Aspects of word order in Russian

The dissertation explores word order phenomena in a ‘free’ word order language, Russian. It has been proposed in the literature that in simple sentences like ‘John sees Mary’, six word orders are equally possible in Russian. The dissertation questions the equal acceptability of these word orders and shows that some of the “felicitous” word orders have a degraded status compared to others. The word order findings are based on experimental evidence from elicitation, perception and grammaticality judgment psycholinguistic studies with 237 adult native speakers of Russian. The results of the experiments demonstrate that Russian speakers have a strong preference for producing some word orders over others. For example, Russian native speakers produce transitive SVO, OVS and SOV felicitous word orders, but consistently do not produce VSO, VOS and OSV felicitous word orders, which they still recognize as acceptable, but as having a degraded grammaticality status. On the basis of the experimental evidence and analysis of the various constituent movements within the Minimalist Program approach, a model of grammar is proposed which adds a pragmatic component responsible for word order permutations. According to this model, the syntactic component of grammar generates only SVO sentences (the basic word order) in Russian. All discourse-dependent sentences result from realignment in the post-syntactic pragmatic component. In contrast to the hierarchical structure of syntax, the pragmatic component of grammar has a linear structure and operates with Optimality Theory-type constraints determining the optimal output word order in a particular discourse structure. The underlying assumption of this model is that this pragmatic component is present in all languages. However, the language specific ranking of the constraints in this component results in word order variations. In contrast to the previous structural approaches to word order permutations in Russian, the proposed model has obvious advantages. The model accounts not only for grammaticality and