Bilingualism and Large-Scale Network Segregation Across Older Adulthood

This study examines how bilingual experience impacts resting-state functional connectivity (RSFC) among the Language Network, Default Mode Network (DMN), and Multiple Demand (MD) network across the adult lifespan. Specifically, the study tests whether bilingual experience moderates age-related differences in large-scale network organization, such that greater bilingual engagement––defined here as the extent to which an individual routinely manages two co-activated linguistic systems in daily life––is associated with preserved functional specialization. Here, specialization is operationalized as network segregation, within and between language-relevant brain networks. More specifically, we assess whether bilingual experience alters age-related patterns of network segregation in the Language, DMN, and MD networks across adulthood. Network segregation is widely used as an index of functional specialization and large-scale network differentiation in the aging brain (e.g., Chan et al., 2014). It is quantified as the relative balance between within-network and between-network functional connectivity (FC). Age-related reductions in network segregation, often described as network dedifferentiation, have been posited to reflect diminished functional specificity of large-scale systems and have been linked to poorer executive functioning and broader cognitive decline across adulthood (e.g., Zhang et al., 2021; Goh, 2011; Rieck et al., 2021). In contrast, higher segregation has been linked to more efficient and specialized neural architecture, supporting coordinated processing within networks while minimizing cross-network interference (Chan et al., 2014). Segregation metrics will be computed for the DMN-Language, Language-MD, and DMN-MD network pairs by comparing average connectivity between networks to average within-network connectivity (averaged across the two networks comprising each pair). Lower values indicate greater functional segregation, whereas higher values indicate reduced segregation, reflecting increased cross-network coupling. Framed within a network organization and neural resilience perspective, the overarching goal of this study is to test whether bilingualism serves as a protective factor against age-related changes in large-scale functional architecture by supporting the maintenance of segregated network organization in systems supporting both language processing and cognitive control. Dataset and planned approaches: This project integrates RSFC and bilingualism data across multiple independent samples that differ in age distributions, geographic location (e.g., United Kingdom, Eastern United States, Macau, Czechia), and language questionnaires. Bilingual experience was assessed using two established instruments: the Language History Questionnaire (LHQ) (Li, 2020) and the Language and Social Background Questionnaire (LSBQ) (Anserson et al., 2018). Although these measures assess related constructs (e.g., language use, switching, proficiency), they differ in scale structure and response format (e.g., hours of usage vs amount of time on a 5-point scale of always to none). To avoid introducing scaling artifacts, all LHQ and LSBQ-based indices will be analyzed separately rather than combined into a single scaled metrics. For the LHQ subset, we will use the Multilingual Language Diversity (MLD) index, which captures the extent and distribution of language use across all reported languages (including L3/L4 where available), reflecting overall diversity of multilingual engagement. For the LSBQ subset, we will use the composite factor score derived from reported language use across contexts, which summarizes patterns of bilingual experience including frequency, balance, and context of use. In addition to these composite measures, we will compute language entropy following Gullifer and Titone (2018). Language entropy is an information-theoretic index that captures the balance and diversity of language use across contexts, with higher values reflecting more balanced multi-language use and lower values reflecting dominance of a single language. In the present analyses, entropy will be used as a quantitative index of the degree to which communicative environments require the active management of multiple languages. Entropy will be calculated within each project prior to aggregation to ensure comparability within measurement subsets. Finally, self-reported L1 and L2 proficiency will be standardized (z-scored) and combined across datasets as a continuous index of bilingual proficiency, allowing us to examine proficiency-related effects on network segregation (see Primary Hypothesis 1c and 2c) Monolingual participants will be retained for baseline age-related and group-level comparisons. Multilingual diversity (including L3/L4 languages where available) will be modeled explicitly in LHQ-based analyses rather than collapsed into a binary bilingual category. In addition to continuous indices of bilingual engagement, we will examine categorical distinctions (e.g., immersed vs. non-immersed bilinguals) to test whether interactional context shapes patterns of network segregation.