Hippocampus and configural-relational information: A relationship confined to memory?

Cashdollar and colleagues (1) describe a powerful combination of behavioral and magnetoencephalography data from healthy participants and patients with bilateral hippocampal sclerosis to support a role for the hippocampus in the active maintenance of configural-relational (C-R) information. This study reveals the importance of hippocampal-dependent theta synchronization of temporal and occipital regions in humans to C-R working memory, and how such coordination may contribute to the integration of visual information represented throughout the rostral-caudal extent of the ventral visual processing stream. These findings, therefore, add to a growing body of literature arguing for an involvement of the hippocampus in both long- and short-term relational memory.

One question arising from this study (1) is whether the involvement of the hippocampus in processing C-R information is restricted to the mnemonic domain. To address this, Cashdollar and colleagues (1) highlight the findings from their task control condition: participants were presented with a sample image and then, following a short delay, were required to determine whether two images unrelated to the earlier sample were identical. These images were either the same or differed by the addition of a new object or the deletion or spatial shifting of an existing object. Critically, patients with hippocampal damage were not impaired on this condition, leading the authors to conclude that their work “does not support a role for the hippocampus in perceptual feature integration” (1).

Although this control condition is a useful comparison with the C-R memory condition, it may not provide a sensitive test of the hypothesis that the hippocampus is involved in the perception of spatial stimuli. To date, the most striking demonstration of spatial processing difficulties in amnesic cases has been in the context of spatial oddity judgment, in which participants select the “odd one out” from an array of simultaneously presented scenes. Hippocampal damaged patients are significantly impaired when all of the presented images depict different viewpoints of the same indoor scene, with the target item containing a feature that has been shifted subtly in position or changed in dimension. This may place a particular demand on the ability to disambiguate multiple features, rather than locating a single feature that is different between images. In contrast, intact performance or milder deficits are observed when the foil images portray the same viewpoint of the same scene and the target item not only contains a spatially shifted or altered feature but also shows a different viewpoint (2, 3), a task that can be solved using a single feature. Thus, the perceptual trials of Cashdollar and colleagues (1) may have placed an insufficient emphasis on those perceptual processes for which the hippocampus may be critical, tempering the authors’ original conclusion that this structure is not critical for perception.

Unsurprisingly, the involvement of the hippocampus in spatial perception remains a contentious issue (4, 5). However, if the boundary between short- and long-term relational memory is no longer as distinct as previously thought, it may be equally difficult to make an absolute distinction between perception and working memory and conclude that the hippocampus is critical for the latter but not the former.