While the loss of subject agreement is often related to general processes of phonological attrition, the processes by which subject agreement emerges remain disputed. This paper tests one of the most influential approaches, the “topic shift” account of Givón (1976 , 2017 ), which identifies a pragmatically-marked construction signaling topic discontinuity as the point from which subject agreement initially spreads. We investigate subject indexing in Hewramî, a West Iranian language, which exhibits differential subject indexing in past transitive constructions. Historically, West Iranian languages lost subject agreement in these constructions, but retained it elsewhere. Most contemporary languages have since restored subject agreement in past transitives, but in Hewramî such innovated subject indexing is not fully grammaticalized. We combine quantitative and qualitative methodologies on a sample of 2005 past transitive clauses from spoken Hewramî in order to identify the factors that influence the presence of an index. The quantitative analysis reveals that subject indexing is overall favoured in clauses that exhibit topic continuity, in particular those otherwise lacking an overt subject NP. In clauses with an overt subject NP, it is features associated with focus, rather than topic continuity, that inhibit indexing. We also find effects of person and animacy, with indexing most frequent with first and second person, and other human subjects. We find no evidence for the claim that a pragmatically marked construction for marking “topic shift” provided a model construction for these developments. Our findings constitute a novel and empirically well-grounded contribution to one of the most widely-researched domains of diachronic morphology.
Mohammadirad et al. (Sun,) studied this question.