LCD monitors (Figure 8A). The rat, therefore, must choose between equally plausible responses associated with the same object (sand-filled jar), and only the visual background context determined which response was appropriate in a given context to achieve goals. Inactivations of the mPFC severely disrupted the performance in this task (Figure 8B), whereas the same mPFC manipulations failed to impair performance in simply discriminating the two visual contexts (Figure 9A) or in performing the same task using tactile cues instead of visual contexts (Figure 9B).

Extrahippocampal areas and contextual object memory Neocortical areas outside the hippocampus in the medial temporal lobe (i.e., extrahippocampal regions) are believed to send critical spatial and nonspatial

information to the hippocampus for normal processing of contextual object information in the hippocampus. In rats, Bussey et al. (2001) tested the roles of the PER in learning contextual object information using a behavioral paradigm similar to the one described above (Lee and Solivan, 2010; Kim et al., 2011). A double Y-maze was used in the study and a pair of objects was presented on opposite sides of the testing room. Rats were required to enter the arm that contained the object associated with a reward in a given context at the end of the stem of the maze. Although the task could produce presumably less overlap in the object-associated contexts (because of using opposite ends of the linear track for presenting objects) compared to our behavioral paradigm, nonetheless, the rats with lesions in the PER were severely impaired in this task and showed no significant improvement throughout the acquisition period. As demonstrated in our studies ( Jo and Lee, 2010a,b), the perirhinal cortical lesions left the capability of visually discriminating objects intact (Bussey et al., 2001), which suggests that the deficits were



A

Percent correct

B

100 80 60 40 20 0 SAL

MUS

Figure 8 Visual contextual response selection (VCRS) task. (A) Cartoon version of the task. The rat was required either to push the sand-filled jar in a certain context (e.g., zebra pattern) in the background or to dig the sand in the same jar in a different context (e.g., pebbles pattern). (B) MUS-inactivations in the mPFC severely disrupted performance in the VCRS task as compared to the SAL condition (2 days of drug conditions combined into one block). Dotted line: chance performance level.

not merely based on impairment in visual recognition for objects after lesions. Bussey and colleagues tested the acquisition of contextual object memory. To test if the PER is also required for normal retrieval of contextual object memory, we

100

B100

80

80

Percent correct

Percent correct

A

trained rats in the similar task using our seven-arm radial maze ( Jo and Lee, 2010a) and tested the rats after lesions in the PER. Excitotoxic lesions in the PER produced initial impairment in performance when the rats were tested after learning (Figure 10A). However, the lesioned rats quickly recovered normal performance after a few days, and no significant differences were found between the control and lesions groups after approximately 4 days of testing. In contrast, as previously demonstrated by the Bussey group, the perirhinal cortical lesions severely disrupted new learning for contextual object information in the same maze, and the deficits sustained for over 8 continuous days. The results suggest that the PER plays critical roles when contextual object representations are formed initially, but other brain areas (presumably as a result of interacting with the PER during the acquisition) appear to be able to replace the role of the PER after learning and/or relearning the task during the post-surgical testing phase. In addition to the more pronounced involvement of the PER in the acquisition of contextual object information, we have shown that the PER becomes important again during performance when interference increases during the retrieval of contextual object representations in the same experimental settings. Specifically, rats were required to alternate between the two arms without entering the start box after each trial was complete in one of our studies ( Jo and Lee, 2010b), as shown in Figure 10B. After choosing one of the objects in one context, for example, the rat was allowed to run continuously toward the other context in order to make another contextual object choice, and so on. The paradigm could have evoked more proactive interference between contextual object choices because of its continuous nature, although normal rats showed impressive performance in the task (SAL, conditions in Figure 10B). However, inactivations of either the

60 40 20

60 40 20 0

0 SAL

MUS

SAL

MUS

Figure 9 Control experiments for the visual contextual response selection (VCRS) task. (A) Left: visual cue discrimination (VCD) task. The rat was rewarded when displacing a washer in the arm in front of the LCD monitor that showed a visual stimulus (e.g., pebbles pattern) associated reward. Right: inactivations of the mPFC with MUS had no significant effect on performance in the VCD task. The same rats used in the VCRS task were used (within-subjects design). Dotted line: chance performance level. (B) Under mPFC MUS conditions, the same rats were also normal in a response selection task that was similar to the VCRS task, except that local tactile cues (floor insert with wire mesh or soft shelf liner) were used instead of visual contextual cues.



A 100 Acquisition group Control PER lesion

Percent correct

90 80 70

Retrieval group Control PER lesion

60 50 40 1

2

3

4 5 Day

6

7

8

B

hippocampus and the PER (by inactivating the hippocampus in one hemisphere and the PER in the other hemisphere at the same time within a subject) also produced equally severe deficits, whereas leaving the dominant interregional communications intact (by inactivating the hippocampus and the PER unilaterally) did not affect the performance (Figure 10B). The results strongly demonstrate that the hippocampus, the PER and their interactions are all necessary in continuously disambiguating the same objects using contextual information.

Summary and conclusion

100

Percent correct

90 80 HP PER HP-PER (contra) HP-PER (ipsi)

70 60 50 40 SAL MUS MUS SAL

Figure 10 Perirhinal cortex (PER) functions in the contextual object memory task. (A) Lesions in the PER severely disrupted the acquisition of the task (Figure 3A) compared to controls throughout 8 days of the testing period (acquisition group), whereas there was a rapid recovery in performance when post-surgical performance was tested after presurgical learning in a separate group of rats (retrieval group). Dotted line: chance performance level. (B) A continuous contextual object memory task. Upper: instead of entering the start box after each trial, the rat was required to alternate between the two arms and to continuously make object choices according to contextual relevance. Possible configurations of objects and correct choices (dotted circles) are shown at the arm ends. Lower: inactivations of the PER, hippocampus (HP), and contralateral hippocampus-PER [HP-PER (contra)] all produce very similar, severe deficits in performance compared to saline (SAL) conditions in the continuous contextual object memory task. The only condition in which such deficits were not observed with mPFC MUS inactivation was when the unilateral hippocampus and PER were inactivated ipsilaterally [HP-PER (ipsi)], thus leaving communications in another hemisphere between the hippocampus and PER still intact. Dotted line: chance performance level.

