| This dissertation brings novel data from Ranmo, an endangered Papuan language, to bear on the phenomenon of split intransitivity, the comparatively understudied type of split ergativity (cf. aspectual and person-based splits). Ranmo is spoken by approximately 300 people in Western Province, Papua New Guinea and belongs to the Morehead-Upper Maro River family. The point of departure for this dissertation is the observation that there are two classes of semantically one-place verbs, unaccusatives and middles, which show distinct patterns of agreement|an apparent case of split intransitivity.;I demonstrate, however, that this "split" is only an illusion: middles, which show a non-ergative pattern of agreement (i.e., S=A), are in fact syntactically transitive, having an NP (as opposed to ϕP) object that is coindexed with and bound by the ϕP external argument. This NP object requires a corresponding functional projection on the clausal spine, XP, which is sandwiched between VP and nuP; this is essentially a new proposal for pseudo-noun incorporation (PNI) (cf. Massam 2001). Under this analysis, middle verbs---a semantically heterogeneous class encompassing reflexives/reciprocals, anticausatives, and agentives---are subsumed under PNI. When nu probes, it cannot agree with the NP object (since it lacks ϕ-features), resulting in the default spell out of object agreement, which is referred to as the 'middle' morpheme; this is an instance of agreement failure in the sense of Preminger (2009, 2011, 2014). No special rules of agreement are required to capture the non-ergative pattern of agreement in Ranmo; therefore, it is entirely ergative rather than split-ergative. This is a significant conclusion especially in light of recent findings showing that aspectual and NP-based splits, too, are epiphenomenal, involving additional clausal structure in the non-ergative portions (Coon 2010, Coon and Preminger 2012).;I further propose that applicative constructions form the "other side of the PNI coin," i.e., their direct object is also an NP, which requires the presence of a clausal correspondent, XP. I argue that the Person-Case Constraint (Bonet 1991, 1994) is evidence for the PNI analysis of applicatives, i.e., only 3rd-person arguments, which are structurally reduced compared to 1st/2nd-person arguments, are licensed in the NP direct object position of applicatives. It is simply that in applicatives, X has the additional function of introducing an applied argument in its specifier and assigning it a theta-role and inherent case.;Another major contribution of this dissertation is that it presents new evidence for the dependent theory of case assignment (Bittner & Hale 1996, Marantz 1991). On this view, case is assigned configurationally on the basis of the c-command relationships between noun phrases themselves; it is an alternative to the standard Chomskyan view that case is assigned as a reflex of agreement/Agree (Chomsky 2000, 2001). From both middle clauses and unaccusative applicative constructions in Ranmo, we have evidence of dependent case assignment: an argument receives ergative case only if it c-commands another noun phrase in the same domain. This also argues against the analysis of Ranmo ergative as inherent case assigned to agents by transitive nu/Voice. New data like those of Ranmo urge us to adopt a more nuanced, perhaps parameterized, view of case/agreement relationship, i.e., whether case is assigned as a reflex of agreement/Agree is a point of cross-linguistic variation, not a universal absolute. |
