ABSTRACT
The core issues of the syntax-phonology interface are what phonology must know about syntax and what makes the phonological behavior of one syntactic unit different from that of another. To be more precise, we want to know which specifi c syntactic properties affect the application of phonological rules and how these syntactic properties are incorporated into phonology. In response to these questions, the two most infl uential but competing theories are proposed. One is proposed by Kaisse, who claims that phonological rules are subject to conditions that are stated directly in terms of syntactic domains. Kaisse distinguishes two types of post-lexical rules (Kaisse 1985: 193) – namely, external sandhi rules (P1 rules) and fast-speech rules (P2 rules). Kaisse thinks that the P1 rules have direct access to the information contained in the labeled bracketing of syntactic structure and apply only under the specifi c syntactic conditions. What is the type of syntactic information that the P1 rules have access to? According to Kaisse, it is neither the syntactic categories such as NP, VP, AP, or PP, nor the syntactic units such as morpheme, word, or phrase. Instead, it is the syntactic relation, i.e., the domain c-command condition. According to Kaisse, the syntactic information needed to delimit sandhi rules at the P1 level is only determined in terms of the m-command condition, which is defi ned by Chomsky (1986a). Following Hyman et al. (1987), this syntax-sensitive approach is termed as the Direct Reference Approach (hereafter DRA). The other theory is proposed by Selkirk (1984, 1986), Nespor and Vogel (1986), and some others, according to whom the surface syntactic structure
is fi rst mapped onto a prosodic structure consisting of prosodic constituents. These prosodic constituents are the domains of phonological rules applying above the level of the word. According to Selkirk and others, the prosodic structure, located between syntactic structure and phonetic representation, consists of different types of domains, including the prosodic word (ω), the clitic group (CG), the phonological phrase (φ), the intonational phrase (ι), and the utterance (υ). These domains are hierarchically organized in a number of layers. Thus a well-formed prosodic representation looks (1).