dorsal hippocampus or the PER (within subjects) equally impaired the performance severely (Figure 10B). In addition, disconnecting the communications between the

A dominant theory for explaining the neural mechanisms of how contextual object information is represented and remembered in the brain currently targets the hippocampus as the main site where critical associations between context information and object information take place (Hargreaves et al., 2005; Knierim et al., 2006; Manns and Eichenbaum, 2006; Kerr et al., 2007). It is hypothesized at the same time that the areas that reside upstream of the hippocampus (e.g., the PER, entorhinal cortex, etc.) represent either object information or contextual (or spatial) information in a rather purer form. As has been reviewed in this article, the hippocampal perturbation studies seemingly support this theory because inactivations or lesions in the hippocampus severely disrupted the contextual object information processing in animal models. If one wants to know more about the neural mechanisms whereby the hippocampus achieves such cognitive tasks, however, further investigations are necessary. For example, if one believes the current theory of considering the hippocampus as the site of association between object and context information, performance deficits associated with hippocampal perturbations may well be considered as supporting evidence for the theory. However, there may be other ways in which this might work without relying on such a theory. For example, if one posits a different scenario in which the hippocampus provides mostly contextual representation only during the processes, and the associations between contextual and non-contextual representations occur at a downstream structure of the hippocampus (e.g., the subiculum, deep layers of the entorhinal cortex, mPFC, etc.), the results from the hippocampal perturbation studies may not be able to tell which theory is correct because both theories would predict the same results in a goal-directed contextual object memory task. On the basis of the results reviewed here and the results from some unpublished observations in contextual behavioral



experiments in our laboratory, we argue that animals with hippocampal perturbations in contextual object memory tasks suffer from not being able to deal with the situation in which physically identical stimuli (e.g., same or similar objects) need to be differentially treated according to background contexts for goal-related purposes. If the background context is associated with those stimuli with similar probabilities, ambiguity in the physical environment would be maximal and the hippocampus and its associated brain regions should perform critical functions to form and retrieve discrete and separable cognitive representations of object-in-context events. As soon as such ambiguity subsides, it appears that less severe impairment in performance is observed. More elaborate manipulations and physiological recordings in the brain regions located upstream and downstream of the hippocampus (as well as in different subfields within the hippocampus) are needed, however, to support such speculations. It may be that the extrahippocampal areas play more significant roles than people have theoretically argued in processing context and object in an associative manner. Specifically, one of the interesting experimental results that should be mentioned from the contextual object recognition experiments is that perturbations in the PER severely disrupted the acquisition (and performance in certain cases) of contextual object recognition ( Jo and Lee, 2010a,b). The same animals were not impaired, however, in visually discriminating objects themselves in a non-contextual fashion in those studies. Considering that dominant theories categorize the PER functionally as one of the extrahippocampal regions in which nonspatial information is processed (Hargreaves et al., 2005; Knierim et al., 2006; Manns and Eichenbaum, 2006; Kerr et al., 2007), the data do not fit those theories well, and further investigations need to be made on such discrepancies. The literature is also mixed in terms of the involvement of the PER in purely spatial tasks such as the Morris water maze task (Wiig and Bilkey, 1994; Glenn and Mumby, 1998; Aggleton et al., 2004). The involvement of the PER in processing contextual information as illustrated in this review is not surprising, however, if one takes a look at the anatomical connectivity of the region. That is, the PER receives heavy projections from the POR in rats (parahippocampal cortex in primates and humans), whereas far less projection is observed in the opposite direction. If a theory posits that the POR dominantly processes contextual or spatial information, it seems inevitable that such functional correlates are observed at the level of the PER, judging from the pronounced feedforward projections from the POR to the PER. Such contextual crosstalk may exist not just at the PER-POR levels but also at multiple

levels in the extrahippocampal cortical regions including the MEC-LEC level (Burwell, 2000; Lavenex and Amaral, 2000; Witter et al., 2000) before the information reaches the hippocampus. Most electrophysiological studies associated with these issues have been mostly done by using foraging paradigms (Fyhn et al., 2004; Hargreaves et al., 2005; Deshmukh and Knierim, 2011), and further investigations are needed with respect to the representational differences among the regions in the medial temporal lobe in goal-directed tasks. Nevertheless, the idea of parallel input streams in the medial temporal lobe has been pervasive in many studies these days. It is important to note that the hypothesized parallel information streams and their association in the hippocampus are mostly based on anatomical studies, which leaves the possibility that the functional segregation in terms of representing contextual versus non-contextual information in the input regions of the hippocampus in goal-directed tasks might be achieved via more complex and dynamic processes between the socalled spatial and nonspatial regions. The likelihood of oversimplification of the functional architecture of spatial and nonspatial information flows in the medial temporal lobe may be gauged by what has happened in the field of visual neuroscience. Specifically, visual neuroscientists once established clear-cut parallelisms of segregated information flows, which have guided several decades of studies in the field by helping to generate research hypotheses and interpret experimental outcomes. However, the functional, and anatomical as well, segregations in a strict sense have now been withdrawn and begun to be replaced by modified views that posit interactive crosstalks between parallel processing streams. The contextual and non-contextual (or spatial and nonspatial) pathways in the medial temporal lobe, the two parallel information flows that are discussed in this review, are analogous to the so-called ‘where’ and ‘what’ pathways in the visual system, respectively. The distinction between these ‘where’ and ‘what’ pathways emphasized the modular architecture of visual information processing streams at a macro (inter-areal) scale (Ungerleider and Mishkin, 1982). The ‘where’ pathway was composed of a set of visual areas constellated along the dorsal direction from the primary visual cortex (V1) and devoted to processing stimulus features associated with spatial information such as motion and depth. In contrast, the ‘what’ information current flows along the ventral stream of visual areas from V1 and process stimulus attributes essential for object recognition such as form and color. This simple dichotomy, however, has run into many problems as accumulating evidence suggests that the stimulus features that presumably belong to the ‘where’



