COORDINATION IN TONGAN: IS SYNTACTIC ERGATIVITY REAL?

A language is said to be syntactically ergative if it shows an ergative pattern with respect to syntactic phenomena, typically relativization and deletion under coordination. There are two puzzling facts about syntactic ergativity. First, it does not seem to be a direct consequence of morphological ergativity. Not all languages with ergative Case marking exhibit syntactic ergativity. Second, even when a language does show syntactic ergativity, it is often restricted to certain syntactic phenomena (Dixon 1979). This leads us to question whether syntactic ergativity is a property of language and can be reduced to a parameter, as suggested by Bittner and Hale 1996 and Ura 2000. In this paper, I argue that such a parameter does not exist, and that an ergative pattern arises in different languages and in different constructions within a language, due to independent reasons. Specifically, I propose that syntactic ergativity is in fact a direct consequence of morphological ergativity in the sense that an ergative pattern arises when the relevant syntactic rule is sensitive to Case. On the other hand, some syntactic rules are sensitive to theta-roles rather than Case. This gives rise to an apparent accusative pattern, in which external arguments are distinguished from internal arguments