Hippocampal Networks and Episodic Memory
Abstract
People routinely organize the flow of events into distinct episodes when dealing with the immense complexity of dynamic real environ ments. Individual episodic memories incorporate information about the identity of items, their spatial locations, and the order in w hich they occurred (what, where, and when). They also include records of actions taken during the episode and the context or s ituation in which the events happened. In all, the brains episodic memory system is a powerful tool for the orderly encoding and su bsequent retrieval of an extraordinary amount of information. It is also critical for planning future actions and for diverse cognit ive operations including inferential thinking. Understanding how the cortical telencephalon processes episodes is thus a basic issue in contemporary neuroscience and one that is central to the development of brain-based artificial intelligence. Experiments using h umans and animals indicate that the hippocampus plays a critical role in the formation and recall of episodes. The studies also iden tified pathways that carry what and when information to the structure, and our recent work led to discovery of a unique hippocam pal subsystem that adds the temporal element. Still missing is a description of network level operations (signal transformations) th at occur in the three basic subdivisions of hippocampus; the absence of such data prevents the construction of realistic simulations and derivation of related computational operations. The first goal of the proposed work is to describe how input signals arriving a t behaviorally relevant frequencies are transformed across polysynaptic hippocampal circuits. Preliminary experiments suggest that n odes in the circuit perform different types of functions (e.g., frequency filtering, on-off switching) and describe novel roles for sparse inputs arriving from lower brain areas. The results from this project will greatly increase the realism and range of our ext ant hippocampal models.Hippocampal output feeds into the subicular-entorhinal complex. Together, the complex and hippocampus form a giant loop that begins and ends in entorhinal cortex, a structure that has extensive two-way connections with association areas of neocortex. Preliminary work led to the surprising conclusion that the subicular nodes massively amplify signals arriving from hippo campus. The second goal is to complete a first network level physiological analysis of the subicular-entorhinal complex and the mann er in which it processes hippocampal output. Completion of this project will enable simulations of the loop and open the way to link age with models of posterior cortex under construction by our collaborators. The third goal is to build relatively simple, anatomica lly based models of prefrontal cortex and then to interconnect these with the hippocampal loop-posterior cortex system. Preliminary studies showed that a model of this type acquires a fundamental form of prefrontal learning (reversal learning) with performance th at compares favorably with a popular version of machine learning. Progress with the models will lead to tests of whether they add ac tions to episodic memories and are able to sort through and segregate such memories.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 22, 2021
- Source ID
- N000142112940
Entities
People
- Gary Lynch
Organizations
- Office of Naval Research
- United States Navy
- University of California, Irvine