Highlights
-
• Two 20-minute sessions of anodal transcranial direct current stimulation (tDCS) over left dorsolateral prefrontal cortex improved L2 listening comprehension.
-
• The effect was large and specific to the treatment group; the sham group showed no gains.
-
• Reading comprehension and verbal working memory (WM) showed no significant improvement.
-
• tDCS appears to selectively benefit more cognitively demanding L2 skills.
-
• tDCS may offer a low-effort alternative to time-intensive WM training.
1. Introduction
Human cognition encompasses a variety of functions, including reasoning, perception, memory, and information processing (Bayne et al., Reference Bayne, Brainard, Byrne, Chittka, Clayton, Heyes, Mather, Ölveczky, Shadlen, Suddendorf and Webb2019). Among these, language occupies a central role because it relies on and shapes general cognitive mechanisms. Acquiring a first language (L1) typically occurs automatically during early development, but acquiring a second language (L2) in adulthood requires conscious effort and greater reliance on domain-general cognitive control (Sanz, Reference Sanz2005). L2 processing is cognitively demanding because bilinguals must manage cross-linguistic interference and resource competition between their languages (Grosjean, Reference Grosjean1989). As a result, individual differences in executive functioning have been found to predict success in L2 processing (Durand-López, Reference Durand-López2024a; Linck & Weiss, Reference Linck and Weiss2015; Wu & Thierry, Reference Wu and Thierry2013).
One cognitive ability associated with L2 processing is verbal working memory (WM), a process with a limited capacity that is in charge of controlling, regulating, and actively maintaining information of many kinds in the face of distracting information (Conway et al., Reference Conway, Jarrold, Kane, Miyake and Towse2007). WM has been found to modulate receptive (Satori, Reference Satori2021; Shin, Reference Shin2020) and productive (Vasylets & Marín, Reference Vasylets and Marín2021) L2 skills as well as structural processing, particularly morphosyntactic processing (Arnold, Reference Arnold2019; Durand-López, Reference Durand-López2024a), such as increased sensitivity to gender agreement violations (Durand-López, Reference Durand-López2024a), faster detection of person and tense violations (Arnold, Reference Arnold2019), and higher accuracy on speaking (Zalbidea & Sanz, Reference Zalbidea and Sanz2020) and reading (Sagarra, Reference Sagarra2017) morphosyntactic tasks.
Given that WM has been associated with L2 processing, recent studies have started to examine whether WM can be enhanced and whether that influences L2 processing. Two recent studies have found that WM cannot only be trained, but it can aid in morphological and morphosyntactic processing (Colflesh et al., Reference Colflesh, Karuzis and O’Rourke2016; Durand-López, Reference Durand-López2024a), particularly in that of more advanced L2 learners (Durand-López, Reference Durand-López2024b). On the contrary, in a study conducted by Hayashi et al. (Reference Hayashi, Kobayashi and Toyoshige2016), researchers found that WM training plus L2 instruction had the most improvements in verbal short-term memory, verbal WM, and visuospatial memory; however, no improvements in L2 proficiency were found. WM training appears to be a tool that may aid L2 processing, but it is time-consuming, as individuals need to focus solely on the WM task completion and are unable to perform other activities. Another logical and more practical route would be to directly stimulate the region of the brain associated with verbal WM. A well-established finding in neurophysiology is that the dorsolateral prefrontal cortex (DLPFC) is associated with executive functioning and WM specifically (Živanović et al., Reference Živanović, Paunović, Konstantinović, Vulić, Bjekić and Filipović2021). In more recent years, brain stimulation through transcranial direct current stimulation (tDCS) has been used to explore whether it can enhance a myriad of cognitive abilities, including WM. Studies have found that anodal (increased neuronal excitation; i.e., depolarization) tDCS over both the left (Baumert et al., Reference Baumert, Buchholz, Zinkernagel, Clarke, MacLeod, Osinsky and Schmitt2020) and right DLPFC (Živanović et al., Reference Živanović, Paunović, Konstantinović, Vulić, Bjekić and Filipović2021) significantly improves WM performance.
Given the above, the present study aims to elucidate whether anodal tDCS can aid reading and listening comprehension in the L2, as well as to further understand the relationship between the prefrontal cortex and L2 processing. This contribution will help advance the fields of cognitive psychology and psycholinguistics by investigating potential avenues to facilitate L2 processing. The following section reviews the existing literature on tDCS in general and its effects on L1 processing and L2 processing.
2. Literature review
2.1. Working memory and L2 proficiency
The link between cognitive abilities and L2 proficiency remains not fully understood, although functions such as WM and inhibitory control appear to play a central role in language processing. L2 cognition can be conceptualized in two complementary dimensions: knowledge and proficiency. L2 knowledge refers to the awareness a learner has about the given L2 in general, such as vocabulary and grammatical rules (Ellis, Reference Ellis2004), whereas L2 proficiency reflects the extent to which a learner can use linguistic knowledge, both implicit and explicit, effectively in comprehension and production (see Ellis, Reference Ellis2008). Evidence suggests that WM primarily affects L2 proficiency rather than knowledge. Regarding general L2 proficiency, Linck and Weiss (Reference Linck and Weiss2015) demonstrated that stronger verbal WM is linked to higher levels of L2 performance. Using a pretest/posttest design, the researchers examined whether WM and inhibitory control predict L2 acquisition. Participants in a Spanish course completed an L2 proficiency pretest at the start of the course and repeated it 2 months later, together with a verbal WM and an inhibitory control task. While there was no relationship between inhibitory control and L2 proficiency, verbal WM was found to be positively correlated to proficiency. Contrary to these findings, Hayashi (Reference Hayashi2019) reported that WM training in Japanese learners of English did not lead to measurable improvements in L2 proficiency. One possible explanation for the discrepancy in Hayashi’s (Reference Hayashi2019) study is that the WM training tasks targeted visuospatial, rather than verbal, stimuli; verbal WM tasks may be more effective in predicting L2 proficiency, as suggested by the current literature (e.g., Linck et al., Reference Linck, Osthus, Koeth and Bunting2014; Service et al., Reference Service, Simola, Metsänheimo and Maury2002).
Within L2 proficiency, WM is associated with writing (e.g., Mavrou, Reference Mavrou2018) and speaking (e.g., Guará-Tavares, Reference Guará-Tavares2013; Perea Irigoyen, Reference Perea Irigoyen2020), and it may also be relevant for reading and listening comprehension, as it enables learners to temporarily hold and manipulate linguistic information while constructing meaning (Luque & Morgan-Short, Reference Luque and Morgan-Short2021).
