Highlights
-
• Monolinguals exhibit significantly higher cerebello-cortical FC than bilinguals.
-
• Acquiring L2 late is associated with increased cerebello-caudate FC.
-
• Increased Immersion-L2 is linked to decreased cerebellar FC.
-
• Using L2 at home more frequently exhibits decreased cerebello-cortical FC.
-
• Increased Usage-L2 in social settings is related to higher cerebello-cortical FC.
1. Introduction
The cerebellum, traditionally associated with motor functions, has been increasingly focused on its involvement in language processes (LeBel & D’Mello, Reference LeBel and D’Mello2023; Turker et al., Reference Turker, Kuhnke, Eickhoff, Caspers and Hartwigsen2023; Yuan et al., Reference Yuan, Li, Du, Dang, Chang, Zhang and Guo2023). The neural activation of the cerebellum was observed during different language tasks in bilinguals (Sulpizio et al., Reference Sulpizio, Del Maschio, Fedeli and Abutalebi2020b; Yuan et al., Reference Yuan, Wu, Zhang, Zhang, Chen, Ding and Guo2021). In particular, the cerebellum plays a role in phonological and grammatical learning, as well as in language control mechanisms (Abutalebi & Green, Reference Abutalebi and Green2016; De Smet et al., Reference De Smet, Paquier, Verhoeven and Mariën2013; Pliatsikas et al., Reference Pliatsikas, Johnstone and Marinis2014). According to the Adaptive Control Hypothesis (ACH) (Green & Abutalebi, Reference Green and Abutalebi2013), the cerebellum is thought to be involved in the adaptation of language control in bilinguals to effectively manage the demands of inhibiting non-target language in different conversational contexts. It seems that there has been a consensus on the role of the cerebellum relating to increased efficiency in language control (DeLuca et al., Reference DeLuca, Rothman and Pliatsikas2019, Reference DeLuca, Segaert, Mazaheri and Krott2020b; Green & Abutalebi, Reference Green and Abutalebi2013; Pliatsikas, Reference Pliatsikas2020). Indeed, compared to monolinguals, bilinguals require additional demand for language control to ensure effective communication. This may induce structural alteration of relevant brain control regions, including the cerebellum, in bilinguals, which reflects the neuroplasticity of learning and using a second language (L2) with a longitudinal effect (Pliatsikas, Reference Pliatsikas2020).
Evidence from neuroimaging studies has shown the structural difference in the cerebellum between monolinguals and bilinguals (Danylkiv & Krafnick, Reference Danylkiv and Krafnick2020; Jin et al., Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023; Schug et al., Reference Schug, Brignoni-Pérez, Jamal and Eden2022). For example, Jin et al. (Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023) found that Cantonese-Mandarin bilinguals showed larger gray matter volume (GMV) in the posterior lobe (bilateral lobules VIIIa, VIIIb and IX), the flocculonodular lobe (right lobule X), and the vermis (VIIIb and IX) compared to Mandarin monolinguals. The increased GMV in the cerebellum can be observed in different stages of learning L2, which may reflect the structural neuroplasticity to adapt in response to increased demands of language processing in bilinguals (Pliatsikas, Reference Pliatsikas2020). In addition, a recent study investigated structural differences in the cerebellum between bilingual and monolingual children (Nguyen et al., Reference Nguyen, Xu, Vaughn and Hernandez2024). The results showed that bilingual children had smaller cerebellar volumes than monolingual children, and cerebellar volume was positively associated with English vocabulary in both groups. The finding of smaller cerebellar volume in bilingual children compared to their monolingual peers may be due to a delayed peak in cerebellar development caused by increased exposure to multiple languages.
It seems that the structural change of the cerebellum was also potentially related to bilingual experiences, such as the age of L2 acquisition (AoA-L2), the immersion of L2 (Immersion-L2), the proficiency level of L2 (PL-L2) and the usage of L2 (Usage-L2) (DeLuca et al., Reference DeLuca, Rothman and Pliatsikas2019, Reference DeLuca, Segaert, Mazaheri and Krott2020b; Pliatsikas, Reference Pliatsikas2020; Wang et al., Reference Wang, Ke, Zhang, Zhou, Li and Yang2020). Previous studies have shown that increased Immersion-L2 led to the cerebellum being more responsive to experience-based restructuring, suggesting that the amount of exposure to L2 influences the structural neuroplasticity of the cerebellum in proficient bilinguals (DeLuca et al., Reference DeLuca, Rothman and Pliatsikas2019). According to the Dynamic Restructuring Model (DRM), bilingual experiences induce structural alterations in the cerebellum that evolve across three distinct stages (Pliatsikas, Reference Pliatsikas2020). In the initial stage, exposure to L2 leads to an expansion of cerebellar GMV due to the increased cognitive demands associated with language control and selection between first language (L1) and L2. During the consolidation stage, cortical expansions undergo renormalization, optimizing neural efficiency by eliminating redundant local connections. However, cerebellar GMV continues to increase, as bilinguals sustain the demands for language control and selection. In the final stage, peak efficiency may be achieved, with cerebellar structural adaptations potentially facilitating optimized bilingual language control and integration, thereby supporting more automatic and fluent L2 processing. Based on the Unifying the Bilingual Experience Trajectories (UBET) framework, structural alterations in the cerebellum resulting from prolonged bilingual experience reflect a shift toward greater efficiency in executive control (DeLuca et al., Reference DeLuca, Segaert, Mazaheri and Krott2020b). As bilinguals become more proficient in managing multiple languages, the cerebellum not only continues to support the coordination of motor functions essential for speech production but also plays a critical role in the cognitive processes underlying language control. However, it should be noted that limited immersion in an L2 environment, varying proficiency levels, and differences in daily language-switching demands may contribute to the absence of detectable cerebellar adaptations. This highlights the importance of considering the interaction between bilingual experiences and environmental context when interpreting neuroanatomical findings (DeLuca et al., Reference DeLuca, Voits, Ni, Carter, Rahman, Mazaheri and Segaert2024). While an increasing number of studies have examined the effects of bilingualism on structural alterations in the cerebellum, the functional neuroplasticity of the cerebellum modulated by bilingualism is not completely clear. Investigating differences in the intrinsic neural activity patterns of the cerebellum between monolinguals and bilinguals may reveal how distinct L1 and L2 experiences influence the functional organization of the cerebellum. Also, this may improve our understanding of the neural mechanisms through which the cerebellum interacts with cortical and subcortical regions to support bilingual language control.
Using functional magnetic resonance imaging (fMRI), previous studies have investigated the effect of bilingualism on resting-state functional connectivity (FC) of the cerebellum in bilinguals (Berken et al., Reference Berken, Chai, Chen, Gracco and Klein2016; Jin et al., Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023). In Jin et al. (Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023)’s study, they found that significantly increased FC between the right inferior posterior lobe of the cerebellar and the orbital part of the left inferior frontal gyrus (IFG) in bilinguals compared to monolinguals. Also, this functional neuroplasticity was positively correlated with the response time of L1 alphanumeric rapid automatized naming. These findings suggested that bilingualism influences the functional alteration of the cerebello-cortical circuit, particularly in the processing of the L1. In addition, Berken et al. (Reference Berken, Chai, Chen, Gracco and Klein2016) applied a seed-based analysis to examine the effect of AoA-L2 on resting-state FC. They observed stronger FC between the bilateral posterior cerebellum and IFG in simultaneous bilinguals compared to sequential bilinguals. This result suggested that early exposure to L2 induces functional neural plasticity of language-related cortical and cerebellar regions, which may potentially enhance language processing ability and cognitive function.
However, there is still a lack of studies to systemically investigate the functional neuroplasticity of the posterior cerebellum that relates to different bilingual experiences, such as AoA-L2, Immersion-L2, PL-L2 and Usage-L2. The posterior cerebellum is anatomically composed of lobules VI, CrusI, CrusII, VIIIb, VIIIa, VIIIb and IX. Recent studies revealed that these cerebellar regions were associated with relevant language processes (D’Mello et al., Reference D’Mello, Centanni, Gabrieli and Christodoulou2020; Li et al., Reference Li, Kepinska, Caballero, Zekelman, Marks, Uchikoshi and Hoeft2021; Yuan et al., Reference Yuan, Li, Du, Dang, Chang, Zhang and Guo2023). For example, the activation of the bilateral cerebellar VIIIa was observed during sentence reading (D’Mello et al., Reference D’Mello, Centanni, Gabrieli and Christodoulou2020), and the activation of left lobule VII extended to lobule VIII negatively correlated with individual reading ability (Li et al., Reference Li, Kepinska, Caballero, Zekelman, Marks, Uchikoshi and Hoeft2021). Whether bilingual experiences influence the FC of the posterior cerebellum remains unclear. Revealing the effects of bilingual experiences on the intrinsic connectivity patterns of the cerebellum may contribute to a more comprehensive understanding of how different experience-based factors shape its functional organization in bilinguals.
