Abstract
<jats:p>Music, which is organized hierarchically (notes, phrases, sections), provides an ideal model for studying fine-grained motor sequence production. However, it is unknown how musicians' brains integrate tonal structure over multiple timescales during real-life performance. Here, we scrambled an unfamiliar Tchaikovsky piano suite at four timescales (every 1/2/8 measures, or fully intact) and asked expert pianists to play (sightread) all four versions in the fMRI scanner. Responses in the motor network, default mode network, and hippocampus were strongly impacted by scrambling, indicating that they integrate tonal structure over relatively long timescales. Additionally, the emergence of functionally connected sub-networks between auditory, visual, motor, and default mode network regions across scramble levels supported this hierarchical integration process. Our results cannot be explained by lower-level cues (tempo, timbre, dynamics; local pitch height or rhythmic density) and instead reflect processing of high-level tonal structure. Our study highlights novel mechanisms of complex auditory-motor action planning during live music performance.</jats:p>