Full Program »
Reinforcement encoding along the hippocampal long axis and individual differences in exploration/exploitation
The human hippocampus binds spatiotemporally adjacent slivers of experience into cognitive maps. Hippocampal place cells and grid cells in the adjacent entorhinal cortex support navigation of longitudinal-latitudinal and abstract spaces. The hippocampal posterior-anterior long axis is organized along a functional gradient from smaller to larger spatial scales, from non-semantic to semantic associations, and from detailed to gist memories. We know less about the differential roles of anterior vs. posterior hippocampus in binding representations of reinforcement into a cognitive map. Employing a reinforcement-based timing task and fMRI, we investigated how anterior vs. posterior hippocampus guides the transition from initial exploration to exploitation. We modeled reinforcement learning with the SCEPTIC model (Hallquist & Dombrovski, 2019). Posterior hippocampus responded to reward prediction errors, particularly in individuals who made more exploratory choices. Anterior hippocampus, by contrast, was active when a global value maximum was found, particularly in individuals whose choices converged on this maximum. Voxelwise deconvolution analyses revealed underlying patterns of offline processing in the anterior hippocampus and striking individual differences in intrahippocampal connectivity graphs. Altogether, our results elucidate how the human hippocampus integrates complex experiences and maps the best action. Individual differences in this process may contribute to maladaptive behavior and psychopathology.