pathway (e.g., motion and depth) are also processed by the visual areas in the ventral stream ( Janssen et al., 1999; Uka et al., 2000; Watanabe et al., 2002), whereas the ‘what’-pathway features are processed by the visual areas in the dorsal stream (Schlack and Albright, 2007). The classical view then underwent important revision in recent years in the following two major aspects. First, the revised view adopted a new concept of functional segregation by positing that the ventral and dorsal streams are distinguished not in terms of which stimulus features are processed, but in terms of which types of ‘computation’ are carried out. According to this computation-oriented view, the ventral and dorsal streams share the same set of stimulus features as inputs, albeit with different weights, but are assigned two different sets of cognitive computations or behavioral goals (Murata et al., 2000; Lehky and Sereno, 2007; Janssen et al., 2008). Specifically, the ventral stream combines various visual features and converts them into representations of invariant properties intrinsic to objects and their relationships. Such representations are used for identifying familiar objects and classifying unfamiliar objects into known object categories. The dorsal stream, however, transforms a similar set of features into dynamic representations of visual space in an either egocentric or allocentric form for the purposes of guiding its own actions for navigation, interacting with other organisms in action or building spatial working memories for on-line reaction to dynamic environments. Second, although the revised view still argues for the presence of parallel processing streams, it posits that parallel modules are not fixed, but instead are gradually redefined at each stage of downstream processing, where they are recombined to form new parallel modules, the computational outputs of which are passed on to the next downstream target areas (Nassi and Callaway, 2009). Returning to the issues of the current review, what can be learned from this recently revised views of visual neuroscientists on parallel information processing pathways? Obviously, more empirical studies should be performed for specifying local parallel modules and their interactions, both anatomically and functionally, in different areas of the medial temporal lobe and their associated regions in the brain. One likely scenario is that, as observed in the visual systems, slightly different forms of computations are achieved through crosstalks between parallel modules at any given stages throughout the cascade of information flows toward the hippocampus. An oversimplified view of clear-cut parallel modular inputs to the hippocampus may misguide future studies on association of contextual and object information flows. More importantly, the transition from the input-based segregations to the computation- or

goal-based segregations observed in the history of visual neuroscience is likely to be applied to other parts of the brain, including the hierarchy of cortical areas devoted to information processing toward the hippocampus. If a set of canonical computations is identified, we then can search the medial temporal lobe or the rest of the brain regions that are associated with it for candidate neural circuits that are likely to implement those computations, as proposed in visual neuroscience recently (Carandini, 2012). One of the important points we want to also deliver in the current review is that it is important to investigate how memory representations of contextual objects are ‘utilized’ to guide actions and choices. This is arguably one of the most important reasons why we prefer to conduct experiments in goal-directed tasks. There are certain brain regions, such as the PFC and the striatum, that are often functionally associated with action or behavioral selection (Granon et al., 1994; Hoshi et al., 2000; Schultz and Dickinson, 2000; Kesner and Ragozzino, 2003; Matsumoto et al., 2003; Rushworth et al., 2004; Voorn et al., 2004; Haddon and Killcross, 2006; Marquis et al., 2007; Graybiel, 2008; Lee and Solivan, 2008; Kimchi and Laubach, 2009; Braun and Hauber, 2011; Horga et al., 2011; Lee and Shin, 2012). The experimental results from our own laboratory strongly suggest that the PFC is indispensable to contextual choice behavior for objects (Lee and Solivan, 2008; Kim et al., 2011; Lee and Shin, 2012 ). The exact functions of the PFC during contextual behavior, however, are largely unknown, and there are different theories assigning various roles to the PFC, including representation of temporal context (Polyn and Kahana, 2008; Jenkins and Ranganath, 2010), resolving conflict in task demands (de Wit et al., 2006; Haddon and Killcross, 2007; Wittfoth et al., 2009; Horga et al., 2011), flexible behavioral selection (Kesner and Ragozzino, 2003; Haddon and Killcross, 2006; Marquis et al., 2007; Lee and Solivan, 2008; Horga et al., 2011)and maintenance of goal-relevant information in working memory (Goldman-Rakic, 1990; Granon et al., 1994; Delatour and Gisquet-Verrier, 1996; Funahashi and Takeda, 2002), to name a few. Anatomically speaking, the PFC has direct connections to the sensory areas contributing to the representation of contextual information in the brain (Uylings et al., 2003). As the PFC has rich connections to sensory cortical areas, it appears to possess the capability of forming its own contextual representations. The PFC may then use the contextual information in multiple situations, most of which require the elimination of ambiguity in behavioral selection or choice behavior for maximizing goals in a top-down manner. In addition to the PFC, the striatum may also play vital roles during the goal-oriented behavioral selection. In comparison to the



dorsolateral striatum that is connected mostly with somatosensory and motor cortices, the dorsomedial part of the striatum is well connected to the cortical areas processing higher associative (e.g., contextual) representations, such as the prelimbic cortex in rodents (Voorn et al., 2004; Graybiel, 2008). The dorsomedial striatum also receives direct visual information from visual cortices and thalamic areas (Saint-Cyr et al., 1990; Serizawa et al., 1994; Lopez-Figueroa et al., 1995; Reep et al., 2003; Schulz et al., 2009). Therefore, the hippocampus and other cortical regions in the medial temporal lobe may well interact with the PFC and the striatum for contextual disambiguation of behavioral

selection in the environment. Further research should be conducted to delineate specific contributions of those areas that may work toward a common functional goal. Acknowledgments: The studies included in the review were supported by the grants from the World Class University program of the Korea Science and Engineering Foundation funded by the Ministry of Education, Science and Technology (R31–10089) and from NIH/NIMH RO1 MH079971.

