Mark Baker (Rutgers University)
Research: Syntactic Theory, Morphology and its relationship to Syntax and Semantics
Title: On Inherent and Dependent Theories of Ergative Case
Ergative languages are languages that use a special case marker on the subjects of transitive verbs, that distinguishes them from the subjects of intransitive verbs and from direct objects. Although about one third of the languages that have overt case marking are ergative in this sense, ergative languages have been controversial and problematic for generative theory because case marking and grammatical functions do not line up as smoothly as they do in more familiar accusative languages (i.e. not all subjects get the same case). The most influential theory of ergative case over the last 10 years has been that ergative is an inherent case, assigned by v to the argument that it theta-marks, on the analogy of how dative case subjects work in languages like Icelandic (e.g. Woolford 2006, Legate 2008).
In this talk, we put forward an alternative view that is gaining momentum: the view that ergative case is "dependent case" assigned to the higher of two NPs in the same spell out domain (Marantz 1991, Baker 2014). We compare the inherent case theory with the dependent case theory in some detail, both conceptually and in terms of their predictions. We show that unaccusative verbs with theme arguments can receive ergative case if there is a second NP in the structure as a result of applicative formation or some similar argument-adding operation, in Shipibo, Greenlandic, and Chukchi. Conversely, we show that transitive verbs with agent arguments cannot receive ergative case if the second NP is rendered syntactically inactive, say by noun incorporation or antipassive. Finally, we show that the inherent case theory actually predicts a so-called active alignment pattern rather than a true ergative alignment pattern. However, the typological record shows that active alignment is at best extremely rare for languages with overt case markers on NPs (Dixon 1995, Comrie 2005), and we show that few that have been mentioned are not really good cases after all (focusing on Northern Pomo and Drehu). Across this range of data, we see that the case an NP gets does not depend on its thematic role (agent vs patient) or on what head theta-marks it (v or V), but rather on how many visible NPs there are in the clause. In other words, the inherent case theory is disconfirmed for most or all languages, whereas the dependent case theory is supported.