Modelling perceived syncopation in popular music drum patterns

Abstract

Recent studies suggest that rhythmic syncopation is a relevant predictor for groove. In order to validate these claims, a reliable measure of rhythmic syncopation is required. This article investigates whether a particular notation-based model for estimating syncopation in Western popular music drum patterns adequately predicts perceived syncopation. A listening experiment was carried out with 25 professional musicians. Six popular music drum patterns were presented to the participants in all 15 pairwise combinations, and the participants chose the pattern from each pair that was more syncopated (win), compared to the other pattern (lose). Perceived syncopation was defined as the proportion of wins for each stimulus. The experiment showed that the model works well in general, but that it overemphasises the weight of syncopes on weak metric positions. This exaggerates the syncopation value of one particular drum pattern and generally leads to inflated syncopation values in the upper syncopation range. In consequence, the fit between the model and perceived syncopation was poor, even when flexible logarithmic functions [Formula: see text], p < .001) or exponential approach functions ([Formula: see text], p < .001) were used to link the model predictions to perceived syncopation. The model was revised and a numeric optimisation process was carried out to improve its fit. The revised model produces syncopation estimates that have a linear relationship with the perceived syncopation measures and a good fit with the data ([Formula: see text], p = .469). However, this revised model is based on only six drum patterns that cover a very limited range of rhythmic phenomena. In order to create a general model of syncopation in popular music drum patterns, further modelling work is necessary that involves a larger number and a wider variety of patterns.