|A Stratal OT Analysis of the Sanskrit Nominal Accent Paradigm
|Stratal Optimality Theory posits constraint evaluation at three distinct levels or strata: stem, word and phrase level. This multi-level variation of the original OT model may provide more economical and elegant explanations of problems involving opacity, derived environment effects and paradigmatic effects than other Parallel OT theories that have been developed to address these issues.
Alternations of pitch accent placement within the Vedic Sanskrit nominal inflectional and derivational paradigms provide evidence supporting Stratal OT. First, two major accentual patterns surface upon inflection; accent may remain fixed on a stem syllable in all cases or may shift to the inflectional ending in "weak" cases only. Second, for all nominals, the vocative is not accented, except when it appears at the beginning of a phrase; in that case, the accent is always on the initial syllable, regardless of the position of the accent on the stem or other inflected forms. Finally, accentual variation of primary and secondary derived stems and words is seen; accent may remain on the root (or primary stem) or may fall on the derivational ending; in either case, accent remains fixed throughout inflection (i.e., no accent shift).
In this analysis, a system of well-grounded morphological, prosodic, identity and alignment constraints are proposed for the stem, word and phrase levels to explain these alternations. It is argued that inflection takes place at the word level while derivation occurs at the stem level. Constraints at the three levels follow a consistent ranking pattern, while remaining relevant to the prosodic and morphological requirements of the specific level. The highest ranked identity constraint at the stem and word levels are substantively identical: MAX-IO(Accent)HEAD. Crucially, however, the head is different at each level, with the derivational suffix being the head at the stem (derivational) level, while the stem or root is the head at the word (inflectional) level. This allows for a principled explanation of the various accentual phenomena and, additionally, provides evidence for the operation of distinct stem and word levels.
Numerous constraint tableaux are presented; in each case, the winner is the correct surface form. These data indicate that Stratal OT effectively captures the Sanskrit nominal accentual paradigms using a small number of well-grounded constraints. Importantly, it is demonstrated that Parallel OT is unable to generate the correct surface forms for all the relevant alternations without the introduction of exotic and otherwise unnecessary constraints. Further cross-linguistic and theoretical research into the Stratal OT model is needed to establish its universality and superiority to other models.