Evaluating the Matrix Language Frame model on the basis of a Russian—Estonian codeswitching corpus

This article focuses on the question to what degree typological characteristics of the languages involved interact with the Matrix Language vs. Embedded Language and Content Morpheme vs. System Morpheme distinctions. In addition, the integration of Estonian items in Russian—Estonian codeswitched sentences is described, because the extent of integration of Estonian items into the Russian matrix is closely related to matrix language determination. Both Estonian and Russian have a complex inflectional morphology. Full integration of an Estonian item into Russian means gender and case assignment, but in fact this does not occur in all codeswitched instances. Russian—Estonian data demonstrate a wider range of possibilities than just fully integrated and completely unintegrated items. Estonian items may be integrated phonologically into Russian (e.g. regarding vowel length or stress), but morphological and phonological integration do not depend on each other: Estonian items may show extensive phonological but no morphological integration into Russian and vice versa. Bilingual data show that the matrix language cannot be determined by analyzing the morphosyntactic level only; the degree of phonological integration of codeswitched words should also be taken into account. Thus, the morphosyntactic bias in the Matrix Language Frame model, meaning that phonological adaptation has no influence at all on matrix language assignment, is criticized.