Research examining the relationship between WM and L2 listening comprehension has yielded mixed findings, suggesting that WM’s contribution may vary as a function of task demands and learner characteristics. In Andringa et al. (Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012), native and non-native speakers of Dutch at different proficiency levels completed a self-paced listening task in Dutch along with digit span and nonword recognition tasks assessing verbal WM. The authors found no significant relationship between WM and listening comprehension in either group, concluding that WM did not meaningfully constrain performance under their task conditions. In contrast, Satori (Reference Satori2021) reported a positive association between verbal WM and L2 listening comprehension. Lower- and upper-intermediate Japanese learners of English completed the listening section of the Test of English for International Communication (TOEIC), listening span tasks in the L1 and L2, digit span tasks in both languages, and a battery of linguistic tasks. Verbal WM, particularly in the L2, was a significant predictor of L2 listening performance, with stronger effects observed among lower-proficiency learners. This pattern of results aligns with Fay and Buchweitz (Reference Fay and Buchweitz2014), who found that verbal WM predicted performance on the listening component of the Cambridge Proficiency Exam among beginner Portuguese learners of English. Taken together, these findings suggest that the role of WM in L2 listening comprehension may be modulated by task design and learner proficiency. The self-paced format used in Andringa et al. (Reference Andringa, Olsthoorn, van Beuningen, Schoonen and Hulstijn2012) may have reduced real-time processing demands, allowing participants to regulate input and thereby diminishing the impact of WM constraints. In contrast, more demanding or externally paced listening tasks appear to reveal stronger associations between verbal WM and comprehension, particularly among lower-proficiency learners who may rely more heavily on controlled processing resources.
Regarding the relationship between WM and L2 reading comprehension, findings are more mixed than for listening comprehension. Some studies have reported positive correlations between these cognitive and linguistic domains (e.g., Joh & Plakans, Reference Joh and Plakans2017; Sagarra, Reference Sagarra2017; Sagarra & Herschensohn, Reference Sagarra and Herschensohn2010), whereas others have not (e.g., Pretorius et al., Reference Pretorius, le Roux and Geertsema2022). Two recent meta-analyses have examined this relationship. Shin (Reference Shin2020) reported a moderate overall correlation between WM and L2 reading comprehension (r = .30), indicating that WM accounts for approximately 9% of the variance in comprehension performance. This suggests that WM plays a role in L2 reading, but it is not the primary determinant. Moderator analyses revealed that features of the reading span task (e.g., scoring procedures, task language, and recall order) influenced effect sizes. In addition, the type of reading comprehension task also affected the magnitude of the association. On the other hand, the meta-analysis conducted by In’nami et al. (Reference In’nami, Hijikata and Koizumi2022) reported a significant but small overall association between WM and L2 reading comprehension (r = .30), identical to the effect size reported by Shin (Reference Shin2020). Substantial heterogeneity indicated considerable variability across studies. Moderator analyses showed that the relationship was influenced by WM task language (L1 vs. L2) and by whether WM task reliability was reported. Overall, both meta-analyses converge on a small-to-moderate correlation, suggesting that WM contributes to L2 reading comprehension but accounts for a limited proportion of variance. In sum, the evidence suggests that WM plays a more consistent role in L2 listening than in L2 reading comprehension, particularly under conditions of higher processing demands and among lower-proficiency learners. In contrast, although WM is reliably associated with L2 reading, its contribution is modest and explains only a limited portion of variance in performance.
2.2. General effects of tDCS
tDCS is a noninvasive neurostimulation device that delivers a low electrical current (i.e., 1.0–1.5 mA for research [see Borodkin et al., Reference Borodkin, Gassner, Ershaid and Amir2022]; 1.0–2.0 mA for cognitive therapy [see Lefaucheur et al., Reference Lefaucheur, Antal, Ayache, Benninger, Brunelin, Cogiamanian, Cotelli, De Ridder, Ferrucci, Langguth, Marangolo, Mylius, Nitsche, Padberg, Palm, Poulet, Priori, Rossi, Schecklmann and Paulus2017]) via electrodes. The electrodes are typically soaked in a saline solution to improve conductivity and reduce skin irritation and held in place over the desired area by a headband. This device contains two electrodes: anodal and cathodal. Anodal tDCS generally increases cortical excitability by depolarizing neuronal resting membrane potentials, whereas cathodal tDCS generally decreases excitability through hyperpolarization. Nonetheless, these effects are not uniform and depend on stimulation parameters and individual variability (Liu et al., Reference Liu, Tong, de Bruin, Li, He and Li2019). Over the years, research has examined the effects of tDCS on psychiatric conditions such as generalized anxiety disorder (de Lima et al., Reference de Lima, Braga, da Costa, Gomes, Brunoni and Pegado2019), anorexia nervosa (Baumann et al., Reference Baumann, Mareš, Albrecht, Anders, Vochosková, Hill, Bulant, Yamamotová, Štastný, Novák, Holanová, Lambertová and Papežová2021), obsessive-compulsive disorder (Bation et al., Reference Bation, Mondino, Le Camus, Saoud and Brunelin2019), substance abuse disorders (e.g., Çabuk & Guleken, Reference Çabuk and Guleken2024), and depressive disorders (see Kekic et al., Reference Kekic, Boysen, Campbell and Schmidt2016, for a systematic review of tDCS in psychiatric disorders). In more recent years, research has begun to examine the potential effects of tDCS on linguistic ability. While different aspects of language have been explored, including language perception and production (Borodkin et al., Reference Borodkin, Gassner, Ershaid and Amir2022; Rodrigues de Almeida et al., Reference Rodrigues de Almeida, Pope and Hansen2019), language acquisition (Balboa-Bandeira et al., Reference Balboa-Bandeira, Zubiaurre-Elorza, García-Guerrero, Ibarretxe-Bilbao, Ojeda and Peña2023), and language rehabilitation in post-stroke aphasia (Feil et al., Reference Feil, Eisenhut, Strakeljahn, Müller, Nauer, Bansi, Weber, Liebs, Lefaucheur, Kesselring, Gonzenbach and Mylius2019), the following paragraphs will first review evidence on tDCS effects on WM before turning to existing studies on language comprehension and production.
In addition to psychiatric and language implications, tDCS has been examined in numerous studies on WM. Many studies have used the n-back task, a measure of WM capacity, to examine the effects of tDCS over the DLPFC. In a study carried out by Baumert et al. (Reference Baumert, Buchholz, Zinkernagel, Clarke, MacLeod, Osinsky and Schmitt2020), anodal tDCS over the DLPFC enhanced accuracy in the n-back task as well as in the Stroop task, but not in the Stroop interference task. These results indicate that the DLPFC is implicated in WM capacity but not in interference control. In three other studies that also used the n-back task, anodal tDCS facilitated verbal WM performance (Živanović et al., Reference Živanović, Paunović, Konstantinović, Vulić, Bjekić and Filipović2021), benefitted training gains and learning rates (Ke et al., Reference Ke, Wang, Du, Kong, Liu, Xu, An and Ming2019), and enhanced the efficacy of WM training in learning (Ruf et al., Reference Ruf, Fallgatter and Plewnia2017). Overall, previous research on the effects of tDCS over the DLPFC has indicated positive enhancements on WM, particularly in verbal WM. Given the well-documented role of verbal WM in L2 processing (e.g., Linck & Weiss, Reference Linck and Weiss2015; Sagarra, Reference Sagarra2017), these findings raise the possibility that tDCS over the DLPFC may also modulate language-related processes that depend on WM resources, a question addressed by the growing body of research reviewed in the following sections.
