Decades of research support the central role of the medial temporal lobes (MTL), particularly the hippocampus, in episodic memory function (Squire et al., 2004
). Although the most profound forms of amnesia are associated with lesions to these structures (Scoville and Milner, 1957
), a variety of other cortical regions also play important roles. For example, the prefrontal cortex contributes to controlled-strategic processing at encoding and retrieval, supporting accurate memory performance (Alexander et al., 2009
, Blumenfeld and Ranganath, 2007
). Working memory, which is in part subserved by frontal and parietal regions (Champod and Petrides, 2007
), also generally modulates episodic memory performance by promoting efficient encoding of information to later be remembered. Additionally, the semantic memory system is involved in modulating the depth and organization of encoding, impacting the strength of a memory trace (Craik and Lockhart, 1972
, Goldblum et al., 1998
). While the left ventrolateral prefrontal cortex has been implicated as critical to semantic access at memory encoding (Logan et al., 2002
, Otten et al., 2001
), such as the high level conceptual representations thought supported by the temporal pole (Lambon Ralph et al., 2009
, Lambon Ralph et al.
, Patterson et al., 2007
), are also likely essential to robust encoding.
Alzheimer's disease (AD) is the most common form of acquired amnesia. Given the early and profound involvement of neuropathology, particularly neurofibrillary tangles, in the MTL of AD patients (Arnold et al., 1991
, Braak and Braak, 1991
, Delacourte et al., 1999
, Price et al., 1991
), much of the work examining the anatomic basis of memory impairment in this condition has focused on the hippocampus and other MTL structures. However, AD pathology, as reflected in the clinical phenotype, involves multiple neural networks supporting other cognitive domains such as executive functioning, language, semantic memory, and visuospatial processing (Arnold et al., 1991
). The influence of these additional domains of impairment is often neglected when considering episodic memory performance measures.
A variety of standardized episodic memory measures, frequently involving verbal list learning tasks, have been utilized for clinical diagnosis and monitoring of disease progression in AD. These tasks typically produce several memory measures, including immediate recall (commonly with repeated exposure), delayed free recall, and often recognition memory. However, much of the work in AD has focused on delayed free recall or treated these different measures as instances of a monolithic process. In fact, a number of studies have combined many of these measures into summary statistics to capture “episodic memory.” While such an approach is not unreasonable in that it may reduce the variability inherent in individual tests and may be useful in screening for any form of memory impairment, it reflects a process purity that is only approximate at best.
For example, a major debate in the memory literature is whether recollection, associated with detailed contextual retrieval of a prior event, is differentially represented in the MTL relative to familiarity, or an acontextual sense of prior encounter (Aggleton and Brown, 2006
, Eichenbaum et al., 2007
, Squire et al., 2007
, Yonelinas, 2002
). A variety of lines of research have supported an anatomic mapping of recollection to the hippocampus and familiarity to extrahippocampal MTL structures, particularly perirhinal and lateral entorhinal cortices [(Bowles et al., 2007
, Davachi et al., 2003
, Fortin et al., 2004
, Montaldi et al., 2006
, Wolk et al., in press
); for review, see (Aggleton and Brown, 2006
)]. As free recall and recognition memory are likely differentially dependent on these memory processes, these measures may reflect differential localization of pathology within the MTL. Moreover, immediate recall after a single trial is likely related more to working memory than to so-called long-term memory processes involved in recall with repeated learning or after a delay. Thus, performance on an initial learning trial(s) would likely be predicted to relate to frontoparietal or other nodes of the working memory system.
Since one major goal of psychometric testing is to reveal specific changes in underlying brain systems, we sought here to determine the neural correlates of the different measures associated with three common verbal memory measures that are typically part of clinical research batteries in AD. Specifically, we focused on correlates of immediate recall (early versus later trials), delayed recall, and recognition memory. These measures were derived from three standardized psychometric memory instruments. Drawing on the concept in both pathology (Arnold et al., 1991
, Braak and Braak, 1991
, Brun and Gustafson, 1976
) and imaging studies (Baron et al., 2001
,Fox et al., 2001
, Frisoni et al., 2002
, Thompson et al., 2001
) that early in its course AD is not a diffuse process but is in fact associated with abnormalities in a relatively stereotyped set of brain regions, we utilized a relatively novel region of interest (ROI) approach in which a set of regions that have been previously demonstrated to be sensitive to mild AD – the “Cortical Signature of AD” (Bakkour et al., 2009
, Dickerson et al., 2009
) – were defined on the cortical surface. These regions, along with the hippocampus, represent structures most likely to be associated with the pathology of AD and serve as a type of lesion model for understanding the brain-behavior relationships of the memory measures.
We predicted that early immediate memory trials would be most dependent on regions supporting working memory and perhaps other aspects of controlled strategic processing and semantic elaboration, but that later trials would be more dependent on MTL structures. Delayed recall and recognition, which require the storage of information over time, were expected to be more dependent on MTL structures than the immediate trials. Finally, following the dual process model, we predicted that recognition memory performance, dependent on a combination of recollection and familiarity, would be more strongly associated with extrahippocampal MTL cortical atrophy while free recall, dependent solely on recollection, would be associated with hippocampal volume.