Ellen Woolford
University of Massachusetts, Amherst
It is common to see descriptions of ergative agreement systems in which the agreement morphemes are labeled ergative and absolutive, even in languages which have no morphological ergative Case. It is also common to see analyses of such agreement systems that crucially refer to covert ergative Case. The main question to be addressed in this paper is whether positing covert ergative Case in order to account for ergative agreement is necessary or desirable.
This paper will argue that when one examines the types of ergative agreement patterns that occur in languages with and without morphological ergative Case, they are different. In particular, the very common type of ergative agreement pattern found in languages like Hindi, where only absolutives agree, is not found in languages without morphological ergative Case. Thus positing entirely covert ergative Case in a language in order to account for an ergative agreement pattern makes the wrong typological predictions. The theory of agreement needs to be made sufficiently restrictive so as not to be able to generate the Hindi-type agreement pattern unless the language has morphological ergative (or dative) Case. That rules out any approach in which agreement is completely independent of Case, such as an account in which agreement can be liked directly to ergative and/or absolutive grammatical relations (independent of Case) or an approach that can set a parameter such that Agr-O must be active (even in languages without morphological ergative Case).
On the other hand, agreement is not completely parasitic on Case. The latter part of this paper addresses the question of how to account for ergative-absolutive agreement patterns in languages without morphological ergative Case. It will be argued that one common pattern involves a clitic that can double any argument, and an agreement morpheme that can only cross-reference the subject. The ergative pattern is produced by a sort of economy effect, wherein the agreement is never used unless it is necessary (i.e. when there are two arguments to cross-reference). The cheaper' clitic is always used if there is only one argument to cross-reference. In languages where the agreement morpheme is attached to the verb, while the clitic is above the verb, this kind of ergative agreement pattern has been interpreted as indicating an upside-down' syntax; but that conclusion is not warranted under this approach.