Hippocampal Network and Episodic Memory
Abstract
The brain, without explicit instructions, routinely partitions the continuous flow of everyday experience into content rich episodes"" (narrative memories), an operation that is fundamental to orderly thinking. Cue identity (~what~), spatial relationships (~where~)," andtemporal order (~when~) constitute the core elements for a sequence. Experiments with humans and rodents indicate the hippocamp"us is critical to the encoding and retrieval of episodes; moreover, there is evidence linking the above three basic elements to spec"ific connections within the structure. But how the hippocampus through interactions with the cortex integrates these diverse factors and then later retrieves the results is very poorly understood. This is one of the major impediments to the development of brain-based theories of cognition and biologically inspired artificial intelligence systems (cognitive architecture AI). The present applica"tiondescribes a network simulation approach, based on past and ongoing work in our laboratory, to analyze the roles played by hippo""campus in the encoding, retrieval, and use of episodic memory. The program is unusual in that it begins with the construction of neu"robiologically realistic models of hippocampal subsystems and then follows with tests of whether these yield diverseaspects of epis"odic memory. Using this strategy, we have found that simulations of hippocampal field CA3 associate temporally spaced inputs (~cues~"") and then retrieve these in their correct order with appropriate relative intervals between them. Moreover, as observed in people," retrieval is time compressed. These effects emerged from an unexpected network level operationinvolving time dependent shifts in populations of active cells. The proposed work builds on these findings to include realistic simulations of the input and output stag"es (dentate gyrus and field CA1, respectively) of hippocampal circuitry; this will allow for first time analyses of how ~what~ and ~"where~ cues interact and of the manner in which hippocampal output is organized prior to delivery to the cortex. The realism of the system will be greatly increased by incorporation of results from proposed experiments on two essential network features: i) freque"ncy facilitation of synapses at dominant hippocampal rhythms, an effect required for robust throughput; and ii) modulation by cholin""ergic inputs. The augmented, three part simulation will then be tested for complex episodic phenomena including partitioning of sequ""ences, reconsolidation duringretrieval, and error detection during re-experience. Further refinement of the model will use data to" be collected on the firing of neurons in each hippocampal subdivision while animals are engaged in episodic learning. These chronic recording studies will be the first to examine neuronal activity associated with such learning in the experiential (unsupervised) c"ontext inwhich it is acquired in humans. Finally, we will develop algorithmic expressions for the complex computations performed by"" the hippocampal simulations, thereby enabling estimates of encoding capacity as a function of number of neurons. The algorithms wil""l be connected to ones previously extracted from network models of cortex, allowing for the construction of a computational version" of the cortico-hippocampal-cortical loop known from neuroanatomy. We intend to test the hypothesis that the hippocampus markedly extends the time frame over which cortical algorithms organize inputs into sequences and hierarchical categories and therebyensures t"hat cortical storage has an episodic quality. Related studies will determine if, as predicted, retrieval of the time expanded cortic"al material is prompted by the hippocampus. The collected set of algorithms will provide means for implementing neurobiologically gr"ounded, cognitively critical memory operations in various AI systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 23, 2018
- Source ID
- N000141812114
Entities
People
- Gary Lynch
Organizations
- Naval Information Warfare Center Pacific
- Office of Naval Research
- United States Navy