Proceedings of the Annual Meeting of the Cognitive Science Society

We present a symbolic, rule-based framework for modeling tonal music composition as the construction and transformation of structured cognitive programs. Building on the Language of Thought hypothesis, our approach treats musical scores as externally observable artifacts that encode hierarchical relations, melodic patterns, and harmonic scaffolds in a reusable symbolic system. Using Minimum Description Length (MDL) principles, we show that tonal abstractions support compression: themes function as shared models, allowing variations to be represented with fewer parameters and transformations, reflecting cognitively plausible reuse and systematicity. We demonstrate the framework on Paganini's Caprice No. 24 and Steve Reich's Clapping Music, showing how surface diversity emerges from a compact set of symbolic operations applied over inherited structures. Unlike data-driven neural models, our method preserves interpretable, inspectable representations aligned with musicological analysis. This work provides a proof-of-concept that tonal music can be formalized as an executable symbolic system, offering both a computational model of composition and a window into the cognitive structures underlying musical creativity.