Received June 5, 2012; accepted September 5, 2012

References Aggleton, J.P., Keen, S., Warburton, E.C., and Bussey, T.J. (1997). Extensive cytotoxic lesions involving both the rhinal cortices and area TE impair recognition but spare spatial alternation in the rat. Brain Res. Bull. 43, 279–287. Aggleton, J.P., Kyd, R.J., and Bilkey, D.K. (2004). When is the perirhinal cortex necessary for the performance of spatial memory tasks? Neurosci. Biobehav. Rev. 28, 611–624. Amaral, D.G. and Lavenex, P. (2007). Hippocampal Neuroanatomy. In: The Hippocampus Book, P. Anderson, R.G.M. Morris, D.G. Amaral, T. Bliss, and J. O’Keefe, eds., (Oxford: Oxford University Press), pp. 37–114. Angeli, S.J., Murray, E.A., and Mishkin, M. (1993). Hippocampectomized monkeys can remember one place but not two. Neuropsychologia 31, 1021–1030. Bachevalier, J. and Nemanic, S. (2008). Memory for spatial location and object-place associations are differently processed by the hippocampal formation, parahippocampal areas TH/TF and perirhinal cortex. Hippocampus 18, 64–80. Bar, M. (2004). Visual objects in context. Nat. Rev. Neurosci. 5, 617–629. Barker, G.R.I. and Warburton, E.C. (2008). NMDA receptor plasticity in the perirhinal and prefrontal cortices Is crucial for the acquisition of long-term object-in-place associative memory. J. Neurosci. 28, 2837–2844. Bartko, S.J., Winters, B.D., Cowell, R.A., Saksida, L.M., and Bussey, T.J. (2007). Perirhinal cortex resolves feature ambiguity in configural object recognition and perceptual oddity tasks. Learn. Mem. 14, 821–832. Baxter, M.G., Gaffan, D., Kyriazis, D.A., and Mitchell, A.S. (2007). Orbital prefrontal cortex is required for object-in-place ccene memory but not performance of a strategy implementation task. J. Neurosci. 27, 11327–11333. Belcher, A.M., Harrington, R.A., Malkova, L., and Mishkin, M. (2006). Effects of hippocampal lesions on the monkey’s ability to learn large sets of object-place associations. Hippocampus 16, 361–367. Biella, G. and de Curtis, M. (2000). Olfactory inputs activate the medial entorhinal cortex via the hippocampus. J. Neurophysiol. 83, 1924–1931. Braun, S. and Hauber, W. (2011). The dorsomedial striatum mediates flexible choice behavior in spatial tasks. Behav. Brain Res. 220, 288–293.

Breese, C.R., Hampson, R.E., and Deadwyler, S.A. (1989). Hippocampal place cells: Stereotypy and plasticity. J. Neurosci. 9, 1097–1111. Brown, M.W., Warburton, E.C., and Aggleton, J.P. (2010). Recognition memory: Material, processes, and substrates. Hippocampus 20, 1228–1244. Browning, P.G. and Gaffan, D. (2008). Impairment in object-inplace scene learning after uncinate fascicle section in macaque monkeys. Behav. Neurosci. 122, 477–482. Buckley, M.J., Booth, M.C.A., Rolls, E.T., and Gaffan, D. (2001). Selective perceptual impairments after perirhinal cortex ablation. J. Neurosci. 21, 9824–9836. Burwell, R.D. (2000). The parahippocampal region: Corticocortical connectivity. Ann. NY Acad. Sci. 911, 25–42. Burwell, R.D., Bucci, W.D.J., Sanborn, M.R., and Jutras, M.J. (2004). Perirhinal and postrhinal contributions to remote memory for context. J. Neurosci. 24, 11023–11028. Bussey, T.J., Dias, R., Amin, E., Muir, J.L., and Aggleton, J.P. (2001). Perirhinal cortex and place-object conditional learning in the rat. Behav. Neurosci. 115, 776–785. Cahusac, P.M., Miyashita, Y., and Rolls, E.T. (1989). Responses of hippocampal formation neurons in the monkey related to delayed spatial response and object-place memory tasks. Behav. Brain Res. 33, 229–240. Carandini, M. (2012). From circuits to behavior: A bridge too far? Nat. Neurosci. 15, 507–509. Charles, D.P., Browning, P.G.F., and Gaffan, D. (2004). Entorhinal cortex contributes to object-in-place scene memory. Eur. J. Neurosci. 20, 3157–3164. Clelland, C.D., Choi, M., Romberg, C., Clemenson, G.D., Jr., Fragniere, A., Tyers, P., Jessberger, S., Saksida, L.M., Barker, R.A., Gage, F.H., et al. (2009). A functional role for adult hippocampal neurogenesis in spatial pattern separation. Science 325, 210–213. Day, M., Langston, R., and Morris, R.G.M. (2003). Glutamatereceptor-mediated encoding and retrieval of paired-associate learning. Nature 424, 205–209. de Wit, S., Kosaki, Y., Balleine, B.W., and Dickinson, A. (2006). Dorsomedial prefrontal cortex resolves response conflict in rats. J. Neurosci. 26, 5224–5229. Delatour, B. and Gisquet-Verrier, P. (1996). Prelimbic cortex specific lesions disrupt delayed-variable response tasks in the rat. Behav. Neurosci. 110, 1282–1298.


