A transparent syntax-phonology interface: Cyclic spellout and phrasal tone in Huave and Luganda
Marjorie Pak,
Thursday, March 13, 2008, 9:00 AM
Theories of the syntax-phonology interface are faced with a basic tension: on the one hand, phrasal phonological rules across a wide range of languages have been shown to be sensitive to general aspects of the syntactic constituent structure; on the other, the phonology has special properties (e.g. rate/rhythm-induced variability, boundary mismatches) that can obscure its relationship to the syntax. The hypothesis developed here is that a direct, transparent relationship between the syntax and the phonology can be maintained under a *cyclic-spellout* model of the grammar - if we adopt a more articulated view of what 'spellout' means.
I present data from two unrelated languages, Huave and Luganda, where tone-spread domains correspond with striking regularity to a certain type of syntactic constituent - essentially a clause minus its top 'edge.' This pattern is naturally accounted for in a model where the relevant phonological domains are spelled-out CP phases. I show furthermore that in utterances containing more than one clause, a new tone-spread domain begins at each CP boundary - again as expected if spellout takes place at each CP in the structure. While there are certain contexts where two clause-like constituents join together to form a single tone domain, I argue that the second 'clause' in these cases is a reduced (non-CP) constituent, so that it does not get spelled out until the next-higher CP is reached. I provide independent syntactic evidence that this reduced-clause analysis is correct: in Huave, reduced clauses cannot be extraposed; and in Luganda, preverbal Spec,CP adverbs are unavailable in just those clausal complement that group together with the main clause phonologically.
The fact that syntactic and phonological evidence converges on the same CP object in Huave and Luganda strongly suggests that the cyclic-spellout approach is on the right track. The phonology in these languages cannot be understood without some reference to the syntax - specifically to clause-hood - and clauses are assumed to be cycles in phase theory and other cyclic-spellout proposals. The broader goal of the current work is to develop an *articulated model of spellout* that could make the transparent-interface approach feasible for phonological alternations cross-linguistically. The current hypothesis is that phrasal phonological rules apply at different stages along the PF branch as syntactic objects are subjected to a series of linearization operations, so that 'early rules' apply to smaller, partially linearized objects and 'late rules' to fully linearized chains. I provide evidence that early and late rules coexist in Luganda, and explore some questions and predictions for syntax-phonology relations in other languages.
Last Modified: March 10, 2008