Electrophysiological correlates of state transition prediction errors

Planning behavior depends crucially on the ability to distinguish between the likely and unlikely consequences of an action. Formal computational models of planning postulate the existence of a neural mechanism that tracks the transition model of the environment, i.e., a model that explicitly represents the probabilities of action consequences. However, empirical findings relating to such a mechanism are scarce. Here we report the results of two electroencephalographic experiments examining the neural correlates of transition model learning. The results implicate fronto-midline theta and delta oscillations in this process and suggest a role of the anterior midcingulate cortex in planning behavior.