Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Influence of language experience on multilingual language processing as revealed by magnetoencephalography.
Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai
This poster is part of the Sandbox Series.
Abigail Licata1, Jose Perez Navarro2, Jim S. Magnuson2,3,4, Valentina Borghesani1, Nicola Molinaro2,3; 1Faculty of Psychology and Educational Sciences, University of Geneva, Geneva, Switzerland, 2BCBL, Basque Center on Cognition, Brain and Language, Donostia, Spain, 3Ikerbasque, Basque Foundation for Science, Bilbao, Spain, 4University of Connecticut, Storrs, CT, USA
How and when does multilingual experience shape the grounding of semantic categories? While recent work has found that earlier age of acquisition (AoA) was associated with greater cross-linguistic similarity in fMRI representational similarity (Nichols et al., 2021), to our knowledge, no studies have examined the influence of AoA on grounded cognition within multilinguals using this multivariate technique. Moreover, recently developed entropy measures of multilingual language proficiency (MLP) can be harnessed alongside AoA to model integrated language experience (Gullifer & Titone, 2020; Kepinska, Caballero, et al., 2023). We therefore examined whether the neural responses to semantic categories align across languages and whether earlier-learned languages show stronger sensorimotor grounding. With data collection ongoing, this unique dataset comprises Spanish-Basque-English trilinguals, who are balanced in proficiency across their early acquired languages (Spanish, Basque) but not their third and later-acquired language (English), listening to single spoken words in each language during magnetoencephalography recording. Representational similarity analysis (RSA) was used to assess cross-language alignment of neural representational dissimilarity matrices and their correspondence to a sensorimotor model derived from the Lancaster norms (Lynott et al., 2020). Exploratory regression analyses were run to assess the effect of language experience on cross-language similarity and neural-sensorimotor alignment across time. In the currently available dataset (n=16), group-level permutation-corrected analyses from 0.1-0.6s post-stimulus onset did not reveal reliable cross-language time-generalization similarity or significant neural-sensorimotor correlations. However, within a restricted time window typically associated with lexical-semantics (0.3-0.6s) and when partialing out effects of word frequency, analyses revealed evidence of positive same-time cross-language similarity for Spanish and English: same-time diagonal analysis revealed one cluster-corrected positive effect (41/60 uncorrected significant time points) and analysis of the full time-generalization matrix showed a significant positive cluster (1963/3600 uncorrected significant cells). By contrast, the other language pairs showed little or no evidence for positive similarity in this window. Pairwise tests further indicated that Spanish-English similarity was stronger than Basque-English in the partial analyses, both on the diagonal, cluster-corrected p = .028, and in the full time-generalization map, cluster-corrected p = .003. Exploratory regressions across the entire time window (0.1-0.6s) using AoA and MLP variables did not provide cluster-corrected evidence that AoA reliably modulated cross-language neural overlap. The clearest language-experience effect concerned neural-sensorimotor model alignment for Basque: higher MLP predicted stronger Basque-sensorimotor model alignment, peak β = .61 at 0.3s, uncorrected p = .01, R² = .38, cluster-corrected p = .005. Later Basque AoA showed a negative association with Basque-sensorimotor alignment, peak β = −.62 at .59s, uncorrected p = .01, R² = .39, cluster-corrected p = .01. These preliminary findings suggest that cross-language semantic alignment is weak at the group level, with the strongest current evidence limited to Spanish-English in the 0.3-0.6s time window. The results also suggest that language experience may shape access to grounded semantic features. However, given the small sample size, the regression analyses are exploratory and require more power to jointly model multiple dimensions of language experience.
Topic Areas: Multilingualism, Meaning: Lexical Semantics