In this study, we aimed to investigate the effect of bilingualism on functional neuroplasticity of the cerebellum. We first examined the difference in whole-brain FC of the cerebellum between bilinguals and monolinguals. Then we constructed the cerebellar network based on intra-cerebellar FC and used the graph theory to analyze the topological properties of the cerebellar network. Graph theory has been widely used to characterize the topological properties of language functional networks in previous studies (Liu et al., Reference Liu, Tu, Wang, Jiang, Gao, Pan and Huang2017, Reference Liu, Jiao, Li, Timmer and Wang2021; Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a). The clustering coefficient (C w), characteristic path length (L w), global efficiency (E glob) and local efficiency (E loc) were used to describe the topological properties of the cerebellar network. The C w reflects the local connectivity within the cerebellum, indicating the strength of connections among neighboring nodes. The L w quantifies the average number of steps required for information transfer across the cerebellar network. The E glob and E loc measure the efficiency of information transfer between different nodes in the entire cerebellar network and within localized regions, respectively. These parameters capture both local and global network characteristics, providing a comprehensive evaluation of how bilingual experiences influence cerebellar neural organization. Given that the cerebellum plays an important role in bilingual language control and motor functions, analyzing its topological properties may help us understand its functional adaptations associated with bilingualism. Both FC and topological properties were used to reflect functional neuroplasticity of the cerebellum induced by bilingualism. Next, we focused on the effect of four bilingual experience-based factors, the AoA-L2, the Immersion-L2, the PL-L2 and the Usage-L2, on FC between each subregion of the posterior cerebellum and whole-brain regions. For the Usage-L2, we investigated its role in shaping cerebellar FC in different contexts, specifically at home and in social settings. This distinction is theoretically rooted in the ACH, which proposes that structural and functional adaptations of bilinguals are shaped by the specific context of Usage-L2 (Green & Abutalebi, Reference Green and Abutalebi2013). Lower scores on Usage-L2 at home would suggest that the home remains primarily an L1 domain, while broader social contexts serve as a (potentially) L2-dominant environment. In contrast, higher Usage-L2 in social settings may better capture a dense code-switching context. Given that Usage-L2 in different contexts may engage distinct control mechanisms, distinguishing between these two aspects of bilingual experience allows us to better isolate their potential effects on cerebellar FC and bilingualism-induced neuroplasticity. We used a seed-based analysis to calculate the cerebello-cortical and cerebello-subcortical FC, and applied a multiple regression analysis to investigate how AoA-L2, Immersion-L2, PL-L2 and Usage-L2 (at home and in social settings) influence cerebellar FC in bilinguals. Most previous studies only examined a single bilingual experience-based factor influencing structural or functional neuroplasticity (Berken et al., Reference Berken, Chai, Chen, Gracco and Klein2016; Liu et al., Reference Liu, Tu, Chen, Wang, Li, Lu and Huang2021; Mouthon et al., Reference Mouthon, Khateb, Lazeyras, Pegna, Lee-Jahnke, Lehr and Annoni2020; Tu et al., Reference Tu, Wang, Abutalebi, Jiang, Pan, Li and Huang2015; Zhao et al., Reference Zhao, Tu, Zhang, Liu, Pan, Wang and Huang2023). To better reveal the synthetical impact of bilingual experiences on functional neuroplasticity of the posterior cerebellum, another two factors, L1 spoken and the exposure to a third (or more) language, were included in the multiple regression model. In addition, previous studies have typically characterized bilingualism using qualitative classifications, categorizing individuals into discrete groups such as early versus late bilinguals based on AoA-L2 (Berken et al., Reference Berken, Chai, Chen, Gracco and Klein2016; Liu et al., Reference Liu, Tu, Wang, Jiang, Gao, Pan and Huang2017; Sheng et al., Reference Sheng, Yang, Rao, Zhang, Li, Wang and Zheng2023) or high- versus low-proficient bilinguals based on PL-L2 (Mouthon et al., Reference Mouthon, Khateb, Lazeyras, Pegna, Lee-Jahnke, Lehr and Annoni2020). However, the criteria for defining early and late bilinguals or high- and low-proficient bilinguals vary across studies, making it difficult to establish clear boundaries or critical periods for classification. This categorical approach may contribute to inconsistencies in the literature, as it overlooks the continuous and dynamic nature of bilingual experiences. To better capture the neural adaptations associated with bilingualism, the present study employed continuous measures of bilingual experiences, enabling a more precise and nuanced analysis. This methodological shift aligns with an increasing number of studies that have adopted continuous variables to examine the effects of bilingual experiences on functional and structural neuroplasticity (DeLuca et al., Reference DeLuca, Rothman, Bialystok and Pliatsikas2019; Reference DeLuca, Rothman, Bialystok and Pliatsikas2020a; Korenar et al., Reference Korenar, Treffers-Daller and Pliatsikas2023a; Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a).
2. Methods
2.1. Subjects
We accessed 52 monolingual subjects (28 females, age: mean ± SD, 22.85 ± 4.66 yrs, range: 18–40) from “The Reading Brain Project L1 Adults” dataset (https://openneuro.org/datasets/ds003974/versions/3.0.0) (Dataset1) (Follmer et al., Reference Follmer, Fang, Clariana, Meyer and Li2018; Li & Clariana, Reference Li and Clariana2019) and 56 bilingual subjects (30 females, age: mean ± SD, 25.14 ± 4.74 yrs, range: 19–38) from “The Reading Brain Project L2 Adults” dataset (https://openneuro.org/datasets/ds003988/versions/1.0.0) (Dataset2) (Table 1). Monolingual subjects (Dataset1) were recruited from Pennsylvania State Hershey Medical Center; they are all native English speakers. Bilingual subjects (Dataset2) are native speakers of Mandarin Chinese, learning English as L2. In order to investigate whether bilingual experiences induced functional neuroplasticity of the cerebellum, we future accessed 64 bilingual subjects (49 females, age: mean ± SD, 31.91 ± 7.60 yrs, range: 18–52) from the “Bilingualism and the Brain” dataset (https://openneuro.org/datasets/ds001796/versions/1.3.0) (DeLuca et al., Reference DeLuca, Rothman, Bialystok and Pliatsikas2019) (Dataset3) (Table 1). Compared to Dataset2, Dataset3 provides more detailed information on bilingual experiences. Specifically, Dataset3 includes measures such as AoA-L2, Immersion-L2, PL-L2, Usage-L2, L1 spoken and exposure to a third (or more) language, which are crucial for examining how different bilingual experiences influence functional neuroplasticity of the cerebellum. In this dataset, bilingual subjects spoke diverse L1, but they all spoke English as their L2, with varying AoA-L2 (mean ± SD, 8.31 ± 4.65 years old, range: 0–22 years old). The majority of them were born in other countries and moved to the United Kingdom at different ages. At the time of the experiment, they were all living in the United Kingdom. The Immersion-L2 was estimated for each subject (mean ± SD, 71.94 ± 73.84 months, range: 0.26–383.85) to reflect how long they had continuously lived in the United Kingdom before the experiment. All bilingual subjects self-reported as proficient and frequent users of English. The PL-L2 was assessed by using the Oxford Quick Placement Test (QPT) for each subject. The results showed that the QPT score was 53.03 ± 6.55 (63 subjects, mean ± SD, range 31–60), suggesting that they were high-intermediate to high-proficiency speakers of English. Meanwhile, they completed the Language and Social Background Questionnaire (LSBQ) to estimate the usage of known language from early childhood to the present day in a range of settings. According to the results of LSBQ, the detailed extents of Usage-L2 at home and in social settings were further derived as weighted aggregate scores by using a factor score calculator. Bilingual subjects with a higher score reflect more usage of the L2, while a lower score indicates more usage of the L1 in these two different settings. The results showed a mean score of 2.55 (SD: 5.09, range: −7.15–16.70) for Usage-L2 at home and a mean score of 51.66 for Usage-L2 in social settings (SD: 11.38, range: 10.77–74.53). In addition, several bilingual subjects (n = 33) reported knowledge of additional languages (L3) beyond their native language and English. The L3 experience was calculated as a percentage of engagement. This was based on responses to four questions related to reading, writing, speaking and listening for each language. The average language exposure of L3 is 0.13 (SD: 0.26, range: 0–1.5). This potential effect of the L3 experience was included as a nuisance covariate in subsequent analysis. For the educational level, all subjects reported they hold at least a postsecondary or diploma degree. More detail information about the subjects can be found in Supplementary Table S1 and the “Methods, Participants and Materials” section of DeLuca et al. (Reference DeLuca, Rothman, Bialystok and Pliatsikas2019)’s study.
Table 1. Descriptive statistics of demographic and linguistic measures
Note: Mean, standard deviation (SD), and range for each measure are listed.
2.2. MRI data acquisition
All MRI data from Dataset1 and Dataset2 were acquired on a 3 T Siemens MAGNETOM Prisma Fit scanner with a 64-channel phased array coil. The resting-state fMRI data were obtained using a single-shot gradient-echo echo-planner imaging (GE-EPI) sequence with the following parameters: repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, flip angle (FA) = 90°, field of view (FoV) = 240 mm × 240 mm, slice thickness = 4 mm without gap, 34 interleaved axial slices with A/P phase encoding direction and a total of 150 time points acquired in 5 mins. During the resting-state fMRI scanning, each subject was requested to look at the cross on the screen, but not think about anything. In addition, high-resolution brain structural images for each subject were acquired by using an MPRAGE sequence (TR/TE = 1,540;ms/2.34 ms, FA = 9°, FoV = 256 × 256 mm, slice thickness = 1.0 mm, voxel size = 1.0 × 1.0 × 1.0 mm3 and 176 sagittal slices covering the whole brain).
Similar to Dataset1 and Dataset2, the MRI data from Dataset3 were acquired on a 3 T Siemens MAGNETOM Prisma Fit scanner but with a 32-channel head matrix coil. The functional images were obtained using a single-shot GE-EPI sequence. Specific parameters of the sequence were as follows: TR = 1,500 ms, TE = 30 ms, FA = 66°, FoV = 192 mm × 192 mm, voxel size 2.1 × 2.1 × 2.0 mm3, 68 transversal slices with 2.0 mm slice thickness and 300 volumes acquired in 7.5 mins. Also, the field maps (gradient echo images, TE: 7.38 and 4.92 ms) were acquired for each subject. Using an MPRAGE sequence, the high-resolution brain structural images for each subject were obtained (TR/TE = 2,400 ms/2.41 ms, FA = 8°, FoV = 250 × 250 mm, slice thickness = 0.7 mm, voxel size = 0.7 × 0.7 × 0.7 mm3 and 256 sagittal slices covering the whole brain). Each subject was asked to lie quietly and open his or her eyes but not to think about anything during the whole scanning.
2.3. MRI data preprocessing
Functional imaging data were preprocessed in SPM 12 (http://www.fil.ion.ucl.ac.uk/spm) and DPABI (version 7.0, http://rfmri.org/dpabi). The preprocessing was consistent across three datasets. The first 10 volumes were discarded to allow the MRI signal to reach a steady state. Then, we performed slice timing for the remaining volumes and realigned them to the first volume for head-motion correction. The mean framewise displacement (FD) was estimated to check the head-motion information for each subject. Subsequently, we regressed out the nuisance covariates including the head-motion profiles derived from the Friston 24-parameter model, white matter signal and cerebrospinal fluid signal, and performed the signal linear detrending within each voxel in the whole brain. In the next step, all function images were spatially normalized to the standard MNI template by using DARTEL normalization and then resampled to a voxel size of 3 × 3 × 3 mm with a kernel of full-width at half maximum (FWHM) of 8 mm. Finally, the data were band-pass filtered (0.01–0.08 Hz).
2.4. Definition of regions of interest (ROIs)
We defined regions of interest (ROIs) for the cerebellum from the probabilistic MR Atlas of the human cerebellum (Diedrichsen et al., Reference Diedrichsen, Balsters, Flavell, Cussans and Ramnani2009). This cerebellar atlas was integrally extracted from the Cerebellar toolbox (SUIT, https://www.diedrichsenlab.org/imaging/suit.htm). We selected cerebellar ROIs from the probabilistic MR atlas of the human cerebellum due to its provision of an unbiased and anatomically detailed reference for assigning lobular labels. This approach minimizes localization errors compared to single-subject anatomical references. The atlas quantifies the certainty of anatomical assignments, thereby enhancing the accuracy in evaluating and integrating empirical evidence. Furthermore, it facilitates ROI analyses through maximum-probability maps, which enable efficient extraction of functional and structural data while supporting hypothesis-driven investigations without necessitating extensive multiple comparison corrections. Importantly, the atlas has been validated for its superior alignment and normalization of cerebellar structures, as cerebellum-specific normalization methods outperform whole-brain approaches. These advantages render the atlas a robust tool for ensuring anatomical accuracy and methodological consistency in cerebellar research. In total, we selected 28 cerebellar subregions, including 10 cerebellar lobules for each hemisphere (lobules I-IV, V, VI, Crus I, Crus II, VIIb, VIIIa, VIIIb, IX and X) and 8 vermis (vermis VI, Crus I, Crus II, VIIb, VIIIa, VIIIb, IX and X). All cerebellar subregions were resampled to a voxel size of 3 mm3 for subsequent analysis. The cortical ROIs were extracted from the Automated Anatomical Labelling (AAL, version 3) atlas (Rolls et al., Reference Rolls, Huang, Lin, Feng and Joliot2020). The AAL atlas includes 86 cortical regions in two hemispheres. The subcortical ROIs were extracted from the Melbourne Subcortex Atlas (Tian et al., Reference Tian, Margulies, Breakspear and Zalesky2020). In total, based on the second hierarchical scale of this atlas, 32 subcortical ROIs including the anterior and posterior hippocampus (aHIP and pHIP), globus pallidus (aGP and pGP), caudate (aCAU and pCAU), putamen (aPUT and pPUT), two subdivisions of nucleus accumbens (NA-shell and NA-core), the lateral and medial amygdala (lAMY and mAMY) and dorso- and ventroanterior and dorso- and ventroposterior thalamus (daTHA, vaTHA, dpTHA and vpTHA) for both hemispheres were selected. In contrast to previous studies that focused on specific brain regions involved in bilingual processing and control, we adopted an exploratory, data-driven approach by including all cortical and subcortical regions to calculate cerebellar FC. Rather than restricting our investigation to predefined regions based on existing bilingual models, we adopted an exploratory, data-driven approach that allowed us to systematically examine how different language experiences influence cerebello-cortical and cerebello-subcortical FC. Additionally, it facilitates the identification of novel regions functionally connected with the cerebellum in relation to bilingual experiences, complementing prior research that has primarily focused on predefined regions. Supplementary Figure S1 shows different anatomical orientations and slices of the cerebellar, cortical and subcortical regions. Supplementary Table S2 lists all cortical and subcortical regions with corresponding abbreviations.
2.5. Whole-brain FC map of the cerebellar subregions
To investigate differences in cerebellar FC between bilinguals and monolinguals, we calculated the whole-brain FC map of each cerebellar subregion with a standard seed-voxel approach for each subject in Dataset1 and Dataset2. The FC analysis was carried out by using Resting-State fMRI Data Analysis Toolkit (REST) V1.8 (https://rfmri.org/REST). Specifically, for a given cerebellar ROI, we took it as a seed region. Then we extracted the averaged time course of all voxels within this seed region and extracted the time course of each voxel in the whole brain for each subject. We estimated FC, i.e., Pearson’s correlation coefficient r, between the selected seed region and each voxel of the whole brain. After this step, we obtained a whole-brain FC map of each cerebellar ROI for each subject. Then Fisher’s r-to-z transform was used to convert the FC map to z-value maps for statistical analysis.
2.6. Constructing cerebellar network (Dataset1 and Dataset2)
We constructed the cerebellar network for each subject in Dataset1 and Dataset2. We first extracted the averaged time series of all voxels within each cerebellar ROI and then calculated Pearson’s correlation coefficient r between any two ROIs to generate the intra-cerebellar FC. These calculations generated a 28 × 28 connectivity matrix for each subject and then, this connectivity matrix was applied in the subsequent analysis. By taking all ROIs as nodes and intra-cerebellar FC as edges, we constructed the cerebellar network for each subject in this study.
