Jessica Coon gave a colloquium talk at MIT last week. The title of her talk was “Two types of ergative agreement: Implications for Dependent Case Theory.” Abstract is below.
A range of literature has shown that agreement is sensitive to morphological case (e.g. Bobaljik 2008, et seq). While the dependence of agreement on case has been robustly demonstrated, the source of morphological case remains controversial. This talk focuses on the assignment of ergative case. Under one line of approach, ergative is an inherent case, assigned by a functional head to external arguments in their thematic position (Woolford 1997; Legate 2008). On another approach, ergative is the mirror image of accusative, assigned configurationally to the higher of two arguments in some local domain (Marantz 1991; Baker & Bobaljik to appear). Through an investigation of ergative agreement systems, I argue that a Dependent Case approach is not only unmotivated for a less-studied type of ergative agreement, but also runs the risk of over-generating.
I argue that ergative-absolutive agreement patterns have two different sources. Type 1: In languages like Hindi-Urdu, agreement comes from T; morphologically case-marked ergative subjects are inaccessible for agreement, resulting in an “ergative” agreement pattern (i.e. absolutive arguments agree; see Bobaljik 2008). Type 2: In languages like Chol and Halkomelem, transitive subjects (i.e. ergative arguments) agree, and the source of this agreement is low: v (Coon to appear; Wiltschko 2006).
This talk has two main goals. First, I provide morphophonological and syntactic evidence for the existence of the less-discussed Type 2 system; specifically, I argue that ergative agreement in Chol has a low source and is the result of a direct relationship between v and the ergative subject. Second, I argue that a Dependent Case analysis––while easily able to handle the Hindi-Urdu-type agreement system––faces problems with the Chol-type agreement system. Not only must the language keep track of two different types of null case, but we are left without a way to rule out languages with nominative-accusative case and ergative-absolutive agreement, a well-known typological gap.
While Dependent Case has achieved a range of empirical coverage (e.g. Baker & Vinokurova 2010; Levin & Preminger 2015; Baker & Bobaljik to appear), the end result is one in which the mechanism of ergative case assignment––inherent or dependent––must minimally be parameterized. Given that Type 2 ergative languages lack morphological case altogether, I suggest that this may not be a bad result.