Co-regulated timing in music ensembles: A Bayesian listener perspective

Abstract

Co-regulated timing in a music ensemble rests on the human capacity to coordinate actions in time. Here we explore the hypothesis that humans predict timing constancy in coordinated actions, in view of timing their own actions in line with the others. An algorithm (BListener) is presented that predicts timing constancy, using Bayesian inference about incoming timing data from the music ensemble. The algorithm is then applied to a timing analysis of real data, first, to a choir consisting of four singers, then, to a dataset containing performances of duet singers. Global features of timing constancy, such as fluctuation and stability, correlate with human subjective estimates of the music ensembles’ quality and associated experienced agency. In future work, BListener could serve as component in an artificial musician that plays along with human musicians in a music ensemble.