2.7. Topological properties of the cerebellar network (Dataset1 and Dataset2)
The topological properties of the cerebellar network (TP cere-net) were estimated for each subject by using the GRETNA software (Wang et al., Reference Wang, Wang, Xia, Liao, Evans and He2015) (http://www.nitrc.org/projects/gretna/). For each subject, the intra-cerebellar FC matrix was initially transformed into an undirected and unweighted matrix, i.e., a binarized matrix, using a sparsity value. The sparsity value represents the ratio between the total number of edges and the maximum possible number of edges for a given network. In line with previous studies (Cao et al., Reference Cao, Zhang and Liu2022; Kim et al., Reference Kim, Criaud, Cho, Díez-Cirarda, Mihaescu, Coakeley and Strafella2017), a range of sparsity values was identified from 0.1 to 0.34 with increments of 0.01 for each subject in this study. Only positive FC was considered in the above analysis. Then, based on each binarized matrix under a specific sparsity value, we calculated TP cere-net including four global parameters: clustering coefficient (C w), characteristic path length (L w), global efficiency (E glob) and local efficiency (E loc) of the cerebellar network, as well as two nodal parameters: the nodal strength and the nodal efficiency of cerebellar regions. In order to avoid the specific selection of a sparsity value, we applied an area under the curve (AUC) approach to estimate the global and nodal parameter value within the defined threshold range. This method was widely used in investigating the topological properties of functional network based on graph theory (Liu et al., Reference Liu, Jiao, Li, Timmer and Wang2021; Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a; Wang et al., Reference Wang, Zuo and He2010). Detailed definitions and mathematical descriptions of these global and nodal parameters are listed in Supplementary Table S3.
2.8. Cerebello-cortical and cerebello-subcortical FC measure (Dataset3)
We estimated cerebello-cortical FC by using a standard ROI-wise approach for each subject in Dataset3. Here, the cortical ROIs were extracted from the Automated Anatomical Labelling (AAL, version3) atlas. As we have mentioned above (2.4. Definition of ROIs), the AAL atlas includes 86 cortical regions in two hemispheres. For each subject in Dataset3, we first extracted the averaged time series of all voxels within each posterior cerebellar ROI (including bilateral VI, CrusI, CrusII, VIIb, VIIIa, VIIIb and IX) and within each cortical ROI. Next, we estimated Pearson’s correlation coefficient r between the time series of each cerebellar ROI and each cortical ROI to obtain a cerebello-cortical FC. Finally, we obtained a dimension of 14 × 86 cerebello-cortical FC matrix for each subject. Similarly, we estimated cerebello-subcortical FC by extracting the averaged time series of all voxels within each posterior cerebellar ROI and each subcortical ROI (Melbourne Subcortex Atlas, the second hierarchical scale of total 32 ROIs, see 2.4 Definition of ROIs), and then calculated the Pearson’s correlation coefficient r between the time series of each cerebellar ROI and each subcortical ROI to obtain a cerebello-subcortical FC matrix for each subject. The dimension of the cerebello-subcortical was 14 × 32.
2.9. Statistical analyses
We applied two sample t-test to detect the difference in whole-brain FC map of cerebellar subregion between monolinguals and bilinguals. We determined the clusters showing statistic between-group difference with the following criteria: (1) significant threshold p < .05 with the false discovery rate (FDR); (2) the number of voxels in each cluster should be more than 100 voxels; (3) peak voxel of the cluster is located in the gray matter. In a single statistical test, the standard significance threshold of p < .05 is commonly used to control the false-positive (Type-I) error rate at 5% (Bennett et al., Reference Bennett, Wolford and Miller2009; Lieberman & Cunningham, Reference Lieberman and Cunningham2009). However, in neuroimaging research, statistical tests are conducted across thousands of voxels throughout the whole brain, raising concerns about multiple comparisons. Since each voxel undergoes a separate statistical test, applying a standard p < .05 threshold without correction could lead to an inflated false-positive rate across the brain. To address this, we applied FDR correction to control for Type-I errors. In addition, we used a nonparametric permutation t-test to determine the difference in TP cere-net between monolinguals and bilinguals. Briefly, for a given parameter (such as C w, L w, E glob, E loc, the nodal strength or the nodal efficiency), we randomly paired its values between monolinguals and bilinguals to generate two new groups. Next, we recalculated the mean value of two new groups and estimated their difference. This permutation was repeated 5,000 times to obtain the empirical distribution of the difference between paired new groups. We then selected a significant level at p < .05 to determine the significant difference between monolinguals and bilinguals at 95% of the empirical distribution in a two-tailed test.
We applied a multiple regression analysis to investigate how the AoA-L2, Immersion-L2, PL-L2 and Usage-L2 (at home and in social settings) modulate the functional neuroplasticity of the cerebellum. We first examined whether four factors of bilingual experiences exhibited multicollinearity by calculating the variance inflation factor (VIF). The result showed that VIFs <5, which means that there is no multicollinearity (Johnston, Reference Johnston1984). Then, we estimated two kinds of models by applying the multiple regression analysis to examine the effects of bilingual experiences on the cerebello-cortical FC (Model 1) and the cerebello-subcortical FC (Model 2). The AoA-L2, Immersion-L2, PL-L2 or Usage-L2 (at home or in social settings) were considered as independent variables; the FC was considered as the dependent variable; and age, sex, L1 spoken, any continued exposure to a third (or more) language and the mean FD parameter of head-motion were run as nuisance covariates in each model. Each bilingual experience (AoA-L2, Immersion-L2, PL-L2 or Usage-L2) was analyzed individually, controlling for the effects of the other factors and nuisance covariates in the multiple regression analysis. The permutation test was used to examine the statistical significance of each model and whether there is a significant main effect of each bilingual experience on cerebellar FC. A statistical test for the main effect will be performed only if the model reaches statistical significance. First, we used the real data of all subjects in the multiple regression analysis to calculate the F-value (F 0) and the t-value (t 0). Then we randomly shuffled the data within the independent variable, the nuisance covariates and the dependent variable across all subjects in each model, and repeated this permutation 5,000 times in the multiple regression analysis to generate the permuted F-value and t-value. We tallied the number of times (SF and St) when the permuted F-value and t-value exceeded the original F 0 and t 0. We computed p-values (p = SF/5000 and p = St/5000) to assess the statistical significance of each model and the main effect of bilingual experience on FC, with a significance level set at p < 0.05 for both. The multiple regression analysis was performed by using the regress() function in the MATLAB software (https://www.mathworks.com/products/matlab.html). Figure 1 shows the flowchart of analysis in this study.
Figure 1. The flowchart for analyzing the functional neuroplasticity of the cerebellum in this study.
3. Results
3.1. Whole-brain FC map of the cerebellum
We examined the difference in whole-brain FC of different cerebellar subregions between monolingual individuals and bilingual individuals. Statistical analysis revealed that monolinguals exhibited significantly higher FC compared to bilinguals (p < 0.05, FDR correction) (Figure 2A and Table 2). More specifically, monolinguals showed significantly higher FC between the right cerebellar lobules V, VI and VIIIa and the right superior temporal gyrus (STG.R) compared to bilinguals. Furthermore, monolinguals also demonstrated significantly higher FC between the left lobule IX, right lobules V and VIIIb and several cortical regions involved in language functions, such as the left inferior frontal gyrus (opercular part, IFGoperc.L), middle frontal gyrus (MFG.L), STG.L and middle temporal gyrus (MTG.L), in comparison to bilinguals. Regarding the bilateral lobules VIIIa, VIIIb, right lobules V, IX and X, these cerebellar subregions exhibited significantly higher FC with the right postcentral gyrus (PoCG.R).
Figure 2. Significant difference in resting-state functional connectivity (FC) map of the cerebellum (A) and topological properties of the cerebellar network (B) between monolinguals and bilinguals. Significant effect of bilingual experiences (AoA-L2, orange; Immerison-L2, yellow; PL-L2, purple; Usage-L2 at home, green; Usage-L2 in social settings, blue) on cerebellar FC (C). The bar corresponds to the mean value and the error bar to the standard deviation. Abbreviations: AoA-L2, age of second language acquisition; PL-L2, proficiency level of second language; C w, Clustering coefficient; L w, characteristic path length; E glob, global efficiency; E loc, local efficiency; L, left; R, right. The corresponding abbreviation and full name of cortical and subcortical regions are listed in Tables 2–4.
Table 2. Cluster locations and peak coordinates corresponding to the resting-state functional connectivity (FC) based on cerebellar region
Note: The statistical significance was set at p < .05 (FDR correction).
Abbreviations: L, left; R, right; STG, superior temporal gyrus; MTG, middle temporal gyrus; PoCG, postcentral gyrus; PCUN, precuneus; PCL, paracentral lobule; CUN, Cuneus; PUT, putamen; ROL, rolandic operculum; PreCG, precentral gyrus; SMA, supplementary motor area; FFG, fusiform gyrus; MFG, middle frontal gyrus; IFGoperc, inferior frontal gyrus (opercular part); ITG, inferior temporal gyrus; SOG, superior occipital gyrus; IPG, inferior parietal gyrus.
