More than half of the global population is bilingual or multilingual and typically acquires and uses two or more languages in daily life [1, 2]. Regardless of the proficiency level, this lifelong language experience profoundly influences individuals’ cognitive processes and neural systems [3, 4]. Neuroimaging studies have demonstrated that bilingualism is associated with widespread structural alterations across cortical, subcortical, and cerebellar regions [5, 6, 7]. These adaptations consistently emerge in core language- and control-related cortical areas such as the left inferior frontal gyrus (IFG.L) [8, 9], inferior parietal lobule (IPL) [10], anterior cingulate cortex (ACC) [11, 12], striatum [13], and anterior cerebellum (CERE) [14]. Additionally, greater white matter integrity has been observed in tracts such as the superior longitudinal fasciculus and corpus callosum in bilingual individuals [15, 16]. However, such adaptations cannot be fully explained by the binary distinction of “bilingual vs. monolingual” but rather reflect the dynamic and continuous nature of bilingual experience [17, 18]. Recent research has emphasized deconstructing “bilingualism” into distinct experiential dimensions, including the age of second language acquisition (AoA-L2), immersion of L2 (Immersion-L2), proficiency level of L2 (PL-L2), and usage of L2 (Usage-L2), each of which modulates brain neuroplasticity in the brain in distinct ways [17, 19, 20, 21]. In particular, AoA-L2 represents a relatively static factor that constrains neuroplasticity through sensitive-period mechanisms: early bilinguals develop neural representations resembling those of their native language, whereas late bilinguals rely more on compensatory prefrontal regions. In contrast, dynamic factors such as Immersion-L2, PL-L2, and Usage-L2 continuously modulate neural adaptation, with higher proficiency linked to an increased gray matter density in the left anterior temporal lobe and immersive language use associated with the reorganization of subcortical structures. Understanding the independent and interactive effects of these factors is crucial for elucidating how bilingual experiences shape individual differences in neuroplasticity.

Accumulating morphological evidence from T1-weighted structural magnetic resonance imaging (sMRI) indicates that bilingual neuroplasticity is not uniform but is differentially shaped by distinct dimensions of bilingual experience. AoA-L2 exerts a particularly robust and enduring influence on the cortical architecture. Early bilinguals typically show increased cortical thickness or gray matter volume in frontotemporal regions such as the inferior frontal gyrus (IFG) and temporal pole [22], reflecting early-established, implicitly learned lexical-conceptual representations. In contrast, later L2 acquisition is associated with a reduced gray matter density in the left IPL [23], which is consistent with greater reliance on explicit learning and compensatory control mechanisms. Immersion-L2 and Usage-L2, in turn, represent dynamic experiential factors that continue to remodel neural systems throughout adulthood. A longer duration and greater intensity of immersion have been linked to volumetric expansion in the putamen (PUT), caudate (CAU), and anterior CERE [24, 25], regions involved in sequencing, articulatory planning, and language control. Prolonged immersion also predicts both expansion and contraction within structures such as the CAU and thalamus (THA) [17], suggesting progressive reorganization toward more efficient linguistic selection and monitoring. Parallel effects are observed for real-world L2 usage: frequent, socially embedded language use is correlated with increased volumes of the CAU, nucleus accumbens (NAc), and THA [17] and with greater cortical thickness in control-related areas, including the ACC, IPL, and superior frontal gyrus [26], consistent with adaptive engagement of control networks. PL-L2, whether achieved through immersion or formal study, further contributes to experience-dependent structural modification. Increased PL-L2 is associated with increased gray matter volumes in prefrontal, temporal, and cerebellar regions [27], as well as longitudinal gains in the IFG and anterior temporal lobe [28]. Importantly, composite indices of bilingual experience reveal both linear and nonlinear relationships with regional morphology [25], underscoring the graded and continuous nature of bilingual neuroplasticity. However, morphological alterations primarily reflect relatively stable, long-term adaptations. Understanding how bilingual experiences dynamically shape neural activity and network integration requires evidence from task-based and resting-state functional magnetic resonance imaging (fMRI).

