Project overview

One of the aims of this two-and-a-half-year project was to adapt conventional analytical techniques and, where appropriate, to develop new approaches relevant to the study of performance. It focused specifically on motivic analysis, which, despite its considerable potential, has always been theoretically and musically problematic because of the synchronic (i.e. non-time-dependent) nature of most analysis in this vein. Motifs are generally defined as short melodic, rhythmic, and/or harmonic patterns repeated either exactly or in varied form; they have long been regarded as important elements of musical structure. This project was based on a related but novel concept referred to as ‘performance motifs’, i.e. discrete expressive patterns used consciously or unconsciously by performers to effect a sense of coherence in their performances. The main thrust of our work has been to strip the concept of motif of the inappropriate assumptions of score-based analysis, and to extend it to those dimensions which are of particular concern to both performers and listeners (whose experiences are as much moulded by performances as by the works that are performed).

To that end, two new analytical techniques for studying performance motifs were developed and then applied in a series of empirical procedures. Although the primary focus was on solo piano repertoire from the nineteenth century, the results are potentially generalisable to music of different periods and for different instruments, whether recorded or live. The outcomes of the project were widely disseminated in the form of international conference papers in Porto, Aveiro, Sydney, Sapporo, Princeton, Dublin etc., as well as through four peer-reviewed articles and a peer-reviewed book chapter. Presentations were also given at a number of British institutions. See here for a detailed profile of the project.

This project was based at Royal Holloway, University of London under the direction of John Rink. Neta Spiro was employed as part-time Research Fellow throughout the project, with collaborative input from Nicolas Gold (King’s College London).


The project’s principal objectives were as follows:

  • 1. demonstration of how analytical techniques developed in relation to scores can be adapted for the purposes of analysing recorded performances
  • 2. a new definition of motif which reflects the temporality of musical performance and perception
  • 3. identification of the expressive strategies by which performance motifs are manipulated
  • 4. determination of the ways in which listeners apprehend performance motifs, and the effect that such motifs have on them.


The project’s main objectives were achieved by carrying out a series of complementary analyses upon a select repertoire and coordinating the results. The basic methodology was as follows:

  • 1. ‘traditional' analysis of select works, based on the score (with reference to published analyses where possible)
  • 2. adaptation of existing performance-analytical techniques and development of altogether new ones relevant to our concept of ‘performance motif’
  • 3. analysis of ‘performance motifs’ within the case-study works, based on contrasting recordings of each and using the techniques described above
  • 4. comparison of the results obtained under 1. and 3. above
  • 5. empirical investigations using groups of listeners who were asked to identify prominent musical features in the same recordings and who thus allowed us to test and evaluate the results of our analyses of ‘performance motifs’.


The first of the two analytical techniques developed in this project combines a simple pattern-matching approach with Formal Concept Analysis (FCA) to show clusters of related repetitions. FCA is a technique for visualising the relationships between objects in terms of their common and distinct attributes; both the objects and the attributes are user-defined. The method used for motivic exploration had three main stages: quaver-timing extraction, pattern matching, and lattice generation. The tool provides facilities for executing multiple analyses with varying parameters of window size and matching thresholds to enable concentration on different sizes of motive and degrees of similarity of repetition. It produces summary results to aid in the refinement of the investigative process and, moreover, to enable global comparison according to window size (i.e. size of motive) and among performances. Further information about this technique can be found in the project publications page.

The second method, using Self-Organising Maps (SOM), enables the systematic analysis and comparison of individual or multiple performances by identifying recurrent expressive patterns and their location within the respective performance(s). It does so by arranging patterns according to similarity. One of the key advantages of the SOM technique is its ability to retain the topological structure of high-dimensional data while reducing the dimensionality; thus, if an item of interest in the domain can be modeled as a point in a multi-dimensional vector space, the SOM will cluster together similar points in similar neighbourhoods of the lower dimensional space. Using this approach allows the discovery of similarities in timing or dynamic shapes without a priori determination of how many similarities should exist or how large each group should be. A very large number of performances can therefore be studied with relative ease and without the data overload typical of many approaches. The results obtained in a series of analytical studies using both the FCA and the SOM methods show that, in general, the structure of the music as performed either emerges from or at least is shaped by the expressive patterns that were identified. For additional information see the the project publications page.

The research culminated in an article in CHARM's second special issue of Musicae Scientiae, a chapter in a collection of essays on Music and Gesture, and a series of conference presentations, some of which appeared as peer-reviewed publications. An article about the SOM method appeared in CHARM’s fourth Newsletter (pdf file), while the fifth Newsletter (pdf file) contains a presentation of the study with listeners described above under Methodology.

The principal outputs are detailed, with links to abstracts, on the project publications page.