Table 3. Size effects (β), t- and p-value of AoA-L2, Immersion-L2, and PL-L2 are reported for each significant result (p < .05, permutation test) in the cerebello-cortical FC
Abbreviations: L, left; R, right; PreCG, precentral gyrus; FFG, fusiform gyrus; STG, superior temporal gyrus; ANG, angular gyrus; ROL, rolandic operculum; PCC, posterior cingulate gyrus; OFCant, anterior orbitofrontal cortex; OFCpost, posterior orbitofrontal cortex; OFClat, lateral orbitofrontal cortex; IFGorb, inferior frontal gyrus (orbital part); LING, lingual gyrus; MFG, middle frontal gyrus; IPG, inferior parietal gyrus; HES, Heschel’s gyrus; IFGtriang, inferior frontal gyrus (triangular part); CAL, calcarine fissure and surrounding cortex; SMG, supramarginal gyrus; TPOmid, temporal pole (middle temporal gyrus); TPOsup, temporal pole (superior temporal gyrus).
Table 4. Size effects (β), t- and p-value of Usage-L2 at home and in social settings are reported for each significant result (p < .05, permutation test) in the cerebello-cortical FC
Abbreviations: L, left; R, right; PreCG, precentral gyrus; FFG, fusiform gyrus; PHG, parahippocampal gyrus; TPOmid, temporal pole (middle temporal gyrus); INS, insula; ANG, angular gyrus; MOG, middle occipital gyrus; ROL, rolandic operculum; OLF, olfactory cortex.
3.2. Topological properties of the cerebellar network
Figure 2B and Supplementary Table S4 show significant differences in global and nodal parameters of the cerebellar networks between monolinguals and bilinguals. Statistical analysis (p < .05, 5,000 permutations) revealed a significantly higher global efficiency (p = .001) in monolinguals compared to bilinguals. In addition, bilinguals demonstrated significantly higher characteristic path length (p = .002) than monolinguals. No significant between-group difference in clustering coefficient (p = .189) and local efficiency (p = .065) was found. For nodal parameters, we found in monolinguals, the nodal strength was significantly higher in the cerebellar lobules V.R, VIIIa.L and bilateral VIIIb, but lower in the CrusII.R compared to bilinguals.
3.3. AoA-L2 and functional neuroplasticity of the cerebellum
Bilinguals who learned L2 earlier (i.e., lower value of AoA-L2) showed an increase in cerebello-cortical FC involving the lobules VI.R, CrusI.R and the right precentral gyrus (PreCG.R); the lobule VIIIa and the right fusiform (FFG.R); the lobule VIIIb and the STG.L; and the lobule IX.R and the right rolandic operculum (ROL.R). However, early bilinguals showed decreased cerebello-cortical FC between the bilateral lobules IX and the angular (ANG) and between the lobule IX.R and the right posterior cingulate gyrus (PCG.R). For the cerebello-subcortical FC, an increase in AoA-L2 was associated with an increase in FC involving the lobule VI.L and the dorsoanterior part of the right thalamus (daTHA.R), the lobule CrusII.R and the posterior part of the right caudate (pCAU.R). In addition, late bilinguals (i.e., higher value of AoA-L2) showed increased cerebello-subcortical FC between bilateral IX and the anterior part of the left caudate (aCAU.L) and the daTHA. Figure 2C and Table 3 show the results of the relationship between the cerebello-cortical FC and the AoA-L2. The results of the relationship between the cerebello-subcortical FC and the AoA-L2 are shown in Table 5.
Table 5. Size effects (β), t- and p-value of bilingual experiences are reported for each significant result (p < .05, permutation test)
Abbreviations: L, left; R, right; daTHA, dorsoanterior thalamus; pCAU, posterior caudate; aCAU, anterior caudate; vaTHA, ventroanterior thalamus; pHIP, posterior hippocampus; pGP, posterior globus pallidus; pPUT, posterior putamen; aGP, anterior globus pallidus; mAMY, medial amygdala; aPUT, anterior putamen; dpTHA, dorsoposterior thalamus; vpTHA, Ventroposterior thalamus.
3.4. Immersion-L2 and functional neuroplasticity of the cerebellum
We found that decreased cerebellar FC was associated with increased Immersion-L2 in bilinguals (Figure 2C and Table 3). In particular, decreased cerebello-cortical FC involves the cerebellar lobules VI, VIIb.L, VIIIa, VIIIb and IX.R and different parts of the orbitofrontal cortex (OFC) (anterior, posterior and lateral: OFCant, OFCpost and OFClat) in bilinguals. Decreased cerebello-subcortical FC between the lobules VIIb, VIIIa.R, VIIIb.R and the ANG.L was related to increased Immersion-L2. Moreover, with Immersion-L2 increased, bilinguals exhibited decreased cerebello-subcortical FC involving the lobules VIIb, VIIIa, VIIIb, and the posterior part of the hippocampus (pHIP). Decreased cerebello-subcortical FC between the lobules CrusI.L, VIIb.L, VIIIa.L, VIIIb, IX.R and the posterior part of the globus pallidus (pGP) also was related to increased Immersion-L2 in bilinguals (Table 5).
3.5. PL-L2 and functional neuroplasticity of the cerebellum
Increased cerebellar FC was associated with higher PL-L2 in bilinguals (Figure 2C and Table 3). Specifically, as PL-L2 increased, bilinguals exhibited increased cerebello-cortical FC between the lobule VI.L and the bilateral posterior cingulate cortex (PCC), as well as between the lobule VIIb.L and the PCC.R. Moreover, higher PL-L2 was related to stronger cerebello-cortical FC between bilateral lobules IX and the bilateral OFCpost and the left temporal pole (superior temporal gyrus, TPOsup). In addition to these cerebello-cortical connectivity patterns, higher PL-L2 was also associated with stronger cerebello-subcortical FC. For example, bilinguals with higher PL-L2 showed increased FC between the lobules CrusI.L and VIIb.R and the left medial amygdala (mAMY.L), as well as between bilateral lobules IX and the bilateral daTHA (Table 5).
3.6. Usage-L2 and functional neuroplasticity of the cerebellum
Decreased cerebello-cortical FC was related to increased Usage-L2 at home in bilinguals (Figure 2C and Table 4). For example, decreased cerebello-cortical FC between the lobule IX.R and the left insula (INS.L), PreCG.R and left TPO (middle temporal gyrus, TPOmid.L) were found in bilinguals who used L2 more frequently at home. Similarly, decreased cerebello-cortical FC between the lobules VI.R and CrusI and the PreCG.R, and between the lobules VI.R, VIIb.L and VIIIa and the FFG.R were associated with increased Usage-L2 at home. For those bilinguals who used L2 less at home, we found increased cerebello-subcortical FC between the lobule IX and the mAMY.L (Table 5).
We found the Usage-L2 in social settings modulated the cerebello-cortical FC in bilinguals (Figure 2C and Table 4). That is, bilinguals who used L2 more frequently during social communications were related to increased cerebello-cortical FC between the lobule IX.R and the ROL.L, right olfactory cortex (OLF.R), INS.L and FFG.R. Also, increased Usage-L2 in social settings was associated with increased cerebello-cortical FC between the lobule VIIIb.L and the right middle occipital gyrus (MOG.R). However, increased cerebello-cortical FC involving the lobules VIIIa.R and IX.L and the ANG.R was found in bilinguals with decreased Usage-L2 in social settings. In addition, increased cerebello-subcortical FC between the lobule IX.L and the bilateral mAMY was related to increased Usage-L2 in social settings (Table 5).
4. Discussion
In this study, we investigated the difference in cerebellar FC between bilinguals and monolinguals, then examined the effect of AoA-L2, Immersion-L2, PL-L2 and Usage-L2 on functional neuroplasticity of the posterior cerebellum in bilinguals. We found stronger cerebellar FC between the right cerebellar lobules including V, VI and VIIIa and the temporal lobe region, in monolinguals compared to bilinguals. Also, monolinguals exhibited stronger cerebellar FC than bilinguals between the lobules V, VIIIb and IX, and several cortical language regions. These results reflected different connectivity patterns of the cerebellum between bilinguals and monolinguals. For the bilingual experiences, we found that AoA-L2 mainly modulated cerebello-cortical FC involving the lobules VI.R, CrusI.R, VIIIa.R, VIIIb.L and bilateral IX. Cerebellar FC between the VI.L, CrusII.R and bilateral IX and several subcortical regions was positively associated with AoA-L2. Bilinguals with increased Immersion-L2 were related to decreased cerebellar FC between the lobules VI, VIIb.L, VIII and IX.R and different parts of the OFC. Higher PL-L2 was associated with stronger cerebellar FC between the lobules VI and VIIb and the PCC. For the Usage-L2, we found that bilinguals who used L2 more frequently at home showed decreased cerebellar FC involving the lobules VI.R, VIIb.L, bilateral VIIIa and the FFG.R. In addition, Usage-L2 in social settings modulated the cerebellar FC between the lobule IX.R and the INS.L. Our findings suggested that AoA-L2, Immersion-L2, PL-L2 and Usage-L2 influence the functional neuroplasticity of the posterior cerebellum in different ways.
