Mapping the functional dynamics of language control and executive functions in the bilingual brain with magnetoencephalography

1 Introduction: bilingualism and brain function

Using more than one language requires the brain to coordinate multiple linguistic systems during everyday communication. Research on bilingual language processing has shown that the languages known by a bilingual remain active even when only one language is immediately relevant, creating competition during both comprehension and production (Marian and Spivey, 2003; Schwartz and Kroll, 2006; Wu and Thierry, 2010). To communicate successfully, bilingual speakers must therefore monitor the linguistic context, select the intended language, manage cross-language competition, and switch between languages when required. The repeated engagement of these processes places sustained demands on language control mechanisms that partly overlap with domain-general executive functions, including attentional control, conflict monitoring, cognitive flexibility, and working memory (Bialystok and Craik, 2022; Green and Abutalebi, 2013). This observation has led to investigations on whether bilingualism may have knock-on effects on domain-general cognition, which in turn have generated some controversy: while some studies have reported performance differences between bilinguals and monolinguals in tasks tapping executive functions (Costa et al., 2008; Prior and Macwhinney, 2010), others have found no consistent or statistically significant group differences (Nichols et al., 2020; Paap and Greenberg, 2013). This controversy has motivated a different line of research which utilizes neuroimaging measures. Neural measures can extend this debate by identifying differences in the engagement of executive functions that are not apparent in reaction times or accuracy. Indeed, bilingual experience can be associated with differences in neural activity during executive control tasks even in the absence of corresponding behavioral differences, suggesting that neural measures may capture experience-related changes that are not yet evident in response times or accuracy (DeLuca et al., 2020a; Luk et al., 2010; Sullivan et al., 2014). Neuroimaging is therefore important for establishing how bilingual language use relates to brain function beyond the presence or absence of behavioral group differences.

Various functional neuroimaging methods have been used to identify bilingualism-related effects in the neural systems supporting executive functions. Functional magnetic resonance imaging (fMRI) studies have reported differences between bilinguals and monolinguals in the recruitment and connectivity of control-related networks during tasks involving interference control, inhibition, conflict monitoring, and cognitive flexibility. Consistent with the dissociation between neural and behavioral findings described above, DeLuca et al. (2020a) found that the duration and extent of bilingual language use predicted neural recruitment during a Flanker task, despite showing no corresponding effects on behavioral performance. Bilingualism-related differences in control networks have also been observed during language-control tasks, reflecting the executive demands involved in managing two languages (for reviews, see Pliatsikas and Luk, 2016; Gallo et al., 2022). Functional near-infrared spectroscopy (fNIRS), which also measures hemodynamic activity, has also been used to investigate the neural correlates of executive control in bilinguals, particularly in populations and experimental settings less suited to MRI, such as children. These studies have reported bilingualism-related differences in pre-frontal recruitment during attentional control and at rest between bilingual and monolingual children (Arredondo et al., 2017; for a review, see Pliatsikas, 2024). Electroencephalography (EEG) studies have also reported bilingualism-related differences in the temporal dynamics of executive control. For example, Kousaie and Phillips (2012) found no significant behavioral differences between bilingual and monolingual young adults across Stroop, Simon, and Flanker tasks, but observed group differences in ERP components associated with conflict monitoring, attentional allocation, stimulus evaluation, and error monitoring. Using pre-task resting-state EEG, Sheehan et al. (2026) further found that bilingual experience non-linearly modulated long-distance and interhemispheric functional connectivity involving frontal, central, temporal, parietal, and occipital electrodes. Such findings indicate that electrophysiological measures can reveal differences in neural processing that are not evident in response times or accuracy (for a review, see Antoniou, 2023).

