Aarts, R. and Jansen, C. (1999) Categorical perception of short rhythms. In Proceedings of the 1999 SMPC, 57. Evanston.

Abstract

Quantization is the process of separating the discrete and continuous time scale from a musical performance. The discrete time scale refers to the rhythmical structure, while the continuous time scale refers to the deviations from this structure (Desain & Honing, 1991). Listeners can seperate the two time scales: they can notate or reproduce the rhythmical structure of an aurally presented performance. A class of performed rhythms notated as the same 'prototype rhythm' may be regarded as a 'rhythm category'. There are several quantization models (Desain & Honing, 1991; Longuet-Higgins, 1987), but they have not been tested systematically. Experiments have been conducted proving the existence of category boundaries in rhythm perception (Clarke, 1989; Schulze, 1989). However, only few stimuli rhythms were used with a fixed set of responses. In our study, a set of 91 rhythms was presented to a subject group of 10 experienced musicians. Each subject was allowed to use any notation commonly encountered in music practice. The goals of the study were: 1) to identify rhythm categories for a wide range of stimulus rhythms and investigate category properties, 2) to use the data to test existing quantization models. Analysis of the data showed that: the number of rhythm categories varied widely over subjects; rhythm category shapes range from fairly regular to irregular; rhythm categories are not centered around the prototype rhythm; certain temporal deviations from the prototype rhythm are more common than others; a simple nearest-neighbor rule (used in commercial quantizers) cannot account for the behavior of the subjects. For this particular data set, the Longuet-Higgins (1987) model correctly predicts a large proportion of responses. The Desain & Honing (1991) model is less successful.

More on time quantization