An explanation for ergative versus accusative languages: An examination of Inuktitut.

This dissertation shows that when specific objects are checked accounts for a language being ergative or accusative. In ergative languages a specific object is checked at Spell-out by the object moving to (Spec, T) with the resultant Abs/Nom case marking of the object. This accounts for the wide scope only reading of Abs/Nom objects and native speakers interpretations of Abs/Nom objects as specific/referential. Since the object moves to (Spec, T) the subject remains inside the VP and is assigned Gen/Erg case by V. In accusative languages a specific object is not checked until after Spell-out and the subject moves to (Spec, T) where it gets Nom case and the object remains inside the VP where it gets Acc case. This accounts for the possibility of wide and narrow scope readings for Acc objects. Ergative languages are also characterized as having "split ergativity" whereby there is also a nominative-accusative case marking/agreement pattern. This is also explained. Specific objects move outside the VP at Spell-out with the resultant ergative case marking, while non-specific objects remain inside the VP are assigned Inst/Acc case through insertion of a postposition with the resultant accusative case marking pattern. This explanation for the existence of ergative and accusative languages eliminates the need for the stipulation in the Minimalist Program (Chomsky 1995, Chapter 4) that arguments have to move to have features checked while features on non-arguments could be checked in situ. It also illustrates the type of feature that is checked, and that features triggering movement for checking can be on the moved item. The analysis of ergativity is based on the North Baffin dialect of Inuktitut and uses field work data on sentences and nominals. The data on sentences shows that a speaker can make specific or non-specific reference to all types of objects: personal names, demonstratives, modified nouns, quantified nouns. It also shows that the audience interprets a specific object as the speaker intending to pick out an entity. Specificity is thus shown to be part of the semantic component (Donnellan 1966, 1978). The data on nominals supports the analysis of case assignment: arguments of derived and non-derived nominals have Erg/Gen case, and the subject and object arguments of gerunds have Erg/Gen and Inst/Acc cases respectively. Finally a discussion of agreement in Inuktitut supports the analysis of agreement being a relation rather than a functional projection, and the checking of specific objects at Spell-out