HEBB MARR NETWORKS THE HIPPOCAMPUS AND SPATIAL MEMORY

点击翻译按钮获取中文摘要

DESCRIPTION (Adapted from applicant's abstract): The aim of this project is to understand the mechanisms and computational principles underlying the encoding of memories that depend on a functional hippocampus (HC), and their subsequent consolidation into HC-independent, long-term storage, presumably in the neocortex (NC). This is a prerequisite to ameliorating memory disorders resulting from developmental abnormalities, disease, trauma, aging or drug abuse. The working hypothesis is that the function of long-term memory is to construct, from past experience, internal representations that reflect the statistical regularities of the environment, and which thus permit adaptive generalization and appropriate responses. Such learning requires repeated, interleaved exposure to the items to be stored, with small adjustments to the synapses on each trial. Attempts to store new information all at once in such a memory lead to "catastrophic interference" with items already stored; however, survival often requires adding new items to the existing categorical structure, with only one or a few experiences. It is hypothesized that HC generates and stores compact representations of the spatial contexts of such events, which become associated with the individual components of the events in their respective NC areas at the time of the experience. By spontaneously reactivating these stored contexts, the HC facilitates the reinstatement of the original patterns in a coherent fashion throughout NC, and thus provides the required "training trials" while the system is "off-line". This provides a framework for understanding the phenomenon of temporally graded retrograde amnesia following HC damage, and why and how memory consolidation takes place. Using simultaneous recordings from 50-150 neurons, it has been shown that population-codes for novel environments develop rapidly during exploration, and that traces of recent events and event-sequences can be observed in the collective activity of neurons in both HC and NC during quiet wakefulness and slow-wave sleep. The proposed recording studies address specific aspects of the theory: whether the robustness of off-line reactivation of memory traces is correlated with subsequent performance; whether more remote memories are interleaved with newer ones during off-line periods; whether there is a retrograde gradient of memory reactivation; which behavioral states are most conducive to reactivation; to what extent reactivation also occurs in NC during the off-line periods; whether NC and HC reactivations reflect the same recent experiences; and whether a functional HC is necessary for reactivation of recent memory in NC. Finally, it is proposed to test the theory by imposing distributed patterns of electrical activity in HC using spatially structured electrical stimuli both during acquisition of spatial information and during off-line periods.