Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Investigating bilingual code-switching comprehension in natural conversation using EEG
Poster Session C, Saturday, September 13, 11:00 am - 12:30 pm, Field House
This poster is part of the Sandbox Series.
James Baybas1, Dorothy Yuxi Lin1, Sharon Lu1, Jonathan Brennan1; 1University of Michigan
Bilinguals frequently engage in code-switching: Seamless alternation between languages in a given conversation without causing miscommunication (Poplack, 1980). Code-switching is governed by sociolinguistic factors such as proficiency, switching frequency, and attitudes towards switching (e.g., Salig et al., 2023). Crucially, these factors are not adequately captured by prior studies with controlled paradigms, such as lexical decision tasks and narrative-listening. Moreover, results thus far are mixed, with some studies showing a cost for processing code-switches (e.g., larger N400 effects; Yacovone et al., 2021), whereas others find no significant effect in more naturalistic contexts (Blanco-Elorrieta & Pylkkänen, 2017). To address this gap, we conducted an electroencephalography (EEG) experiment where bilingual participants actively engaged in a natural, bilingual conversation. We evaluate the potential cost of intra-sentential Mandarin-to-English code-switches within a free-flowing conversational paradigm and examine how language proficiency may modulate this cost. 30 Mandarin-English bilingual participants (Mage = 21.6) were recorded with EEG during an hour-long conversation with a Mandarin-English bilingual research assistant (RA). Most participants were L1 Mandarin, L2 English speakers. Utterances of each interlocutor were recorded with headset microphones during EEG recording. Conversations were Mandarin dominant with spontaneous English code-switches. RA speech was annotated for Mandarin-to-English code-switches; the onset and duration of intra-sentential code-switches were used to identify EEG epochs of interest. On average, we find ~70 usable code-switches per conversation (range: 40-130) after applying exclusions (e.g., proper nouns and repetitions), providing adequate data for analyzing time-locked responses. We average waveforms time-locked to code-switch onset that are analyzed relative to responses to non-switch Mandarin-only sentences. Data analysis is currently ongoing. If there is a cost associated with comprehending code-switches in natural conversation, we predict a centro-parietal negative deflection 250-500 ms post-switch onset, reflecting increased processing difficulty or prediction error derived from the matrix language (LAN/N400; Tanner & Van Hell, 2014). We also predict a late positive component (LPC) 600-1000 ms post-switch onset with a parietal scalp distribution, with a larger magnitude for participants with lower proficiency in English, suggesting reanalysis or increased integration effort (Yacovone et al., 2021). We expect an inverse correlation between the magnitude of each component and the duration of conversation as participants become more familiar with the RAs; they may begin to predict and anticipate code-switches more accurately alongside the matrix form (non-switched form; Moreno et al., 2002) We further explore multivariate decoding approaches as “cost” may not be monotonic as well as between-subject variation in terms of interlocutors’ respective linguistic backgrounds and code-switching practices. This study enhances the external validity of neurolinguistic code-switching research by utilizing a free-flowing, natural conversational paradigm. By embracing participant variability and naturalistic input, we aim to more accurately capture the neural dynamics underlying code-switching comprehension and the role that language proficiency plays in such processing. The findings of the current study may further support the development and critical analysis of naturalistic experimental paradigms that better reflect real-world bilingual language use.
Topic Areas: Multilingualism,