4.1. Difference in cerebellar FC and topological properties of the cerebellar network between monolinguals and bilinguals
We found significantly stronger cerebellar FC between the right cerebellar lobule VI and right STG, and between right lobule VIIIa and bilateral STG in monolinguals compared to bilinguals. These results reflected that different language experiences modulated cerebello-cortical FC. The lobules VI and VIIIa have been implicated in language-related functions. For example, a previous meta-analysis study showed that the right lobule VI was related to phonological processing during English tasks (Tan et al., Reference Tan, Laird, Li and Fox2005). For the right lobule VIIIa, its greater activation was observed in the semantic processing task (D’Mello et al., Reference D’Mello, Centanni, Gabrieli and Christodoulou2020; Wu et al., Reference Wu, Ho and Chen2012). In addition, both lobules VI and VIIIa are involved in verbal working memory tasks but play two distinct roles; the lobule VI is involved in articulatory rehearsal, while lobule VIIIa is implicated in the maintenance and storage of information (Chen & Desmond, Reference Chen and Desmond2005; Mariën et al., Reference Mariën, Ackermann, Adamaszek, Barwood, Beaton, Desmond and Ziegler2014). The STG plays a fundamental role in different aspects of speech processing, such as acoustic-phonetic analysis, prosodic interpretation and integration of multimodal input for speech comprehension (Yi et al., Reference Yi, Leonard and Chang2019). Additionally, the STG is essential for encoding temporal cues in speech signals and generating context-dependent phonological representations for coherent perception of syllables, words and phrases. It has been shown that the lack of using spoken language may induce decreased FC involving the STG. Accordingly, the intrinsic FC between the lobules VI and VIIIa and the STG may be related to phonological and semantic processing during speech expression. Previous studies have shown differences in neural activation involving these relevant language processes between monolinguals and bilinguals (Brice et al., Reference Brice, Salnaitis and MacPherson2023; Kovelman et al., Reference Kovelman, Baker and Petitto2008; Marian et al., Reference Marian, Chabal, Bartolotti, Bradley and Hernandez2014). In line with these previous findings, our results further suggested that different L1 and L2 experiences modulated intrinsic connectivity patterns of cerebellar and cortical regions relating to phonological and semantic processing.
Compared to monolinguals, bilinguals showed significantly weaker FC between the lobules VIIIb.R and IX.L and the IFGoperc.L. The VIIIb and IX have been traditionally linked to balance control and response to tactile stimulation (Bushara et al., Reference Bushara, Wheat, Khan, Mock, Turski, Sorenson and Brooks2001; Koppelmans et al., Reference Koppelmans, Hoogendam, Hirsiger, Mérillat, Jäncke and Seidler2017). However, the structural study found that increased GMV of cerebellar lobules VIII (consisting of VIIIa and VIIIb) and IX were related to better cognitive performance in vocabulary, reading, working memory and set-shifting (Moore et al., Reference Moore, D’Mello, McGrath and Stoodley2017). In addition, the GMV of cerebellar lobule VIIIb showed a negative correlation with the language composite scores (D’Mello et al., Reference D’Mello, Moore, Crocetti, Mostofsky and Stoodley2016). A recent structural study found that significant difference in GMV of the VIIIb and IX between monolinguals and bilinguals suggested the effect of bilingual experiences on the structural neuroplasticity of these cerebellar regions (Jin et al., Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023; Schug et al., Reference Schug, Brignoni-Pérez, Jamal and Eden2022). These results reflected that the lobules VIIIb and IX are not only implicated in motor functions but also in relevant language functions. For the IFGoperc.L, it plays an important role in language processes. Cytoarchitecturally, this region is known as Brodmann area 44 (BA44), which is a posterior part of Broca’s area involving speech production. The IFGoperc is also related to recognizing the tone of voice during spoken in L1 (Schremm et al., Reference Schremm, Novén, Horne, Söderström, van Westen and Roll2018). Stronger cerebello-cortical FC between lobules VIIIb.R and IX.L and IFGoperc.L may reflect that monolinguals are more sensitive to spoken in L1 than bilinguals. These results were similar to a previous study showing stronger FC between the dorsal anterior cingulate cortex (dACC) and other cortical spoken regions (STG.L and ROL.L) in monolinguals compared to bilinguals (Li et al., Reference Li, Abutalebi, Zou, Yan, Liu, Feng and Ding2015).
We observed significantly higher E glob in the cerebellar network for monolinguals compared to bilinguals. This finding was similar to previous findings indicating higher E glob in monolinguals compared to bilinguals (Amoruso et al., Reference Amoruso, García, Pusil, Timofeeva, Quiñones and Carreiras2024), suggesting that bilingual experience consistently influences brain structure and function. Monolinguals typically rely on a more unified cerebellar network dedicated to a single language, facilitating more efficient global integration of cerebellar regions. In contrast, bilinguals exhibit different neural activation of the cerebellum for L1 and L2 processing (Pillai et al., Reference Pillai, Allison, Sethuraman, Araque, Thiruvaiyaru, Ison and Lavin2004), which may result in a more complex and less integrated cerebellar network. The increased specialization of neural activation for two languages may induce a trade-off, where bilinguals exhibit stronger internal modular organization but at the cost of the E glob of the cerebellar network. Similar patterns have been observed in training-induced neuroplasticity, where long-term expertise leads to more selective and specialized neural organization. For example, elite gymnasts and professional dancers showed reduced global topological properties, reflecting increased automaticity and efficiency within task-relevant networks (Amoruso et al., Reference Amoruso, Pusil, García and Ibañez2022; Wang et al., Reference Wang, Lu, Fan, Wen, Zhang, Wang and Huang2016). Similarly, in bilinguals, the demand for processing two languages may drive a reorganization of cerebellar connectivity, leading to decreased global integration of the cerebellar network (Amoruso et al., Reference Amoruso, García, Pusil, Timofeeva, Quiñones and Carreiras2024).
4.2. Cerebellar FC and AoA-L2
We found cerebello-cortical FC involving subregions of the posterior cerebellum and several cortical regions that was negatively associated with the AoA-L2. For example, early bilinguals exhibited stronger FC between the lobules VI.R, CrusI.R and the PreCG.R. Previous studies employing fluency-based tasks, such as verb generation, verbal fluency and verbal working memory (Desmond et al., Reference Desmond, Gabrieli, Wagner, Ginier and Glover1997; Frings et al., Reference Frings, Dimitrova, Schorn, Elles, Hein-Kropp, Gizewski and Timmann2006; McDermott et al., Reference McDermott, Petersen, Watson and Ojemann2003; Stoodley & Schmahmann, Reference Stoodley and Schmahmann2009; Vias & Dick, Reference Vias and Dick2017), have identified the involvement of the VI.R and CrusI.R in phonological and semantic processing, while the PreCG.R has been implicated in language switching relative to single-language production (Luk et al., Reference Luk, Green, Abutalebi and Grady2012; Yuan et al., Reference Yuan, Wu, Zhang, Zhang, Chen, Ding and Guo2021). During early bilingual acquisition, individuals frequently switch between the phonological and semantic representations of two languages. This continuous linguistic alternation may impose greater demands on neural resources, which induces stronger FC between these cerebellar and cortical regions. These results were similar to a previous study that found significantly higher intra-FC of “phonological module” and “semantic module” in early bilinguals than in late bilinguals (Liu et al., Reference Liu, Tu, Wang, Jiang, Gao, Pan and Huang2017). Combined with these findings, the observed stronger cerebello-cortical FC between the VI.R and CrusI.R and PreCG.R in early bilinguals may reflect the neural coordination required for phonological and semantic processing. This increased cerebellar FC may facilitate the rapid formation of phonological-semantic associations, which is crucial for new language learning in early bilinguals. Martin et al. (Reference Martin, Strijkers, Santesteban, Escera, Hartsuiker and Costa2013) found that early bilingual learners of an L3 exhibited different neural activation patterns compared to late bilingual L3 learners, which may underlie some of the advantages early bilinguals demonstrate in learning a new language. In addition, a previous study has shown stronger FC between the cerebellum and the IFG in simultaneous bilinguals (who acquire L1 and L2 from birth) compared to sequential bilinguals (who acquire L2 after age 5) (Berken et al., Reference Berken, Chai, Chen, Gracco and Klein2016). Early bilinguals exhibit higher dynamic FC between the cerebellum and the orbital part of the IFG compared to late bilinguals (Liu et al., Reference Liu, Tu, Chen, Zhong, Niu, Zhao and Huang2020). These findings highlight the potential impact of early bilingual experience on cerebello-prefrontal FC. However, previous studies have not specified which subregion of the cerebellum is involved in neural pathways related to AoA-L2 in bilinguals. In this study, we examined the effect of AoA-L2 on cerebello-cortical FC between each subregion of the posterior cerebellum and cortical regions. Our findings indicate that AoA-L2 modulates the cerebello-cortical neural circuit, involving not only the IFG but also the posterior part of the frontal lobe (i.e., PreCG). These results enhance our understanding of how AoA-L2 shapes the neural coupling between the posterior cerebellum and language-related cortical regions.
