Principal Components Analysis of Musicality in Pitch Sequences

Musicality can be thought of as a property of sound that emerges when specific organizational parameters are present. We hypothesize that this property is not binary (where an auditory object is or is not a musical object), but rather exists on a continuum whereby some auditory objects may be considered more or less musi- cal than other auditory objects. We suggest that identification of an auditory object as being more musical than another begins with a modularized analysis of features that coheres into a holistic interpre- tation. To explore this, we designed two experiments. In the first, 30 subjects evaluated 50 ten-tone sequences according to how musical they thought they were. A special stimulus set was designed that controlled for timbre, pitch content, pitch range, rhythm, note and sequence length, and loudness. Mean z-scored stimulus ratings showed significantly distinct groupings of musical versus non- musical sequences. In the second, a Principal Component Analysis (PCA) of the ratings yielded three components that explain a statisti- cally significant proportion of variance in the ratings. The stimuli were analyzed in terms of parameters such as key correlation, range, and contour. These values were correlated with the eigenvalues of the significant PCA components in order to determine the dominant strategies listeners use to make decisions about musicality.