No abstract

No abstract

In foraging, behavior is guided by action values that are stored in memory and updated depending on the history of choices and rewards. What is the neural mechanism for action value maintenance and reward-dependent update? Here we explore two contrasting network models: neural integration and synaptic learning of action value. We show that both mechanisms reproduce existing experimental data, but they depend on distinct and falsifiable assumptions about the biological neural circuits. We demonstrate neural dynamical signatures and perturbations that could be carried out experimentally to differentiate the two contrasting scenarios.