Different functions for different medial temporal lobe structures?

Abstract

Since the landmark study of the globally amnesic patient H.M. by Scoville and Milner (1957), the essential role of the medial temporal lobes (MTL) in long-term memory has been well established. Considerable success has been made dissociating declarative memory processes (explicit memory for events and facts), which rely on the integrity of the MTL, from nondeclarative memory processes, which rely on other cortical, subcortical, and cerebellar structures (Gabrieli 1998). There has also been progress in distinguishing frontal-lobe and MTL contributions to declarative memory. Determining the precise contributions of the structures within the MTL to declarative memory, however, has proved challenging. The MTL is comprised of multiple structures including the hippocampal formation (the dentate gyrus, CA fields, and the subiculum), the amygdala, entorhinal cortex, and surrounding perirhinal and parahippocampal cortices. The focal role of the amygdala in the emotional modulation of memory appears clear (Phelps and Anderson 1997), but there is little consensus about how to characterize the differential contributions of specific MTL structures to declarative memory. Isolating the function of each of these structures, however, is central to understanding declarative memory processing and how damage or disease disrupts such processing. Different MTL structures must be mediating different component processes underlying declarative memory, but the unanswered question is what are those different processes? One emerging idea about differential functions within the MTL is that the hippocampal formation is necessary for declarative memory tasks that require the processing of relations between multiple stimuli, whereas surrounding cortices mediate performance on tasks that rely on stimulus familiarity (or the converse, novelty). This hypothesis is consistent with the hierarchical connectivity of this region (Fig. 1A). Information from unimodal and polymodal association cortices enters the MTL through the perirhinal and parahippocampal cortices, which project to the entorhinal cortex, which in turn provides the major input to the hippocampal formation. This connectional anatomy suggests that the representational capacity of the perirhinal and parahippocampal cortices may be highly related to the incoming sensory information. These cortices may mediate familiarity-based memory processes linked to sensory aspects of items. In contrast, the hippocampus may generate a more abstract representation that integrates or binds multiple cortical sources of information, which is an important aspect of relational memory processes. According to this view, the degree to which a memory task recruits relational versus familiarity processes would determine its relative reliance on hippocampal versus nonhippocampal MTL structures. Tasks that rely greatly on relational processing, such as paired-associate learning, would be greatly impaired by damage to the hippocampal formation. In contrast, tasks that can be performed on the basis of stimulus familiarity, such as tests of item recognition, could be supported by the function of the surrounding MTL cortices in the absence of hippocampal function. Several convergent findings have supported this hypothesis, including evidence from lesion and electrophysiological studies in animals (for review, see Brown and Aggleton 2001; Eichenbaum 2001), and from neuropsychological (Vargha-Khadem et al. 1997), and neuroimaging (Eldridge et al. 2000; Yonelinas et al. 2001) studies in humans. A provocative challenge to this idea is set forth in a study by Stark et al. (2002) examining amnesic patients with bilateral MTL lesions thought to be limited to the hippocampal formation. By using patients with well-characterized lesions, the authors add an important contribution to the ongoing efforts to characterize MTL function in declarative memory. Examining patients with focal hippocampal damage allows the authors to test how damage to this structure affects memory tasks that differentially require relational and familiarity processing. In this study, hippocampal patients performed both single-item and associative recognition tasks. For single-item recognition, patients studied pictures of single faces or houses and after a delay, performed a yes/no recognition memory judgment. For associative recognition, patients studied face-house paired associates and after a delay, were presented with pairs of items. In this associative task, patients had to discriminate items that had been studied together (intact pairs) from items that had not been studied together (recombined pairs). These tasks differentially focus on the processing of relations between test items compared with the processing Corresponding author. E-MAIL gabrieli@psych.stanford.edu; FAX (650) 725-5699. Article and publication are at http://www.learnmem.org/cgi/doi/ 10.1101/lm.54702.