Conditions on Case
In some dependent-marking languages of the Tibeto-Burman area, Differential Argument Marking (DAM) (Bossong 1985, Aissen 2003, de Hoop & de Swart 2008, Malchukov 2008) is manifested through the presence of case morphology conditioned by information structure (see Chelliah and Hyslop 2011). Our corpus-based work on three Tamangic languages of Manang District, Nepal demonstrates that in spontaneous discourse, various grammatical correlates of topichood constitute key factors in determining the presence of ergative (rather than absolutive) subjects in transitive clauses (Bond, Hildebrandt & Dhakal 2010, 2014, 2015). Yet, while the relationship between information structure and DAM is well known (e.g. Nikolaeva 2001, Dalrymple & Nikolaeva 2011), determining which factors are the best predictors of particular case-forms within a statistical model of a conditional system remains open to debate. Little work has been attempted so far to take a statistical approach to this problem using corpus-based methods in under-described languages (although see Morey 2011). In this talk I explore some of the ways to model conditions on case in a probabilistic way.
Dr Oliver Bond, University of Surrey