ReviewRelationship between hippocampal volume and memory ability in healthy individuals across the lifespan: review and meta-analysis
Introduction
Most, if not all, of the cerebral cortex is likely to subserve the learning and retrieval of facts and events, with the engagement of specific regions dependent on the nature of the material learned (see Damasio, 1989; McClelland, McNaughton, & O’Reilly, 1995 for theoretical accounts, and Köhler, Moscovitch, Wincocur, Houle, & McIntosh, 1998; Nyberg, Habib, McIntosh, & Tulving, 2000; Senkfor, Van Petten, & Kutas, 2002; Wheeler, Peterson, & Buckner, 2000 for empirical reports). Despite general acceptance of the idea that memory is not localized to one neural structure, there is overwhelming evidence that the medial temporal lobe (MTL) plays a central role in episodic memory for diverse material, via its interactions with distributed cortical regions (Scoville & Milner, 1957). The MTL consists of multiple cortical areas, including the hippocampal formation, surrounding entorhinal, perirhinal, and posterior parahippocampal regions within the parahippocampal gyrus in humans, and part of the temporal pole (Insausti et al., 1998a, Insausti et al., 1998b). Although many cases of human domain-general amnesia have damage to multiple regions of the medial temporal lobe, or to the diencephalon (Aggleton & Brown, 1999; Brown & Aggleton, 2001; Mayes, 2000; Smith & Bigel, 2000), lesions restricted to the hippocampus alone result in memory deficits in both humans and nonhuman primates (Rempel-Clower, Zola, Squire, & Amaral, 1996; Zola & Squire, 2000). The latter conclusion was drawn from postmortem examination of the affected brains, but the availability of high-resolution magnetic resonance images has led to a more widespread research effort to tie hippocampal pathology to memory impairment over the last decade.
MR scans are used to identify and/or confirm the location of damage after a frank neural insult such as stroke, hypoxia, or closed head injury. But an additional advantage of a noninvasive method is the possibility of identifying subtle neural damage when the insult is not as obvious. Recent studies have examined hippocampal volumes in individuals with childhood seizures, posttraumatic stress disorder, borderline personality disorder, depression, high risk of schizophrenia due to affected relatives, an ApoE-4 allele, and high estrogen levels (Bremner et al., 1995; Cohen, Small, Lalonde, Friz, & Sunderland, 2001; den Heijer et al., 2003, Driessen et al., 2000; Fennema-Notestine, Stein, Kennedy, Archibald, & Jernigan, 2002; Lawson et al., 2000, O’Driscoll et al., 2001, Plassman et al., 1997, Seidman et al., 2002; Sheline, Sanghavi, Mintun, & Gado, 1999; Simpson, Baldwin, Burns, & Jackson, 2001; Stein, Koverola, Hanna, Rochia, & McClarty, 1997; VanLandingham, Heinz, Cavazoa, & Lewis, 1998). Particularly strong attention has focused on the possibility of detecting Alzheimer’s disease before its clinical onset, when interventions might be more effective (see Chetelat & Baron, 2003; Kantarci et al., 2002, Wolf et al., 2003 for recent reviews).
Some clinical investigations have included measures of memory performance in additional to MR measures. When pathology is clearly present, these studies have largely succeeded in demonstrating quantitative relationships between hippocampal volumes and memory. Positive correlations between hippocampal volume and memory performance have been frequently reported in Alzheimer’s patients, for instance, as well as in other varieties of dementia, and in amnesic patients of mixed etiology (Barber, McKeith, Ballard, Gholkar, & O’Brien, 2001; Cahn et al., 1998, de Toledo-Morrell et al., 2000, Deweer et al., 1995; Jernigan et al., 2001, Jernigan et al., 2001; Köhler, Black, et al., 1998; Kopelman et al., 2001; Mizuno, Wakai, Takeda, & Sobue, 2000; Mungas et al., 2002, Petersen et al., 2000, Wilson et al., 1996). Such correlations have also been observed in samples for which memory performance is at the borderline of clinical impairment, and a pathological process is suspected, as in older adults with mild cognitive impairment (Jack et al., 2000, Soininen et al., 1994).1
These clinical studies, and the expanding application of MR volumetry to diverse populations, raise a fundamental question: what is the relationship between hippocampal volume and memory ability in the normal brain? Given that the measured size of hippocampi varies a good deal across healthy individuals, and memory abilities also vary, is there a detectable structure–function relationship? When this question is posed bluntly, the response of many cognitive neuroscientists is likely to be “no”, in part because macroscopic size is the crudest of neurobiological metrics, and in part because the question elicits memories of the failures of phrenology a century ago. In response, it might be argued that the cognitive psychology of the 21st century is more sophisticated than what the phrenologists had to work with, so that it may be more plausible to imagine size/function relationships for “episodic memory” than for the various “faculties” proposed by Gall and Spurzheim (but see Uttal, 2001 for a thoughtful and critical discussion of the history and current status of cognitive taxonomies and the limits of localization). More to the point, a survey of the literature indicates that a modern exercise in searching for size/function relationships is already underway. Numerous recent studies report significant correlations between hippocampal volumes and memory across individuals without a neurological or psychiatric diagnosis. However, numerous studies have also failed to find such relationships; these null results are less frequently cited, perhaps because they often occur in brief descriptions of results from a control group and are overshadowed by results from a patient group.
