Abstract

We argue that Onsets can be licensed not only by Nuclei (the universal case), but also (parametrically) directly by prosodic domains such as Word or Phonological Phrase. The surfacing or deletion of Coda/Word-final consonants follows automatically if we allow that (1) a Coda contains no more than one consonant, (2) an Onset consists of a consonant or a nan-strident obstruent + a liquid, (3) an Onset is licensed by (a) a Nucleus or (b) the right edge of a Phonological phrase, (4) unlicensed elements are subject to Stray Erasure. Of these principles, (1), (2), (3a) and (4) are uncontroversial in GU; (4) corresponds to a parameter setting. This somewhat abstract syllabification allows a simple statement of the distribution of several phonological processes.