Neural dynamics reveal foraging-like computations in the frontal cortex

Abstract

The prefrontal cortex has a central role in flexible behavior. Computations for weighing alternatives, committing to an option, integrating evidence, and abandoning an option are dominantly formalized in reinforcement learning (RL). Yet, ecological considerations showcase foraging as a competing account. Identifying how the brain actually solves efficient weighing of options requires the identification of the neural representation of this abstract yet precise computation. We recorded neural populations from three areas of the primate frontal cortex involved in decision making to characterize neural dynamics while animals performed a standard RL task. By combining behavioral and neuron population analysis, we found a neural subspace in midcingulate (MCC) and ventrolateral prefrontal cortex (vLPFC) that precisely tracked value computations. We demonstrated that MCC dynamics showed signatures of foraging-like computation but conflicted with RL requirements. Our study thus provides a detailed neural account of foraging-like flexibility in the primate frontal cortex.