Task-based fMRI studies provide important insights into how bilingual experience modulates the neural systems that support language processing and executive control. AoA-L2 is associated with the neural circuitry supporting phonological and syntactic processing: late bilinguals consistently show increased activation in the IFG.L, premotor cortex, fusiform gyrus (FFG), and basal ganglia during L2 production or syntactic manipulation, reflecting a greater reliance on articulatory planning and executive control mechanisms [29, 30]. PL-L2 is strongly linked to the neural efficiency of semantic processing. Low-proficiency bilinguals recruit widespread frontal and temporal regions, including the left middle frontal gyrus (MFG.L) and FFG.R, while exhibiting reduced engagement of canonical semantic regions such as the left IFG, indicating increased cognitive effort and less specialized processing pathways [31]. Beyond AoA-L2 and PL-L2, the dynamics of everyday Usage-L2 also induce rapid and reversible changes in activation profiles. Short-term reductions in exposure to a nondominant language led to increased activation in core control regions (e.g., the left pars opercularis, bilateral BA46 and BA9, and marginally the CAU) during L2 production, which is consistent with increased monitoring demands when a temporarily weakened language system is managed [32]. More recent work further demonstrated that the neural implementation of bilingual language processing is shaped by interactions between experiential factors and linguistic modality: early sign language exposure increases the recruitment of the IPL.R during sign processing, whereas modality-specific demands reliably engage the IPL.L across both early and late signers [33]. In addition, Jia [34] investigated the neural correlates linking PL-L2 to executive control, specifically inhibitory processing, in late Chinese–English bilinguals performing a Simon task. They found that higher L2 vocabulary proficiency was correlated with both superior behavioural performance and, critically, reduced neural activation in key brain regions integral to general cognitive control, including the ACC.R, INS.L, and left superior temporal gyrus (STG.L). Together, these findings indicate that task-evoked neural activation in bilinguals reflects a dynamic balance between maturational constraints, language-specific experience, and control demands.

Task-evoked activations reflect transient brain states and are therefore constrained by task-specific demands. In contrast, resting-state functional MRI (rs-fMRI) measures spontaneous BOLD fluctuations in the absence of explicit tasks, providing a privileged window into the intrinsic functional architecture of the brain shaped by long-term language experience [35, 36, 37]. From rs-fMRI time series, researchers can estimate resting-state functional connectivity (RSFC), which captures the temporal covariation of spontaneous BOLD signals between predefined regions of interest (ROIs) or voxels and thus characterizes the intrinsic activity patterns of the brain [38]. Pairwise connectivity can be represented as edges linking network nodes (ROIs or voxels), allowing the construction of a specific functional network (e.g., language, executive control, or default mode network) or a whole-brain network. Through graph theory analyses, the topological organization of functional network(s) can be further characterized using measures such as global efficiency (E_glob), local efficiency (E_loc), and modularity [39]. This network-level analysis provides a comprehensive framework for evaluating how bilingual experience reshapes large-scale brain topology, extending beyond regional coupling strengths to reveal system-level adaptations in functional organization [40, 41]. Recent studies have reported that bilingual experience is related to altered RSFC within the language, control, default mode, subcortical, and cerebellar networks in bilinguals [20, 40, 41, 42, 43, 44]. These findings suggest that bilingual experience optimizes the intrinsic coordination of distributed brain systems, reflecting enduring neuroadaptive mechanisms beyond transient task demands.

Despite the growing interest in the neuroplasticity of bilingualism, evidence regarding how different bilingual experiences reshape RSFC and the intrinsic organization of large-scale brain networks remains highly fragmented. The existing findings from different studies remain fragmented, likely because of heterogeneity in the operationalization of bilingual experience, analytical approaches, and network definitions. This fragmentation highlights the need for a systematic synthesis of rs-fMRI evidence to clarify the neural mechanisms by which bilingual experience influences intrinsic functional architecture. Hence, the present review synthesizes rs-fMRI findings on how different bilingual experiences, specifically AoA-L2, PL-L2, and Usage-L2, shape intrinsic RSFC patterns and the topological organization of the functional network(s). These experience-based factors have been selected because they represent theoretically central and empirically well-established indices of bilingual variability, capturing complementary aspects of developmental timing, attained competence, and ongoing language usage. Moreover, they are among the most consistently examined variables in neurocognitive studies of bilingualism, enabling more systematic cross-study comparison. Importantly, these factors are sufficiently represented in RSFC and brain network topology research to allow structured synthesis of brain-behavioural associations. While other experiential factors, such as immersion-L2 and exposure to L2, are also relevant, they are less consistently examined in resting-state neuroimaging studies. Accordingly, the present review selected AoA-L2, PL-L2, and Usage-L2, as these dimensions allow for theoretically meaningful and methodologically consistent comparisons across studies. We first review alterations in RSFC associated with these experience-based factors, followed by an examination of network-level findings based on graph theory, with a focus on network segregation, integration, and E_glob. Finally, we discuss methodological challenges, conceptual inconsistencies, and future research directions aimed at advancing a mechanistic understanding of bilingual neuroplasticity within the resting-state framework.

