THE MEDIAL TEMPORAL LOBE

emotional memory (Bechara et al., 1995; Cahill, Babinsky, Markowitsch, & McGaugh, 1995). Here emotional memory refers to the formation of affective representations that is not necessarily available for explicit retrieval but can be implicitly expressed in for example attraction- and avoidance behavior, as well as in the modulation of autonomic nervous system responses. However, the amygdala appears to have a broader role in human long-term memory. In particular, the amygdala, which is a part of the anterior MTL, has prominent recurrent connections with the hippocampus and the MTL memory system.

Thus, it seems that the amygdala is well placed anatomically to modulate declarative memory. For example, a time-varying learning rate that changes with the relevance of the information being processed, opens up for the possibility to control learning rate by various relevance or 'print-now' signals. This mechanism can be used to make the memory selective and modulated by relevance (cf., appendix 2.2). Several functional neuroimaging studies have investigated the role of the amygdala in enhancing declarative memory for emotional experiences and suggested a correlation between amygdala activation during encoding and subsequent memory. For example, the degree of activity in the left amygdala during encoding was predictive of subsequent memory (Canli, Zhao, Brewer, Gabrieli, & Cahill, 2000). Furthermore, it has also been suggested that the amygdala may play a role in modulating the strength and consolidation of memories in other memory systems (Cahill et al., 1995, cf. chapter 2 and appendix 2.2).

immediate short-term memory appears relatively intact; and (3) the memory impairment appears to occur against a background of intact perceptual, cognitive and motor abilities as well as intact non-declarative memory. Thus, important aspects of declarative memory are dissociated from general perceptual, cognitive, and motor function. Furthermore, the evidence suggest a particular role for the MTL memory system in encoding-storage-consolidation processes, while the role of the MTL in retrieval processes is less clear (cf., Simons & Spiers, 2003; Squire et al., 2004). One way to understand this is that the MTL creates a long-term storage format, making the stored information suitable for effective retrieval; when the MTL is damaged, immediate short-term memory representations in neocortex are not adequately processed from the perspective of long-term retrievability. If the MTL is not functional at the time of learning, declarative memory is not established in a proper way and is therefore not readily available for later retrieval. However, this does not necessarily imply that the MTL is a (permanent) repository of memory, and because remote memory is spared in patients with large MTL lesions, it appears that long-term information would have to be stored elsewhere.

In addition to the phenomenon of anterograde amnesia, damage to the MTL often results in partial loss of memory for information acquired before the damage occurred, so-called retrograde amnesia. Squire and colleagues (2004) suggest that when the MTL lesion is limited to the hippocampus, entorhinal cortex, and/or fornix, the retrograde memory impairment is temporally graded. This implies that more recently formed memories are relatively more impaired than more remotely acquired information. Temporal gradients of retrograde amnesia have also been described in patients with damage limited to the hippocampal region (Kapur & Brooks, 1999; Manns, Hopkins, & Squire, 2003).

Furthermore, the remote memory for facts (semantic memory) is intact and it seems that remote episodic memory for autobiographical events can also be preserved (Bayley &

Squire, 2003).

The consolidation view of temporally graded retrograde amnesia begins with the principle that long-term memory is stored as outcomes of MTL processing in interaction with the regions in the neocortex that are specialized for what is to be remembered (Squire et al., 2004). On this view, the MTL initially works together with the neocortex to allow memory to be formed into a retrievable format. Through a gradual process of integration

and re-organization (cf., Figure 2.2), it is suggested that the connections among neocortical regions are progressively strengthened until the neocortical memory can be retrieved independently of the hippocampus.

The Medial Temporal Lobe Memory System

Prefrontal cortex

Hippocampus Neocortical association areas

Parahippocampal cortex

Perirhinal cortex

Entorhinal cortex

Subcortical structures

[Figure 4.4] The medial temporal lobe memory system. Adapted from Simons and Spiers 2001. Whereas the medial temporal lobe has been associated with the encoding, storage and retrieval of long-term declarative memory, the prefrontal cortex, and the neocortex more generally, also plays an important role in declarative memory (Eichenbaum

& Cohen, 2001; Nyberg, Cabeza, & Tulving, 1996; Simons & Spiers, 2003; Squire et al., 2004). The prefrontal cortex has been linked to short-term working memory, language processing, and various cognitive control processes such as selection, monitoring, manipulation and inhibition of information (Baddeley, 1992, 2000, 2003; Bookheimer, 2002; Fletcher & Henson, 2001). Here working memory refers to short-term online representations of information that are explicitly available for further processing (e.g., active rehearsal or manipulation/processing). Encoding, consolidation, and retrieval from declarative memory are thought to depend on the interaction between the medial temporal

lobe and the prefrontal cortex, as well as posterior neocortical regions. Other brain regions, which are also important for long-term declarative memory, include the thalamus (e.g., the anterior and mediodorsal nuclei), the mamillary bodies, and the basal forebrain nuclei, as well as the retrosplenial cortex.

Electrophysiological studies have demonstrated the importance of the parahippocampal and perirhinal cortices in memory consolidation in animal models. In rats, perirhinal cortex activation appears to promote enhancement in cortico-cortical pathways (Ivanco, Michelin,

& Racine, 1996). In monkeys, Higuchi and Miyashita (1996) demonstrated that lesions of the entorhinal and perirhinal cortices prevented the formation of neuronal memory responses to visual paired associates in the inferotemporal cortex. The parahippocampal region seems to mediate the extended persistence of these cortical representations and processing within the neocortex may take advantage of lasting parahippocampal representations, and come to reflect complex associations between events that are processed in different neocortical regions or occur sequentially in the same or different areas (Eichenbaum, 2000). An alternative proposal states that the hippocampus and related structures are always necessary for recalling the richness of detail available in autobiographical recollections (Nadel & Moscovitch, 1997).

