What differentiates Serbo-Croatian verbal theme vowels: content or markedness?

Abstract

We examine two hypotheses regarding the role of theme vowels (ThVs) in Serbo- Croatian (SC): (i) that the various ThVs attested in SC are markedness-based realizations of the same syntactic feature specification, and (ii) that different ThVs carry different syntactic features. We focus on the two SC ThVs occurring with the highest number of bases: <a, a> and <i, i> (the ordered pair specifies the infinitive-stem and the present-tense-stem realization of the ThV). We show that if these ThVs are to be distinguished by feature specification, the best fitting analysis has <a, a> bearing only the categorial verbal feature, while <i, i> is additionally specified for the feature [SCALE], which contributes scalarity to the verbal predicate (Hay et al. 1999; Kennedy & Levin 2008). A corpus-based exploration shows that the stronger hypothesis (ii) encounters problems, the most obvious being that the regularities are only tendential, with a significant number of exceptions. If the ThVs carried different features, they would be expected to yield systematic patterns. We conclude that the weaker alternative (i) provides an empirically more accurate account and propose a specific model where at the interface with phonology, the aggregate degree of markedness of the context in which the ThV is realized is computed from a set of markedness hierarchies of the relevant phonological and semantic properties of that context (the latter mediated by the corresponding syntactic features). A mapping of the aggregate degree of markedness onto the morphological markedness hierarchy of ThVs determines the realization.