20 I. Lee and S.-H. Lee: Contextual object memory Deshmukh, S.S. and Knierim, J.J. (2011). Representation of nonspatial and spatial information in the lateral entorhinal cortex. Front. Behav. Neurosci. 31, 1905-1918. Dore, F.Y., Thornton, J.A., White, N.M., and Murray, E.A. (1998). Selective hippocampal lesions yield nonspatial memory impairments in rhesus monkeys. Hippocampus 8, 323–329. Fenske, M.J., Aminoff, E., Gronau, N., and Bar, M. (2006). Top-down facilitation of visual object recognition: Object-based and context-based contributions. Prog. Brain Res. 155, 3–21. Funahashi, S. and Takeda, K. (2002). Information processes in the primate prefrontal cortex in relation to working memory processes. Rev. Neurosci. 13, 313–345. Fyhn, M., Molden, S., Witter, M.P., Moser, E.I., and Moser, M.B. (2004). Spatial representation in the entorhinal cortex. Science 305, 1258–1264. Gaffan, D. (1994). Scene-specific memory for objects: a model of episodic memory impairment in monkeys with fornix transection. J. Cogn. Neurosci. 6, 305–320. Gaffan, D. and Parker, A. (1996). Interaction of perirhinal cortex with the fornix-fimbria: Memory for objects and ‛object-in-place’ memory. J. Neurosci. 16, 5864–5869. Gilbert, P.E. and Kesner, R.P. (2002). Role of the rodent hippocampus in paired-associate learning involving associations between a stimulus and a spatial location. Behav. Neurosci. 116, 63–71. Gilbert, P.E. and Kesner, R.P. (2003). Localization of function within the dorsal hippocampus: The role of the CA3 subregion in paired-associate learning. Behav. Neurosci. 117, 1385–1394. Gilbert, P.E. and Kesner, R.P. (2004). Memory for objects and their locations: the role of the hippocampus in retention of object-place associations. Neurobiol. Learn. Mem. 81, 39–45. Glenn, M.J. and Mumby, D.G. (1998), Place memory is intact in rats with perirhinal cortex lesions. Behav. Neurosci. 112, 1353–1365. Goh, J.O.S., Siong, S.C., Park, D., Gutchess, A., Hebrank, A., and Chee, M.W.L. (2004). Cortical areas involved in object, background, and object-background processing revealed with functional magnetic resonance adaptation. J. Neurosci. 24, 10223–10228. Goldman-Rakic, P.S. (1990). Cellular and circuit basis of working memory in prefrontal cortex of nonhuman primates. Prog. Brain Res. 85, 325–335. Granon, S., Vidal, C., Thinus-Blanc, C., Changeux, J.P., and Poucet, B. (1994). Working memory, response selection, and effortful processing in rats with medial prefrontal lesions. Behav. Neurosci. 108, 883–891. Graybiel, A.M. (2008). Habits, rituals, and the evaluative brain. Annu. Rev. Neurosci. 31, 359–387. Guzowski, J.F., Knierim, J.J., and Moser, E.I. (2004). Ensemble dynamics of hippocampal regions CA3 and CA1. Neuron 44, 581–584. Haddon, J.E. and Killcross, S. (2006). Prefrontal cortex lesions disrupt the contextual control of response conflict. J. Neurosci. 26, 2933–2940. Haddon, J.E. and Killcross, S. (2007). Contextual control of choice performance. Behavioral, neurobiological, and neurochemical influences. Ann. NY Acad. Sci. 1104, 250–269. Hargreaves, E.L., Rao, G., Lee, I., and Knierim, J.J. (2005). Major dissociation between medial and lateral entorhinal input to dorsal hippocampus. Science 308, 1792–1794.

Hetherington, P.A. and Shapiro, M.L. (1997). Hippocampal place fields are altered by the removal of single visual cues in a distance-dependent manner. Behav. Neurosci. 111, 20–34. Hirsh, R. (1974). The hippocampus and contextual retrieval of information from memory: a theory. Behav. Biol. 12, 421–444. Holland, P.C. and Bouton, M.E. (1999). Hippocampus and context in classical conditioning. Curr. Opin. Neurobiol. 9, 195–202. Horga, G., Maia, T.V., Wang, P., Wang, Z., Marsh, R., and Peterson, B.S. (2011). Adaptation to conflict via context-driven anticipatory signals in the dorsomedial prefrontal cortex. J. Neurosci. 31, 16208–16216. Hoshi, E., Shima, K., and Tanji, J. (2000). Neuronal activity in the primate prefrontal cortex in the process of motor selection based on two behavioral rules. J. Neurophysiol. 83, 2355–2373. Janssen, P., Vogels, R., and Orban, G.A. (1999). Macaque inferior temporal neurons are selective for disparity-defined threedimensional shapes. Proc. Natl. Acad. Sci. USA 96, 8217–8222. Janssen, P., Srivastava, S., Ombelet, S., and Orban, G.A. (2008). Coding of shape and position in macaque lateral intraparietal area. J. Neurosci. 28, 6679–6690. Jenkins, L.J. and Ranganath, C. (2010). Prefrontal and medial temporal lobe activity at encoding predicts temporal context memory. J. Neurosci. 30, 15558–15565. Jo, Y.S. and Lee, I. (2010a). Perirhinal cortex is necessary for acquiring, but not for retrieving object-place paired association. Learn. Mem. 17, 97–103. Jo, Y.S. and Lee, I. (2010b). Disconnection of the hippocampalperirhinal cortical circuits severely disrupts object-place paired associative memory. J. Neurosci. 30, 9850–9858. Kennedy, P.J. and Shapiro, M.L. (2004). Retrieving memories via internal context requires the hippocampus. J. Neurosci. 24, 6979–6985. Kerr, K.M., Agster, K.L., Furtak, S.C., and Burwell, R.D. (2007). Functional neuroanatomy of the parahippocampal region: The lateral and medial entorhinal areas. Hippocampus 17, 697–708. Kesner, R.P. and Ragozzino, M.E. (2003). The role of the prefrontal cortex in object-place learning: a test of the attribute specificity model. Behav. Brain Res. 146, 159–165. Kesner, R.P., Hunsaker, M.R., and Warthen, M.W. (2008). The CA3 subregion of the hippocampus is critical for episodic memory processing by means of relational encoding in rats. Behav. Neurosci. 122, 1217–1225. Kim, J.J. and Fanselow, M.S. (1992). Modality-specific retrograde amnesia of fear. Science 256, 675–677. Kim, J., Delcasso, S.B., and Lee, I. (2011). Neural correlates of object-in-place learning in hippocampus and prefrontal cortex. J. Neurosci. 31, 16991–17006. Kim, S., Lee, J., and Lee, I. (2012). The hippocampus is required for visually cued contextual response selection, but not for visual discrimination of contexts. Front. Behav. Neurosci. 6, 66. Kimchi, E.Y. and Laubach, M. (2009). Dynamic encoding of action selection by the medial striatum. J. Neurosci. 29, 3148–3159. Knierim, J.J., Lee, I., and Hargreaves, E.L. (2006). Hippocampal place cells: parallel input streams, subregional processing, and implications for episodic memory. Hippocampus 16, 755–764. Kobayashi, T., Nishijo, H., Fukuda, M., Bures, J., and Ono, T. (1997). Task-dependent representations in rat hippocampal place neurons. J. Neurophysiol. 78, 597–613.



