Investigating community norms and linguistic mechanisms in codeswitching: Bridging linguistic theory and psycholinguistic experimentation

Restricted (Penn State Only)
Halberstadt, Lauren Perrotti
Graduate Program:
Doctor of Philosophy
Document Type:
Date of Defense:
March 17, 2017
Committee Members:
  • Paola Dussias, Dissertation Advisor
  • Rena Torres Cacoullos, Committee Chair
  • John Lipski, Committee Member
  • Matthew Carlson, Committee Member
  • Carrie Jackson, Outside Member
  • Damian Vergara Wilson, Special Member
  • Rena Torres Cacoullos, Dissertation Advisor
  • Paola Dussias, Committee Chair
  • linguistics
  • psycholinguistics
  • sociolinguistics
  • codeswitching
  • bilingualism
  • language science
  • Spanish
  • corpus
  • eye tracking
Research into how bilinguals process codeswitched language is based largely on non-codeswitching bilinguals (i.e. Kootstra, van Hell, & Dijkstra, 2010), although there are exceptions (Guzzardo Tamargo, Valdés Kroff, & Dussias, 2016; Valdés Kroff, Dussias, Gerfen, Perrotti, & Bajo, 2016). Given links between language usage and language processing (Bybee, 2010; MacDonald, 2006), there is a glaring need to examine codeswitches in speakers who regularly engage in codeswitching. The present work helps fill this lacuna by examining both production and comprehension data by Spanish-English bilinguals for whom codeswitching is part of their linguistic repertoire. To study production, a corpus of naturalistic spoken codeswitched samples was created; to study comprehension, two experiments were conducted. Participants were recruited from Albuquerque, New Mexico, an area where bilingual communities engage regularly in Spanish-English codeswitching (Travis & Torres Cacoullos, 2013). Crucially, participants (N=16) took part in both the production and comprehension studies, providing evidence on the processing of switched language from bilingual speakers for whom codeswitching is a discourse mode. Four codeswitched syntactic structures were targeted for investigation – estar ‘be’ + gerund, hacer ‘do’ + verb, NP + verb, and pronoun + verb. In Experiment 1, participants listened to a sentence containing a target codeswitch, completed a memory load task, and then repeated the sentence. The memory load component, consisting of a film description task, added a cognitive load, under the expectation that the changes participants made to the repetitions would be more aligned to their natural linguistic tendencies (Gullberg, Indefrey, & Muysken, 2009). Innovative in this research is the additional use of the film descriptions for the creation of a speech corpus consisting of spontaneous narratives. The corpus was then analyzed to determine the real usage patterns of the participants. Crucially, the codeswitches found in the corpus were used as the source materials in Experiment 2. Participants’ eye movements were recorded while they read paragraphs aloud that contained the target codeswitched structures together with no-switch counterparts. The driving question is whether habitual codeswitchers are sensitive to variation in bilingual speech. If so, switches that are preferred in production should be more easily processed than those that are not part of the linguistic repertoire of the speakers. Indeed, codeswitches that adhered to the usage patterns found in the spontaneous speech corpus were more easily repeated in the sentence repetition task, supporting the growing need to interpret lab-based findings by accounting for natural speech data. In the eye tracking study, while codeswitches led to longer reading durations compared to their no-switch counterparts, when dysfluencies were found in production, they were associated with only two of the target syntactic structures, suggesting that codeswitching preferences are modulated by syntactic structure. The findings indicate that codeswitching cost does not manifest in perceptible difficulties in real-life language and thus the construct should be reconsidered as gradient and multi-dimensional, to account for nuanced differences in language use. Findings are consistent with the conclusion that codeswitches that are part of the linguistic repertoire of the community are less costly to the processing mechanism during comprehension. Looking ahead, the data collection protocols put forward prove to be effective in eliciting valid codeswitching data. In sum, findings across the two experiments support a usage-based approach to the study of language in which production and comprehension are indelibly linked.