Korean Prosodic Domains and Phrase Structure

1. Introduction Selkirk (1986) argues for the existence of a level, which she calls P-structure, that mediates between syntactic structure and phonetic representation. She claims that this P-structure is made of prosodic structures which are hierarchically ordered. That is, a phonological representation consists of a hierarchy of prosodic constituents or categories. From the bottom-up, the hierarchy consists of syllable, foot, phonological word, phonological phrase, intonational phrase, and utterance. These categories may be determined by syntactic structure, although they may not be isomorphic to syntactic constituents. The importance of prosodic structure comes from the fact that postlexical rules (which are the subject of sentence phonology) applies within certain prosodic domains. For instance, Chi Mwi:ni Vowel Shortening applies within the domain of phonological phrase, and French liaison applies within the domain of phonological word. The two representative approaches to deriving prosodic domains are relation-based approach and end-based approach. In a relation-based approach, the prosodic domains of rules are defined by referring to syntactic relations of the sentence. In direct relation-based approach, the domains are directly expressed in syntactic terms such as 'c-command' to which the phonological rule refers (Kaisse 1985). In indirect relation-based approach, domains can be defined syntactically, with phonological rules referring only to these domains. That is, a phonological rule itself does not care about the syntax, but syntactic relations are taken into account in deriving the prosodic domains of the rule (Nespor & Vogel 1986). Notions such as head-complement, adjacency and linear order are used to derive the domains. In end-based approach, domains are derived by just referring to the left or right edges and projection levels (XP, or X) of syntactic constituents, without paying attention to their relation to other constituents (Selkirk 1986). The parametric difference in the derivation of prosodic domain is determined by the difference in the setting for the category and the setting for right or left end. In general, left-headed languages tend to have right end setting, and right-headed languages tend to have left end setting. In this paper, I consider a postlexical rule in Korean which voices unaspirated untensed stops that occur between voiced segments. I refer to this rule as Stop Voicing Rule. I present the relevant data in section 2, and I review two approaches (relation and end-based approach) in determining the prosodic domain for this rule in sections 3 and 4. In determining the prosodic domain for this rule, Cho …