Trade-offs in Coordination Strategies for Networked Jazz Performances

Abstract

Coordination between participants is a necessary foundation for successful human interaction. Group musical improvisation, for example, requires the synchronization of timing between performers to be effective. Networked mediation can disrupt the coordination process by introducing a delay between the production and reception of sounds. This can result in significant deteriorations in temporal asynchrony and stability between performers. Here we show that pairs of musicians adopt diverse strategies when coordinating performances over a network. What appear to be two alternatives involve (i) one musician being led by the other, tracking the timings of the leader's performance, or (ii) both musicians accommodating to each other, mutually adapting their timing. During networked performance, these two strategies favor different sides of the trade-off between, respectively, asynchrony and tempo stability; in the absence of delay, both achieve similar outcomes. Our research highlights how remoteness presents new complexities and challenges to successful interaction.

Rhythmic Qualities of Jazz Improvisation Predict Performer Identity and Style in Source-Separated Audio Recordings

“…One of the best established is the linear phase correction model implemented by Wing et al (2014), where the duration of a performer's upcoming inter-beat interval is predicted from both the duration of their prior inter-beat interval and the asynchrony with their partner(s) at the previous beat. In an earlier study, we demonstrated good results using phase correction to model the interaction between a jazz pianist and drummer (Cheston et al, 2023). As in this study, when we applied the phase correction model to our dataset, we expressed every quarter-note inter-beat interval by a performer in terms of its difference from the preceding interval, to control for any global drift in performance tempo (Figure 2d).…”

“…We added three further features to these 15, under the category "tempo": ( 16) the average tempo of the recording, as estimated by the beat tracking algorithm; (17) the tempo slope, specifically the signed overall tempo change per second in a recording, equivalent to the slope of a linear regression of piano beats against inter-beat intervals such that positive values imply net acceleration and negative deceleration (Cheston et al, 2023); and (18) the tempo stability, calculated by taking the standard deviation of pianist inter-beat intervals along a sliding four-measure window, then calculating the median of all windows (Cheston et al, 2023). The mean values for average tempo, tempo slope, and tempo stability were 197.38 beats-per-minute, +0.03 beats-per-minute-per-second, and 198.46 milliseconds, respectively.…”

“…We then cleaned the data by setting outlying values for a feature (e.g., coefficients obtained from phase correction models with fewer than 30 observations) to the mean value obtained for that feature, across all pianists and recordings. For the "tempo" category, outliers were considered to be those lying outside the range of values obtained from a previous experimental study of jazz piano performance (Cheston et al, 2023). We fitted a random forest classification model to predict the identity of the pianist in each track of our corpus, using these features as predictors.…”