Komorowski, R.W., Manns, J.R., and Eichenbaum, H. (2009). Robust conjunctive item-place coding by hippocampal neurons parallels learning what happens where. J. Neurosci. 29, 9918–9929. Lavenex, P. and Amaral, D.G. (2000). Hippocampal-neocortical interaction: a hierarchy of associativity. Hippocampus 10, 420–430. Lee, I. and Kim, J. (2010). The shift from a response strategy to object-in-place strategy during learning is accompanied by a matching shift in neural firing correlates in the hippocampus. Learn. Mem. 17, 381–393. Lee, I. and Shin, J.Y. (2012). Medial prefrontal cortex is selectively involved in response selection using visual context in the background. Learn. Mem. 20, 247–250. Lee, I. and Solivan, F. (2008). The roles of the medial prefrontal cortex and hippocampus in a spatial paired-association task. Learn. Mem. 15, 357–367. Lee, I. and Solivan, F. (2010). Dentate gyrus is necessary for disambiguating similar object-place representations. Learn. Mem. 17, 252–258. Lee, I., Yoganarasimha, D., Rao, G., and Knierim, J.J. (2004). Comparison of population coherence of place cells in hippocampal subfields CA1 and CA3. Nature 430, 456–459. Lee, I., Hunsaker, M.R., and Kesner, R.P. (2005). The role of hippocampal subregions in detecting spatial novelty. Behav. Neurosci. 119, 145–153. Lee, I., Griffin, A.L., Zilli, E.A., Eichenbaum, H., and Hasselmo, M.E. (2006). Gradual translocation of spatial correlates of neuronal firing in the hippocampus toward prospective reward locations. Neuron 51, 639–650. Lehky, S.R. and Sereno, A.B. (2007). Comparison of shape encoding in primate dorsal and ventral visual pathways. J. Neurophysiol. 97, 307–319. Leutgeb, S., Leutgeb, J.K., Treves, A., Moser, M.-B., and Moser, E.I. (2004). Distinct ensemble codes in hippocampal areas CA3 and CA1. Science 305, 1295–1298. Leutgeb, J.K., Leutgeb, S., Treves, A., Meyer, R., Barnes, C.A., McNaughton, B.L., Moser, M.-B., and Moser, E.I. (2005). Progressive transformation of hippocampal neuronal representations in ‛morphed’ environments. Neuron 48, 345–358. Leutgeb, S., Leutgeb, J.K., Moser, E.I., and Moser, M.B. (2006). Fast rate coding in hippocampal CA3 cell ensembles. Hippocampus 16, 765–774. Lopez-Figueroa, M.O., Ramirez-Gonzalez, J.A., and Divac. I. (1995). Projections from the visual areas to the neostriatum in rats. A re-examination. Acta Neurobiol. Experiment 55, 165–175. Malkova, L. and Mishkin, M. (2003). One-trial memory for object-place associations after separate lesions of hippocampus and posterior parahippocampal region in the monkey. J. Neurosci. 23, 1956–1965. Manns, J.R. and Eichenbaum, H. (2006). Evolution of declarative memory. Hippocampus 16, 795–808. Maren, S., Aharonov, G., and Fanselow, M.S. (1997). Neurotoxic lesions of the dorsal hippocampus and Pavlovian fear conditioning in rats. Behav. Brain Res. 88, 261–274. Markus, E.J., Qin, Y.L., Leonard, B., Skaggs, W.E., McNaughton, B.L., and Barnes, C.A. (1995). Interactions between location and task affect the spatial and directional firing of hippocampal neurons. J. Neurosci. 15, 7079–7094. Marquis, J., Kilcross, S., and Haddon, J.E. (2007). Inactivation of the prelimbic, but not infralimbic, prefrontal cortex impairs the

contextual control of response conflict in rats. Eur. J. Neurosci. 25, 559–566. Marr, D. (1971). Simple memory: a theory for archicortex. Philos. Trans. R. Soc. Lond. B Biol. Sci. 262, 23–81. Matsumoto, K., Suzuki, W., and Tanaka, K. (2003). Neuronal correlates of goal-based motor selection in the prefrontal cortex. Science 301, 229–232. McClelland, J.L. and Goddard, N.H. (1996). Considerations arising from a complementary learning systems perspective on hippocampus and neocortex. Hippocampus 6, 654–665. McNaughton, B.L. and Morris, R.G.M. (1987). Hippocampal synaptic enhancement and information storage within a distributed memory system. Trends Neurosci. 10, 408–415. Muller, R.U. and Kubie, J.L. (1987). The effects of changes in the environment on the spatial firing of hippocampal complex-spike cells. J. Neurosci. 7, 1951–1968. Muller, R.U., Kubie, J.L., and Ranck, J.B., Jr. (1987). Spatial firing patterns of hippocampal complex-spike cells in a fixed environment. J. Neurosci. 7, 1935–1950. Mumby, D.G. (2001). Perspectives on object-recognition memory following hippocampal damage: lessons from studies in rats. Behav. Brain Res. 127, 159–181. Murata, A., Gallese, V., Luppino, G., Kaseda, M., and Sakata, H. (2000). Selectivity for the shape, size, and orientation of objects for grasping in neurons of monkey parietal area AIP. J. Neurophysiol. 83, 2580–2601. Nadel, L., Wilner, J., and Kurz, E.M. (1985). Cognitive Maps and Environmental Context. In: Context and Learning, P.D. Balsam and A. Tomie A, eds. (Hillsdale, NJ: Lawrence Erlbaum Associates), pp. 385–406. Nassi, J.J. and Callaway, E.M. (2009). Parallel processing strategies of the primate visual system. Nat. Rev. Neurosci. 10, 360–372. Norman, G. and Eacott, M.J. (2005). Dissociable effects of lesions to the perirhinal cortex and the postrhinal cortex on memory for context and objects in rats. Behav. Neurosci. 119, 557–566. O’Keefe, J. (1976). Place units in the hippocampus of the freely moving rat. Exp. Neurol. 51, 78–109. O’Keefe, J. and Nadel, L. (1978). The Hippocampus as a Cognitive Map (Oxford: Clarendon Press). O’Reilly, R.C. and McClelland, J.L. (1994). Hippocampal conjunctive encoding, storage, and recall: avoiding a trade-off. Hippocampus 4, 661–682. Parkinson, J.K., Murray, E.A., and Mishkin, M. (1988). A selective mnemonic role for the hippocampus in monkeys: memory for the location of objects. J. Neurosci. 8, 4159–4167. Polyn, S.M. and Kahana, M.J. (2008). Memory search and the neural representation of context. Trends Cog. Sci. 12, 24–30. Prusky, G.T., Douglas, R.M., Nelson, L., Shabanpoor, A., and Sutherland, R.J. (2004). Visual memory task for rats reveals an essential role for hippocampus and perirhinal cortex. Proc. Natl. Acad. Sci. USA 101, 5064–5068. Rainer, G., Asaad, W.F., and Miller, E.K. (1998). Selective representation of relevant information by neurons in the primate prefrontal cortex. Nature 393, 577–579. Rajji, T., Chapman, D., Eichenbaum, H., and Greene, R. (2006). The role of CA3 hippocampal NMDA receptors in paired associate learning. J. Neurosci. 26, 908–915. Rao, S.C., Rainer, G., and Miller, E.K. (1997). Integration of what and where in the primate prefrontal cortex. Science 276, 821–824.


