A movement theory of ergativity

In this thesis, I propose a theory of ergativity in which NP arguments are checked for Case by moving to projections of agreement at LF. The Case-marking pattern of an ergative language arises when transitive subjects move to the projection of agreement usually associated with objects (AGR.o), while transitive objects and intransitive subjects move to the projection of subject agreement (AGR.s). While this proposal assigns the same underlying structure to clauses in an ergative language (unlike Marantz, 1984), it does have distinctive syntactic effects. In this it contrasts with a purely morphological approach to ergativity, such as that of Anderson (1976).Arguments can move to the specifier position of agreement, or adjoin to its maximal projection. Movement cannot take place across the same kind of position as the landing site, which leads us to predict that transitive subjects cannot undergo grammatical extraction in an ergative language. This prediction turns out to be correct in a number of languages, including Chamorro, Mam, and other members of the Mayan group. Our theory also allows for a plausible account of split ergativity--non-canonical patterns in an otherwise ergative language where transitive and intransitive subjects are marked the same, but behave differently under extraction.The proposal that NPs are not checked for Case until LF entails that they remain in their base positions at S-structure. Evidence for this claim is adduced from the distribution of empty pronoun arguments whose contents must be identified. Our prediction is that transitive subjects in an ergative language will interfere in the identification of an empty object pronoun, since it is closer to the pronoun than its legitimate identifier, AGR.s. This is also shown to be the case