A Substance-Free approach to harmony and underspecification

Proposed features to account for post-velar phonology in Shira Yugur is arguably [tre]. However, the acoustic correlates of [tre] in this language are inconsistent across different harmonic pairs. Vowel harmony may have had a phonetic motivation; in the synchronic phonology of Shira Yugur, however, [tre] lacks a unified phonetic gesture. This supports the claim from radical substance-free phonology (Odden 2022)--that how segments pattern together is determined by computation, rather than their phonetic substance. Consequently, underspecification becomes undefinable, such that disharmony in this languages cannot be addressed. I proposed a string-based solution, in which the tier consists of segment-defined harmonic classes and morphemic boundary (+).