Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Learning a new language shapes the cortical encoding of linguistic information: Longitudinal evidence from Spanish native speakers learning Basque
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
Jose Pérez-Navarro1,2, Noemi Bonfiglio1,3, Jose Antonio Gonzalo-Gimeno1,3, Jiaqio Mao1,3, Giada Antonicelli1,6, Romain Pastureau1,3, Li-Chuan Ku1, Guobin Xie1,3, Mikel Lizarazu1, David Hernández-Gutiérrez4, Nicola Molinaro1,5; 1Basque Center on Cognition, Brain and Language (BCBL), 2University of Geneva, 3University of the Basque Country, 4University of Rochester, 5Ikerbasque, Basque Foundation for Science, 6University of California San Francisco
During language learning, brain mechanisms for language processing get refined to extract increasingly abstract and meaningful information from the acoustic speech signal, going from an initial acoustic processing, through phonological, towards syntactic and semantic processing. 3 recent studies on bilingual cortical encoding, across children and adults and different bilingual combinations, converge on two findings: cortical encoding of acoustic information decreases as a function of proficiency, whereas encoding of linguistic information (i.e., phonological, lexical, and semantic) strengthens (Tezcan et al., 2023; Brodbeck et al., 2024; Pérez-Navarro et al., 2024). However, these studies compared groups with different proficiency cross-sectionally; longitudinal evidence on when and how these changes occur is lacking. In this study, we ask how learning a second language shapes its cortical encoding during continuous speech listening. We use temporal response function (TRF) modeling of the magnetoencephalographic (MEG) signal to capture the time-resolved cortical encoding at different levels of the language hierarchy with milliseconds resolution (Crosse et al., 2016; Brodbeck et al., 2023). We assessed 19 adult native speakers of Spanish enrolled in formal Basque instruction, who were tested at the beginning and end of the academic year. At both time points, we measured proficiency in both languages and recorded MEG responses to continuous natural speech in each, estimating source-level TRFs with individual MRI structural maps. Critically, our TRF predictors tap into acoustic features (i.e., spectrotemporal development of the speech signal) and lexico-semantic representations (i.e., contextual surprisal and entropy extracted from multilingual MARCO-LLM; Ming et al., 2024), allowing us to differentiate whether neurocognitive resources are allocated to encoding the acoustic signal or abstract linguistic information from it as a function of proficiency. Our findings reveal significantly stronger encoding of acoustic information in Basque, the language being learned, than Spanish, the language in which participants are already proficient (reflected by larger prediction accuracy in left and right superior-temporal clusters, cluster-corrected statistics: p<.01). For lexico-semantic information (i.e., contextual surprisal), the pattern is reversed: stronger encoding in the more proficient language (left superior-temporal: p<.01). This acoustic–linguistic tradeoff aligns with the literature reviewed above. Crucially, our longitudinal design revealed a language-by-testing-time interaction in the amplitude of the response to contextual surprisal (observable in the TRF weights 180-380 ms after word onset, right superior-temporal: p=.011): while TRFs in Spanish did not change across the year, the amplitude of response to contextual surprisal in Basque increased after instruction, coming to resemble the Spanish one. Overall, we show that second language learning is tractable in the cortical encoding of linguistic information after a year of formal language instruction. This has a relevant impact on neurobiological theories of second language learning as well as on broader language processing research, by showing that even a relatively short but sustained language experience reorganizes what information the brain extracts from the speech input in order to facilitate comprehension.
Topic Areas: Multilingualism, Speech Perception