Abstract
The purpose of this study was to (a) objectively measure and analyze performance deviations from mechanical regularity for three jazz pianists via MIDI-based "groove quantize" procedures and (b) measure how experts rate musical examples incorporating these deviations as being representative of the swing style.The "groove quantize" software procedure was used to measure performance deviations from mechanical regularity for (a) note placements (timings), (b) note durations (articulations), and (c) note velocities (dynamics) contained in 281 sample measures taken from 33 performances by three jazz pianists. The data were analyzed for performance differences among the performers and for relationships between the performance variables and tempo. Four "derived" performance models were developed as "grooves," representative of each performer's performance style and the jazz swing style at large. For comparison, three "mechanical" performance models were constructed based on mathematical ratios representative of traditional notation.Forty-two expert judges were asked to rate the "swing representativeness" of an unaltered melody from each pianist and seven variations of each, four based on the derived performance models and three based on the mechanical performance models. The data were analyzed for rating differences between the derived performance model variations and the mechanical performance model variations, and for rating differences between each performer's unaltered melody and the variations based on his or her own derived performance model.Significant differences $(\alpha=.05)$ were observed among performers' downbeat note placements, upbeat note placements, downbeat note velocities, upbeat note velocities, and downbeat note durations, but not for upbeat note durations. Significant positive correlations were observed between tempo and downbeat duration and between tempo and upbeat duration.Analysis of the judges' ratings revealed that the four derived performance model variations were rated significantly more representative of the jazz swing style $(\alpha=.05)$ than the three mechanical performance model variations. There were no significant differences in swing representativeness ratings between a performer's unaltered melody and the variations based on his or her own derived performance model for two of the three performers, suggesting that those two derived performance models were, in fact, representative grooves.