Neural tracking of melodic prediction is pre-attentive

Music’s ability to modulate arousal and manipulate emotions relies upon formation and violation of predictions. Music is often used to modulate arousal and mood while individuals focus on other tasks, suggesting that sophisticated musical prediction may be implicit and obligatory, operating regardless of attention. Here we investigated the role of attention in musical prediction by presenting participants simultaneously with a musical passage in one ear and an audiobook in the other. Participants were asked to attend to either speech or music to perform a recognition memory task. Musical prediction was investigated using a computational model of music which was trained on a corpus of Western music (long-term learning) and/or incrementally on each stimulus (short-term learning). While neural tracking of the acoustics of music, including encoding of note onset and pitch interval size, was enhanced by attention, musical prediction tracked the neural signal robustly regardless of attention. However, although neural tracking of long-term musical prediction based on previously-learned statistical patterns was unaffected by attention, tracking of short-term musical prediction learned from the structure of each stimulus was present only when the music was attended. Thus, attention is crucial for short-term musical learning and prediction but is not required for musical prediction based on long-term stylistic exposure.