2.3. tDCS effects on language processing, production, and comprehension
Recent research suggests that tDCS can modulate several aspects of language processing (but see Balboa-Bandeira et al., Reference Balboa-Bandeira, Zubiaurre-Elorza, García-Guerrero, Ibarretxe-Bilbao, Ojeda and Peña2023; Pu, Reference Pu2015, for null effects). While a number of studies have focused on L1 comprehension (Hussey et al., Reference Hussey, Ward, Christianson and Kramer2015; Mitchell et al., Reference Mitchell, Vidaki and Lavidor2016), others have examined L2 processing, which is more cognitively demanding due to reduced automatization (Vejnović et al., Reference Vejnović, Milin and Zdravković2010). Stimulation sites across studies have included the DLPFC, the left posterior superior temporal gyrus, and the left inferior frontal gyrus, among others (e.g., Bolling et al., Reference Bolling, King, Enam and McDonough2021; Borodkin et al., Reference Borodkin, Gassner, Ershaid and Amir2022; Pu, Reference Pu2015; Radman et al., Reference Radman, Britz, Buetler, Weekes, Spierer and Annoni2018).
The DLPFC, part of the lateral prefrontal cortex and located roughly around areas F3/F4 (Mylius et al., Reference Mylius, Ayache, Ahdab, Farhat, Zouari, Belke, Brugières, Wehrmann, Krakow, Timmesfeld, Schmidt, Oertel, Knake and Lefaucheur2013), is a central hub for executive control, attention, WM, and inhibition, which are mechanisms essential for bilingual language regulation (Hertrich et al., Reference Hertrich, Dietrich, Blum and Ackermann2021). Although not a language-specific area, its connections with frontal and temporal networks allow it to influence sentence processing, discourse comprehension, prosodic integration (Hertrich et al., Reference Hertrich, Dietrich, Blum and Ackermann2021), and semantic integration (Mitchell et al., Reference Mitchell, Vidaki and Lavidor2016). The DLPFC has been found to be involved in language comprehension, seen in studies on comprehension of idioms (Mitchell et al., Reference Mitchell, Vidaki and Lavidor2016) and garden-path sentences (Hussey et al., Reference Hussey, Ward, Christianson and Kramer2015), as well as language production, observed in studies on picture naming (Fertonani et al., Reference Fertonani, Rosini, Cotelli, Rossini and Miniussi2010; Wirth et al., Reference Wirth, Rahman, Kuenecke, Koenig, Horn, Sommer and Dierks2011). However, the exact neural underpinnings of the DLPFC’s role in language processing are still unknown (Klaus & Schutter, Reference Klaus and Schutter2018). In contrast, posterior temporal regions, including Wernicke’s area, are more directly associated with speech perception and lexical-semantic processing (Pu, Reference Pu2015; Yi et al., Reference Yi, Leonard and Chang2019), whereas the left inferior frontal gyrus (Broca’s area) contributes to phonological and semantic encoding as well as inhibitory control (Swick et al., Reference Swick, Ashley and Turken2008). Together, these regions provide a network-based framework in which tDCS can modulate both domain-general control and domain-specific linguistic functions.
2.4. L1 studies
Research on native speakers indicates that the DLPFC plays a role in language processing. Studies administering tDCS to the DLPFC found that it can enhance sentence comprehension (Hussey et al., Reference Hussey, Ward, Christianson and Kramer2015), semantic processing (Mitchell et al., Reference Mitchell, Vidaki and Lavidor2016), and production (Klaus & Schutter, Reference Klaus and Schutter2018). Furthermore, studies that have used other forms of stimulation over the DLPFC found the stimulation facilitated auditory sentence comprehension (Cotelli et al., Reference Cotelli, Calabria, Manenti, Rosini, Zanetti, Cappa and Miniussi2010) and semantic processing (Sarubbo et al., Reference Sarubbo, Tate, De Benedictis, Merler, Moritz-Gasser, Herbet and Duffau2020), respectively. These findings highlight the prefrontal cortex’s contribution to complex linguistic integration in the L1.
2.5. L2 processing studies
Studies examining the effects of tDCS on language abilities in bilinguals have examined L2 processes such as vocabulary learning (e.g., Balboa-Bandeira et al., Reference Balboa-Bandeira, Zubiaurre-Elorza, García-Guerrero, Ibarretxe-Bilbao, Ojeda and Peña2023), word production (e.g., Borodkin et al., Reference Borodkin, Gassner, Ershaid and Amir2022), and language switching (e.g., Li et al., Reference Li, Liu, Pérez and Xie2018). Most studies applied 1.0–1.5 mA of stimulation for 20–25 minutes across two to three sessions. While they did not measure identical constructs, they consistently concluded that tDCS can modulate these language processes. For clarity, these L2 processing studies can be grouped into two domains: acquisition, and production and language switching.
2.5.1. L2 acquisition studies
Research on the effects of tDCS on L2 acquisition has produced mixed results. Pu (Reference Pu2015) applied anodal stimulation over the left posterior superior temporal gyrus during Spanish vocabulary learning and found no significant gains. In contrast, Bolling et al. (Reference Bolling, King, Enam and McDonough2021) observed long-term improvements in Swahili vocabulary recall after stimulating the DLPFC, suggesting that this region may facilitate learning through enhanced executive control and memory consolidation. However, Balboa-Bandeira et al. (Reference Balboa-Bandeira, Zubiaurre-Elorza, García-Guerrero, Ibarretxe-Bilbao, Ojeda and Peña2023) reported no effects when stimulating the left inferior frontal gyrus during English vocabulary learning, possibly due to a between-groups design. Overall, these findings indicate that tDCS may modulate L2 vocabulary acquisition, although effects vary depending on stimulation site, task demands, and linguistic distance between the L1 and L2. This variability underscores the need to examine other domains of L2 processing, such as production and comprehension.
2.5.2. L2 production and language switching studies
Research on L2 production has also explored the effects of tDCS over prefrontal areas. Radman et al. (Reference Radman, Britz, Buetler, Weekes, Spierer and Annoni2018) administered anodal stimulation over the left DLPFC in French learners of English to test its role in lexical processing. While stimulation improved nonverbal fluency, it did not enhance verbal tasks, suggesting that the DLPFC may play a limited role in lexical retrieval. In contrast, Borodkin et al. (Reference Borodkin, Gassner, Ershaid and Amir2022) found that stimulation over the left posterior superior temporal cortex improved both speech perception and production in Hebrew learners of English, particularly in phonetic discrimination and imitation tasks. These findings indicate that while the DLPFC may contribute to executive control mechanisms supporting production, regions more directly involved in auditory or phonological processing may yield stronger effects. This pattern reinforces the need to examine comprehension processes, where both executive and perceptual components interact.
While a substantial body of research has examined L2 acquisition, production, and bilingual language control, L2 comprehension remains comparatively underexplored. With the exception of Borodkin et al. (Reference Borodkin, Gassner, Ershaid and Amir2022), which examined both production and perception, no tDCS studies, to the best of our knowledge, have directly targeted sustained L2 comprehension. Studies of language switching, often grouped under bilingual processing, instead target production-side control mechanisms via picture naming paradigms (Li et al., Reference Li, Liu, Pérez and Xie2018; Liu et al., Reference Liu, Tong, de Bruin, Li, He and Li2019; Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019) and have predominantly stimulated the right DLPFC.
