Reward changes the use and structure of predictive spatial maps

Abstract

Reward-related changes in hippocampal place-cell activity may reflect either different use of an existing spatial map or plasticity that reshapes the map. We separated these contributions in a two-pathway model: a recurrent network learned to predict sensory input and developed a spatial representation, while a reinforcement-learning controller combined this representation with an explicit goal signal to guide behavior. Replaying identical trajectories before and after learning isolated representational plasticity. Goal-directed behavior strongly concentrated place fields near reward, whereas changes to the representation were smaller and more variable. A reward-proximity teaching signal made these changes larger and more consistent, and the change was strongly attenuated when recurrent plasticity was restricted. After reward relocation, goal-related directional tuning adapted more fully than place-field organization, while withdrawing reward and proximity teaching reversed the learned field concentration. Thus, reward affects predictive spatial maps through separable mechanisms: goal information changes how an existing map is used, whereas reward-related teaching signals reshape the map itself. Our results provide a concrete framework for future experiments to disentangle how reward information, plasticity and behavior jointly shape hippocampal activity.