Harmony Triggering as a Contrastive Property of Segments

In some languages, segments bearing a potential harmonizing feature trigger harmony while other segments bearing that feature do not. This paper examines such a case in Classical Manchu tongue root harmony, in which some advanced tongue root (ATR) vowels idiosyncratically trigger a process of ATR harmony while others do not. I propose that this pattern and others like it are best accounted for by an analysis in which a segment’s status as a trigger of harmony is encoded within its subsegmental representation, and a phonological grammar shapes a language’s inventory to include both harmony-triggering and non-triggering segments. This analysis is implemented within the Gestural Harmony Model, in which the assumed units of phonological representation are dynamically-defined gestures.