Slide Slam S12 Sandbox Series
Longitudinal Structural Plasticity of the Language Network by Second Language Learning
Xuehu Wei1, Helyne Adamson1, Matthias Schwendemann1, Tomás Goucha1, Angela D. Friederici1, Alfred Anwander1; 1Max Planck Institute for Human Cognitive and Brain Sciences
Introduction: Languages of the world strongly differ from one another in all dimensions (sound, lexicon, syntax, and orthography)(Evans and Levinson, 2009), and each language relies on a particular neural network adapted to its processing demands (Ge et al., 2015; Goucha, 2019; Paulesu et al., 2000). Similar to the specialization of the brain to the characteristics of the mother tongue, efficient processing of novel structures in adult foreign language learning was previously related to brain plasticity in various gray matter and white matter regions (Bialystok et al., 2012; Li et al., 2014; Qi and Legault, 2020). Here we analyzed the longitudinal structural changes of the white matter language connectome during adult second language learning in a large and well controlled cohort. Method: We recruited 60 healthy right-handed Arabic native speakers (mean age, 25.9 years; range, 19-34) for an intensive German course (5h/day, 5days/week) over a 6-months period. We acquired high spatial resolution diffusion MRI data from each participant at the beginning (time point 0: TP0), after 3 months (TP1), and after 6 months (TP2) of language learning. Using probabilistic tractography, we computed the structural network between all the language-related areas in both hemispheres. We first analyzed overall connectivity change (sum of all weighted connections in language network) by testing the brain lateralization at each time point in a paired t-test analysis. In a longitudinal analysis, we then tested the learning-induced change of the intra- and inter-hemispheric connectivity using a Linear Mixed Effects (LME) model with each time point as fixed effects. To localize the learning-induced connectivity change across each time point to specific connections and subnetworks, we used the Network-Based R-statistic (NBR) mixed-effects models (p-threshold = 0.01, K = 3000 permutations)(Zeus Gracia-Tabuenca et al., 2020). Results: The lateralization test showed leftward lateralization of the network for the initial and the middle timepoint (TP0: Left >> Right, t= 3.11, p= 0.003; TP1: Left > Right, t = 2.04, p= 0.046; TP2: Left ≯ Right, t = 1.79, p= 0.08). The longitudinal analysis in a LME model statistic showed a significant dynamic decreased inter-hemispheric connectivity during leaning with the strongest effect in the second half of the learning period (TP0-TP1: t = -1.1, p = 0.27 (n.s.)；TP1-TP2: t = -6.2, p = 1.4e-08，TP0-TP2: t= -8.1, p =1.2e-12). Finally, the NBR showed increased intra-hemispheric connectivity in sub-networks, including the bilateral parietal-temporal system and the right IFG mainly in the second half of the learning period. Additionally, the connectivity of sub-networks including connections of orbital IFG–aSTG, parahippocampal- lateral temporal lobe and inter-hemisphere were decreased. Conclusion: The present study showed a dynamic reorganization of multiple sub-networks during second language learning. During the initial learning period, the intra-hemispheric connectivity of phonological-semantic related sub-networks increased, while subnetworks related with lexical retrieval and long-term memory showed a decrease in connectivity. In addition, we found a crucial role of the right hemisphere in second language learning and reduced transcallosal connectivity between hemispheres related to a stronger intra-hemispheric specialization of both hemispheres.