The discrepancy among published reports may result from random variability around a mean of zero correlation. Alternatively, both significant and null results across studies could arise from a subtle but genuine relationship. By analogy, the impact on cognitive function of white matter hyperintensities (WMH) observed in MR images from healthy older adults on cognitive function was subject to debate for some time, as the first several studies variously reported a negative relationship with cognitive performance (Rao, Mittenberg, Bernardin, Haughton, & Leo, 1989; Schmidt et al., 1991), or no detectable relationship (Hunt et al., 1989, Mirsen et al., 1991; Tupler, Coffey, Logue, Djang, & Fagan, 1992). In that case, pooling results from multiple studies to obtain a larger combined N was successful in demonstrating a moderate negative impact of WMHs on cognitive performance (Gunning-Dixon & Raz, 2000). The present paper evaluates whether pooling results from multiple studies in a meta-analysis will similarly show a relationship between hippocampal volumes and memory ability. Below, I first describe alternative hypotheses about this relationship, which guide the subsequent analyses.
Section snippets
Bigger is better
Three general hypotheses about the possible relationship between hippocampal volume and memory ability can be imagined. The first and simplest is the bigger is better (BIB) hypothesis, that regardless of the causal factors underlying the size of a structure, a larger structure should result in stronger function. This hypothesis seems to be implicit in some reports, but the investigators who have explicitly considered the BIB hypothesis have rejected it, because they observed significant
Search and inclusion criteria
Papers reporting relationships between hippocampal volumes and memory measures were located by a search of the Medline and PsycLit databases using the keyword combination (hippocampus OR hippocampal OR medial temporal) AND (memory OR recognition OR recall OR neuropsychological) AND (volume OR volumetric OR atrophy), and by examining the references of those papers. Results were included when they met all of the following criteria: (1) publication in English; (2) participants without a
Common correlation across studies, and heterogeneity of the results
The common correlation between hippocampal volume and memory performance across all 33 studies (weighted by sample size) was 0.071 (Ztest=3.00, P<0.005). Although the estimated correlation across studies was significantly positive, Table 1 shows a great deal of variability across studies: reported correlations range from positive 0.55 to negative 0.55. Fig. 1 shows the distribution of reported correlations, which appears to deviate from normal. The heterogeneity test confirmed the visual
Limitations of meta-analyses
A potential hazard of any meta-analysis is the so-called “file drawer problem”: studies that are unpublished due to null or difficult-to-interpret results. Some protections against this problem are built into the present topic. The first stems from the expense, effort, and relative novelty of relating structural MR data to cognition, so that investigators are likely to be strongly motivated to publish, and editors likely to show a similar interest. The second protection applies to a subset of
Acknowledgements
I am grateful to Patrick Davidson and Elizabeth Glisky for discovering some of the relevant papers in the literature, and to Trudy Kuo for comments on the manuscript. Four anonymous reviewers also provided extremely constructive comments on a previous version of this paper. Financial support was provided by the National Institute of Aging (AG14792).
References (151)
- et al.
Longitudinal decline of the neuronal marker N-acetyl aspartate in Alzheimer’s disease
Lancet
(2000) - et al.
Sexual dimorphism and asymmetries in the gray–white composition of the human cerebrum
Neuroimage
(2003) - et al.
Problems with ratio and proportion measures of imaged cerebral structures
Psychiatry Research: Neuroimaging
(1991) - et al.
A characterization of performance by men and women in a virtual Morris water maze task: A large and reliable sex difference
Behavioural Brain Research
(1998) - et al.
Is there a negative correlation between explicit memory and hippocampal volume?
NeuroImage
(1999) - et al.
Early diagnosis of Alzheimer’s disease: Contribution of structural neuroimaging
Neuroimage
(2003) Time-locked multiregional retroactivation: A systems-level proposal for the neural substrates of recall and recognition
Cognition
(1989)- et al.
Brain morphometry in female victims of intimate partner violence with and without posttraumatic stress disorder
Biological Psychiatry
(2002) - et al.
Childhood-onset schizophrenia: Progressive brain changes during adolescence Biological Psychiatry
(1999) - et al.
A voxel-based morphometric study of aging in 465 adult human brains
Neuroimage
(2001)