|Abstract:||While New Indo-Aryan languages are a common example of morphological ergativity, the range of variation in ergative marking and agreement among these languages has not been examined in detail. The goals of this paper are twofold. We first present a typology of ergative marking and agreement in Indo-Aryan languages, demonstrating that a progressive loss of ergative marking has occurred to varying degrees in different systems. This process is manifested in two distinct strategies of markedness reduction: loss of overt subject marking in the nominal domain and loss of marked agreement in the verbal domain. Using the framework of Optimality Theory (OT; Prince & Smolensky 1993), we account for the typology in terms of universal subhierarchies of markedness (Aissen 1999; Woolford 2001). Extending the analysis to dialect variation in one language, Marathi, we show that the dialect typology parallels the cross-linguistic typology, but only within the range permitted by changes already present in the parent language (Old Marathi). Furthermore, the dialect typology includes additional hybrid case-agreement systems predicted by our analysis.