Background. It has been shown that listeners are sensitive to pitch distributional information in music, which is used in categorisation of musical styles by statistical means. However, it remains uncertain whether these methods simulate human classification process and a behavioural categorisation task using identical material is therefore needed.
Aims. This study examines how listeners classify folk melodies in an experiment aimed at determining the effectiveness of classification of music by statistical means.
Method. 17 undergraduate music students rated the similarity of pairs of melodies on a 9-point scale. Three melodies of five distinct musical styles were used, producing fifteen melodies and 105 paired comparisons. The musical styles represented North Sami yoiks, Finnish Spiritual folk hymns, Irish hornpipes, German folksongs and Greek folksongs. The melodies were chosen by native experts as typical examples of the style. The similarity ratings were analysed using multidimensional scaling, which is able to produce spatial distances between the melodies comparable to those obtained from frequency-based classification methods.
Results. Multidimensional scaling of similarity ratings showed that melodies seemed to form groups when the distances were mapped onto a two dimensional plot. Moreover, no clear differences emerged between German folksongs and Finnish Spiritual folk hymns, which was reasoned to occur because of the common elements and history they share. Cluster analysis of statistical frequencies of the individual melodies showed also a similar tendency to group melodies of certain styles together.
Conclusions. Categorising melodies using similarity ratings can be used to substantiate categorisation by means of a statistical style analysis that acknowledges the importance of the frequency data for perception. The results of the study suggest that both methods are able to capture some aspects of the structures which portray common salient dimensions to which listeners pay attention whilst categorising melodies.
Key words. classification, similarity, categorization, style, melody, folksongs