However, fMRI and fNIRS rely on relatively slow hemodynamic responses, limiting their ability to distinguish rapidly unfolding neural processes, whereas the cortical sources of EEG signals can be difficult to localize. Magnetoencephalography (MEG) is an alternative neuroimaging method with the added benefit of combining millisecond temporal resolution with source estimation and frequency-resolved analyses, providing very good spatial and temporal resolution of the effects of interest (Baillet, 2017). Consequently, MEG provides a valuable tool for examining how bilingual language use relates to brain function, particularly the neural systems supporting executive functions. This paper provides an overview of the method before reviewing evidence from task-based and resting-state MEG studies on bilingualism. It concludes with reflections on how MEG can clarify the timing, neural mechanisms, and network organization of functional brain differences associated with bilingual experience, and on its value for refining theories on the effects of bilingualism on the brain.

2 MEG as a window into neural oscillations and functional brain networks

MEG is a non-invasive electrophysiological method that records the weak magnetic fields generated primarily by synchronous postsynaptic currents in populations of neurons. Because these magnetic signals arise directly from neuronal activity rather than from the associated hemodynamic response, MEG can track changes in brain function on a millisecond timescale (Baillet, 2017; Hämäläinen et al., 1993; Kim and Davis, 2021). Signals are recorded using sensors positioned around the head, and source-estimation procedures can be used to infer the cortical generators most likely to have produced them. MEG can therefore characterize both the timing and likely neural sources of cognitive processes. Event-related fields characterize responses that are consistently time-locked to stimuli or behavioral events, whereas time-frequency analyses reveal changes in the power and phase of neural oscillations. These oscillations are commonly divided into delta (approximately 1–4 Hz), theta (4–8 Hz), alpha (8–13 Hz), beta (13–30 Hz), and gamma (>30 Hz) frequency bands, which have been associated with partly distinct aspects of cognition. Of particular relevance to executive functions, theta activity has been associated with conflict monitoring and cognitive control (Cavanagh and Frank, 2014), alpha activity with attentional selection and inhibition (Klimesch, 2012), and beta activity with the maintenance of ongoing cognitive states (Engel and Fries, 2010). Functional connectivity analyses can additionally estimate the temporal coordination of activity across brain regions. Together, these approaches allow researchers to examine the sequence of cortical recruitment, transient changes in neural activity, and the dynamic coordination of distributed functional networks (Baillet, 2017; Gross et al., 2013). This information is particularly relevant to bilingualism research because executive processes such as attentional allocation, conflict detection, response selection, and cognitive switching may unfold in rapid succession and may not be distinguishable using behavioral measures alone. Figure 1 provides a schematic overview of MEG recording and analysis, including preprocessing, source estimation, spectral and time-frequency analyses, functional connectivity, and network topology.

MEG complements rather than replaces other functional neuroimaging methods. Like EEG, it records electrophysiological signals generated by neuronal activity. However, magnetic fields are generally less affected than scalp electrical potentials by variations in the conductivity of the skull and surrounding tissues, which can facilitate the modeling of cortical sources under appropriate recording and analytical conditions (Kim and Davis, 2021; Gross et al., 2013; Yang et al., 2024). As already mentioned, fMRI and fNIRS measure hemodynamic changes associated with neural activity. fMRI provides detailed anatomical coverage, whereas fNIRS offers greater portability and flexibility in participant positioning and testing environments. However, hemodynamic responses evolve over seconds and cannot be linked to neural events separated by tens or hundreds of milliseconds. MEG is therefore particularly well-suited to examining the temporal ordering, oscillatory organization, and rapid coordination of cognitive processes, while providing some anatomical precision of the source of these effects.

MEG has been used to investigate both language processing and executive functions. In language research, it has been used to identify neural responses associated with phonemic, lexical, phrasal, and syntactic processing during continuous speech comprehension (Brodbeck et al., 2022; Ten Oever et al., 2022; Weissbart and Martin, 2024). In executive function research, MEG has identified temporally and spectrally distinct neural responses during different stages of working-memory processing, including encoding and maintenance (Otstavnov et al., 2024). Resting-state MEG spectral power has also been associated with individual differences in executive functions and language performance (Oswald et al., 2022). These applications highlight MEG as an appropriate method to examine how bilingual language experience relates to the neural processes supporting both language processing and executive functions. The following paragraphs review the relevant evidence both from task-based and from resting-state paradigms.

