My dissertation draws on a corpus of bilingual speech which I recorded amongst Russian-speaking communities in Germany. In code-mixing, stems from language A with affixes from language B alternate with fully-fledged constituents of language A used in the morphosyntactic context of language B. Although researchers widely report on this variation, explanations for it in the literature are controversial and remain unsupported by systematic empirical evidence. In this study, I provide empirical evidence on how frequency of co-occurrence of words and recency determine the choice between the competing structures. I predict the outcome of this competition by utilizing corpus-linguistic methodology and statistical modeling. I argue that an analysis that takes linguistic, usage-related and processing factors into consideration and allows for their interplays is robust in predicting variation in code-mixing.

The dissertation investigates switch placement within a syntactic phrase and at its boundary as well as plural marking of German noun insertions in Russian sentences. In the context of the noun phrase modified by an adjective, the switch is placed before the adjective or between the adjective and the noun. German nouns that are frequently used with certain adjectives in German are inserted into Russian sentences together with these adjectives. If the frequency of an adjective-noun combination in German is low, the realized German noun is usually preceded by the corresponding Russian adjective. In the context of the prepositional phrase, I found that, apart from the frequencies of preposition-noun combinations in German, recency, or repetition priming, also influences the choice of the preposition in the mixed phrase. In other words, exposure to a preposition in prior discourse increases the chance of using the same preposition in the same language in the target structure. The final study examines determinants of marking plural on German nouns inserted into Russian sentences. The choice between Russian and German plural marking turned out to depend on the frequency distribution of singulars and plurals in German, the phonetic shape of inserted German nouns as well as the morphosyntactic context in which they occurred in mixed sentences. As such, the work unearths effects of frequency and word repetition in bilingual production and accounts for the variation in commonly reported code-mixing patterns.