This thesis investigates how fluent bilinguals make use of the grammar of their two languages when they construct verb- and pronoun-agreement only in one language (monolingual mode) or in both their languages (bilingual mode). We are particularly interested in the impact of the non-response language in sentence processing on the response language. Bilingual research has provided evidence for language integration in bilingual speech (e.g., Hartsuiker, Pickering, & Veltkamp, 2004) which is also consistent with the phenomenon of code-switching whereby speakers can use elements of each language in producing mixed-language utterances (e.g., Myers-Scotton, 2002). So far, studies at the lexical level have provided support for parallel language activation (e.g., Colomé, 2001), yet the issue of whether activation of either language can be strong enough to influence the workings of the other is still in dispute (e.g., Hermans, Bongaerts, de Bot, & Schreuder, 1998, but see Costa, La Heij, & Navarrete, 2006). In three separate sections of the thesis we employ a sentence-completion paradigm widely used in monolingual agreement literature (Bock & Miller, 1991) to examine language interaction effects in the monolingual and the bilingual modes of speech (Grosjean, 2000). English-Greek and Greek-English fluent bilinguals produced completions to singular or plural subjects when the number of the translation was either the same or different, and when their completion either did or did not switch languages. The first section investigates whether there is influence of the divergent number properties of the nonresponse native language (L1) on verb-agreement in the response second language (L2). The results of Greek-English bilinguals show influence of the underlying number of the L1 on completions in the L2. We interpret this in terms of a markedness account (e.g., Eberhard, 1997) whereby parallel activation and competition between an L2 singular subject noun and its L1 plural translation results in plural verbagreement because the singular form is more vulnerable to the marked plural form. English-Greek bilinguals who perform on the same monolingual mode do not show influence of their L1 when speaking in the L2 (Greek). We attribute this finding to a difference of morphological/inflectional properties between the two languages which renders a language that displays fewer overt markings (English) easier to control when utterances are produced in a language that displays more overt markings (Greek) (e.g., Vigliocco, Butterworth, & Semenza, 1995).

This thesis examines the bilingual discourse of a cohort of Belizean Spanish speakers who engage in robust language mixing between Spanish, English and to a lesser extent Belize Kriol, an English-lexified creole. The speakers selected for the current study have been identified as the “highest language mixers” in a corpus of 51 interviews conducted in northern and western Belize, areas which have been classified as two distinct dialect regions (Cardona Ramírez 2010; Hagerty 1979). While an abundance of research exists on Spanish-English bilingualism in the U.S. (e.g. Torres Cacoullos and Travis 2014; Silva Corvalán 1994; Roca and Lipksi 1993) there is less research on non-U.S. varieties of Spanish in contact with English, in particular, Belizean varieties of Spanish. Thus, by appealing to the comparative variationist framework (Poplack and Tagliamonte 2001), the major aims of the study are: (i) to describe the major patterns of use among those speakers of Belizean varieties of Spanish who engage in language mixing and, (ii) to determine the status of the single and multiword English-origin fragments which comprise the majority of non-native material in Spanish discourse. In determining the status of the English-origin material with regard to borrowing and code-switching, not only are the specific linguistic mechanisms used by these speakers elucidated, but insights are gained as to whether code-switching and borrowing are distinct linguistic phenomena. Diagnostics of subject position and gender and number agreement on English-origin nouns and verbal morphology and variable clitic placement for English-origin verbs revealed both these categories to pattern with Spanish suggesting that they are borrowings. The remaining one-third of the data, comprised of multiword fragments, consisted primarily of intrasentential and intersentential code-switching and a large category of multiword fragments which initially appeared to be neither code-switches nor borrowings. A comparative quantitative analysis revealed these items to be integrated into Spanish suggesting that they may be treated as single units of meaning. Results, for the most part, are consonant with the literature on bilingual speech. Data consists mainly of lexical borrowing (Thomason 2001; Pfaff 1979; Berk-Seligson 1986), specifically nouns, the most borrowed category cross-linguistically (Muysken 2000; Poplack et al. 1988). Speakers engage in “skilled” or equivalence intrasentential code-switching consistent with other Spanish-English data (Poplack 1980). In addition, only those speakers who reported being equally dominant in the respective languages exhibited robust intrasentential code-switching, thus, concurring with the prevailing assertion that code-switching is the domain of fluent bilinguals (Bullock and Toribio 2012; Lipski 1985; Poplack 1980). Some distinctive features of the language mixing employed by these speakers include the frequent and productive use of bilingual compound verbs (BCV) and a near categorical preference for BCVs as the mechanism for borrowing English-origin verbs. With regard to Spanish determiner marking on English-origin nouns, the masculine default is used almost exclusively, unlike the variability reported elsewhere (Dubord 2004; Smead 2000). By analyzing data from both dialect regions of Belize, this study provides insight both into the global picture of language mixing practices in Belize as well as regional patterns insofar as they are instantiated by the cohort analyzed.