Neuroimaging evidence has indicated that AoA-L2 reshapes functional network organization[36, 40, 41, 45] (see Table 1, Ref. [17, 20, 36, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49]). Using resting-state fMRI, Berken et al. [36] examined how AoA-L2 affects RSFC in bilinguals. They found that early bilinguals exhibited stronger interhemispheric RSFC involving the language control network than late bilinguals did. Specifically, early bilingualism is associated with increased RSFC between the left and right IFG, as well as with regions involved in cognitive control, which correlates with increased neural efficiency during language processing. Conversely, late bilinguals rely more on left-lateralized circuits, suggesting that different neuroplastic adaptations depend on the timing of language learning. These findings highlight the effect of sensitive developmental periods on brain connectivity and underscore the role of early language experience in shaping neural networks for bilingual language control. This study was the first to investigate the relationship between AoA-L2 and RSFC in bilinguals. Similarly, Liu et al. [40] investigated the effect of AoA-L2 on the neural network of the language network in Cantonese–Mandarin bilinguals with high PL-L2. Early bilinguals showed more efficient and integrated neural networks characterized by higher clustering coefficients (C_w), E_glob, and E_loc, and stronger intra- and inter-RSFC within language-related modules. In graph-theoretical analyses, the C_w quantifies the extent to which a node’s neighbors are interconnected, reflecting the degree of local segregation or cohesiveness within the network [39]. The characteristic path length (L_p) represents the average shortest path between all pairs of nodes and indexes the number of intermediate steps required for information to travel across the network, thereby reflecting the overall efficiency of long-range communication. The E_glob reflects the efficiency of information transfer across the entire network, whereas the E_loc indexes the efficiency of information exchange within local neighborhoods or subnetworks. These differences persisted despite similar proficiency levels, indicating that AoA-L2 significantly influences the topological properties and modular organization of the language network. This study represents an early application of graph theory to characterize how AoA-L2 modulates the functional architecture of the language network, providing network-level evidence for age-related differences in bilingual neuroplasticity. Given that neural representations of L1 and L2 differ between early and late bilinguals [50], Liu et al. [46] further investigated how AoA-L2 shapes the intrinsic organization of both L1 and L2 networks. The results showed that early bilinguals exhibited stronger frontoparietal and frontotemporal RSFC and higher C_w, as well as increased E_glob and E_loc within the L2 network. Notably, the influence of AoA-L2 was more pronounced in the L2 network than in the L1 network, suggesting that early language exposure promotes neuroplasticity and optimized network organization in the L2 system of the brain. A complementary structural connectivity study showed stronger white matter or tract-based connectivity between left frontal regions (including the IFG) and parietal (supramarginal gyrus) and temporal (STG) cortices in early bilinguals, a pattern that is consistent with optimized L2-related pathways [51]. However, the evidence regarding the specificity of L2 versus L1 networks remains mixed and requires direct within-subject comparisons.

Accumulating evidence on RSFC suggests that AoA-L2 also modulates domain-general networks, including control, attention, default mode, visual, subcortical, and cerebellar systems. Late bilinguals exhibited increased RSFC between the PUT.L and the right rolandic operculum, as well as between the left supplementary motor area (SMA.L) and the CAU.L [42]. Additionally, AoA-L2 modulates the RSFC between the CAU.L and other control-related regions, suggesting that later AoA-L2 may lead to compensatory or more integrated control network configurations. These results highlight the dynamic interaction between AoA and functional control network organization, especially in individuals with balanced language usage. This property is manifested as efficient synergy between the core frontoparietal network and specific subcortical pathways, a characteristic that shows a robust association with the early optimization of white matter structure [52, 53, 54, 55]. Early bilingual experience strengthens frontoparietal connectivity (e.g., bilateral IFG, dorsolateral prefrontal cortex, and IPL), which is linked to superior reactive control and neural efficiency [36, 47]. This pattern suggests that functional reorganization may represent an early or sensitive marker of bilingual neuroplasticity, potentially preceding or accompanying structural adaptations [56]. However, because current evidence is cross-sectional, the temporal sequencing and mechanistic primacy of RSFC remain to be established through longitudinal investigation. Dash et al. [48] reported that earlier AoA-L2 was associated with stronger RSFC in the alerting attention network, indicating that early bilingual exposure increases intrinsic connectivity related to alertness at rest. These results suggest that early L2 acquisition may contribute to cognitive resilience by strengthening attentional control networks, thereby supporting the notion of bilingualism as a form of cognitive reserve. In another study, Gold [43] observed stronger RSFC within the default mode network in early bilinguals than in later bilinguals, suggesting that AoA-L2 may influence intrinsic processes supported by the default mode network, such as internally directed attention and memory-related integration. In contrast to earlier studies that adopted a binary classification of bilinguals (e.g., early and late bilinguals or high-proficiency and low-proficiency bilinguals), DeLuca et al. [17] considered experience-based factors as continuous variables, enabling a more fine-grained assessment of how bilingualism affects brain structure and function. They found that earlier AoA-L2 was associated with increased RSFC of the visual network involved in orthographic processing. These results suggest that early L2 acquisition may facilitate more efficient neural networks for language and literacy functions, reflecting greater neuroplasticity in sensory processing pathways. In addition, Liu et al. [44] reported that earlier AoA-L2 was associated with stronger RSFC within the subcortical network, whereas later AoA-L2 was linked to reduced intrasubcortical connectivity across regions such as the basal ganglia, amygdala (AMY), and NAc. These differences in connectivity suggested that the timing of L2 acquisition may differentially shape subcortical circuits supporting language regulation, cognitive control, and affective processing. A recent study also revealed that AoA-L2 affects functional cerebellar neuroplasticity. Liu et al. [20] observed that bilingual individuals who acquired L2 later exhibited increased cerebello-subcortical RSFC. This pattern suggests that late L2 learners engage cognitive control processes to a greater extent, reflecting higher demands for language selection and interference resolution. Overall, these findings highlight that AoA-L2 shapes the intrinsic functional organization of both domain-specific and domain-general networks, with early L2 acquisition promoting more efficient and integrated connectivity and later acquisition eliciting compensatory or resource-demanding network configurations.