3 Task-based MEG

Task-based MEG studies of bilingualism have primarily examined language selection and switching. These processes involve contextual monitoring, selective attention, inhibition, and cognitive switching, and therefore provide insight into the operation of executive-control mechanisms during bilingual language use. This section first reviews MEG evidence from bilingual language-control tasks before considering more limited evidence from non-linguistic tasks tapping executive functions.

Blanco-Elorrieta and Pylkkänen (2016) used MEG to compare language switching during production and comprehension with switching between semantic categories on 19 bilingual participants. Language switching during production recruited dorsolateral pre-frontal regions that were also involved in non-linguistic switching, whereas switching during comprehension engaged the anterior cingulate cortex. The neural implementation of language control therefore differed between production and comprehension. Subsequent studies on 19 bilingual individuals showed that these responses also depend on the context and specific requirements of switching. Specifically, experimentally imposed switches elicited greater pre-frontal and anterior cingulate activity than switches occurring naturally during conversation (Blanco-Elorrieta and Pylkkänen, 2017). In bimodal bilinguals, Blanco-Elorrieta et al. (2018) found that control-related activity was primarily associated with disengaging from an active language rather than activating an additional language.

Zhu et al. (2020) identified two temporally distinct stages of control during cued language switching in 20 unbalanced Mandarin-English bilinguals. Switching into the non-dominant language elicited increased left inferior frontal activity following the language cue, whereas naming in the non-dominant language elicited greater right inferior frontal activity following the target. These responses were interpreted as reflecting interference resolution during preparation and inhibition of competing lexical representations during production, respectively. Zhu et al. (2022), in a study with 24 participants, further showed that allowing participants to select the language freely and consistently name each item in the same language reduced behavioral switching costs and eliminated the right inferior frontal switching response observed under forced switching. These findings demonstrate that the neural demands of language control vary across processing stages and switching contexts.

Despite the otherwise growing literature on language control, direct MEG evidence concerning bilingualism and domain-general executive functions remains extremely limited. To our knowledge, Bialystok et al. (2005) remains the only published MEG study directly comparing neurologically healthy bilingual and monolingual participants during a non-linguistic task explicitly engaging domain-general executive control. This claim is limited to such between-group comparisons and does not encompass within-bilingual designs, linguistic-processing paradigms, or studies of clinical populations. Bialystok and colleagues (2005) compared 10 Cantonese-English bilinguals, 10 French-English bilinguals, and 10 English monolinguals performing a non-verbal Simon task while whole-head MEG activity was recorded. Behaviorally, the Cantonese-English bilinguals responded faster than both the French-English bilinguals and the monolinguals, whereas the latter two groups did not differ. This group difference applied across the control, congruent, and incongruent conditions, with no group-by-condition interaction. It therefore reflected an overall response-time difference rather than a specifically reduced Simon interference effect. Nevertheless, the MEG analyses revealed differences in how neural activity was related to performance. In the 4–8 Hz band, all three groups showed a broadly similar distinction between congruent and incongruent trials, although this pattern was more strongly expressed in the bilingual groups. In the 8–15 Hz band, faster responses in both bilingual groups, across congruent and incongruent trials, were associated with greater activity in superior and middle temporal, cingulate, and superior and inferior frontal regions, predominantly in the left hemisphere. For monolinguals, faster responses were principally associated with activity in the left middle frontal region. The findings therefore indicated different brain-behavior relationships across language groups, including between the French-English bilingual and monolingual groups, whose behavioral performance was comparable.

Taken together, task-based MEG studies show that bilingual experience is associated with differences in the neural implementation of language and cognitive control. During language tasks, these differences have been observed in the timing, spatial distribution, and oscillatory organization of processes involved in language selection, switching, and interference resolution. Evidence from the non-verbal Simon task further indicates that bilinguals and monolinguals may recruit different neural systems even when their behavioral performance is comparable. The current evidence therefore supports differences in how control is implemented in the bilingual brain, but direct MEG research on domain-general executive-function tasks remains limited.

