Evidential, raised possessor, and the historical source of the ergative construction in Indo-Iranian

This paper argues (i) that the source of the ergative construction of the transitive verb in Indic and Iranian languages was anticausative but not passive as has widely been assumed, (ii) that it functioned as a modally marked evidential which indicated that the event in question was inferred or reported rather than directly witnessed, and (iii) that the agent was by origin a genitive-marked adnominal possessor raised out of its noun phrase and later reanalysed as the syntactic subject, its uniform instrumental-marking in Sanskrit being an innovation. In view of the fact that the possessive modifier precedes its head this analysis can account naturally for the position of the transitive agent at the beginning of the clause, preceding the object. It is, finally, suggested that the construction originated with non-agentive intransitive verbs and that it spread to transitives through the intermediary of ergative (ambitransitive) verbs which can have both intransitive-spontaneous and transitive-causative forms, a hypothesis which creates a diachronic link between lexical and structural ergativity.