Although region- and network-specific findings provide important insights into how AoA-L2 shapes distinct functional systems, a whole-brain perspective is essential for understanding how different timings of L2 acquisition influence global network topology, including segregation, integration, and efficiency across the functional connectome. Li et al. [49] demonstrated that AoA-L2 modulates the whole-brain topological organization. Compared with monolinguals, both early and late bilinguals showed increased E_glob and E_loc efficiency, indicating increased global information integration and more efficient local processing within the whole-brain functional network. Specifically, early bilinguals exhibited an intermediate level of E_loc between late bilinguals and monolinguals, indicating that acquiring an L2 later in life may shift the functional connectome toward a more efficient topological architecture, complementing the region-specific results observed in earlier studies. Notably, although Liu et al. [40] reported that early bilinguals showed more efficient and integrated networks within the language system, as indexed by higher C_w and greater E_glob and E_loc, these findings are not inconsistent with the whole-brain patterns observed by Li et al. [49]. Rather, they reflect differences in analytical scale, with Liu et al. [40] focusing on language-specific subnetworks and Li et al. [49] examining large-scale network organization across the entire brain. This distinction may indicate different patterns of neuroplasticity associated with the timing of L2 acquisition. Early bilinguals may develop more specialized and efficiently organized language-related networks due to prolonged exposure during sensitive developmental periods. In contrast, later L2 learners may rely more on distributed large-scale networks to support second-language processing. From this perspective, early acquisition may promote network specialization within language systems, whereas later acquisition may involve greater engagement of broader integrative networks. In addition, AoA-L2 is developmentally intertwined with other experiential factors, particularly proficiency and cumulative language usage. In naturalistic settings, earlier acquisition is typically associated with longer exposure duration, greater practice intensity, and higher attained proficiency. Although some studies statistically control for proficiency, such adjustments may not fully disentangle sensitive-period effects from cumulative experience. Therefore, reported AoA-related differences in RSFC should be interpreted cautiously, as they may reflect timing-dependent mechanisms, experience-dependent plasticity, or an interaction between the two.

Generally, early L2 exposure is typically associated with higher proficiency; AoA-L2 and PL-L2 are deeply interrelated dimensions of bilingual experience [57, 58]. This interdependence raises an important question for bilingual neurocognition: To what extent do neural differences attributed to AoA-L2 reflect the timing of acquisition itself, and to what extent do they instead reflect variability in PL-L2? Consequently, recent studies have begun to investigate PL-L2 as an independent predictor of functional and structural connectivity, providing a more nuanced understanding of bilingual neuroplasticity [20, 48, 59] (see Table 2, Ref. [20, 27, 41, 42, 48, 49, 59]). Similar to the influence of AoA-L2, previous studies have shown that PL-L2 is a key determinant of how the functional architecture of the brain of bilingual individuals is reorganized at rest.