To explore the efficacy of language switching in late bilinguals, Li et al. (Reference Li, Liu, Pérez and Xie2018), Liu et al. (Reference Liu, Tong, de Bruin, Li, He and Li2019), and Tong et al. (Reference Tong, Kong, Wang, Liu, Li and He2019) administered tDCS over the right DLPFC in Chinese L2 learners of English. After a tDCS session (anodal, cathodal, or sham), participants completed a picture naming task. The results of the three studies were mixed, with some findings indicating cathodal stimulation enhanced language control during switching (Li et al., Reference Li, Liu, Pérez and Xie2018) and other findings indicating cathodal stimulation increased switching costs (Liu et al., Reference Liu, Tong, de Bruin, Li, He and Li2019; Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019). Results on anodal tDCS varied as well. One study found tDCS modulated language control when switching from the L1 to the L2 (Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019); however, Liu et al. (Reference Liu, Tong, de Bruin, Li, He and Li2019) indicated that anodal tDCS increased switch costs. Examined together, the findings leave inconclusive results on the role of executive functioning in language switching.
In sum, although research on noninvasive brain stimulation in L2 processing has expanded in recent years, findings remain heterogeneous across linguistic domains and stimulation parameters (see Pandža, Reference Pandža, Morgan-Short and van Hell2023, for a review). Taken together, L1 and L2 studies have found significant effects of tDCS stimulation on language processes, most notably when administering tDCS over the DLPFC. Studies examining the effects of tDCS on L2 processing have been found to effectively aid acquisition (e.g., Bolling et al., Reference Bolling, King, Enam and McDonough2021), production (e.g., Radman et al., Reference Radman, Britz, Buetler, Weekes, Spierer and Annoni2018), and language control during switching (e.g., Li et al., Reference Li, Liu, Pérez and Xie2018), whereas comprehension remains understudied.
3. The study
Previous studies highlight the role of the left DLPFC in language processing, as this region serves as a key neural substrate of verbal WM (Hertrich et al., Reference Hertrich, Dietrich, Blum and Ackermann2021). Verbal WM, in turn, supports receptive, productive, and structural aspects of language, including L2 comprehension and the processing of syntactic information. Evidence from neurocognitive research further underscores this link: studies using tDCS have shown that anodal stimulation of the DLPFC enhances bilingual performance, particularly in tasks involving language switching (Li et al., Reference Li, Liu, Pérez and Xie2018; Liu et al., Reference Liu, Tong, de Bruin, Li, He and Li2019; Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019). Language switching has been a primary focus because it relies heavily on executive functioning, especially WM and inhibition, and because it represents a core feature distinguishing bilinguals from monolinguals (Costa & Sebastián-Gallés, Reference Costa and Sebastián-Gallés2014). However, despite this progress, the effects of tDCS on other central aspects of bilingual processing, such as L2 comprehension, remain understudied. Since language comprehension provides the foundation for more complex abilities, including language switching, investigating whether stimulation of the DLPFC can facilitate L2 comprehension is essential.
As mentioned above, only one study has explored stimulation effects via electrophysiological techniques, such as tDCS, on L2 processes other than language control. Borodkin et al. (Reference Borodkin, Gassner, Ershaid and Amir2022) found that tDCS over the left posterior superior temporal cortex improved L2 perception and production in native Hebrew L2 learners of English, who had learned the language in a classroom setting. The discrimination task used in the study is informative about phonological processing in bilinguals but not L2 comprehension as a whole; it measures the structure of words and not the meaning. The present study aims to further the findings of Borodkin et al. (Reference Borodkin, Gassner, Ershaid and Amir2022) by focusing on facilitating L2 comprehension holistically.
We still have no data on whether late bilinguals undertaking a brain stimulation regimen can comprehend aural and written information in their L2 better, especially before they reach high proficiency in the L2. This is important because in the early stages of L2 acquisition, learners are challenged with cognitive overload due to the complexity of the incoming L2 information they process. During periods of cognitive overload, learning and retention are hindered, fluency in the L2 is reduced, and more errors are observed, both in production and comprehension (Chandrika & Ambedkar, Reference Chandrika and Ambedkar2020). Once higher L2 proficiency is achieved, greater automatization occurs, and costs decrease as L2 processing becomes more automatic (Vejnović et al., Reference Vejnović, Milin and Zdravković2010). In addition, exploring whether we can improve L2 reading and listening comprehension via anodal tDCS may help better understand the relationship between the brain and language processing. The present study aims to further research L2 comprehension by examining electrical activity in the left DLPFC, an area of the brain associated with verbal WM, which in turn supports multiple aspects of L2 processing (Hertrich et al., Reference Hertrich, Dietrich, Blum and Ackermann2021). More specifically, we aim to investigate whether anodal stimulation of the left DLPFC improves L2 listening and reading comprehension in English intermediate learners of Spanish using a randomized controlled design. The research questions are as follows:
RQ1: Does anodal tDCS improve L2 listening comprehension in intermediate L2 learners of Spanish?
RQ2: Does anodal tDCS improve L2 reading comprehension in intermediate L2 learners of Spanish?
RQ3: If improvements in L2 listening or reading comprehension are observed, are they attributable to verbal WM gains, or does tDCS facilitate these skills independently of WM enhancement?
Based on the existing literature on tDCS and language processing (e.g., Balboa-Bandeira et al., Reference Balboa-Bandeira, Zubiaurre-Elorza, García-Guerrero, Ibarretxe-Bilbao, Ojeda and Peña2023), we hypothesize that the left DLPFC modulates comprehension of the L2. In addition, manipulating activity in this region via tDCS will result in increased comprehension for both L2 listening and reading. We predict that intermediate L2 learners will show improved comprehension on the reading and listening tasks on their posttest in comparison with their pretest and the sham group.
By investigating whether direct neuromodulation of the left DLPFC can enhance L2 comprehension, the present study has the potential to advance both theoretical and applied knowledge. On a theoretical level, it may clarify ongoing debates regarding the extent to which domain-general executive resources, such as verbal WM, are recruited during language comprehension. On a practical level, it may identify new ways to support L2 learners at intermediate proficiency, who often face persistent difficulties in comprehension due to cognitive overload. The results may help delineate the role of the prefrontal cortex in receptive language processes, thereby refining our understanding of the neural mechanisms underlying bilingualism. Ultimately, this line of research has implications for cognitive psychology, psycholinguistics, and L2 pedagogy, as it opens potential avenues for neurocognitive interventions that could complement traditional forms of language instruction.
4. Method
4.1. Participants
Thirty right-handed participants between the ages of 18 and 23 (M = 20.3, SD = 1.41) were randomly assigned to either the sham (n = 15) or neurostimulation (n = 15) group. Participants were recruited through 300–400-level Spanish courses at a liberal arts college in the Southeastern United States, where they majored or minored in Spanish. Their degree requirements included completing an Oral Proficiency Interview (OPI) before graduation. All participants took the OPI within one semester of participating in the present study and achieved a proficiency level of Intermediate Mid. On the American Council on the Teaching of Foreign Languages (ACTFL) scale, Intermediate Mid corresponds to A2 in productive skills (e.g., OPI) and to B1.1 in receptive skills (ACTFL, 2024); the latter aligns with the proficiency level of the Diploma de Español como Lengua Extranjera (DELE) B1 receptive materials used in the comprehension tasks (see the Materials section). Participants had no metal implants in their head, were not taking psychiatric medication, were not pregnant, had no open wounds or inflammatory skin conditions, and had no known history of epilepsy or seizures. Finally, participants received monetary compensation for their time.
