Typologies of agreement: Some problems from Kayardild

In this paper I describe a number of agreement-type phenomena in the Australian language Kayardild, and assess them against existing definitions, stating both the boundaries of what is to be considered agreement, and characteristics of prototypical agreement phenomena. Though conforming, prima facie, to definitions of agreement that stress semantically based covariance in inflections on different words, the Kayardild phenomena considered here pose a number of challenges to accepted views of agreement: the rich possibilities for stacking case-like agreement inflections emanating from different syntactic levels, the fact that inflections resulting from agreement may change the word class of their host, and the semantic categories involved, in particular tense/aspect/mood, which have been claimed not to be agreement categories on nominals. Two types of inflection, in particular - 'modal case' and 'associating case' - lie somewhere between prototypical agreement and prototypical government. Like agreement, but unlike government, they are triggered by inflectional rather than lexical features of the head, and appear on more than one constituent; like government, but unlike agreement, the semantic categories on head and dependent are not isomorphic. Other types of inflection, though unusual in the categories involved, the possibility of recursion, and their effects on the host's word class, are close to prototypical in terms of how they fare in Corbett's proposed tests for canonical agreement.