Characterizing Statistical Learning of Music
statistical learning , music perception , EEG
Understanding how we gain representations of music is crucial for a comprehensive understanding of music perception. Several studies suggest that listeners extract statistical regularities in music via passive exposure, i.e., statistical learning. Statistical learning could thus underlie our ability to learn about unfamiliar music. While research shows that musical short-term knowledge can be gained via this process, it remains unclear whether longer lasting knowledge can be gained this way. This thesis aims to address this issue. In Study 1, participants gave probe-tone ratings before and after exposure to an artificial music genre defined by its tone distribution. Participants’ responses to these tones revealed that they are sensitive to both the tone distribution of the preceding tone sequence and tone distributions encountered earlier in the experiment. This suggests that musical knowledge can be gained via statistical learning over and above short-term knowledge. Participants also distinguished the exposed music genre from another artificial genre after exposure. In Study 2a, I ascertain that this result is not due to prior familiarity with one of the artificial music genres. In Studies 2b and 2c, I show that participants can differentiate the two music genres based on pitch cues. In Studies 3a and 3b, I compared participants’ brain activity following congruent and incongruent tones to familiar or unfamiliar tone distributions using EEG. Activity at frontal electrodes from 380 to 450 ms post stimulus onset was larger for incongruent probe tones, but only when the preceding probe-tone context was based on the familiar tone distribution. This event-related potential can be considered an index of musical long-term knowledge because behavioral results suggest that participants were influenced not just by short-term knowledge when presented with familiar tone distribution probe-tone contexts. In Study 4, participants completed Study 1’s paradigm. Key behavioral results from Study 1 were replicated. Further, EEG data suggest that musical long-term knowledge may form simply through encountering an unfamiliar music genre in short tone sequences, without extended exposure. These findings are discussed regarding their implications for the statistical learning of music, the influence of music training, and the field of statistical learning research in general.