4.2. Procedure
All participants completed two 45-minute sessions (one pretest session and one posttest session). L2 learners in both conditions completed two listening comprehension tasks, a running memory task, and two reading comprehension tasks. Participants took the posttest session between 6 and 9 days apart (M = 7.27, SD = 1.30). In the pretest session, participants first completed a consent form and a health questionnaire adapted from Sreeraj et al. (Reference Sreeraj, Arumugham and Venkatasubramanian2023). The questionnaire screened for exclusionary conditions, such as a history of epilepsy or seizure disorders, the presence of metal implants or devices in or around the head (e.g., cochlear implants, brain stimulators, and aneurysm clips), and skin conditions or scalp injuries that could interfere with electrode placement. This is because tDCS could potentially pose safety risks, such as for skin conditions, metallic implants, or seizure disorders, or interfere with interpretation of results, such as with psychiatric medication (Nitsche et al., Reference Nitsche, Cohen, Wassermann, Priori, Lang, Antal, Paulus, Hummel, Boggio, Fregni and Pascual-Leone2008). Then, for the first 25 minutes, participants completed the tasks, followed by 20 minutes of anodal tDCS stimulation at 2.0 mA over the left DLPFC. Before administering tDCS, participants’ heads were measured with a 10–20 system cap to locate the F3 region, traditionally associated with the left DLPFC (Bludau et al., Reference Bludau, Eickhoff, Mohlberg, Caspers, Laird, Fox, Schleicher, Zilles and Amunts2014; Bruno et al., Reference Bruno, Lothmann, Bludau, Mohlberg and Amunts2024). The anode was placed on the F3 region, whereas the cathode was placed on the O2 region, as this area does not seem to intervene in language processing (Friederici, Reference Friederici2011). During the posttest session, participants’ heads were measured again to locate the F3 area, and then anodal tDCS stimulation was delivered for 20 minutes. While being stimulated by tDCS, participants played an action game involving no verbal information on a computer. In the game, a character is running through a location and must jump or duck to avoid obstacles while collecting coins. Participants controlled the character using a computer keyboard. This procedure is consistent with previous studies using video games during tDCS sessions to avoid participant attrition (e.g., Gold & Ciorciari, Reference Gold and Ciorciari2019). Participants then completed the linguistic tasks for the remaining 25 minutes. The order in which the comprehension tasks were presented was reversed for the posttest to avoid potential confounds.
The procedure for the sham group was conducted the same as that of the treatment group, except tDCS was delivered for only 30 seconds at the beginning and 30 seconds at the end of each 20-minute stimulation period. Given that the equipment used did not include a sham function, we administered a current for 30 seconds, then manually turned the device off, as previous studies have also done in a similar manner (Liu et al., Reference Liu, Tong, de Bruin, Li, He and Li2019; Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019). Finally, all stimulation procedures adhered to current safety guidelines outlined by Loo et al. (Reference Loo, Martin, Alonzo, Gandevia, Mitchell and Sachdev2011), ensuring compliance with recommended standards for current density, duration, and electrode preparation.
4.3. Materials
4.3.1. Equipment
tDCS machine. Stimulation was administered using the ActivaDose II device. This machine has a current range of 0–2 mA and a dose range of 0–80 V. There are two push-knob controls: one that operates the current and one that operates the dose. The device has automatic ramp-up and ramp-down, with changes occurring in 0.1 mA/minute increments. The automatic ramp-up and ramp-down can be manually overridden. In addition, it includes visual and auditory indicators for resistance limit, dose and current limits, and electrode rejection. Time calculations are performed automatically depending on the current and dose but can be changed during use by pausing or restarting. The device uses a 9 V DC Alkaline battery. Current was delivered through two 3 in. × 3 in. electrodes, one anodal and one cathodal, which held sponges soaked in a 0.9% NaCl isotonic solution as the conductive medium. The electrode area was 58.06 cm2, resulting in a current density of 0.034 mA/cm2. The device automatically ramped the current from 0 to 2.0 mA at the start of each session and ramped down at the end of stimulation. If contact issues arose, then a highly conductive electrode gel specifically designed for neuroimaging was applied on the center of the anodal sponge to improve sponge-to-skin contact for better conductance or to reduce irritation a participant may have felt.
4.3.2. Running memory task
This verbal WM updating task was originally designed by Pollack et al. (Reference Pollack, Johnson and Knaff1959) and was computerized for a study conducted by Durand-López (Reference Durand-López2024a). The present study employed the same computerized task. Participants viewed a sequence of digits, presented one by one, of a length they were unaware of. After the sequence ended, participants were asked to recall the last N digits they saw, which required them to continuously update the digits held in memory as new ones appeared. The minimum sequence was N (the number of digits participants were asked to recall), and the maximum sequence was 13. The N value increased adaptively based on participant performance every two trials: at Level 1, sequences ranged from 1 to 13 digits with N = 1; at Level 2, sequences ranged from 2 to 13 digits with N = 2; subsequent levels followed the same logic up to Level 13. Digits within each sequence were randomly generated with two constraints: no two consecutive digits could be identical (e.g., 5 and 5), and no two consecutive digits could be sequentially adjacent (e.g., 7 and 8). Each digit was displayed for 1,125 ms, with a 600-ms interstimulus interval. No time limit was imposed on responses. Before the experimental trials, participants completed three practice trials with feedback. The task was completed once two consecutive trials were answered incorrectly. The task took approximately 5 minutes. Scoring was determined by the number of correct trials completed.
4.3.3. Listening comprehension task
Participants listened to two audio recordings, both of which were taken from DELE exams administered by the Instituto Cervantes (2012). The first audio, which was 3 minutes long, featured an Argentine woman discussing her love for skiing and how she had made a career out of it in Spain. This audio was the second task in the listening comprehension section of the 2012 B1 DELE exam. The second audio was 1 minute and 27 seconds long and featured a conversation between a man and a woman about their friend who is getting married soon. This audio was the fifth task in the listening comprehension section of the B1 DELE exam administered by the Instituto Cervantes (2012). Each audio was followed by five multiple-choice questions, with three answer choices each (see Supplementary Appendices A and B). Each question was initially presented for 5 seconds without the answer choices, which then appeared afterward. Responses were untimed, and the questions focused on general comprehension of the audio recordings, including details such as locations, occupations, and the speakers’ actions.
4.3.4. Reading comprehension task
Participants read two passages, both of which were taken from the DELE exam. The first passage discussed the importance of seashells in pre-Hispanic Mexico and how an archeologist spent many years researching the production of the seashells through his workshop. This passage was the second passage of the reading comprehension section of the 2012 B1 DELE exam. The second passage covered the life of a famous singer, director, and actress, Ana Belén. This passage was the second passage of the 2013 B1 DELE exam. Each passage was followed by six questions, with three answer choices each, pertaining to the readings (see Supplementary Appendices C and D). Participants had unlimited time to answer the questions, and the questions and answer choices all remained on the screen for the entire time. The questions focused on general comprehension of the passage, such as information about people or objects.
