Temporal Trajectories of Mismatch Negativity Reveal Dynamics of Auditory Perceptual Learning

Theories of perceptual learning, such as predictive processing, describe learning as a dynamic and continuous process of updating hierarchical internal models in response to environmental regularities. However, most studies reduce this process to static condition-based averages, making them insensitive to how prediction error signals unfold over time. The current study addressed this gap by investigating the temporal trajectories of precision-weighted prediction errors-indexed by mismatch negativity (MMN)-across stable and volatile auditory oddball sequences. EEG was recorded for 45 young adults as they passively listened to Traditional (stable) and Alternating (volatile) sequences, presented in counterbalanced order. Consistent with theoretical predictions, MMN amplitude declined over time in the stable environments and increased across volatile contexts following changes in environmental regularity, reflecting higher-order model refinement and ongoing lower-order model building, respectively. Crucially, MMN trajectories were modulated by prior experience. MMN was attenuated in the stable sequence when preceded by volatility, suggesting that elevated beliefs about environmental uncertainty may constrain the precision of subsequent perceptual inferences. In contrast, prior exposure to stability had no impact on learning in volatile environments, indicating an asymmetry in how prior context shapes perceptual learning. These findings demonstrate the theoretical and methodological value of assessing MMN trajectories, revealing dynamic learning processes often masked by conventional averaging. This approach offers a promising framework for investigating altered learning mechanisms in clinical populations. Together, our results underscore the importance of temporal resolution, prior experience and environmental structure in understanding the dynamic nature of perceptual inference.