On detecting the differences in jazz: a reassessment of comparative methods of measuring perceptual veridicality

Abstract

This article reports an extension of the jazz studies of Holbrook and Huber, which compared individuals' perceptions of music with its objective characteristics. One of their major findings was that a compositional approach to the data, i.e., one that used multiple discriminant or factor analysis provided a better statistical fit to objective features such as style, and key, than did a decompositional method which used a map derived by multidimensional scaling. They concluded that the compositional approach enables a more accurate assessment of perceptual veridicality. However, there is a confounding in their studies of the type of statistical analysis employed with the response mode allowed the subjects, which draws into question the validity of their conclusions. The study reported here assesses the relative merits of the type of statistic and mode of response separately. Its findings support the use of repertory grid as a viable tool in musical perception studies in which under appropriate conditions multiple discriminant analysis and multidimensional scaling are equally viable statistical procedures.