4.3.5. Edinburgh Handedness Inventory
The Edinburgh Handedness Inventory (EHI) is a 20-item assessment of handedness as used in daily life. The items are self-rated, in that the individual indicates which hand they prefer to use for a given task. Some of the tasks include writing, using a toothbrush, using a hammer, striking a match, and dealing cards (Oldfield, Reference Oldfield1971). In the present sample, all participants demonstrated a right-hand preference, with EHI scores ranging from 50–100 (M = 88.66, SD = 14.55). In tDCS studies targeting the left DLPFC, handedness control is essential because right- and left-handed individuals may differ in cortical organization and functional lateralization. These differences can influence baseline activation patterns and stimulation response, potentially confounding the interpretation of outcomes. In addition, research has demonstrated that handedness impacts language processing in electrophysiological tasks (Grey et al., Reference Grey, Tanner and Van Hell2017; Ogunniyi et al., Reference Ogunniyi, Abugaber, Finestrat, Luque and Morgan-Short2021). While handedness serves as a convenient proxy for hemispheric dominance in language functions, it remains imperfect. Left-hemisphere language dominance appears in 80%–96% of right-handed people but only 70%–80% of left-handed individuals, with the remainder showing right-hemisphere or bilateral dominance (Knecht et al., Reference Knecht, Deppe, Dräger, Bobe, Lohmann, Ringelstein and Henningsen2000; Mazoyer et al., Reference Mazoyer, Zago, Jobard, Crivello, Joliot, Perchey, Mellet, Petit and Tzourio-Mazoyer2014). The ideal approach, using functional magnetic resonance imaging (fMRI) language lateralization tasks to precisely determine individual hemispheric dominance, is typically impractical in most research contexts due to high costs, limited access, and logistical difficulties associated with large sample scanning.
4.4. Statistical analysis
All analyses were conducted in R (R Core Team, 2024). To address whether tDCS enhances listening and reading comprehension, we fit Poisson regression models predicting the number of correct answers in the respective tasks. Poisson regression was selected because the dependent variable consisted of count data (number of correct responses) and the relatively small number of trials per task limited the practical advantage of binomial logistic regression for these comparisons. The baseline (intercept) for the models was the treatment group at pretest. The model then predicted the change in the predicted number of correct answers for the other three combinations per one unit change in the outcome (treatment group-post, sham group-pre, and sham group-post). A Poisson regression was also fit for WM scores as a function of the same predictors. In all models, alpha was set to .05. Post hoc comparisons were estimated using the emmeans package (Lenth, Reference Lenth2025) with Tukey correction for comparing a family of four estimates.
5. Results
Overall, the results provide evidence that tDCS enhances L2 listening comprehension but not reading. Likewise, there was no evidence for improved verbal WM.
5.1. Listening comprehension
Figure 1 shows the quantity of correct answers on the listening test by both groups at two test times. Table 1 shows the model effects relative to the baseline (treatment at the pretest). There was a large effect for the treatment group (d = 1.34) going from pretest to posttest, which was significant according to the model (estimate = 0.50, SE = 0.17, p = .004). For the sham group, the data showed a small descriptive effect for going from pretest to posttest (d = 0.21). This effect was not significant according to the model in post hoc comparisons (estimate = 0.10, SE = 0.17, p = .93).
Number of correct responses on the listening tasks by each group at both test times.
Figure 1. Long description
The x-axis is labeled Test with two categories: pre and post. The y-axis is labeled Listening Score, ranging from 2.5 to 7.5. For the pre-test, the control group (red) has a median near 4, interquartile range from about 2.5 to 5.5, and whiskers from about 1.5 to 7. The treatment group (blue) at pre-test has a median near 3.5, interquartile range from about 2.5 to 5, and whiskers from about 1.5 to 7. At post-test, the control group has a median near 5.5, interquartile range from about 3.5 to 7, and whiskers from about 2 to 8. The treatment group at post-test has a median near 7.5, interquartile range from about 6 to 8, and whiskers from about 3.5 to 8.5. The legend at the right identifies red as control and blue as treatment. The main trend is that the treatment group shows a substantial increase in listening scores at post-test, while the control group shows a smaller increase.
Model effects for listening comprehension
Table 1. Long description
From the top row downward, the table lists four conditions: Intercept, Sham posttest, Treatment posttest, and Sham pretest. Each row contains five columns: Estimate, S E, z, and p. The Intercept row shows Estimate 1.44, S E 0.13, z 11.05, p less than .001. Sham posttest has Estimate 0.13, S E 0.18, z 0.74, p .46. Treatment posttest displays Estimate 0.50, S E 0.17, z 2.91, p .004. Sham pretest shows Estimate 0.03, S E 0.18, z 0.15, p .88. The Treatment posttest row is the only condition with a statistically significant effect, indicated by p .004.
The estimated power for the treatment group was high (0.94), indicating adequate sensitivity to detect the observed effect. In contrast, statistical power for the sham group was very low (0.09). This outcome is consistent with the absence of meaningful pre–post differences in the sham group. To reliably detect an effect of d = 0.21 in the sham group, a sample size of 366 participants per group would have been required. With the actual sample size of 15 per group, a power level of 0.80, and α = 0.05, the minimum detectable effect size in the present experiment was d = 1.06.
5.2. Reading comprehension
Figure 2 shows the quantity of correct answers on the reading test by both groups at two test times. Table 2 shows the model effects relative to the baseline (treatment at pretest). Unlike listening, there was no strong evidence that reading improved in any comparison.
Number of correct responses on the reading tasks by each group at both test times.
Figure 2. Long description
X axis labels are pre and post, representing test times. Y axis is labeled Reading Score, ranging from 0 to 10. For pre-test, the control group (red) has a median near 5, interquartile range from 4 to 6, and whiskers from 3 to 8. The treatment group (blue) at pre-test has a median near 4, interquartile range from 3 to 6, and whiskers from 2 to 8. At post-test, the control group median is about 6, interquartile range from 5 to 7, whiskers from 4 to 8. The treatment group at post-test has a median near 7, interquartile range from 5 to 8, whiskers from 3 to 10. The legend at the top right identifies red as control and blue as treatment. The treatment group shows a greater increase in median and range from pre to post compared to the control group.
Model effects for reading comprehension
Table 2. Long description
From the top row, the intercept has an estimate of 1.57, S E of 0.12, z value of 12.82, and p value greater than point zero zero one. The next row, sham posttest, shows estimate 0.16, S E 0.16, z 0.96, p 0.34. Treatment posttest follows with estimate 0.26, S E 0.17, z 1.49, p 0.14. The final row, sham pretest, has estimate 0.04, S E 0.17, z 0.26, p 0.79. All values are aligned in columns labeled estimate, S E, z, and p.
The data showed a medium descriptive effect for the treatment group going from pretest to posttest (d = 0.68), which was not significant according to the model (estimate = 0.26, SE = 0.17, p = .14). The data showed a small-to-medium descriptive effect for the sham group going from pretest to posttest (d = 0.43). This effect did not reach significance in the pairwise comparisons (estimate = 0.11, SE = 0.16, p = .89).