4 Resting-state MEG

Where as task-based MEG relates neural activity to specific cognitive events, resting-state MEG records spontaneous brain activity while participants rest without performing a particular task. It therefore provides a complementary perspective by examining whether bilingual experience is associated with relatively persistent differences in frequency-specific connectivity and network topology, rather than only with transient responses evoked by language or executive-control tasks.

The earliest resting-state MEG study of bilingualism was conducted by De Frutos-Lucas et al. (2020), who examined whether lifelong bilingual experience was associated with functional brain connectivity in 38 healthy older adults with postsecondary education. Source-level functional connectivity was estimated using the phase-locking value, which measures the consistency of phase relationships between signals from different cortical regions. Compared with monolinguals, bilinguals showed significantly greater functional connectivity across five clusters of connections located predominantly in parietal and occipital regions. These clusters included connections involving the supramarginal gyrus, postcentral gyrus, superior parietal cortex, lingual gyrus, and several occipital areas, some of which have established roles in language processing. Within the bilingual group, connectivity strength in each of the five clusters was positively associated with the volume of the caudate nucleus, measured using structural MRI. The authors suggested that stronger posterior connectivity and its association with the caudate nucleus could represent a neural pattern through which lifelong bilingual experience helps maintain cognitive functioning in older age. These findings extend previous task-based functional evidence by showing that lifelong bilingualism is also associated with the organization of spontaneous functional brain networks in older adults. More specifically, the study demonstrates that resting-state MEG can detect differences in phase-based cortical connectivity and relate these functional patterns to structural variation within the language-control system.

More recently, Amoruso et al. (2024) examined whether bilingual experience could be identified from the frequency-specific organization of resting-state brain networks. The study recorded resting-state MEG activity from 20 highly proficient Spanish-Basque bilinguals and 20 Spanish monolinguals. Functional connectivity and graph-theoretical measures were calculated across frequency bands and entered into a machine-learning classifier. The classifier distinguished bilinguals from monolinguals with an accuracy of 84.8% (AUC = 0.91). Network strength in the beta (15–30 Hz) and delta (2–4 Hz) bands contributed most strongly to the classification and was lower in bilinguals. Beta-band differences primarily involved temporal regions associated with lexico-semantic processing, whereas delta-band differences involved cingulate and frontoparietal regions associated with large-scale integration, cognitive control, and the default mode network. Within the bilingual group, lower beta and delta network strength was also associated with higher second-language proficiency. The authors interpreted these patterns as reflecting more selective and efficient network organization following sustained bilingual experience. The study therefore shows that resting-state MEG can capture both group differences and variation in bilingual experience through frequency-specific network organization. The latest is also supported by recent complementary resting state EEG findings that show bilingual experience was non-linearly associated with long-distance and interhemispheric connectivity involving frontal, central, temporal, parietal, and occipital electrodes (Sheehan et al., 2026). Such findings motivate the use of continuous measures of bilingual experience and the examination of non-linear associations in future resting-state MEG studies.

Together, the available MEG studies show that bilingual experience is associated with phase-based cortical connectivity, frequency-specific network topology, and relationships between network features and second language proficiency. The analogous EEG findings extend this evidence by identifying non-linear associations with bilingual experience and differences between pre-task and post-task resting-state connectivity patterns. Resting-state MEG therefore extends the study of the bilingual brain beyond task-evoked activation, allowing researchers to characterize how bilingual experience relates to spontaneous cortical connectivity and its organization across frequency bands.

