Temporal regularity does not drive neural and behavioural tracking of musical phrases.

When listening to music or speech, people naturally divide continuous sound streams into segments to process incoming information. The segmentation boundaries are not random. Listeners agree on the points of segmentation, which are consistent with arbitrary rules, e.g. established by music theory, and often occur at regular time intervals. It is thus unclear whether phrase tracking rely on structural knowledge or merely on temporal predictability. To address this, we examined how non-musicians process both regular (temporally predictable) and irregular musical phrases derived from J.S. Bach’s compositions. This approach preserved authentic musical structure while manipulating temporal predictability. We also included control stimuli matched in acoustic properties but lacking musical structure. Behavioural and EEG measures revealed that listeners could accurately detect phrase boundaries in both regular and irregular conditions. Neural activity, indexed by an increase in low-frequency EEG power, tracked structural boundaries regardless of temporal regularity. These findings demonstrate that musical segmentation depends fundamentally on implicit understanding of musical structure, rather than on temporal predictability alone.