Code-Switching in Context

Bilingual code-switching (CS), the alternation between two languages within a conversation, is ubiquitous as a communicative practice in Spanish-English bilingual communities, and yet often incurs comprehension difficulties or costs during lab-based psycholinguistic measures of language processing. These costs are sensitive to language-internal properties, including syntactic acceptability and constraints on speech planning. However, there is mounting evidence that the magnitude of these costs is likely overestimated; switch costs may be better understood as artifacts of a misalignment between experimental task and the naturalistic contexts in which CS occurs (i.e., linguistic ecology). When experimental contexts approximate the conditions in which CS is ordinarily encountered, habitual code-switchers exhibit a reduction in comprehension cost commensurate with a switched structure’s frequency in their bilingual input. This finding aligns with experience-based models of language processing, which propose that the distributional frequency of linguistic structures predicts the effort required for their real-time comprehension.

Leveraging Spanish-English as the target language pairing across ecologies boasting distinct sociohistorical language contact situations, the present dissertation adjudicates between syntactically and socially driven accounts of switch costs. Under a syntactically driven account, switch costs emerge from violations of grammatical constraints and should remain stable across communities sharing the same language pairing. For example, grammatically permissible switches between the Spanish auxiliary estar (‘be’) and an English progressive participle (estar + VProg; están testing, ‘are testing’) elicit less comprehension cost than impermissible switches between haber (‘have’) and an English past participle (haber + VPerf; han tested, ‘have tested’). By contrast, a socially driven account predicts that switch costs are context-dependent, reflecting community-specific patterns even across speakers of the same languages. For instance, switches between the semantically empty light verb hacer (‘do’) and an English infinitive (hacer + VInf; hacen test, ‘do test’) are productive in Northern Belize and the U.S. Southwest but generally rejected in South Florida and Puerto Rico. Because this construction is permissible under theoretical syntactic analyses, its divergent distribution across Spanish-English communities provides a critical test of how grammatical and experiential constraints jointly contribute to language processing.

The dissertation probes the breadth (i.e., community-level variation) and depth (i.e., individual experience) of predictions made by experience-based models of processing by establishing a field-to-cognition pipeline that maps ecological patterns of bilingual speech to measures of real-time comprehension. The first study analyzes Spanish-English bilingual speech corpora from Miami, Florida, and El Paso, Texas, to characterize the community-level distributions of three target structures: grammatical and attested (estar + VProg), ungrammatical and unattested (haber + VPerf), and grammatical but community-specific (hacer + VInf) switches. The second study brings these structures into a naturalistic reading paradigm with Florida bilinguals, using eye-tracking to measure costs associated with grammatical and community-specific violations. It further examines whether sensitivity to community-level distributions varies with individuals’ CS engagement, spanning incidental, ambient exposure among adult second-language learners immersed in bilingual contexts to active, habitual use among early Spanish-English bilinguals. The third study uses electroencephalography (EEG) to determine whether grammatical and community-specific violations elicit dissociable neural responses during comprehension.

The dissertation tests whether sensitivity to CS reflects grammatical constraints alone or also tracks the distribution of linguistic structures within an individual’s community and personal experience. Progressive and perfect switches are predicted to exhibit comparable distributions across Miami and El Paso, consistent with their grammatical status, whereas light verb switches are expected to be more frequent in El Paso than Miami, reflecting previously documented community-specific variation. In experimental comprehension, both grammatical and community-specific violations are expected to elicit measurable processing difficulty, reflected in longer reading times and larger late positive event-related potentials. Critically, sensitivity to community-specific violations, but not grammatical violations, is expected to vary with the nature and extent of individuals’ engagement in CS. By integrating sociolinguistic corpus analysis with behavioral and neural measures, this dissertation refines experience-based models of sentence processing to account for variation at both community and individual levels. More broadly, it establishes a replicable field-to-cognition framework for translating the multidimensional structure of bilingual experience into measurable predictors of language processing.

This work is funded by the Threadgill Doctoral Dissertation Fellowship and a Language Learning Dissertation Grant.