Modeling pattern importance in Chopin's mazurkas

This study relates various quantifiable characteristics of a musical pattern to subjective assessments of a pattern's salience. Via score analysis and listening, twelve music undergraduates examined excerpts taken from Chopin's mazurkas. They were instructed to rate already-discovered patterns, giving high ratings to patterns that they thought were noticeable and/or important. Each undergraduate rated thirty specified patterns and ninety patterns were examined in total. Twenty-nine quantifiable attributes (some novel but most proposed previously) were determined for each pattern, such as the number of notes a pattern contained. A model useful for relating participants' ratings to the attributes was determined using variable selection and cross-validation. Individual participants were much poorer than the model at predicting the consensus ratings of other participants. While the favoured model contains only three variables, many variables were identified as having some predictive value if considered in isolation. Implications for music psychology, analysis, and information retrieval are discussed