22 I. Lee and S.-H. Lee: Contextual object memory Reep, R.L., Cheatwood, J.L., and Corwin, J.V. (2003). The associative striatum: Organization of cortical projections to the dorsocentral striatum in rats. J. Comp. Neurol. 467, 271–292. Rolls, E.T. and Kesner, R.P. (2006). A computational theory of hippocampal function, and empirical tests of the theory. Prog. Neurobiol. 79, 1–48. Rolls, E.T. and Treves, A. (1998). Neural Networks and Brain Function (Oxford: Oxford University Press). Rolls, E.T., Miyashita, Y., Cahusac, P.M., Kesner, R.P., Niki, H., Feigenbaum, J.D., and Bach, L. (1989). Hippocampal neurons in the monkey with activity related to the place in which a stimulus is shown. J. Neurosci. 9, 1835–1845. Rolls, E.T., Xiang, J., and Franco, L. (2005). Object, space, and object-space representations in the primate hippocampus. J. Neurophysiol. 94, 833–844. Rudy, J.W. (2009). Context representations, context functions, and the parahippocampal-hippocampal system. Learn. Mem. 16, 573–585. Rushworth, M.F.S., Walton, M.E., Kennerley, S.W., and Bannerman, D.M. (2004). Action sets and decisions in the medial frontal cortex. Trends Cog. Sci. 8, 410–417. Saint-Cyr, J.A., Ungerleider, L.G., and Desimone, R. (1990). Organization of visual cortical inputs to the striatum and subsequent outputs to the pallido-nigral complex in the monkey. J. Comp. Neurol. 298, 129–156. Save, E., Poucet, B., Foreman, N., and Buhot, M.C. (1992). Object exploration and reactions to spatial and nonspatial changes in hooded rats following damage to parietal cortex or hippocampal formation. Behav. Neurosci. 106, 447–456. Schlack, A. and Albright, T.D. (2007). Remembering visual motion: neural correlates of associative plasticity and motion recall in cortical area MT. Neuron 53, 881–890. Schulz, J.M., Redgrave, P., Mehring, C., Aertsen, A., Clements, K.M., Wickens, J.R., and Reynolds, J.N.J. (2009). Short-latency activation of striatal spiny neurons via subcortical visual pathways. J. Neurosci. 29, 6336–6347. Schultz, W. and Dickinson, A. (2000). Neuronal coding of prediction errors. Annu. Rev. Neurosci. 23, 473–500. Schwerdtfeger, W.K., Buhl, E.H., and Germroth, P. (1990). Disynaptic olfactory input to the hippocampus mediated by stellate cells in the entorhinal cortex. J. Comp. Neurol. 292, 163–177. Serizawa, M., McHaffie, J.G., Hoshino, K., and Norita, M. (1994). Corticostriatal and corticotectal projections from visual cortical areas 17, 18 and 18a in the pigmented rat. Arch. Histol. Cytol. 57, 493–507. Smith, D.M. and Mizumori, S.J. (2006). Learning-related development of context-specific neuronal responses to places and events: The hippocampal role in context processing. J. Neurosci. 26, 3154–3163. Smith, D.M., Barredo, J., and Mizumori, S.J.Y. (2012). Complimentary roles of the hippocampus and retrosplenial cortex in behavioral context discrimination. Hippocampus 22, 1121–1133. Staresina, B.P., Duncan, K.D., and Davachi, L. (2011). Perirhinal and parahippocampal cortices differentially contribute to later recollection of object- and scene-related event details. J. Neurosci. 31, 8739–8747. Suzuki, W.A., Miller, E.K., and Desimone, R. (1997). Object and place memory in the macaque etorhinal cortex. J. Neurophysiol. 78, 1062–1081. Talpos, J.C., Dias, R., Bussey, T.J., and Saksida, L.M. (2008). Hippocampal lesions in rats impair learning and memory for

