Codeswitching in Spanish Heritage Bilinguals

Codeswitching is commonly practiced by Spanish heritage bilinguals in the United States (Fairchild & Van Hell, 2017). Further, intrasentential codeswitching is the use of more than one language within a clause, as shown in (1) (Sankoff & Poplack, 1981: 11-12). (1) There was this guy, you know que he se montó, he started playing with congas… ‘There was this guy, you know that he got up, he started playing with congas…’ This dissertation studied the perception and production of switches between a pronominal subject and inflected verb in Spanish-English codeswitching. Two of the leading theories on codeswitching, the Minimalist Program (MP) approach (e.g., MacSwan, 2000, 2009) and the Matrix Language Frame (MLF)/4-M models (e.g., Myers-Scotton, 1993; Myers-Scotton & Jake, 2001), make conflicting predictions of the grammaticality of these codeswitches. The MP approach proposes that codeswitches after a pronominal subject are ungrammatical. For the MLF/4-M models, however, their grammaticality depends on the type of pronouns involved. These models posit that null pronouns (e.g., those used in broad focus in Spanish) are outsider late system morphemes, which are prohibited from codeswitching. Overt pronouns (e.g., English pronouns and those used in contrastive focus in Spanish) are categorized as content morphemes and can be codeswitched. To test these theoretical predictions, this dissertation examined codeswitches after pronominal subjects in broad and contrastive focus. Focus was manipulated by presentation of a drawing of one person performing two actions (broad focus) or two people each performing one action (contrastive focus). An example of an English-Spanish target stimulus is presented in (2a) and in (2b) for Spanish-English. (2) a. Alicia reads a book while she descansa en el sofá. b. Alicia lee un libro mientras ella relaxes on the couch. ‘Alicia reads a book while she relaxes on the couch.’ Participants were 55 Spanish heritage bilinguals (aged 18-24 years old; 43 female) who were frequent codeswitchers. All reported learning Spanish as a first language and demonstrated high proficiency on Spanish and English written exams. The Bilingual Language Profile (Birdsong et al., 2012) examined participants’ language dominance and portions of the Assessment of Code-Switching Experience Survey (Blackburn, 2013) assessed their codeswitching experience. A concurrent memory-loaded repetition task (adapted from Lipski, 2017) elicited production: it required participants to listen to four numbers, look at a drawing, hear a sentence, and repeat back the numbers and sentence. Written and aural acceptability judgement tasks (AJTs) measured participants’ ratings of stimuli on a seven-point Likert scale. All participants also completed these tasks in unilingual Spanish to assess each participant’s null and overt pronoun distribution, as variation in overt pronoun use in Spanish in contract with English has been documented (e.g., Shin & Montes-Alcalá, 2014). These results were used to select subgroups of participants for each task that showed the expected Spanish pronoun distribution for broad and contrastive focus contexts. Results from the production task showed that codeswitches in contrastive focus were produced significantly more accurately than those in broad focus contexts. These findings demonstrated support for the MLF/4-M models and went against the predictions of the MP approach. Results from the written and aural judgement tasks showed that the targeted codeswitches in broad focus contexts were rated significantly higher than contrastive focus contexts. These findings were not predicted by either model. Further analyses revealed that target stimuli were rated significantly higher in broad focus than contrastive focus contexts in unilingual Spanish as well, suggesting a task effect. However, target sentences were not rated at the low end of the scale (or below the fillers’ ratings), which would have shown support for the MP approach. Finally, the (linear) direction of the switch and effects of sociolinguistic variables (language dominance and codeswitching frequency) were also examined. English-Spanish codeswitches were reproduced more accurately than Spanish-English in the production task. This may have been influenced by the frequency and obligatory nature of English pronouns. Linear regressions revealed that higher English dominance scores resulted in greater accuracy at producing codeswitches at the correct site and higher ratings of target stimuli in both AJTs. Further, greater codeswitching frequency resulted in more accurate production of codeswitches at the correct site in English-Spanish sentences and higher ratings for Spanish-English codeswitches in the written AJT. In conclusion, this dissertation provided empirical evidence in support of the MLF/4-M models and contributed to research on the role that switch direction and sociolinguistic variables play in Spanish heritage bilinguals’ production and perception of codeswitches after pronominal subjects.