Vedic Post-Lexical Retroflexion: Opacity and Diachrony
Introduction Sanskrit possesses a well-known rule by which alveolar n becomes retroflex n . if r, r /r " / or s . /ù/ precedes it and no coronal obstruent intervenes. In Vedic, Epic and Classical Sanskrit, this rule is completely regular within word boundaries. In Vedic, the oldest variety of Sanskrit, a process of n-retroflexion also occurs across word boundaries, albeit variably. This paper investigates the phonological factors conditioning Vedic post-lexical retroflexion of n (henceforth
... f n (henceforth PLR), and the extent to which PLR conforms to theoretical claims regarding lexical phonology and morphology (LPM). This paper demonstrates that Vedic PLR is an opaque, phonetically unnatural phenomenon. While it can be represented to some extent by models of LPM that allow for ordered levels of post-lexical phonology, this phenomenon presents problems for the assumption that lexeme-internal, structure-sensitive rules must precede post-lexical ones. This paper advocates a diachronic explanation of PLR. I show that a model sensitive to diachronic factors better predicts the distribution of retroflex clitics than a model which assumes phonetic naturalness in prosodic phonology. At the same time, it is highly likely that the opaque and phonetically unnatural distribution of retroflex clitics led to the gradual under-generalization and loss of this already gradient phenomenon. * This paper has benefited greatly from the input of audiences at CLS 50 and SCIHS 3. Special thanks are due to Gary Holland for comments on an early draft. All errors are my own.