Based on a sample of 15 per group, a power level of 0.8 and an alpha of 0.05, the minimum detectable effect was d = 1.06. The power of the pre–post comparison for both groups was low (sham = 0.21; treatment = 0.44). To achieve a statistically powered sample, 84 participants per group would have been needed in the sham group, and 35 participants would have been needed in the treatment group. As a result, it is possible that there is a true effect of reading, but that it is smaller than listening, and more data were needed to detect it with confidence.
5.3. Working memory
Figure 3 illustrates the number of correct responses on the WM task across both groups at the two testing periods. Model effects in relation to the baseline (treatment at pretest) are presented in Table 3. Similar to the reading task, but unlike the listening task, neither group demonstrated significant improvement.
Number of correct responses on the working memory task by each group at both test times.
Figure 3. Long description
The x-axis is labeled Test with two categories: pre and post. The y-axis is labeled Working Memory Score, ranging from 0 to 15. For each test time, two boxplots are shown: control group in red and treatment group in blue-green. At pre-test, the control group has a median near 9, interquartile range from about 7 to 12, and whiskers from about 5 to 14. The treatment group at pre-test has a median near 6, interquartile range from about 5 to 8, whiskers from about 4 to 10, and one outlier near 10. At post-test, the control group has a median near 8, interquartile range from about 6 to 11, and whiskers from about 4 to 14. The treatment group at post-test has a median near 8, interquartile range from about 6 to 10, and whiskers from about 4 to 13. The legend on the right identifies red as control and blue-green as treatment.
Model effects for working memory
Table 3. Long description
From the top row, the intercept has an estimate of 1.88, S E of 0.11, z value of 17.31, and p value less than .001. The next row, model factor post control, shows estimate 0.15, S E 0.15, z 1.01, p .31. The following row, model factor post treatment, has estimate 0.13, S E 0.15, z 0.89, p .37. The final row, model factor pre control, displays estimate 0.25, S E 0.14, z 1.78, p .07.
The treatment group exhibited a modest descriptive effect from pretest to posttest (d = 0.37), although this effect did not reach statistical significance in the model (estimate = 0.13, SE = 0.15, p = .37). The sham group showed a small negative descriptive effect from pretest to posttest (d = −0.27), which was also nonsignificant according to pairwise comparisons (estimate = −0.10, SE = 0.13, p = .85).
Statistical power was limited for both comparisons. For the treatment group, power was 0.16, suggesting that 116 participants per group would have been required to achieve a power level of 0.80. Similarly, the sham group comparison had a power of 0.11, indicating that 215 participants per group would have been needed to reach adequate statistical power. Given the actual sample size of 15 per group, with a power target of 0.80 and alpha of 0.05, the minimum detectable effect size was d = 1.06.
6. Discussion
The present study investigated whether anodal tDCS over the left DLPFC could improve L2 reading and listening comprehension in intermediate Spanish learners. Previous cognitive training research has shown that WM training can enhance L2 processing (e.g., Colflesh et al., Reference Colflesh, Karuzis and O’Rourke2016; Durand-López, Reference Durand-López2024a), whereas tDCS has been found to improve language switching (Li et al., Reference Li, Liu, Pérez and Xie2018) and L2 production (Borodkin et al., Reference Borodkin, Gassner, Ershaid and Amir2022). Given that WM is typically associated with the left DLPFC and that WM training is time-consuming and effortful, we aimed to directly stimulate this brain region to determine whether it could facilitate L2 comprehension more efficiently. The DLPFC is involved in a wide range of cognitive functions that support language comprehension, making it a suitable target for stimulation rather than more narrowly specialized areas. We predicted that anodal stimulation over the left DLPFC would improve L2 comprehension in both reading and listening. Results partially supported this hypothesis, showing significant improvements in listening but not reading. The following sections discuss possible interpretations of these findings.
6.1. Reading comprehension
Results indicated no significant effects of anodal tDCS on L2 reading comprehension. The absence of significant effects on reading comprehension performance raises several important considerations. One possibility is that our stimulation parameters did not effectively target the neural regions most critical for written language processing. Research suggests that typical reading engages left-hemisphere frontal and temporal areas, whereas struggling readers show bilateral activity across frontal, temporal, parietal, and occipital regions (Pollack et al., Reference Pollack, Luk and Christodoulou2015). In the context of L2 reading, which is more effortful than reading in the L1, learners may similarly recruit additional right-hemisphere regions. Because only the left DLPFC was stimulated, it remains unclear whether stimulating right-hemisphere cortical regions might differentially affect L2 reading comprehension.
Reading comprehension in an L2 likely engages different cognitive processes than listening comprehension, relying more on orthographic decoding and mapping text to meaning. Reading typically allows learners to control the pace of information processing and revisit content as needed (Lund, Reference Lund1991). In this study, the untimed reading task may have reduced WM demands, facilitating transfer from short-term to long-term memory during the comprehension questions. Since we targeted the left DLPFC, an area associated with WM, it is possible that the lower cognitive load in this task limited the potential for tDCS to enhance performance. In addition, orthographic similarities between English and Spanish, such as cognates, could have supported comprehension (Sun-Alperin & Wang, Reference Sun-Alperin and Wang2011). Despite these advantages, pretest reading scores averaged only 60%, suggesting that the task remained challenging. Because participants had room for improvement, other factors beyond pacing and orthographic support likely influenced reading performance.
Another possible explanation for the results is that reading comprehension depends in part on visuospatial WM. Reading requires processing letter shapes, line placement, and spatial organization of text, yet the present study did not assess visuospatial WM. Neuroimaging studies have shown activation in right-hemisphere prefrontal, occipital, parietal, and premotor regions during visuospatial processing (Jonides et al., Reference Jonides, Smith, Koeppe, Awh, Minoshima and Mintun1993). The absence of right-hemisphere stimulation may therefore help explain the lack of effects observed, given the essential involvement of visuospatial mechanisms in written comprehension. At the same time, verbal WM has been shown to modulate reading comprehension (Durand-López, Reference Durand-López2024a, Reference Durand-López2024b; Sagarra, Reference Sagarra2017). Reading thus relies on both WM subsystems, verbal and visuospatial, and future studies could investigate whether concurrent stimulation of both systems enhances reading performance.
6.2. Listening comprehension
Results indicate significant effects of anodal tDCS on listening comprehension. The significant improvement in listening comprehension following tDCS is consistent with the possibility that anodal stimulation of the left DLPFC modulated neural networks supporting auditory language processing. Auditory language processing has been found to be processed in multiple areas of the brain, not solely in the DLPFC. Wernicke’s area has traditionally been associated with auditory speech perception and comprehension (Okada & Hickok, Reference Okada and Hickok2006; Yi et al., Reference Yi, Leonard and Chang2019). In addition, a ventral processing route that connects the temporal and frontal lobes has been found to be important in speech recognition and comprehension (Mirman et al., Reference Mirman, Chen, Zhang, Wang, Faseyitan, Coslett and Schwartz2015). Finally, it is also important to note that we administered stimulation to a location associated with executive functioning; it is possible language comprehension could involve executive functioning, as some previous studies have indicated (Wang, Reference Wang2015). These findings together suggest that the brain is operating in a holistic manner during L2 processing, recruiting resources from multiple areas, possibly hosting other mental abilities. This should be taken into consideration when investigating the efficacy of cognitive training programs used to facilitate L2 processing.
