Overview of the KTH rule system for musical performance

A musical performance within Western classical music typically consists of the combined artistic contribution of the composer and the performer. The composer’s part, the score, has long been studied and scrutinized in terms of its structural aspects such as harmony, melody, form, instrumentation, as well as of studying the composer’s intention or the inherent emotional expression (Hevner, 1937). The performer’s expressive contribution has certainly been acknowledged but has until recently been more difficult to study. The contribution from the performer involves the shaping of all audible performance parameters such as tempo, sound level, articulation and vibrato with respect to individual notes, phrases and the whole piece. These parameters can not be studied as readily as the notation in the score. However, they did become easier to study when new acoustic measurement methods were developed (Seashore, 1938). Since then, considerable knowledge has been gained about how performance parameters contribute to structural, motional, and emotional communication (Gabrielsson, 1999; Friberg & Battel, 2002; Juslin & Sloboda, 2001). With further developments in computer-based tools, it also became possible to formulate and run algorithmic models for musical performance. Thus, for the first time theories of music performance could be verified by listening to model-generated performances. Coinciding with the introduction of personal computers around 1980, a number of such models were developed. One computational model concerned the metrical patterns in the Viennese waltz by Bengtsson and Gabrielsson (1983). They found that the three beats in the measure were performed in a short-long-intermediate pattern. They extended these measurements into a model for the performance of metrical patterns of the Viennese waltz taking into account several metrical levels and performance parameters. Clynes (1983) proposed a rather speculative model suggesting that each composer has a specific “pulse” describing rhythmic and dynamic changes related to the metrical structure (see also Widmer & Goebl, 2004). The large tempo changes reflecting the phrase structure often found in Romantic music were modeled by Todd (1985, 1989). Based on measurements of piano performances, he proposed a general model of tempo as a function of phrase structure. Clarke (1988) suggested a set of nine ABSTRACT