Given that L2 proficiency is closely linked to the development of more efficient cognitive control, neuroimaging research has increasingly examined how PL-L2 modulates the functional and structural organization of control-related brain networks. Sun et al. [59] demonstrated that compared with low-proficiency bilingual individuals, high-proficiency bilingual individuals exhibited reduced RSFC in regions involving cognitive control, such as the ACC and MFG. Additionally, high-proficiency bilinguals exhibited shorter reaction times and superior inhibitory control, with the strength of neural connectivity negatively correlated with inhibitory performance. These findings suggest that greater L2 proficiency modulates neural efficiency within cognitive control networks, emphasizing the role of language experience in shaping the neural mechanisms underlying executive functions. In another study, Wang et al. [27] reported that higher PL-L2 was positively correlated with increased engagement of the salience network during the resting state, suggesting that enhanced cognitive flexibility is associated with L2 learning. Furthermore, during language processing tasks, individuals with greater proficiency exhibited increased activation in cognitive control regions, such as the bilateral insula (INS) and ACC, and increased activation of the FFG.R, which is crucial for the orthographic processing of Chinese characters. Structurally, higher proficiency was linked to an increased gray matter volume within a widespread corticocerebellar network, including regions involved in language and cognitive control. These findings highlight the neuroplastic potential of adult L2 acquisition and emphasize the integration of functional and structural adaptations in bilingual neural network, providing important insights into the neural mechanisms underlying the development of L2 proficiency.

Although both Wang et al. [27] and Sun et al. [59] examined the neural correlates of PL-L2 during the resting state, their findings appeared to diverge because of differences in the analytical focus and neural indices. Sun et al. [59] reported that higher proficiency was associated with weaker RSFC among key cognitive control regions such as the ACC and MFG, a pattern interpreted as increased neural efficiency. In contrast, Wang et al. [27] observed that higher PL-L2 predicted an increased amplitude of low-frequency fluctuations (ALFF), reflecting stronger low-frequency intrinsic fluctuations, in regions of the salience and visuoperceptual networks, including the anterior INS, FFG, parahippocampal gyrus, and PUT. These findings are not contradictory. Rather, they capture complementary aspects of resting-state neural dynamics. RSFC quantifies the synchronization between regions, whereas ALFF reflects the amplitude of intrinsic neural oscillations within a region, which are often linked to responsiveness, baseline excitability, or network readiness. The combined evidence suggests that with increasing proficiency, cognitive control regions may require less synchronized coupling at rest (lower RSFC), whereas salience-related and perceptual regions show greater intrinsic fluctuation strengths (higher ALFF), potentially enhancing their ability to detect salient cues and dynamically coordinate switching between networks. Wang et al. [27] explicitly proposed this dissociation, suggesting that elevated intrinsic fluctuations in salience regions may co-occur with decreased regional connectivity within the control network as L2 proficiency increases.

The effects of PL-L2 on attention, language-related control, and cerebellar functional systems have been observed in previous studies. Dash et al. [48] investigated the relationship between PL-L2 and the intrinsic functional organization of attentional networks. They found that higher L2 proficiency was associated with increased RSFC across alerting, orienting, and executive control networks, suggesting that greater language mastery enhances the functional integration of attentional networks. These findings suggest that higher L2 proficiency is associated with enhanced functional integration within attentional control networks. Given the established role of attentional and executive systems in supporting cognitive flexibility, such neural adaptations may contribute to more efficient control processes [60]. However, the implications for cognitive reserve, particularly in aging populations, remain indirect and warrant further longitudinal investigation. As previously mentioned, Sun et al. [59] primarily examined connectivity within core cognitive control regions, where reduced RSFC may indicate increased neural efficiency as language control becomes more automatized with greater proficiency. In contrast, Dash et al. [48] focused on large-scale attentional networks, where stronger connectivity may reflect enhanced functional integration supporting flexible allocation of attentional resources during language processing. These findings suggest that proficiency-related neuroplasticity may involve both increased efficiency within specialized control regions and stronger integration across broader attentional networks. Sheng et al. [41] further showed that higher PL-L2 was linked to reduced dynamic FC between the dorsal attention network and the cognitive control network, specifically in a weakly connected dynamic state (state 2). This negative association suggested that more proficient bilinguals exhibit less frequent or weaker interactions between attention- and control-related systems when these networks are already in a low-interaction state. Such a pattern may reflect more efficient, selectively engaged cross-network interactions, enabling skilled L2 users to modulate network interactions in a more specialized and context-dependent manner during language processing. Sulpizio et al. [42] found that higher L2 proficiency was associated with increased E_loc within the language-related control network, indicating enhanced functional segregation and potentially greater neural efficiency. Notably, PL-L2 interacts with other experiential factors, such as AoA-L2 and Usage-L2, to influence connectivity patterns, suggesting that L2 mastery contributes to neuroplastic changes that support more efficient language control and processing [42]. In addition, Liu et al. [20] observed that higher PL-L2 was associated with increased RSFC between the posterior cerebellar regions and the posterior cingulate gyrus. This result suggested that greater language proficiency increases the integration between the CERE and neural networks involved in language processing, reflecting proficiency-dependent neural refinement. Taken together, these findings demonstrate that PL-L2 has a wide influence on multiple functional systems, including attentional, control, and cerebellar networks. Higher L2 proficiency is consistently associated with more integrated attentional circuitry, greater segregation and efficiency within language-related control networks, and refined cerebello-cortical coupling, supporting linguistic and executive processes.

