Highlights
-
• Mediators of bilingualism–EFL learning link are examined in kindergarteners.
-
• Phonological awareness, EF and motivation are assessed as mediators.
-
• Bilinguals outperformed monolinguals in noncognate receptive vocabulary.
-
• No bilingual advantage was observed for cognate words or English grammar.
-
• Suggested factors did not act as mediators of bilingualism–EFL learning link.
1. Introduction
Bilingualism has profound cognitive and linguistic consequences, shaping how individuals process and acquire languages. This phenomenon, known as the bilingual effect, refers to the human brain’s ability to adapt to complex linguistic environments in multilingual contexts (Bialystok, Reference Bialystok2017; D’Souza & D’Souza, Reference D’Souza and D’Souza2021; Leivada et al., Reference Leivada, Westergaard, Dunabeitia and Rothman2021; Masullo et al., Reference Masullo, Dentella and Leivada2024; Yee et al., Reference Yee, Yap, Korenar, Saddy and Pliatsikas2023). Bilingual effect extends to foreign language (FL) learning in school contexts, with a growing body of research showing that bilingual children often outperform their monolingual peers in this domain (e.g., Abu-Rabia & Sanitsky, Reference Abu-Rabia and Sanitsky2010; Cenoz & Valencia, Reference Cenoz and Valencia1994; Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Maluch et al., Reference Maluch, Kempert, Neumann and Stanat2015; Rauch et al., Reference Rauch, Naumann and Jude2011; Salomé et al., Reference Salomé, Casalis and Commissaire2022; Sanz, Reference Sanz2000; Tribushinina & Boz, Reference Tribushinina and Boz2025; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). Although several mechanisms have been proposed to explain the effect of bilingualism on FL learning, these mediators have not been systematically studied. Suggested mechanisms include enhanced metalinguistic awareness, improved cognitive abilities and positive attitudes toward learning new languages (see reviews by Cenoz, Reference Cenoz2003; Festman, Reference Festman2021; Hirosh & Degani, Reference Hirosh and Degani2018). Moreover, most research in this area has focused mostly on adults and secondary school children (e.g., Hofer & Jessner, Reference Hofer and Jessner2019; Lorenz et al., Reference Lorenz, Rahbari, Schackow and Siemund2020, Reference Lorenz, Toprak-Yildiz and Siemund2023, Reference Lorenz, Toprak-Yildiz and Siemund2024; Rauch et al., Reference Rauch, Naumann and Jude2011; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023), with a few studies investigating upper primary school children (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Maluch et al., Reference Maluch, Neumann and Kempert2016; Tribushinina & Boz, Reference Tribushinina and Boz2025; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). Thus, there is a significant gap in understanding how bilingualism influences FL learning during earlier developmental stages, such as the kindergarten years. The current study addresses this gap by investigating metalinguistic awareness, cognitive abilities and attitudes toward learning new languages as potential mediators in the relationship between bilingualism and FL proficiency. Specifically, it examines these factors in the context of English as a foreign language (EFL) learning among kindergarten children in the Netherlands.
1.1. The bilingual effect in FL learning
Bilingualism is associated with unique cognitive (Bialystok, Reference Bialystok2017; Chamorro & Janke, Reference Chamorro and Janke2022) and socioemotional processes (Han, Reference Han2010), which may facilitate FL learning. Supporting this notion, a growing body of evidence suggests that bilingual children often outperform their monolingual peers in FL performance (Abu-Rabia & