Harmonic Grammar & Musical Syntax

Just as natural language follows grammatical rules, tonal music follows a harmonic syntax. Rohrmeier (2011) proposed a generative grammar for tonal harmony that models chord progressions as hierarchical structures, not just linear sequences. A V–I cadence isn't simply "dominant followed by tonic" — it's a nested dependency where the dominant creates an expectation that the tonic resolves. Lerdahl & Jackendoff (1983) formalized this in their Generative Theory of Tonal Music (GTTM), showing that listeners parse music into hierarchical trees of tension and resolution, much like parsing a sentence into subject, verb, and object.

Steedman (1984) extended generative grammar to jazz, demonstrating that the ii–V–I pattern and its substitutions (tritone subs, backdoor dominants) follow recursive rewrite rules. A complex jazz progression like |Dm7 G7 | Cmaj7| can be generated from a single rule: "approach the target chord with its ii–V." This recursive structure explains why jazz musicians can navigate unfamiliar chord changes in real time — they're applying grammatical rules, not memorizing every possible sequence. MusicMind's harmonic analysis uses these grammatical models to score harmonic sophistication and predict chord resolutions.

Burgoyne et al. (2011) created the McGill Billboard corpus — expert annotations of 740 pop songs — revealing that just 4 chord functions (I, IV, V, vi) account for over 80% of all harmony in popular music. White & Quinn (2018) extended this work to show that harmonic function depends heavily on context: the same chord (e.g., a C major triad) plays different roles depending on what precedes and follows it. These corpus statistics power MusicMind's chord transition probability model, enabling it to quantify how surprising each harmonic move is.