Although several studies have shown that PL-L2 contributes to the reorganization of a special functional network, the evidence at the whole-brain topological level remains mixed. On the one hand, Li et al. [49] showed that higher PL-L2 is positively correlated with increased topological efficiency, small-world properties, and strengthened connectivity within rich- and diverse-club regions. These results suggest that greater L2 proficiency facilitates more efficient and integrated brain network organization, emphasizing the importance of language proficiency as a key factor in mediating the neuroplasticity induced by bilingual experience. On the other hand, Gracia-Tabuenca et al. [61] reported no significant association between PL-L2 and global graph theory metrics such as E_glob or small-worldness, with only AoA-L2 predicting large-scale organizational features of the functional connectome. Several factors may account for these discrepancies. First, whole-brain graph-theoretical metrics are highly sensitive to analytical decisions, including node definition (atlas selection), edge construction, and thresholding strategies. Variations in these parameters, as well as differences in whether studies examine static or dynamic connectivity, can influence efficiency estimates and statistical outcomes, potentially obscuring proficiency-related effects when global metrics are computed under different modeling frameworks. Consequently, discrepancies across studies may reflect methodological variation rather than the absence of biologically meaningful neural adaptations. Second, PL-L2 effects may be more localized or network-specific (e.g., within language, attention, or control systems) than truly global, meaning that the effects of proficiency may not be sufficiently large or widespread to alter whole-brain topology unless the sample characteristics or analytical strategies amplify these effects. Third, proficiency often covaries with other experiential factors, including AoA-L2, daily language use, and immersion history. Without explicitly modeling these interrelated variables, the unique contribution of proficiency may be difficult to disentangle from earlier acquisition or broader language experience. Consequently, neural differences attributed to PL-L2 or AoA-L2 may reflect their combined influence rather than isolated effects. Finally, the global network organization may be more strongly shaped by the developmental timing (AoA-L2), whereas proficiency exerts its effect primarily through finer-scale specialization and dynamic coordination among specific functional systems. These considerations collectively highlight the need for methodologically harmonized, large-sample, and multilevel analyses to clarify whether and under what conditions PL-L2 robustly influences the whole-brain network topology.

Although AoA-L2 and PL-L2 are associated with the maturational and competence-related reorganization of resting-state functional networks, respectively, recent studies have increasingly focused on neural adaptation induced by the use of L2 (Usage-L2) in bilinguals [17, 44, 47] (Table 3, Ref. [20, 42, 44, 47]). Usage-L2 is typically operationalized through detailed self-report language history instruments, such as the Language and Social Background Questionnaire (LSBQ), which assess the frequency, context, and distribution of L2 use across home, work, and social environments. Participants are commonly asked to rate how often they use L1 and L2 in different contexts using Likert scales or to estimate daily exposure time, from which proportional usage indices or weighted composite factor scores are derived [17, 44]. Some studies further quantify usage diversity through entropy-based measures, calculating the balance and variability of L1 and L2 use across contexts [42, 47]. As a dynamic and experience-sensitive measure reflecting real-world language engagement, Usage-L2 serves as a proximal force that may continuously reshape intrinsic connectivity patterns through experience-dependent neuroplasticity.

Previous studies first investigated how everyday language practices shape intrinsic connectivity and highlighted the role of variation in L2 engagement. Gullifer et al. [47] specifically examined how social diversity in language use influences RSFC in highly proficient bilinguals. They found that greater social diversity in daily language use correlates with increased connectivity between the ACC and the PUT bilaterally, as well as an increased dependence on proactive control strategies. These findings indicated the distinct neural pathways through which static language learning history (e.g., AoA-L2) and dynamic social language practices shape neural network in bilinguals, highlighting the importance of social context in modulating neurocognitive control mechanisms. In their study, language entropy, derived from Shannon entropy, was used to quantify the social diversity of language use by capturing the distributional unpredictability of an individual language choices across social contexts. Specifically, the proportional use of each language (e.g., L1 and L2) is calculated within different social spheres, such as home, work, and broader social settings, and these proportions are incorporated into the Shannon entropy formula to generate a single value reflecting the degree of language-use diversity. Lower entropy values indicate highly predictable and compartmentalized language use in which one language dominates particular contexts, whereas higher entropy values reflect more balanced and integrated language use across contexts. Importantly, Shannon entropy captures the distributional unpredictability of language choice across contexts rather than the cognitive load or emotional intensity associated with specific interactions; for example, balanced language use in informal family settings may impose different cognitive and affective demands compared with language mixing in high-stakes professional environments. Entropy should therefore be interpreted as a structural index of interactional diversity rather than a direct measure of cognitive effort, while still providing a useful metric for examining how patterns of language-use dispersion relate to neural and cognitive aspects of bilingual experience. In another study, Sulpizio et al. [42] examined how Usage-L2 shapes the RSFC of language and control networks in bilinguals. The results showed that an interaction between Usage-L2 and AoA-L2 modulated RSFC within the control network, with higher language entropy increasing coupling between PUT.L and SMA.L, and between the CAU.L and SMA.L. Notably, the increased RSFC within the control network was more pronounced in late bilinguals. Usage-L2 also interacted with PL-L2, modulating RSFC between the THA.R and right lingual gyrus. These findings indicate that dynamic, context-sensitive Usage-L2 shapes RSFC in ways that are conditioned by other bilingual experiences.

