Effects of feedback delay and agency on feedback-locked beta and theta power during reinforcement learning.

Abstract

Feedback-based learning initiated by dopamine (DA) cell firing is crucial for adaptive behavior. The nature and context of feedback can vary, however, affecting how feedback is processed. For example, the feedback-related negativity (FRN) in the ERP in humans, which has been linked to the DA system, is reduced for delayed feedback and for observational compared to active learning. Recent research suggested that oscillations in the theta and beta band over the medio-frontal cortex reflect distinct feedback processing mechanisms. In this study, we hypothesized that the power in both frequency bands is affected by feedback delay and agency. We thus investigated effects of feedback delay (1s vs. 7s) on induced theta and beta band power and the FRN in a probabilistic feedback learning task in two participant groups, one learning actively and one by observation. For theta and beta, a larger power difference between negative and positive feedback for immediate than delayed feedback was found, driven by positive feedback for beta and by negative feedback for theta, while no differential modulation by agency was seen for theta or beta power following positive and negative feedback. These results indicate that feedback-locked beta and theta both reflect neural processes that are specific for the integration of feedback and recently preceding events, possibly linked to cognitive control and memory. With respect to the FRN amplitude, we could replicate previous findings of both delay and agency modulations, suggesting that the neural processes underlying feedback-locked ERPs and theta and beta power modulations differ.