Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Cross-language differences in the cortical encoding of continuous speech in trilinguals: Evidence from MEG
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
This poster is part of the Sandbox Series.
Jose Pérez-Navarro1,2, José Antonio Gonzalo Gimeno1,3, Ihintza Malharin1,3, Phoebe Gaston4, Kevin S. Brown5, Emily B. Myers6, James S. Magnuson6,1,7, Nicola Molinaro1,7, Christian Brodbeck4; 1Basque Center on Cognition, Brain and Language, 2University of Geneva, 3University of the Basque Country, 4McMaster University, 5Oregon State University, 6University of Connecticut, 7Ikerbasque, Basque Foundation for Science
As we listen to speech, our brains rapidly (within ~100 ms) transform the acoustic input to linguistic representations, supporting comprehension. This mechanism, termed here cortical encoding of linguistic information, has been shown to track linguistic information beyond the acoustic speech signal, namely phonological, lexical, syntactic and semantic representations (e.g., Di Liberto et al., 2015; Brodbeck et al., 2022; Heilbron et al., 2022). Relatedly, recent studies with different bilingual combinations have shown that the link between proficiency in each language and the cortical encoding of linguistic information is not straightforward, although there are results suggestive of a tradeoff between the cortical encoding of acoustic features and higher-order linguistic features as proficiency increases (Tezcan et al., 2023; Brodbeck et al., 2024; Pérez-Navarro et al., 2024). However, to date, no study has gone beyond this “acoustic-to-linguistic” tradeoff and comprehensively tested relations between information tracked at acoustic vs. phonological, lexical, and semantic levels. That is the aim of the current study, to resolve whether proficiency is linked to the cortical encoding of linguistic information at all levels uniformly or it differentially modulates each of them. The current study tests whether and how proficiency differences in three languages are related to the cortical encoding of acoustic and these higher linguistic levels. We capitalize on a sample of Spanish-Basque-English trilingual adults (N=35). This language combination offers useful natural cross-language contrasts, e.g., Spanish-Basque share phonology, Spanish-English share syntactic structure, and Basque-English share neither. In our magnetoencephalography (MEG) experiment, participants listened to 15 minutes of speech narratives per language. This allowed us to model the source-localized MEG responses, constrained by individual MRI anatomical maps, to acoustic and linguistic speech features with encoding models. In particular, we fit temporal response function models (TRFs; Crosse et al., 2016; Brodbeck et al., 2023) of the source-localized MEG signal in response to: acoustic (operationalized through the envelope and acoustic onsets), phonological (surprisal and entropy extracted from a phoneme 5-gram model), lexical (surprisal and entropy based on word frequency), and contextual-semantic (surprisal and entropy based on Latxa-Llama large-language model next word probabilities; Etxaniz et al., 2024). We predict that, given group-level proficiency Spanish>Basque>English (BEST assessment; De Bruin et al., 2017), cortical encoding of higher-order linguistic features will scale with proficiency, while encoding of acoustic features will show the inverse pattern, extending the tradeoff reported in bilinguals. Beyond group-level effects, we test whether individual lifetime proficiency predicts within-language encoding, and whether multilingual entropy (Gullifer & Titone, 2020) — indexing how distributed vs. compartmentalized language use is — modulates encoding across languages. The trilingual design also partially dissociates which linguistic level drives cross-language effects, since differences in phonological encoding should pattern with the Spanish-Basque similarity vs. English dissimilarity, whereas syntactic effects should pattern with the Spanish-English vs. Basque differences. In addition, cross-linguistic coactivation (e.g., whether English contextual-semantic representations are active while listening to Spanish) will be tested in subsequent analyses. Together, these analyses will help clarify how language experience, both lifelong and current, shapes the cortical encoding of speech, a foundational mechanism for comprehension.
Topic Areas: Multilingualism, Speech Perception