Drivers of word order variation in Sanskrit nominal expressions

Abstract

This paper investigates variation in the linearization of noun phrases in Vedic and Post-Vedic Sanskrit. Employing a large set of structural, information-theoretic, and complexity-related features, we develop a Bayesian model assessing which of these features drives continuous versus discontinuous linearization of noun phrases. Results show that variation in word order patterns is largely systematic, with pronominal dependents being the only word class that significantly favors discontinuous linearization. Contrary to previous assumptions, diachronic differences largely disappear once other linguistic factors are controlled for, suggesting that synchronic determinants such as genre and style play a more central role than previously recognized. Individual texts, among them the Paippalāda Saṃhitā of the Atharvaveda, show idiosyncratic behavior that remains unexplained by our model, and may point to dialectal differences.