'Good', 'Fair', and 'Bad' chord progressions: A regression-analysis of some psychological chord progression data obtained in an experiment by J. bharucha and C.krumhansl

Abstract

This paper attempts to justify the psychological ratings determined in an experiment reported by Bharucha and Krumhansl (1983) and Krumhansl (1990a, pp. 192-195) for the 42 possible triad-progressions of the major key. The paper shows how a set of six musical factors can be used in a regression formula to generate an extremely close fit to this experimental data. Of these six factors, the most important is the psychological weight of the individual chords this dimension alone accounts for.541 of the explained variance in the data. The remaining five factors pertain to chord-order, root progression by an upward or downward fifth, the possibility of a progression being misread in the relative minor, the treatment of the leading tone, and a condition concerning the juxtaposition of a subdominant chord (IV or II) with a dominant chord (V or VII). The paper ends with a derivation of the chord weights generated by the regression; this derivation is based on a series of hypothesized tone weights, and assumes that the psychological weight of a chord is dependent on the summed psychological weights of that chord's component tones. It is shown that the proposed tone weights exhibit a high correlation with two sets of empirically-determined tone weights, one reported by Krumhansl (1990a) and the other by Cuddy and Thompson (1992).