Mixture Modeling and Spatial Representations in the Rat Hippocampus

Mark Fuhs

Abstract

Place cells throughout the rat hippocampus show location-specific firing fields that topologically reorganize, or ``remap,'' when the rat moves to a different environment. In some cases, however, remapping occurs only after extensive exposure to one or both environments, indicating that it is experience-dependent and not purely sensory driven. We propose that this construction of a multi-map representation of the world can be understood as a mixture modeling process, where the degree of remapping between environments reflects the perceived statistical likelihood that the features observed are derived from distinct sources. We simulated the construction of mixture models for several different training paradigms where the environments differed (square vs. circular arenas, or different room locations, or different rooms), or the animals had different amounts of pretraining. We found that the observed time course of remapping could be explained by the degree of similarity between environments and the amount of experience the rat had in each one. Our computationally explicit mixture modeling theory can also be used to predict the time course of remapping in as-yet untested training paradigms.

Back to the Main Page