003 Theories of hippocampal function in anterograde amnesia
- Psychological Sciences, University of Manchester
- Cognition and Cognitive Neuroscience, University of Manchester
Abstract: In anterograde amnesia, following an encoding encounter, memory for facts and episodes is impaired after distraction has occurred. In contrast, intelligence and working memory along several kinds of procedural memory are preserved. This problem is caused not just by damage to the hippocampus or other medial temporal lobe (MTL) structures, but also to basal forebrain, midline diencephalon, retrosplenial cortex, and ventral midbrain structures as well as to fibre tracts, such as the fornix, that connect these regions together. Memory is disrupted in very similar ways by damage to each of these structures, so it is clear that they work together as a system to create, store and later retrieve declarative memories. Despite the relatively unified working of the system, however, some evidence suggests that lesions to components of the MTLs and their connections may have dissociable effects. The MTLs comprise a hierarchical convergence zone. Processed object information converges in the perirhinal cortex, spatial/contextual information converges in the parahippocampal cortex, and these and other kinds of information converge in the hippocampus at the top of the hierarchy. Following further processing, there is back-projection to the neocortical regions that sent the information. A dominant view is that the hippocampus represents, consolidates and stores high-level associations through a rapid learning process. This creates a hippocampal index, which when triggered by a cue reactivates the memory by backprojections to the neocortical regions that represent the memory's components. Although the MTLs are highly interconnected structures, they receive and process different informational projections and the perirhinal/parahippocampal cortices' neural architecture differs from that of the hippocampus, so they probably perform qualitatively different processing operations. Neural network modelling suggests that the perirhinal cortex rapidly creates relatively un-sparse, rather unspecific object memories whereas the hippocampus rapidly creates sparse, very specific, high-level associative memories. Perirhinal cortex memories support item familiarity memory well, but cued recall very poorly whereas hippocampal memories provide powerful support for recollection, but are not needed for item familiarity. This is consistent with some human and animal lesion evidence that finds recognition (familiarity)/recall (recollection) double dissociations with hippocampal vs perirhinal/entorhinal cortex lesions. For over a decade, there has been no consensus about whether this hypothesis is broadly correct, whether all MTL lesions have qualitatively similar effects on declarative memory, or whether such lesions always disrupt recall and recognition equally, but with different lesions disrupting memory for different information. This disagreement will be discussed.