A review of the current state of work in computer modelling in music reveals underlying problems of completeness and variability in the application of principles, which are indicative of central problems in systematic music theory generally. Solutions suggested by current work, again in computer modelling, require consideration of the pragmatics of listening to music leading to hypotheses which might act as bases both for variation in the application of principles and for the inference of new principles. Proposed hypotheses are derived from discussion of analyses of two pieces of music (the first of Webern's Five Pieces for String Quartet, Op. 5, and the third movement of Mozart's piano sonata in D major, K.284). Some conclusions are drawn both for potential application in computer modelling and for theories of musical meaning.