Ergative case attrition in Central Indo-Aryan: NP-splits and the Referential Hierarchy

Differential case marking is often determined on the basis of inherent semantic properties associated with core arguments of the verb. This frequently results in a hierarchical split in which certain types of NPs are more or less likely to be case marked when in the role of agent/patient. The Referential Hierarchy (RH) (see Silverstein 1976; Comrie 1981) models this phenomenon in terms of the markedness of agent vs. patient roles, based on the semantic parameters of animacy/definiteness. Yet recent studies have raised doubts as to the constistency of the RH in predicting split-ergative marking (e.g. Filimonova 2005; Bickel 2008). This paper explores an Indo-Aryan dialect with an NP-split in ergative marking that appears to contradict the RH: Kherwada Wagdi. It examines the possible historical scenarios that could result in a reverse NP-split, suggesting that such historical transitions tend to follow a non-linear course and are frequently left incomplete.