Sources of timing variations in music performance: A psychological segmentation model

Abstract

Systematic timing variations observed during music performance have usually been attributed to a musical expression hypothesis, related to relatively highlevel processes, by which musicians emphasize certain events in order to transmit a particular musical interpretation to the listener. We propose, in addition, a perceptual hypothesis, related to lower-level processes, in which some observed variations would be related to functional constraints of the auditory system. (Some intervals would be heard shorter and thus played longer as a phenomenon of perceptual compensation.) We present a psychological model of temporal organization proposing two types of process (regularity extraction and segmentation into groups) operating parallel that allow listeners to parse complex auditory sequences such as music. Each type of process operates at both a low processing level (beat extraction and segmentation into basic groups) and a higher processing level (hierarchical metric organization and hierarchical segmentation organization).The analysis of musical and mechanical performances of Schumann's TraÈumerei demonstrated performance variations in relation to both hierarchical segmentation and hierarchical metric organizations, and to rhythmic groups. Variations were not systematically observed in relation to melodic groups. Regression analyses quanti®ed these eects and demonstrated that hierarchical segmentation and rhythmic groups accounted for approximately 60% of the variance for musical and mechanical performances, leaving room for the description of other, as yet unidenti®ed, processes. The percentage of variance explained by high-level processes (hierarchical segmentation) decreased from musical to mechanical performances, whereas the percentage of variance explained by lower-level processes (rhythmic groups) increased.We conclude that it is important to go beyond the traditional approach of describing performance variations in relation to musical structure and to adopt the approach of studying performance variations in relation to the psychological processes that allow the musician to perceive the musical structure. Finally, we adapt the psychological model of temporal organization to expressive timing: similar psychological processes operate at multiple hierarchical levels ± namely, those of segmentation and grouping ±, and these similar processes result in the same pattern of performance variations (an accelerando/ritardando pro®le).

Accenting and detection of timing variations in tone sequences: Different kinds of accents have different effects

“…Drake and Palmer (1993) found that melodically accented notes were separated from the preceding notes by longer IOIs in the performance of simple sequences and real music. Penel and Drake (1998) replicated the finding in cases in which pitch jumps did not coincide with conflicting structural features. Repp (1992b) did not find a consistent relationship between pitch change and timing in musical performance.…”

“…Structural cognition is assumed to be under voluntary control, whereas auditory processing and structural perception are not (Repp, 1998). The fact that performers use some expressive variation even when they are instructed to perform mechanically (Drake & Palmer, 1993;Gabrielsson, 1974;Penel & Drake, 1998) indicates that structural cognition may be only partially responsible for the variations in perception and performance of music.…”

“…This points to an interesting possibility about expressive variations in performance. It is known that some variation in the timing of notes is observed even when performers are instructed to play without any expression (Drake & Palmer, 1993;Gabrielsson, 1974;Penel & Drake, 1998). This may be because performers deviate from a uniform tempo, in part, because they are unable to detect these deviations.…”

“…A sequence of sounds with perfectly equal beat periods can still have variations in other properties of sound, such as pitch, intensity, duration, and harmonic function. Such variations would create specific expectations in terms of timing, because expressive temporal variations in actual performance are correlated with other aspects of musical structure (Penel & Drake, 1998;Repp, 1990Repp, , 1992bRepp, , 1995Repp, , 1996Repp, , 1998Shaffer, Clarke, & Todd, 1985;Todd, 1985;Windsor & Clarke, 1997). One very simple and general expectation is that consecutive sounds that belong to different perceptual groups should be separated by longer time intervals, as compared with consecutive sounds that belong to the same perceptual group (Handel, 1989, pp.…”

Structural Communication

“…The first computational model was presented by Todd (1985Todd ( , 1989; see also Windsor & Clarke, 1997;Penel & Drake, 1998) and accounted for the variation of measure duration over phrases. Todd later developed a revised model based on a different mathematical function, which he argued was more closely related to a metaphor of physical motion (Todd, 1992(Todd, , 1995.…”