locations on a touch-sensitive computer screen: The ‛ASAT’ task. Behav. Brain Res. 192, 216–225. Tse, D., Langston, R.F., Kakeyama, M., Bethus, I., Spooner, P.A., Wood, E.R., Witter, M.P., and Morris, R.G.M. (2007). Schemas and memory consolidation. Science 316, 76–82. Uka, T., Tanaka, H., Yoshiyama, K., Kato, M., and Fujita, I. (2000). Disparity selectivity of neurons in monkey inferior temporal cortex. J. Neurophysiol. 84, 120–132. Ungerleider, L.G. and Mishkin, M. (1982). The Analysis of Visual Behavior (Cambridge: MIT Press). Uylings, H.B., Groenewegen, H.J., and Kolb, B. (2003). Do rats have a prefrontal cortex? Behav. Brain Res. 146, 3–17. Vazdarjanova, A. and Guzowski J.F. (2004). Differences in hippocampal neuronal population responses to modifications of an environmental context: evidence for distinct, yet complementary, functions of CA3 and CA1 ensembles. J. Neurosci. 24, 6489–6496. Voorn, P., Vanderschuren, L.J.M.J., Groenewegen, H.J., Robbins, T.W., and Pennartz, C.M.A. (2004). Putting a spin on the dorsalventral divide of the striatum. Trends Neurosci. 27, 468–474. Watanabe, M., Tanaka, H., Uka, T., and Fujita, I. (2002). Disparityselective neurons in area V4 of macaque monkeys. J. Neurophysiol. 87, 1960–1973. Weible, A.P., Rowland, D.C., Monaghan, C.K., Wolfgang, N.T., and Kentros, C.G. (2012). Neural correlates of long-term object memory in the mouse anterior cingulate cortex. J. Neurosci. 32, 5598–5608. Wiig, K.A. and Bilkey, D.K. (1994). The effects of perirhinal cortical lesions on spatial reference memory in the rat. Behav. Brain Res. 63, 101–109. Winters, B.D., Forwood, S.E., Cowell, R.A., Saksida, L.M., and Bussey, T.J. (2004). Double dissociation between the effects of peri-postrhinal cortex and hippocampal lesions on tests of object recognition and spatial memory: heterogeneity of function within the temporal lobe. J. Neurosci. 24, 5901–5908. Winters, B.D., Saksida, L.M., and Bussey, T.J. (2008). Object recognition memory: neurobiological mechanisms of encoding, consolidation and retrieval. Neurosci. Biobehav. Rev. 32, 1055–1070. Wirth, S., Avsar, E,, Chiu, C.C., Sharma, V., Smith, A.C., Brown, E., and Suzuki, W.A. (2009). Trial outcome and associative learning signals in the monkey hippocampus. Neuron 61, 930–940. Witter, M.P. and Amaral, D.G. (2004). The Hippocampal Formation. In: The Rat Nervous System, 3rd ed. G. Paxinos G, ed. (San Diego, CA: Academic Press), pp. 635–704. Witter, M.P., Wouterlood, F.G., Naber, P.A., and Van Haeften, T. (2000). Anatomical organization of the parahippocampalhippocampal network. Ann. NY Acad. Sci. 911, 1–24. Wittfoth, M., Schardt, D.M., Fahle, M., and Herrmann, M. (2009). How the brain resolves high conflict situations: double conflict involvement of dorsolateral prefrontal cortex. Neuroimage 44, 1201–1209. Wood, E.R., Dudchenko, P.A., and Eichenbaum, H. (1999). The global record of memory in hippocampal neuronal activity. Nature 397, 613–616. Wood, E.R., Dudchenko, P.A., Robitsek, R.J., and Eichenbaum, H. (2000). Hippocampal neurons encode information about different types of memory episodes occurring in the same location. Neuron 27, 623–633. Yoon, J., Seo, Y., Kim, J., and Lee, I. (2012). Hippocampus is required for paired associate memory with neither delay. Learn. Mem. 19, 1–8.



Dr. Inah Lee received his PhD in Neuroscience under the supervision of Dr. Raymond Kesner from the University of Utah in 2002. Dr. Lee published several seminal papers investigating the functional significance of different subfields of the hippocampus in learning and memory using behavioral paradigms combined with lesion and inactivation techniques during his doctoral training period. He spent 2 years in Dr. James Knierim’s lab at the University of Texas Medical School at Houston as a postdoctoral researcher, during which time he published a seminal article in Nature showing the neural firing correlates of hippocampal computational processes based on an electrophysiological experiment using freely moving animals. He worked as a postdoctoral researcher for another 1.5 years in the Center for Memory and Brain with Dr. Howard Eichenbaum and Dr. Michael Hasselmo, further investigating how neural firing patterns dynamically changed when rats performed a goal-directed memory task. Dr. Lee spent 3 years in the Department of Psychology at the University of Iowa as an assistant professor and is currently an associate professor of the Department of Brain and Cognitive Sciences at the Seoul National University. His research interest is in understanding the neural mechanisms of context memory-dependent choice behavior involving the hippocampus and its associated areas, such as the prefrontal cortex, striatum and cortical regions in the medial temporal lobe.

Dr. Sang-Hun Lee is currently an associate professor in the Department of Brain and Cognitive Sciences at the Seoul National University. Dr. Lee received his PhD in Visual Neuroscience from Vanderbilt University in 2001. From 2001–2004, he spent three years at Stanford University and New York University as a postdoctoral fellow in David Heeger’s computational neuroimaging laboratory at the Center for Neuroscience. Dr. Lee’s early studies focused on the behavioral aspects of human visual perception; he published several influential papers on binocular vision, temporal vision and visual grouping using methods of pychophysics. He later gained expertise in brain imaging and functional magnetic resonance imaging (fMRI). He has tried to establish links between perceptual dynamics in human vision and neural dynamics in the cerebral cortex by combining diverse data collected from psychophysical, brain imaging and computational modeling experiments. He received the William James Young Investigator Award from the Association for the Scientific Study of Consciousness in 2006 and was chosen as one of the Frontier Scientists by the Korean Ministry of Science and Technology (2009) in recognition of his recent pioneering work on visual awareness and brain activity. He has also recently launched a new department, the Department of Brain and Cognitive Sciences (http://bcs.snu.ac.kr), at the Seoul National University by leading a team of international scholars. He is an acting chair of this new department and also directs the SNU Brain Imaging Center, which houses a 3-T research-devoted MRI scanner.


Putting an object in context and acting on it: neural ... - CiteSeerX

Putting an object in context and acting on it: neural ... - CiteSeerX

Suggest Documents

Technology and Education: Putting it in context

Technology and Education: Putting it in context

Putting conservation's professional qualification in context - CiteSeerX

Putting "it" in Context: Diversity in Composition Classes

Putting the Context in Context-Driven Testing (an ... - Cem Kaner

The Neural Basis of Object-Context Relationships on Aesthetic ... - PLOS

Putting Dementia Into Context

Putting Retrieval-Induced Forgetting in Context: An ... - Tanya Jonker

Putting it into Practice

Putting Sportswear In Context - Durham University Community

Putting Student-Centered Education in Context

Neural Substrates of Tactile Object Recognition: An fMRI ... - CiteSeerX

Putting MAIDs in Context - UBC Computer Science

Putting Web Services in Context - Semantic Scholar

An Object-Oriented Version Model for Context

Color-Shape Context for Object Recognition - CiteSeerX

Neural Networks Learning transform invariant object ... - CiteSeerX

"Putting the 'I' in IT Education" - Educause

It makes sense: Using an autoassociative neural network ... - CiteSeerX

Object Recommendation based on Usage Context

Putting Gender Into Context: An Interactive Model of Gender-Related ...

(College and Career) Ready? Putting Common Core in Context

(College and Career) Ready? Putting Common Core in Context

Putting It All Together