In the following section we will outline a position on human memory systems that is closely related to that of Squire et al. (2004) and Eichenbaum and Cohen (2001).

However, it should be noted that there are several alternative perspectives (cf., section 4.3), suggesting that distinct sub-regions of the MTL support dissociable functions (e.g., Aggleton & Brown, 1999; Murray & Mishkin, 1986; Murray & Bussey, 1999; Simons &

Spiers, 2003; Tulving & Markowitsch, 1998; Yonelinas et al., 2002). For example, one model suggests that a system involving the hippocampus (as well as the thalamus, mamillary bodies and retrosplenial cortex) subserve recollection, while parts of the parahippocampal cortex (perirhinal cortex) support familiarity-based recognition. In addition, Squire and colleagues (2004) as well as Eichenbaum and Cohen (2001) argue for a time-limited role in remote declarative memory and a central role of the MTL in memory consolidation.

Afferent information to the MTL originates from most neocortical association areas (Suzuki & Amaral, 1994a, 1994b). These neocortical regions project to one or more of the parahippocampal subdivisions, which include the parahippocampal, the perirhinal, and the entorhinal cortices. The subdivisions of the parahippocampal region are strongly interconnected and send efferent projections to several parts of the hippocampus itself, including the dentate gyrus, the CA1-3 (Cornu Ammonis) fields, and the subiculum. Within the hippocampus, there are divergent and convergent connections, supporting plasticity mechanisms that participate in the rapid encoding of information (Amaral, 1993; Amaral &

Witter, 1989; Bliss & Collingridge, 1993). In particular, the CA3 has the basic architecture of a generic recurrent network. The outcome of hippocampal processing is returned, via the parahippocampal region, to the same brain regions from where the information originated (Burwell, Witter, & Amaral, 1995; Suzuki, 1996). Several additional structures, including for example the mamillary bodies, the anterior and mediodorsal thalamic nuclei, the basal forebrain nuclei along with other subcortical nuclei, interact with the hippocampus through a major fiber bundle the fornix (Lavenex & Amaral, 2000; Squire, 1992; Squire & Zola-Morgan, 1991; Suzuki, 1994).

A prominent feature of the structures forming the neocortical-MTL loop is their organization into hierarchical association networks (Felleman & Van Essen, 1991; Lavenex

& Amaral, 2000). The connections within the parahippocampal, entorhinal, and perirhinal cortices as well as the convergence/divergence of inputs at the different levels of the neocortical-MTL loop enable a significant amount of integration before information reaches the hippocampus proper. Lavenex and Amaral (2000) suggest that the level of integration and complexity of the information increases when moving from the neocortex to the hippocampal complex; unimodal information becomes polymodal/amodal, and reaches the highest level of abstraction within the hippocampal complex, before it is returned to the neocortex. This suggests a significant contribution of the neocortical-MTL loop to declarative memory formation, consolidation, and memory retrieval. Furthermore, the hippocampal output can also influence the processing of incoming information through the feedback projections from the hippocampal complex to the neocortex. The output of the MTL system is ultimately distributed, via these feedback projections, to much of the neocortex. Neocortical regions have specific perceptual, cognitive, and motor processing

functions required to complete a given memory task, including the important aspects of short-term working memory. It has been suggested that the parahippocampal region is critical in extending the persistence of information over brief periods (Eichenbaum, 2000), while the ultimate structure along the neocortical-hippocampal loop, the hippocampal complex, participates at the highest level of integration and organization of information (Felleman & Van Essen, 1991; Lavenex & Amaral, 2000). Thus, it is suggested, that the parts of the parahippocampal region, which receives convergent inputs from the neocortical association areas and return projections to all of these areas, might mediate the extended persistence of neocortical representations. If this is correct, the interactions between neocortical regions and the MTL can utilize briefly lasting parahippocampal representations, which reflect complex associations between events that are processed separately in different cortical regions or occur sequentially in the same or different areas (Eichenbaum, 2000). It is interesting to note an early suggestion of David Marr (1971), who first suggested that the hippocampal formation creates indexes or pointers of incoming information for rapid storage. These pointers are thought to participate in the consolidation and integration/reorganization of neocortical representations during sleep. In line with this suggestion, Squire and colleagues (2004) suggest that the MTL interacts with the neocortex in order to establish, maintain, and retrieve long-term memory, and that ultimately, declarative memory becomes (relatively) independent of the MTL through a process of consolidation. Finally, it is likely that the principal component of the declarative memory system, including the neocortex, contributes differentially to declarative memory and the interactions between these components are essential (Simons & Spiers, 2003). However, given the neuroanatomic characteristics of the neocortical-MTL system (weak hierarchical organization, high level of associativity, and recurrent connectivity), it might be difficult to experimentally distinguish the different functional properties of some of the sub-structures.

Although, it has been suggested that neurophysiological, neuroimaging, and neuroanatomic data indicate a division of labor within the MTL (Tulving & Markowitsch, 1997), Squire and colleagues (2004) suggest that the available data do not support simple dichotomies between the functions of the hippocampus and the adjacent MTL structures (e.g., associative vs. non-associative memory, episodic vs. semantic memory, recollection vs.

familiarity).