EXTENDING ALIGN CONSTRAINTS TO NEW DOMAINS

Alignment theory proposes that constraints on the edges of morphologic al and prosodic categories interact with each other and with phonologi cal constraints (McCarthy and Prince 1993a). Ito and Mester (1994) ext end the family of alignment constraints to relate prosodic categories (syllables) and subprosodic categories (segments and features). This p aper further investigates the sensitivity of phonological constraints to edges and argues for extending alignment constraints to prosodic ca tegories larger than the syllable and word (Nespor and Vogel 1986; Sel kirk 1986, 1995; Inkelas 1989), in order to constrain phrase edges and their interaction with other domain edges and subprosodic units. I ar gue for such alignment constraints based on syllabification that cross es word boundaries, where words; cannot be correctly syllabified in is olation from their phrasal context. Data from the phonologies of Caire ne Arabic and Tamil show that the constraints on syllables, words, and phrases are evaluated in parallel, accounting for the locations of ep enthetic, deleted, and assimilated segments. Syllabification must be s ensitive to the simultaneous evaluation of constraints on all these do mains, in order to provide the context for alternations within domains and at their edges. Parallel evaluation in multiple domains dispenses with stages of syllabification and eliminates derivationalism from th e distinction between lexical and postlexical levels of analysis. The analysis is developed within the optimality theory of Prince and Smole nsky (1993) and the alignment theory of McCarthy and Prince (1993a).