Feedback reliability modulates learning-related dynamics of midfrontal error monitoring

Abstract

Learning from feedback requires the brain to evaluate both performance outcomes and the reliability of the information used for learning. This study examined the dynamics of neural performance monitoring during deterministic (100% reliable) and probabilistic (75% reliable) associative learning using two within-subject tasks with different stimulus materials. Behavioral performance improved under both feedback regimes, but learning was consistently more successful under deterministic feedback. Error-related negativity, feedback-related negativity, and phase-locked midfrontal theta became progressively more pronounced across deterministic learning, particularly in the image-image task, whereas corresponding changes were attenuated under probabilistic feedback. Induced theta showed the opposite pattern, decreasing across deterministic learning while remaining comparatively stable under probabilistic learning. These opposing theta trajectories suggest that phase-locked and induced activity reflect distinct components of learning-related neural adaptation that may be obscured when total theta power is considered alone. Overall, the findings indicate that feedback reliability shapes the progressive tuning of midfrontal performance-monitoring responses during learning.