The Role of Semantic Features in Scrambling Roksolana Mykhaylyk Stony Brook University 1. Introduction Previous studies have examined semantic/pragmatic correlates of word order change in a number of languages (see Thrainsson 2001 for an overview). They attested that leftward movement of a direct object affects its interpretation: scrambled position is usually associated with a specific semantic feature. This correlation has been shown to exist in L1 and L2 grammars of learners of Swedish, Dutch, Norwegian, Serbian, Russian, Ukrainian, and other languages (Josefsson 1996, Deen 2004, Brun 2005, Unsworth 2005, Westergaard 2008, Mykhaylyk and Ko 2008, Anderssen et al (to appear), inter alia). The relevant semantic feature, however, has not been clearly defined. It has 1 . The earlier studies, then, raise two specific research questions: i) which semantic feature plays the most important role in object scrambling; and ii) do effects of specificity, definiteness and partitivity vary cross-linguistically? This paper provides novel evidence teasing apart the role of semantic features in scrambling, using new experimental data from Ukrainian child language acquisition. Examination of monolingual Ukrainian language acquisition sheds light on the first question, while investigation of English- Ukrainian bilingual development addresses the second question. 37 bilingual English-Ukrainian children and 41 monolingual Ukrainian children from 2 to 6 were tested in an oral elicitation task which triggered the use of scrambled or basic structures in different semantic contexts. The results show that both groups of children are aware of correlation of semantic features and syntactic movement: they scramble optionally, but not randomly (confirming Mykhaylyk and Ko 2008). The role of semantic features varies by language and age group, but the general pattern is clear: there is a highly significant effect of specificity-partitivity on scrambling in Ukrainian. The data also suggest that the implementation of features might differ cross-linguistically since bilingual children use less scrambling in specific-referential contexts than monolingual children. The paper is structured as follows. First, relevant language facts from Ukrainian and English are discussed in Section 2. Hypotheses and predictions are presented in Section 2.4. Section 3 provides a detailed description of the experimental study, and in section 4, the results are summarized. The paper concludes with a discussion of the findings and their implications for language acquisition theory. 2. Background Ukrainian and English differ in the way they employ scrambling and encode semantic features. Ukrainian is an article-less language that uses scrambling as its chief means of encoding specificity-presuppositionality (Mykhaylyk & Ko 2008). English does not have scrambling and uses articles mainly as definiteness or indefiniteness markers, but in L1 and L2 acquisition the choice of articles might also depend on specificity and/or partitivity (Maratsos 1976, Ionin et al 2004, Ko et al 2008). * I would like to thank Richard Larson, Natalie Batmanian, Heejeong Ko, Dan Finer, John Bailyn, and the audience at Gasla 10 for their comments and help at different stages of this work. All shortcomings are my own. This research was supported by the Shevchenko Scientific Society of America. 1 See Section 2.1 for the definitions. © 2009 Roksolana Mykhaylyk. Proceedings of the 10th Generative Approaches to Second Language Acquisition Conference (GASLA 2009), ed. Melissa Bowles et al., 157-167. Somerville, MA: Cascadilla Proceedings Project.