To Compound or not to Compound? A Diachronic Bayesian Analysis of Compounds and Equivalent Constructions in Vedic Sanskrit

A characteristic trait of Vedic as well as Classical Sanskrit is the use of nominal compounds. Diachronic linguistic studies have observed an increasing use of compounds in Vedic texts. It is also generally accepted that compounds should be read as syntactic phrases and that they can be equivalent to subordinate clauses. However, it has not been studied so far whether and to which degree compounds replaced competing syntactic structures such as relative clauses or participial constructions over time. Using data from a syntactic treebank of Vedic and early Classical Sanskrit, this paper addresses the questions whether compounding replaced equivalent constructions and which textual and sociolinguistic factors may have driven this process. The paper studies compounds used as adnominal and adverbial modifiers, and compares their frequency distributions with those of subordinate clauses, adjectives, converbs, and participial constructions. Since the number of relevant cases is limited and the sociolinguistic factors driving the use of compounds are not well understood, the observed distributions are modeled with a hierarchical Bayesian framework that extracts an optimal subset from a set of possible explanatory factors (chronology, geography, poetry/prose alternation, genre, and school affiliation of Vedic texts).