Whole-brain computation of cognitive versus acoustic errors in music

Previous studies have evidenced how the local prediction of physical stimulus features may affect the neural processing of incoming stimuli. Less known are the effects of cognitive priors on predictive processes, and how the brain computes local versus cognitive predictions and their errors. Here, we determined the differential brain mechanisms underlying prediction errors related to high-level, cognitive priors for melody (rhythm, contour) versus low-level, local acoustic priors (tuning, timbre). We measured with magnetoencephalography the mismatch negativity (MMN) prediction error signal in 104 adults having varying levels of musical expertise. We discovered that the brain regions involved in predictive processes for local priors were primary and secondary auditory cortex and insula, whereas cognitive brain regions such as cingulate and orbitofrontal cortices were recruited for melodic errors in cognitive priors. The involvement of higher-level brain regions for computing cognitive errors was enhanced in musicians, especially in cingulate cortex, inferior frontal gyri, and supplementary motor area. Overall, the findings expand knowledge on whole-brain mechanisms of predictive processing and the related MMN generators, previously mainly confined to the auditory cortex, to a frontal network that strictly depends on the type of priors that are to be computed by the brain.