Another factor that could have led to significant results is that listening comprehension in an L2 may pose different and possibly more difficult challenges compared with reading. These challenges include processing speaker accent, speech rate, and prosodic features (Bloomfield et al., Reference Bloomfield, Wayland, Rhoades, Blodgett, Linck and Ross2011), which may explain why there was greater potential for improvement in this task. With an average pretest score of around 45%, there was greater potential for development, whereas the opposite was true for the reading task.
6.3. Working memory
Interestingly, we did not observe corresponding behavioral improvements on the running memory task, despite the known association between the DLPFC and WM function. One possibility is that tDCS may have modulated WM at a neurophysiological level without producing measurable behavioral gains, a pattern reported in previous research. For example, Salmi et al. (Reference Salmi, Vilà-Balló, Soveri, Rostan, Rodríguez-Fornells, Lehtonen and Laine2019) showed that although WM training did not lead to far transfer in behavioral tasks, it produced measurable changes in neural dynamics, including modulation of the P2–N2–P3 complex, with reduced load effects after training, and task-dependent changes in slow-wave activity during maintenance. These neural changes were interpreted as reflecting more efficient processing under low demand and improved active maintenance under high demand. These findings suggest that neurocognitive adaptations within WM-related networks may precede or even occur independently of overt behavioral change. In the present study, similar neural-level modulation cannot be ruled out; however, because no electrophysiological measures were collected, this possibility cannot be directly evaluated.
It is also possible that while tDCS did not result in behavioral improvements in WM, it may have enhanced inhibitory control, a cognitive ability essential for minimizing interference from a learner’s L1. By suppressing automatic native-language responses, learners can focus on L2 structures and process the target language more effectively (Luque & Morgan-Short, Reference Luque and Morgan-Short2021). As mentioned above, tDCS has been found in some instances to facilitate inhibitory control when language switching (Li et al., Reference Li, Liu, Pérez and Xie2018; Tong et al., Reference Tong, Kong, Wang, Liu, Li and He2019). If inhibitory control is being enhanced, it may in turn allow for more accurate listening comprehension in the L2. Given that no measures of inhibitory control were included in the present study, this possibility cannot be discarded.
Taken together, the pattern of results observed in this study can be cautiously interpreted within the bilingual adaptive control framework proposed by Green and Abutalebi (Reference Green and Abutalebi2013; see also Calabria et al., Reference Calabria, Costa, Green and Abutalebi2018). Within this model, the left prefrontal cortex is primarily implicated in conflict resolution through response selection, a process that remains relevant even in the absence of explicit language switching. Importantly, prior work within this framework has shown that activity in the left DLPFC is modulated by L2 proficiency and exposure, with greater engagement observed for less proficient or less practiced linguistic systems (Calabria et al., Reference Calabria, Costa, Green and Abutalebi2018). In the present study, participants’ pretest performance was higher in reading than in listening comprehension, suggesting that reading, overall, might have been a more established skill than listening in this L2 population. From this perspective, anodal stimulation of the left DLPFC may have been associated with greater improvement in the comprehension task that placed greater demands on response selection mechanisms due to lower skill-specific proficiency (i.e., L2 listening). In contrast, L2 reading comprehension, being relatively more automatized, may have required less engagement of left DLPFC resources, rendering it less susceptible to neuromodulatory enhancement. Under this interpretation, the observed effects might not be driven by modality per se or by global L2 proficiency but by differential control demands associated with skill-specific proficiency levels in the L2.
6.4. Limitations
While this study provides a first approximation of whether anodal tDCS might enhance L2 listening comprehension, there are a few factors that may limit the findings. The sample size was small, likely influenced by factors such as the requirement to complete two sessions, the potentially intimidating nature of tDCS, and exclusion criteria like psychiatric medication use or inflammatory skin conditions. These factors may have contributed to challenges in recruiting more participants. Despite the small sample, the within-subjects design and the inclusion of a control group may help support the reliability of the findings. A potential drawback of the within-subjects design is learning effects; however, the pre- and posttest sessions were spaced an average of 7.27 days apart, which likely reduced the chance that participants remembered material or developed strategies between sessions. Moreover, the power analysis for listening comprehension suggests that significant effects could potentially be detected even with a small sample, given that the effects are strong.
Another consideration is the study’s limited ecological validity. While the listening and reading tasks were drawn from an official L2 exam, real-world reading and conversations are typically longer and more engaging. For example, Tai and Chen (Reference Tai and Chen2021) found that participants in immersive virtual reality conversations showed better comprehension, retention, and engagement than those who only watched a video. This suggests that real-life interactions may facilitate listening comprehension more effectively than traditional tasks, and a more realistic approach could potentially yield different results.
A final limitation of the present study concerns the interpretation of the differential effects observed across the listening and reading comprehension tasks. Although both tasks were selected from the same proficiency level (B1) of the DELE examination and from the same examination year, and were comparable in length, the materials were not linguistically matched on fine-grained properties such as lexical frequency, syntactic complexity, discourse structure, or inferential demands. Consequently, differences observed across the two tasks might not be attributed exclusively to modality-specific processing. Future research may address this issue by employing more tightly matched materials across comprehension modalities (e.g., parallel auditory and written versions of the same content); however, potential repetition effects that may arise when identical content is presented within the same experimental session should also be considered.
7. Conclusion
This study examined the effects of anodal tDCS over the left DLPFC on L2 reading and listening comprehension in English intermediate learners of Spanish. Two 20-minute sessions of stimulation produced significant gains in listening but not reading comprehension, despite the absence of measurable improvement in verbal WM. One tentative interpretation is that listening was the less established skill in this sample (as reflected in lower pretest performance) and may therefore have recruited prefrontal control resources more heavily, rendering it more responsive to neuromodulation than the more automatized reading task. Modality-specific factors may have additionally played a role, as listening engages distributed temporo-frontal networks that interface with the DLPFC, whereas reading additionally draws on visuospatial components subserved by right-hemisphere regions that were not targeted in the present montage; however, these possibilities are speculative and require direct empirical testing. The absence of behavioral gains in verbal WM is informative rather than contradictory: WM-related networks can undergo neurocognitive adaptation without producing detectable changes in standard behavioral measures, and stimulation may have additionally modulated cognitive control mechanisms, particularly inhibitory control, that support L2 listening by suppressing L1 interference under real-time processing pressure.
The present study offers an initial step toward understanding how electrophysiological methods can be applied to L2 processing. Our findings suggest that tDCS may serve as a low-effort intervention to enhance certain aspects of L2 comprehension, such as listening, providing benefits without the extensive time demands required by approaches such as WM training. At the same time, these results should be interpreted with caution, as further studies with larger samples and varied methodologies are needed to confirm and extend the present findings.
Supplementary Material
The supplementary material for this article can be found at http://doi.org/10.1017/S136672892610145X.
Data availability statement
The data that support the findings of this study are openly available at https://osf.io/e8trg/files/osfstorage?view_only=7ff0b2d1cf394a4481d2a255c7942157.
Author contribution
The authors are listed in alphabetical order. Both authors contributed equally to the development of this study and article.
Competing interests
The authors declare no competing interests.
Open access
