Differential replay for reward and punishment paths predicts approach and avoidance

Abstract

Planning is thought to involve neural replay, where states relevant to a task goal are rapidly reactivated in sequence. It remains unclear if, during planning, replay of a path relates to an actual prospective choice. Here, using magnetoencephalography (MEG), we studied participants while they planned to either approach or avoid an uncertain environment that contained paths leading to reward and to punishment. We show significant planning-related forward sequential replay with state-to-state transitions in the range of 20 to 90 ms. Replay of rewarding paths was boosted prior to a decision to avoid, and attenuated prior to a decision to approach. Crucially, a trial-by-trial bias in replaying punishing paths predicted an irrational choice to approach when a prospective environment was more risky, an effect that was particularly marked in more anxious participants. The findings reveal a coupling between the content of forwards replay and rational choice behaviour, such that replay prioritises an online representation of potential reward.