Proficient bilinguals often code-switch in the midst of speaking with other bilinguals, and the linguistic principles that govern the observed switches have been a focus of debate. Although code-switching performance has been analyzed primarily from the perspective of bilingual speakers, there are critical consequences for comprehension because, unlike production, which is under the control of the speaker, comprehension may be less predictable.

To investigate the relationship that exists between the frequency of occurrence of code-switches in naturalistic data (i.e., spontaneous production) and the ease with which the comprehension mechanism processes code-switches, my students and I have used eye-tracking during reading, pupillometry, and ERP methods to examine the processing of code-switched sentences. Initial studies, which received funding from a grant proposal by the National Science Foundation in 2008 (PI, Dussias; Co-PI Gerfen) show that sentences containing switches frequently found in corpora are easier to process than sentences containing switches that are infrequent. These results not only inform the debate in the monolingual literature about the relationship between comprehension and production mechanisms, but also enrich the theoretical foundation of code-switching research by providing a measure of the predictive accuracy of grammatical and psycholinguistic models, and by suggesting critical and potentially neglected variables in the study of code-switching (Valdés Kroff, J. R., & Dussias, P. E., 2023; Johns, M. A., & Dussias, P. E. 2021; Johns, M. A., Valdés Kroff, J. R., Dussias, P. E., 2019; Beatty-Martínez, A. L., & Dussias, P. E. 2017; Valdés Kroff, J. R., Dussias, P. E., Gerfen, C., & Perrotti, L., 2017; Guzzardo Tamargo, R. E., Valdés Kroff, J. R., & Dussias, P. E. 2016).

A common finding involving cases of multilingualism where a heritage language is acquired is that the heritage grammar is divergent from that of the baseline (or family) language. This divergence is thought to be the result of restricted input in the heritage language, as can be the case, for example, with heritage speakers of Spanish in the U.S., where emergent bilinguals are often exposed to the heritage language only through a small social circle of family and friends, leaving fewer opportunities for the development of the heritage language in ways that mirror the baseline language. In cases such as this, a typical finding is that the language spoken by the broader community develops as expected, while the heritage language develops in ways that deviate from the family language. At the same time, we know that abundant input in the heritage language over a long period of time can lead to more balanced bilingualism. In our lab, we have investigated linguistic knowledge among heritage speakers of Spanish who are members of the latter group. Our focus has been predominantly on the processing of the subjunctive mood because its acquisition, including the semantic and pragmatic contexts in which it is used, represents a challenge for language learners, including Spanish-speaking children. Our pupillometry studies have shown that heritage speakers experienced an increased pupillary dilation in conditions where the indicative was used but the subjunctive was expected, much like Spanish-dominant bilinguals. In addition, high-frequency governors and irregular subordinate verbs boost participants’ sensitivity to the presence of the subjunctive. In production, we find no significant differences between heritage speakers and Spanish-dominant bilinguals when producing the subjunctive with non-variable and negated governors.

In Spanish locative constructions, a different form of the copula is selected in relation to the semantic properties of the grammatical subject. Broadly speaking, sentences that locate objects require estar, while those that locate events require ser (both translated in English as ‘to be’). Using ERP methodology, we have examined whether second language (L2) speakers of Spanish are sensitive to the selectional restrictions that the different types of grammatical subjects impose on the choice of the two copulas. Our findings from a grammaticality judgment task showed that native-speaking participants correctly accepted object + estar and event + ser constructions. In addition, while ‘object + ser’ constructions were considered grossly ungrammatical, ‘event + estar’ combinations were perceived as unacceptable to a lesser degree. For these same participants, ERP recording time-locked to the onset of the critical word showed a larger P600 for the ser predicates when the subject was an object than when it was an event (*La silla es en la cocina vs. La fiesta es en la cocina). This P600 effect is consistent with syntactic repair of the defining predicate when it does not fit with the adequate semantic properties of the subject. For estar predicates (La silla está en la cocina vs. *La fiesta está en la cocina), the findings showed a central-frontal negativity between 500–700 ms. 

Grammaticality judgment data for the L2 speakers of Spanish showed that beginning learners were significantly less accurate than native speakers, and advanced L2 speakers differed from the natives in the event + ser and event + estar conditions. For the ERPs, the beginning learners did not show any effects in the time windows under analysis. However, the advanced speakers showed a pattern similar to that of native speakers: (1) a P600 response to ‘object + ser’ violation more central and frontally distributed, and (2) a central-frontal negativity between 500–700 ms for ‘event + estar’ violation. Findings for the advanced speakers suggest that behavioral methods commonly used to assess grammatical knowledge in the L2 may underestimate what L2 speakers actually know.

Research findings with monolingual speakers indicate that verbal information influences the way in which syntactic structure is built during online sentence processing. Two types of verbal information that become available immediately during processing are the verb’s subcategorization frame and the verb’s preferences or biases (e.g., Trueswell, Tanenhaus, and Kello, 1993). A verb’s subcategorization frame denotes the type of complement that is permissible or required following a verb (i.e., direct object, prepositional phrase, etc.). Verb bias, however, refers to the verb’s preferential subcategorization frame and plays a fundamental role in the case of verbs that subcategorize for more than one complement type. 

My students and I have explored whether the high cross-linguistic activation resulting from the existence of cognates in the bilingual’s two languages (Elston-Güttler, 2000; Jared & Kroll, 2001; Schwartz and Kroll, 2006; Van Hell, 1998) allows for the transfer of verbal information from the L1 system, resulting in differences between the way in which L2 readers parse L2 written input. Findings indicating that lexical information from the bilingual’s more dominant language is highly activated and used during L2 sentence processing would be congenial with constraint-based models of language comprehension because they would suggest that the parsing mechanism of bilingual speakers is permeable in the sense that information in one language affects processing in the other language.  The results of these studies have been published in Acta Psychologica (Dussias & Cramer, 2008) and Behavior and Research Methods (Dussias, Marful, Gerfen, & Bajo, 2010). Some of this work was also published in a recent article (Román, Kaan, & Dussias, 2021)