5 Future directions: theoretical and methodological considerations

Findings from current MEG studies of bilingualism are broadly consistent with those obtained using other neuroimaging methods, particularly in showing that bilingual language use is associated with functional effects in brain networks related to language and executive functions. However, the application of MEG in this field remains limited and could be extended considerably. The evidence base remains small, with generally modest samples and heterogeneous tasks and analyses. Studies differ in source-estimation methods—including synthetic aperture magnetometry and minimum-norm estimation—connectivity measures, and preprocessing procedures; source estimates also depend on head models (Cho et al., 2015). These choices, together with differences in statistical testing and correction across sensors, sources, time points, and frequencies, limit cross-study comparability. Larger samples, transparent reporting, appropriate multiple-comparison correction, and independent replication are therefore needed (Gross et al., 2013; Pernet et al., 2020). More task-based studies are needed to examine non-linguistic executive functions, while resting-state MEG could provide complementary information about intrinsic oscillatory activity and functional network organization. Combining task-based and resting-state recordings within the same participants may further clarify whether neural differences observed during specific cognitive tasks are related to more persistent patterns of brain activity. Crucially, evidence from MEG could also help refine theoretical accounts of bilingual neuroplasticity. The Bilingual Anterior-to-Posterior and Subcortical Shift model (BAPSS) proposes that increasing bilingual experience is accompanied by reduced reliance on frontal control regions and greater recruitment of posterior and subcortical systems (Grundy et al., 2017). Resting-state MEG could provide further evidence for this account by testing whether bilingual experience is associated with the predicted redistribution of intrinsic functional connectivity from frontal to posterior networks. The Unified Bilingual Experience Trajectories model (UBET) further proposes that neural adaptation varies according to the duration, intensity, diversity, and switching demands of bilingual language use (DeLuca et al., 2020b). MEG could test these experience-dependent predictions using both task-based and resting-state measures regressed against detailed measures of bilingual experiences.

Research should also address the limited age range represented in the existing literature. Most task-based bilingual MEG studies have examined young adults, while evidence from children is largely absent and, to our knowledge, De Frutos-Lucas et al. (2020) provides the only investigation focused on older adults. Extending MEG research across childhood, adulthood, and older age would establish whether bilingualism-related functional patterns emerge during development and change across the lifespan. Recent developments in optically pumped magnetometer-based MEG (OPM-MEG), which uses wearable quantum sensors placed close to the scalp, may facilitate this expansion. Unlike conventional systems with fixed sensor arrays, OPM-MEG can be adapted to different head sizes and permits greater movement during recording, facilitating recordings in participants for whom remaining still is difficult. Because OPM systems avoid liquid-helium cooling, they may also reduce cooling-related costs; however, these advantages do not remove the barriers of limited availability, cost, and magnetic shielding requirements (Brookes et al., 2022).

MEG also has limitations that are particularly relevant to bilingualism research. It is primarily sensitive to cortical activity and is less effective at detecting signals from deep brain structures, despite the central roles of the basal ganglia and thalamus in theories of language selection and executive control. MEG alone may therefore provide an incomplete account of the cortical–subcortical systems involved in bilingual language control. Its source localization also depends on modeling assumptions and the quality of anatomical and recording data. Combining MEG with structural MRI, fMRI, and behavioral measures could provide a more complete account of cortical–subcortical interactions and clarify how the temporal and oscillatory dynamics detected by MEG relate to brain structure, spatially defined functional networks, and cognitive performance.

Practical constraints also limit wider use. Conventional MEG is costly to purchase and operate and requires dedicated magnetically shielded facilities and specialist support, often beyond the resources of language-science departments. Long sessions increase participant burden, and head motion can reduce data quality. Collaboration with established MEG facilities and shared acquisition and analysis resources could broaden access. Thus, MEG is a valuable addition to bilingualism research, but its wider contribution depends on both rigorous methods and access to the necessary infrastructure.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. The MEG dewar illustration in Figure 1A was generated using OpenAI's gpt-image (version 2.0), accessed through ChatGPT (https://chatgpt.com). The author(s) reviewed and edited these illustrations and took responsibility for the final figure.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Greg Poarch, University of Groningen, Netherlands

Reviewed by: Muyu Chen, Army Medical University, China