||In the standard generative perspective, English is a mora-sensitive language, permitting syllables with one or two moras (or in some perspectives, strictly two). However, structures are readily available in the lexicon which seem to have three moras in a single syllable. Lavoie and Cohn 1999 argue that these structures are unstable because they violate the mora-counting condition of English, and this violation motivates speaker-dependent variability. Feature instability engenders a skepticism for speakers' own intuition about the syllabification of these so-called unstable structures. Thus, an alternative approach is necessary, whereby linguistic representation may be inferred from observable behavior beyond speech or writing. Through an exploratory study on pitch of performed music, and an experimental study on musical text-setting, I find evidence for the use of musical features and behavior as indicators of linguistic representation. Notably, findings from both research designs lend support to the claim that highly moraic structures in English are unstable and exhibit a dichotomy of syllabification between individual speakers.