In addition, we found that cerebello-subcortical FC between bilateral IX and the aCAU was positively associated with AoA-L2. The cerebellar lobule IX is part of the default mode network (DMN) (Diedrichsen et al., Reference Diedrichsen, King, Hernandez-Castillo, Sereno and Ivry2019). The DMN plays a crucial role in complex cognitive functions such as memory, abstract thought and self-referential processing, while typically exhibiting reduced activity during attention-demanding tasks (Smallwood et al., Reference Smallwood, Bernhardt, Leech, Bzdok, Jefferies and Margulies2021). A previous study found higher connectivity within the DMN in late bilinguals compared to early bilinguals, which suggests that late bilinguals require greater cognitive effort for L2 learning than early bilinguals (Gold, Reference Gold2018). The caudate has been implicated in both cognitive and language control (Grahn et al., Reference Grahn, Parkinson and Owen2008; Green & Abutalebi, Reference Green and Abutalebi2013). It contributes to the selection of the appropriate language and inhibition of irrelevant ones in bilinguals. Its involvement in managing lexico-semantic sets based on task demands indicates its role in overarching task-level control in language processing (Green & Abutalebi, Reference Green and Abutalebi2013). Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a found that the modulation of FC between the CAU.L and the right cerebellum emerges from the AoA-L2 by PL-L2 interaction, showing that the earlier individuals become bilinguals, the more a higher level of proficiency is associated with weaker cerebello-caudate FC. These findings suggested that early proficient bilinguals may have an enhanced ability to minimize cross-linguistic interference efficiently. In this study, we found stronger FC between the lobule IX and the aCAU, which may reflect the increased cognitive demands of L2 learning and the greater involvement of cognitive control required for language selection and interference inhibition in late bilinguals.
4.3. Cerebellar FC and Immersion-L2
We found that cerebellar FC was negatively associated with Immersion-L2 in bilinguals. Previous studies have reported structural differences in the cerebellum between bilinguals and monolinguals, with several studies indicating greater GMV and density in bilinguals, suggesting adaptations related to the increased cognitive demands of managing two languages (Jin et al., Reference Jin, Fan, Xu, Pan, Jing, Song and Li2023; Schug et al., Reference Schug, Brignoni-Pérez, Jamal and Eden2022). According to the DRM, cerebellar GMV is proposed to consistently increase across three stages of bilingual experience: initial exposure, consolidation and peak efficiency (Pliatsikas, Reference Pliatsikas2020). During the second stage (consolidation), cortical expansions are reversed and renormalized, eliminating redundant local connections and optimizing lexical learning and control. This process leads to more efficient neural circuits for language processing (Marin-Marin et al., Reference Marin-Marin, Costumero, Ávila and Pliatsikas2022). In our study, we observed decreased cerebello-cortical FC between cerebellar lobules VI, VIIb.L, VIIIa, VIIIb and IX.R, and the OFC, which was negatively associated with increased Immersion-L2. This may reflect the cortical renormalization process that induces decreased cerebellar-cortical FC. The OFC, particularly the lateral OFC (OFClat), has strong anatomical and functional connections with key language regions, including Broca’s area (BA 44/45) within the IFG, as well as the MTG and STG, which support semantic processing, language comprehension and executive control (Du et al., Reference Du, Rolls, Cheng, Li, Gong, Qiu and Feng2020; Jiang et al., Reference Jiang, Ma, Sanford and Li2024). As Immersion-L2 increases in bilinguals, the brain reorganizes its neural circuits to enhance language processing efficiency, which may lead to a decreased involvement of neural connectivity between cerebellar and cortical regions. Our results were in line with previous findings, which suggest that adaptations related to the length of L2 immersion reflect an increased automation in language control processing with prolonged, intensive exposure to L2 (DeLuca et al., Reference DeLuca, Rothman, Bialystok and Pliatsikas2019; Linck et al., Reference Linck, Kroll and Sunderman2009). This increased automation may underlie the observed negative association between cerebellar FC and Immersion-L2 in our study.
4.4. Cerebellar FC and PL-L2
Our findings revealed that higher PL-L2 was associated with stronger FC between the cerebellar lobules VI.L and VIIb.L and the PCC in bilinguals. These two cerebellar regions play a crucial role in lexical-semantic processing for non-native languages, indicating their involvement in the more demanding cognitive control and language monitoring required when retrieving words in later-learned languages (Mariën et al., Reference Mariën, van Dun, Van Dormael, Vandenborre, Keulen, Manto and Abutalebi2017). Anatomically, lobules VI and VIIb are closely connected to prefrontal cortical regions, suggesting that activation in these cerebellar regions supports the executive control mechanisms necessary for fluent L2 word retrieval and helps manage interference from other languages during multilingual language production. Similarly, previous studies have indicated that the PCC is involved in lexical-semantic processing (Price, Reference Price2010) and may contribute to the integration of semantic and episodic memory, as well as self-referential processing (Binder et al., Reference Binder, Desai, Graves and Conant2009). Importantly, several studies have demonstrated that engagement of the PCC during L2 language processing is associated with individual PL-L2 (Grant & Li, Reference Grant and Li2019; Palomar-García et al., Reference Palomar-García, Bueichekú, Ávila, Sanjuán, Strijkers, Ventura-Campos and Costa2015; Zhang et al., Reference Zhang, Wang, Lin, Turk-Browne and Cai2023). For example, Grant and Li (Reference Grant and Li2019) reported that highly proficient L2 learners exhibited greater PCC activation and more organized FC during the semantic decision task, reflecting a shift from effortful, top-down control to more precise, integrated processing. Zhang et al. (Reference Zhang, Wang, Lin, Turk-Browne and Cai2023) found that proficient L2 learners showed stronger PCC synchronization during narrative comprehension, approaching native-like patterns of semantic integration and high-level contextual processing. In our study, stronger FC between the lobules VI and VIIb and the PCC in bilinguals with high PL-L2 may reflect the coordinated engagement of these regions to support precise lexical-semantic retrieval, executive monitoring and semantic-contextual integration. Specifically, the lobules VI and VIIb may contribute to fine-grained control and lexical access, while the PCC provides a hub for semantic convergence and the integration of episodic or contextual cues. The functional integration of neural activity between these regions may therefore reflect a neural adaptation that supports more precise and native-like language use as proficiency increases.
In addition, our results aligned with previous evidence indicating that higher PL-L2 was associated with increased FC and neural activity involving regions implicated in language processing and cognitive control (Mondt et al., Reference Mondt, Balériaux, Metens, Paquier, Van de Craen, Van den Noort and Denolin2009; Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a; Wang et al., Reference Wang, Ke, Zhang, Zhou, Li and Yang2020). Wang et al. (Reference Wang, Ke, Zhang, Zhou, Li and Yang2020) reported that bilinguals with higher PL-L2 exhibited stronger resting-state and task-based FC among regions supporting language selection, articulation and cognitive control, indicating enhanced integration between semantic and executive systems. Similarly, Sulpizio et al., Reference Sulpizio, Del Maschio, Del Mauro, Fedeli and Abutalebi2020a found that PL-L2 was positively correlated with the strength of FC between the left posterior STG and the precuneus during L2 switching, suggesting refinement of lexico-phonological representations and optimization of control processes. In this study, our results provided further evidence that higher PL-L2 was associated with stronger FC involving cerebellar and cortical regions. This pattern may reflect proficiency-dependent network refinement involving cerebello-cortical circuits, supporting more precise lexical-semantic processing in bilinguals.
