Dr. Eliot Handelman: Recursively structured pattern, shape and motion asymmetries generate complex musical features

Recursion is an essential feature of the human language and music faculties. Generative models of (or with) musical recursion have been proposed and are widely deployed in music cognition research. But available models are either not fully computational, or do not predict complex musical structures, or are too underspecified to produce music.  Consequently they are difficult to evaluate.

The aim of this research is an account of musical recursion serving both analysis of existing music and automatic composition of original music. Based on the intuitive concept of ``shape'' together with the idea of patterns of development, it describes how things in music become bigger and smaller in a structured network of hierarchical orientations. A double model was developed that theorizes and generates abstract patterns as an algebra of occurrence, and "z-shape" -- the patterns of all orientable musical objects that are themselves either simple shapes (like scales, crescendi and accelerandi) or, recursively, patterns of z-shapes. The goal is to synthesize models of high-level musical objects.

The model provides a framework for the Jack & Jill composition/listening architecture, which generates music at an engagingly high level, spontaneously structured in elaborate and original ways, sometimes evoking existing styles. Analysis is a transformation that attempts to expose underlying _simplicity_, hypothesizing music as variations of "canonical" (i.e., maximally simple) music.  Simplicity is determined by a variety of straightforward metrics applied to variations that promote regularity, symmetry, and descriptive parsimony over the z-shape/pattern complex. For viewing, annotated scores are conveniently generated.

The ability of the model to generate music with complex formal features without explicit guidance strongly suggests that there is something right about the underlying theory.

The model offers a large number of testable hypotheses concerning the music-perceptive and cognitive relevance of simplicities. The model is orthogonal to harmony-oriented analysis---indeed, most other types of analysis--- and could easily operate alongside other theoretical approaches. The structures it presents, in which music is transformed with respect to hypothesized pattern-breaks, can serve as the basis for research in musical prediction and expectancy. It offers technical machinery easily adapted for machine-learning in problems of style analysis and formal classification---and would seem to have strong potential as an instrument of thought in further musical research.