Temporal Progression of Prior Precision, Surprisal, Model Updating and Tone Adaptation in Human Auditory Cortices

Abstract

Auditory perception emerges from the continuous integration of incoming acoustic signals with statistical predictions, a process complemented by simpler local adaptation. Yet, how all these concurrent mechanisms unfold in time remains underexplored. Using magnetoencephalography (MEG) in participants listening to stochastic tone sequences with periodically shifting inferential statistics, we modelled the temporal, spatial, and spectral dynamics of stimulus-specific repetition suppression, surprisal, model updating, and prior precision. By concurrently tracking these computational parameters within this dynamic auditory environment, we revealed a distinct chronological hierarchy governing predictive and adaptive processing. Prior precision states - reflecting expectations - emerged earliest, followed by rapid and concurrent repetition suppression and surprisal - reflecting local acoustic attenuation and global mismatch detection. These initial dynamics culminated in subsequent, distributed model updates within fronto-parietal networks. Spectrally, these components mapped onto diverging oscillatory fingerprints, ranging from theta-alpha modulation during adaptation to beta suppression during surprisal and broad spectral dampening during precision weighting. Together, these findings demonstrate that the brain navigates uncertain auditory environments through a cascade of neural processes, where fast, local sensory gating and mismatch signals precede the hierarchical revision of internal models.