4.5. Cerebellar FC and Usage-L2
We found a negative relationship between cerebello-cortical FC and the frequency of L2 use at home in bilinguals. Specifically, bilinguals who used L1 more frequently at home exhibited higher cerebellar FC between the lobules VI.R, VIIb.L, and bilateral VIIIa and the FFG.R. As discussed above, lobules VI and VIIIa are implicated in phonological and semantic processing, respectively. Additionally, lobule VIII is involved not only in language-related functions but also in executive functions (Stoodley & Schmahmann, Reference Stoodley and Schmahmann2009). The FFG.R has also been implicated in language processing, particularly in the visual recognition of written words. This region is more sensitive to the visual appearance of words rather than their intrinsic linguistic information (Dehaene et al., Reference Dehaene, Naccache, Cohen, Bihan, Mangin, Poline and Rivière2001, Reference Dehaene, Jobert, Naccache, Ciuciu, Poline, Le Bihan and Cohen2004; Qu et al., Reference Qu, Zhang, Chen, Xie, Li, Liu and Mei2019). It responds to visual features such as letter case but does not exhibit sensitivity to phonological or semantic information. In the context of bilingualism, the FFG.R appears to be more engaged in processing the L1. (Giraud & Truy, Reference Giraud and Truy2002; Li et al., Reference Li, Zhang and Ding2021; Mei et al., Reference Mei, Xue, Lu, Chen, Wei, He and Dong2015; Suh et al., Reference Suh, Yoon, Lee, Chung, Cho and Park2007). For instance, Suh et al. (Reference Suh, Yoon, Lee, Chung, Cho and Park2007) found that Korean-English bilinguals showed greater activation in the FFG.R during L1 processing. Additionally, a meta-analysis by Li et al. (Reference Li, Zhang and Ding2021) confirmed that the FFG.R consistently activates in response to L1 reading. These findings suggest that the FFG.R may play a role in processing familiar orthographic forms, contributing to the automatic and efficient recognition of L1 words. In the current study, we found that higher cerebellar FC involving L1 processing was related to the increased frequency of L1 usage in bilinguals, which may reflect more efficient or automatized processing of L1-related visual and linguistic information. Furthermore, the negative association between cerebello-cortical FC and L2 home usage suggested that reduced L1 exposure might lead to weaker connectivity patterns, potentially reflecting a shift in neural engagement as bilinguals allocate more cognitive resources to L2 processing at home.
Cerebellar FC between the lobule IX.R and the INS.L was positively correlated with Usage-L2 in social situations. A previous study has shown that the insula plays a crucial role in language conflict monitoring in bilinguals (Teubner-Rhodes et al., Reference Teubner-Rhodes, Bolger and Novick2019). Bilinguals have been found to exhibit greater activation of the INS during conflict detection tasks. This region is involved in the controlled retrieval of semantic information when external cues are insufficient to support retrieval (Badre et al., Reference Badre, Poldrack, Paré-Blagoev, Insler and Wagner2005). It is engaged in tasks requiring the maintenance and retrieval of task goals, especially when dealing with multidimensional stimuli associated with multiple response rules. Moreover, bilinguals tend to rely more on the neural resources involved in language switching, such as the left ventrolateral prefrontal cortex (VLPFC)/insula, to detect conflict accurately. This increased activation in the left VLPFC/insula during conflict monitoring tasks may indicate that bilinguals use top-down control to retrieve goal-relevant information, leading to improved accuracy following congruent trials and potentially slower response time. We observed higher functional connectivity between the IX.R and the INS.L, which was associated with increased Usage-L2 in social settings. This may be because, as L2 usage increases in dynamic, real-world social interactions, bilinguals enhance their conflict monitoring of different languages. Our result was in line with previous findings that increased Usage-L2 in social settings is associated with heightened demands for language control (DeLuca et al., Reference DeLuca, Rothman, Bialystok and Pliatsikas2019).
5. Limitations and future directions
It should be noted that this study had several limiting factors. First, we only focused on static but not dynamic cerebellar FC induced by bilingual experiences. Recent studies showed that significant difference in functional network dynamics between early and late bilinguals suggested that early language experiences may affect the dynamic reorganization of neural networks in bilinguals (Liu et al., Reference Liu, Tu, Chen, Zhong, Niu, Zhao and Huang2020; Sheng et al., Reference Sheng, Yang, Rao, Zhang, Li, Wang and Zheng2023). However, previous studies have demonstrated the need for more than 10 minutes to acquire resting-state fMRI data when the data was applied in dynamic analysis (Hindriks et al., Reference Hindriks, Adhikari, Murayama, Ganzetti, Mantini, Logothetis and Deco2016; Tomasi et al., Reference Tomasi, Shokri-Kojori and Volkow2017). Given the limitation of the current data, we cannot analyze the cerebellar network dynamics in this study. Future studies should lengthen the scanning time and then investigate the effect of bilingualism on the dynamic properties of the cerebellar network. Second, we investigated the effect of bilingual experiences on cerebellar FC using multiple regression analysis, without considering the non-linear nature of neuroplasticity in the brain. The DRM suggests that bilingualism-induced neuroplasticity follows a dynamic trajectory, with structural adaptions occurring at different stages of L2 learning. Recent studies have increasingly applied non-linear analyses to examine how bilingual experiences shape brain structural changes (DeLuca & Voits, Reference DeLuca and Voits2022; DeLuca et al., Reference DeLuca, Voits, Ni, Carter, Rahman, Mazaheri and Segaert2024; Korenar et al., Reference Korenar, Treffers-Daller and Pliatsikas2023b). These findings highlight the importance of capturing non-linear developmental patterns in the bilingual brain. Future studies should investigate whether the functional neuroplasticity of the cerebellum follows similar non-linear trajectories in bilinguals. Last but not least, we examined differences in cerebellar functional neuroplasticity between monolinguals and bilinguals, which may involve an inherent limitation associated with this comparative framework (Dash et al., Reference Dash, Joanette and Ansaldo2022; De Houwer, Reference De Houwer2023; Rothman et al., Reference Rothman, Bayram, DeLuca, Di Pisa, Duñabeitia, Gharibi and Kupisch2023). This framework risks reinforcing biases regarding monolinguals and bilinguals, potentially implying that monolingualism constitutes a normative baseline or that bilinguals are expected to demonstrate equal proficiency in both languages. Such assumptions overlook the multidimensional nature of bilingualism, which emerges from complex and dynamic interactions between language acquisition history, proficiency levels, usage patterns and sociocultural contexts. Future studies should adopt more comprehensive frameworks, such as the network model of bilingualism (Kałamała et al., Reference Kałamała, Chuderski, Szewczyk, Senderecka and Wodniecka2023), to investigate how language background and bilingual experience-related factors influence cerebellar functional neuroplasticity during L2 acquisition (i.e., the transition from monolingualism to bilingualism).
6. Conclusion
In this study, we examined the effect of bilingualism on functional neuroplasticity of the cerebellum. We found that monolinguals exhibited higher FC between cerebellar regions and temporal lobe regions. AoA-L2 modulated cerebello-cortical FC, particularly involving the VI, Crus I, VIII and IX, along with relevant cortical language regions. Also, increased AoA-L2 was related to increased cerebellar FC between bilateral IX and subcortical regions (caudate and thalamus). This pattern suggested that late bilinguals may rely more on cerebellar and subcortical regions to support L2 learning. Immersion-L2 was negatively related to cerebellar FC, which may reflect that with increased immersional duration of L2, the bilingual brain reorganizes the neural circuits of the cerebellum to increase language processing efficiency. Higher PL-L2 in bilinguals was associated with stronger cerebellar FC, supporting more precise L2 processing. Increased Usage-L2 at home was associated with decreased cerebellar FC between the IX and the FFG, while increased cerebellar involving the IX and INS was observed in bilinguals who use L2 frequently in social settings. Our results highlighted the potential impact of bilingualism on the cerebello-cortical and cerebello-subcortical neural pathways and suggested that AoA-L2, Immersion-L2, PL-L2 and Usage-L2 shape the cerebellar FC in different ways. These findings improved our understanding of how bilingual experiences influence the functional neuroplasticity of the cerebellum.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1366728925100734.
Data availability statement
The data comes from these open datasets: (1) “The Reading Brain Project L1 Adults” dataset (https://openneuro.org/datasets/ds003974/versions/3.0.0); (2) “The Reading Brain Project L2 Adults” dataset, (https://openneuro.org/datasets/ds003988/versions/1.0.0); (3) “Bilingualism and the Brain” dataset (https://openneuro.org/datasets/ds001796/versions/1.3.0). The corresponding processed MRI images data and analysis codes related to this publication will be available upon request with a legitimate reason.
Acknowledgments
This work has received funding from the National Natural Science Foundation of China (Grant number: 62407005), the Scientific Research Foundation for Talented Scholars, Beijing Normal University (Grant numbers: 28817-310432101 and 28817-312200502506) and the Excellence Enhancement Program of BNU First-class Education Discipline Project (grant number: YLXKPY-XSDW202213).
Author contribution
Xiaojin Liu conceived the research. Xiaojin Liu analyzed the data. Xiaojin Liu, Xin Tong, Ying Yang and Yuqi Liang wrote the paper. Shan Jiang, Yongqiang Jiang and Naiyi Wang organized the data and results. Xiaojin Liu, Xin Tong, Ying Yang, Yuqi Liang, Shan Jiang, Yongqiang Jiang and Naiyi Wang revised the paper.
Ethical approval
All procedures performed were following the ethical standards of the appropriate institutional research boards and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Competing interests
The authors declare none.
Open access
