ABSTRACT
Research on musical performance over the past decade or so has focused on how it is that performers are able to produce a stable and coherent performance from the notation of a piece of a music, and in particular they expressively transform a piece in performance. In addressing this issue, the authors of empirical research have converged virtually unanimously on a single model of the origin and control of expression in music. This is a generative model, in which the expressive aspects of a performance (the deviations in timing, dynamic, articulation, etc. from the instructions of the score) are generated from a structural description of the music in the performer’s mind. There are three primary features of the empirical data collected by different researchers that have influenced the general adoption of this model. The first is the remarkable stability that has been observed in musical performances: performers clearly have an impressive ability to replicate the expressive profile of a piece in performance, with a degree of variability in the timing properties of a performance of one percent or less, sometimes across performances that are years apart (Shaffer, Clarke & Todd, 1985; Clynes and Walker, 1982). Quite apart from begging the question as to the origin of the expressive profile in the first place, a theory that proposed that a performer constructed some kind of literal record of all the timing data required to specify the entire performance (often many thousands of events) would put completely unrealistic demands on the memory system, and goes against all the evidence of research on the organisation of memory. The second critical piece of evidence is the exact converse of replicability: performers have the ability to change spontaneously (or occasionally on demand) their interpretation of a piece and to produce a different expressive treatment at very little or no notice, and with no obvious period of experimentation and memorisation. Evidence for this can be found in Clynes & Walker, 1982 and in recent work by Shaffer (Shaffer, 1990) who discusses data in which performers give two or three different interpretations of the same piece of music in immediate succession, some of which are at the request of the experimenter and are entirely unprepared. The third piece of critical evidence is the existence of expression in sight-read performances: once again Shaffer has shown that a pattern of expression exists in sight-read performances (Shaffer, 1981) which is replicated in subsequent performances. It is clearly, therefore, not an arbitrary expressive profile, nor is it something that the performer can have learned previously. The most convincing way to account for these three kinds of evidence is to regard the expressive profile as being generated from the performer’s conception of the structure of the music at the time of its performance. The extraordinarily high level of long-term stability that has been found is thus the consequence of a performer retaining a very stable conception of how the music is organised. Likewise, the ability to produce different expressive versions of the same piece at a moment’s notice is the result of being able to imagine a number of different structural interpretations of the piece, each with its own expressive consequences. Finally, sight-read performances have a non-arbitrary expressive profile because even when first encountering a piece of music, a performer is obliged to form some kind of structural interpretation of it as s/he goes along, and provided that the music is in a familiar style this interpretation is likely to contain many of the same features as would
be present after a period of study and practice, albeit in a somewhat more rough and ready form. (Because there is still very little empirical work that has investigated what happens to an interpretation in the course of practice, it is not possible to be more precise about the extent or ways in which a first performance and a highly practised performance by the same performer differ).
