This article proposes a new analysis of yers in Polish and addresses the issue of how yers should be represented. Reviving the debate started by Kiparsky (1973), the article argues that diacritic use of phonological features is superior to phonological use of diacritic features. Since diacritic representation of yers misses generalizations, yers are better represented as floating melodic segments (Rubach 1986, Kenstowicz & Rubach 1987). The patterns of Yer Vocalization and Yer Deletion are derived without recourse to syllable structure constraints such as *COMPLEX-Coda. Yer Deletion applies at an earlier level than Yer Vocalization and is enforced by a distribution-based constraint. Yer Vocalization, on the other hand, is a context-free process that is driven by the need to parse segments in output representations.
