Does expressive timing in music performance scale proportionally with tempo?

✍ Scribed by Peter Desain; Henkjan Honing

• Publisher: Guilford Publishing Inc
• Year: 1994
• Tongue: English
• Weight: 838 KB
• Volume: 56
• Category: Article
• ISSN: 0340-0727
• DOI: 10.1007/bf00419658

✦ Synopsis

Evidence is presented that expressive timing in music is not relationally invariant with global tempo. Our results stem from an analysis of repeated performances of Beethoven's variations on a Paisiello theme. Recordings were made of two pianists playing the pieces at three tempi. In contrast with the relational-invariance hypothesis (see Repp, 1994), between-tempo correlations were in general lower than within-tempo correlations. Analyses of variance of log-transformed inter-onset intervals (IOIs) showed significant interactions between tempo and IOI, i.e., evidence against a proportional relation between timing and tempo. Complex, but consistent, nonproportional patterns were shown in the analyses of the timing of the grace notes in the piece. The analysis suggests that timing aspects of music are closely linked to the musical structure and can be studied and manipulated only respecting this relation -not as a global timing pattern or tempo curve. Finally, it is shown that methodological issues of data collection and analysis had a significant influence on the results. variation, or even simply changing the key from major to minor without modifying the tempo curve -produced results judged unmusical, and we argued against their use on this basis. Instead, we suggested that the timing aspects of music are linked to its structural aspects and should therefore be studied and manipulated respecting this relation and as a global transformation based on a tempo curve. Repp (1994), prompted by our work, explored the scaling of expressive aspects of music systematically with regard to global tempo, for performances of Schumann's "Tr~iumerei." In his work, which will be referred to as the "Trgumerei" study, there was relational invariance: expressive timing scaled "approximately" proportionally with tempo. Consequently, a tempo curve for a different tempo can be obtained if the original is plotted on a log-duration y axis and shifted vertically.

To gain further understanding on this issue, we present a new analysis. We follow the lines as set out in the "Trfiumerei" study to allow easy comparison. However, we shall restrict ourselves to the examination of the patterns of onset timing.