The goal of this work is to help clarify the contribution of the hippocampus to memory by testing the implications of a recently-developed theoretical framework for understanding the complementary roles of the hippocampus and neocortex in learning and memory. This framework centers around a set of principles derived from computational neural network models, and we have instantiated it in a specific biologically-based model. Many prominent theories of the hippocampus assume that it contributes to memory by storing conjunctive representations of the features that make up a particular experience. However, this idea is too general because there are a number of reports that animals with severe damage to the hippocampus solve problems that require conjunctive representations. The present model recognizes that the cortical system alone can store stimulus conjunctions. However, it is posited that the hippocampal and cortical systems differ along two important dimensions: (a) learning rate-the hippocampal system rapidly stores conjunctions whereas the cortical system learns more slowly, and (b) the task demands under which the two systems are driven to develop conjunctive representations-the hippocampal system has a natural bias to automatically store conjunctions just as a function of the organism sampling the environment, whereas the cortex is not naturally biased to store conjunctions, and only does so when forced to by the demands of a problem (e.g., reinforcement contingencies). As implemented in the present model, these ideas provide a good account of many phenomena in the literature. By comparing rats with damage to the hippocampus and intact rats, and conducting additional simulations, the goal is to evaluate the proposed theoretical framework by testing its implications for: (a) when the hippocampus will contribute to complex nonlinear discrimination problems, (b) understanding how the hippocampus contributes to contextual fear conditioning, (c) the role of the hippocampus in tasks where conjunctions are automatically learned versus those where they must be learned to satisfy the contingencies of reinforcement, and (d) how the hippocampus mediates what some call logical reasoning (transitive inferences). By pursuing these aims, the proposed research will provide a more complete understanding of the role of the hippocampus in memory.
Showing the most recent 10 out of 23 publications
