Bilingualism Modulates White Matter–Cognition Associations: A Tract-Level Cross-Sectional and Longitudinal Analysis

Bilingualism has been proposed as a cognitively enriching experience that may influence brain structure and cognitive aging across the lifespan (Xia, Pineiro, et al., 2026; Xia, Ren, et al., 2026). Beyond structural differences, emerging evidence suggests that bilingualism may modulate the relationship between brain measures and cognitive function (Xia, Pineiro, et al., 2026). However, it remains unclear whether bilingualism alters how white matter (WM) microstructure supports cognition, particularly at the level of specific white matter tracts. White matter (WM) microstructure, indexed by diffusion tensor imaging (DTI) metrics such as fractional anisotropy (FA) and mean diffusivity (MD), plays a critical role in supporting cognitive function (Peter et al., 2025). However, it remains unclear whether bilingualism alters how tract-specific WM microstructure relates to cognition, particularly over time. The present study investigates whether bilingualism modulates the associations between tract-specific white matter (WM) microstructure (i.e., FA and MD) and cognition. Given that bilingualism contributes to cognitive reserve by allowing individuals to maintain cognitive performance despite structural declines in WM integrity (Anderson et al., 2021; Berkes et al., 2021; Gold et al., 2013), we examine these relationships both cross-sectionally and longitudinally to capture the dynamic, experience-dependent nature of structural neuroplasticity (Pliatsikas et al., 2015). Analyses are conducted at the level of individual tracts to address the localized rather than universal impact of bilingualism on the brain. Hemisphere is explicitly modelled to account for the more bilateral structural organization of language-related tracts and the recruitment of right-hemisphere areas for language and executive control systems characteristic of the bilingual brain (Gold et al., 2013; Grundy et al., 2017; Luk et al., 2020; Nichols & Joanisse, 2016; Singh et al., 2018).