A theoretically grounded distinction between Usage-L2 at home and Usage-L2 in social settings is essential for understanding how this experience-based factor shapes the intrinsic functional organization. According to the adaptive control hypothesis (ACH) [62], the neural adaptations associated with bilingual experience are driven by the specific interactional contexts in which the L2 is deployed. Different environments impose distinct cognitive and control demands; home settings often involve predictable, routine exchanges, whereas social contexts typically require more dynamic monitoring, conflict resolution, and language selection. Incorporating this distinction into the present review allows for a more mechanistic understanding of how experience-dependent neuroplasticity arises from context-sensitive patterns of bilingual language use. In a previous study, Liu et al. [44] observed that higher Usage-L2 at home correlates with decreased connectivity within subcortical regions, including the AMY, globus pallidus, hippocampus, and NAc, possibly reflecting neural efficiency or a reduced reliance on certain subcortical pathways. In contrast, increased Usage-L2 in social settings is positively related to subcortical–cortical functional connectivity, indicating that contextual factors of language use modulate distinct neural mechanisms. Similarly, when Usage-L2 is higher at home, it is associated with decreased cerebello-cortical FC, indicating that frequent L2 use in the home setting may lead to neural reorganization characterized by reduced connectivity between the CERE and cortical language or sensorimotor regions [20]. Reduced RSFC in predictable home environments may reflect functional specialization or automatization, but it may also indicate reduced engagement of monitoring or integrative pathways due to lower contextual demands. Without longitudinal evidence or task-based validation, it remains unclear whether such decreases represent adaptive optimization or diminished network integration. Therefore, interpretations of decreased connectivity should remain cautious and acknowledge the potential for multiple functional explanations. Conversely, increased Usage-L2 in social settings is linked to increased cerebello-cortical FC, particularly involving lobule IX and regions such as the olfactory cortex, INS, and FFG, suggesting that social language use promotes stronger cerebellar interactions with areas involved in social communication and multisensory integration. Together, these contrasting results highlight that the context of L2 usage differentially influences subcortical and cerebellar neural plasticity, with social interactions promoting connectivity related to social and multisensory processing, whereas frequent home use may be related to less engagement of these neural pathways.

Despite increasing interest, studies directly examining how Usage-L2 relates to the intrinsic organization of functional networks remain limited, with only a small number of investigations published to date and many originating from a relatively small group of research teams. Existing findings suggest that different dimensions of L2 use may shape subcortical- and cerebello-cortical RSFC in distinct ways. However, the field remains at an early stage, and the emerging patterns have not yet undergone broad independent replication across laboratories and populations. Consequently, conclusions regarding usage-based neural plasticity should be considered preliminary. Establishing robust and generalizable patterns will require larger multi-site collaborations, harmonized analytic pipelines, and cross-cultural validation to ensure that observed connectivity effects are not sample- or laboratory-specific.

A central unresolved issue in bilingualism research concerns whether neural adaptations are consistently associated with measurable behavioural performance. The studies reviewed here collectively reveal a patterned relationship rather than a simple correspondence: neuroimaging and behavioural findings converge in some domains, diverge in others, and occasionally show inverse associations. Importantly, these patterns are systematic and theoretically interpretable rather than contradictory. This issue is particularly salient given ongoing debates regarding the replicability of behavioural bilingual advantages.

As highlighted in the preceding sections of this review, recent resting-state fMRI research on bilingualism has increasingly adopted more refined, multidimensional, and methodologically integrative approaches for understanding how bilingual experience shapes intrinsic brain network. Several converging trends can be identified and are described below.

First, a growing movement toward conceptualizing bilingualism as a continuum rather than a categorical distinction exists. Instead of relying on distinctions such as monolinguals vs. bilinguals, early bilinguals vs. late bilinguals, or high-proficient bilinguals vs. low-proficient bilinguals, recent work models bilingual experiences, including language use, exposure, proficiency, and interactional demands, as continuous predictors of neural outcomes. Studies adopting this approach [17, 20, 42, 44] demonstrate that different dimensions of bilingual experience vary substantially in their predictive validity and exert dissociable effects on the brain structure and RSFC. This trend reflects a broader theoretical shift away from the comparative fallacy inherent in binary designs and toward a more ecologically valid understanding of bilingual neuroplasticity.

Second, a methodological trend in resting-state fMRI research is the adoption of multiscale analytical frameworks that capture bilingualism-related neuroplasticity across voxel-level, regional, and whole-brain network levels [63, 64]. Rather than relying solely on traditional seed-based RSFC, recent studies have increasingly incorporated comprehensive metrics, including dFC [41], ALFF [27], regional homogeneity (ReHo) [65], degree centrality (DC) [49], and graph theory indices [49, 61], to characterize distinct dimensions of intrinsic neural network related to bilingualism. These measures illuminate different facets of resting-state activity: dFC captures temporal flexibility and state-dependent dynamics of bilingual neural networks; ALFF is a measure of the intensity of spontaneous oscillations; ReHo quantifies local coherence within neighboring voxels; DC identifies highly connected hubs within the network; and graph theory metrics describe global properties such as modularity, integration, and efficiency. The increasing use of these multilevel metrics reflects a trend toward capturing the hierarchical and distributed nature of bilingual neuroplasticity. This diversification is theoretically motivated: bilingual experience induces changes that may be manifested locally, regionally, or across large-scale networks, and no single analytical approach can fully capture these multilevel adaptations. For instance, localized fluctuations may remain stable even when network-level coordination reorganizes, depending on the intensity, context, and variability of bilingual experience.

Third, an increasing number of studies integrate multimodal neuroimaging to provide a more comprehensive account of neuroplasticity in bilingual individuals. Although the present review focuses on resting-state fMRI studies, current research trends increasingly combine functional and structural modalities, such as Diffusion tensor imaging (DTI), sMRI, or magnetoencephalography (MEG), to more comprehensively characterize how bilingual experience shapes both functional neural interactions and anatomical pathways [17, 21, 27, 52, 66, 67]. This integration allows researchers to examine not only functional reorganization but also accompanying structural adaptations, providing converging or diverging evidence for experience-dependent plasticity. Multimodal approaches are particularly valuable for dissociating adaptations in cortical-subcortical or cortico-subcortico-cerebellar circuits, long-range white matter tracts, and large-scale functional networks, as well as for linking the microstructural properties of the brain to functional network dynamics. By capturing complementary aspects of neuroplasticity, such approaches enable a more mechanistic understanding of how bilingual experience influences hierarchical neural network, from local circuits to distributed systems, and support more precise modeling of structure–function relationships in the bilingual brain.

Despite growing interest in how bilingual experience shapes intrinsic functional organization, several limitations in the literature constrain theoretical integration and empirical progress. These limitations span conceptual inconsistencies, sample and methodological issues, insufficient contextual and linguistic diversity, and gaps in analytical frameworks.

In light of the limitations mentioned above, several lines of inquiry are needed to advance a more unified, mechanistic, and generalizable understanding of how bilingual experience shapes the intrinsic functional architecture. Future research should prioritize the directions described below.

This review demonstrates that bilingual experience systematically reshapes the brain’s intrinsic functional organization, with RSFC providing a sensitive window into experience-dependent neuroplasticity. Age of L2 acquisition, L2 proficiency, and language usage emerge as partially dissociable while interacting dimensions that modulate both domain-specific language networks and domain-general control, attentional, subcortical, and cerebellar systems. Importantly, the current evidence indicates that bilingual neuroplasticity cannot be captured by a single or uniform neural signature, but instead reflects dynamic reconfigurations across multiple functional networks shaped by developmental timing, achieved competence, and contextual language engagement. By integrating findings across network-specific and whole-brain perspectives, this review underscores the need to move beyond binary categorizations of bilingualism toward multidimensional, network-level, and mechanistically informed models. Such an approach will be essential for advancing a coherent understanding of how bilingual experience dynamically organizes the human brain.

YJ and LZ conceived the research and wrote the paper. YZ, YY, and NW retrieved the literature and organized the results. XL conceived the research and wrote the paper. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Not applicable.

Not applicable.

This work has received funding from the Scientific Research Foundation for Talented Scholars, Beijing Normal University (310432101 and 312200502506), Guangdong Philosophy and Social Science Foundation (GD25YJY37), the National Natural Science Foundation of China (Grant numbers: 62407005), the Humanities and Social Sciences Research Program of the Ministry of Education of China (Grant No.: 25YJC190030), and the Natural Science Foundation of Jiangxi Province (Grant No.: 20252BAC240703).

The authors declare no conflict of interest.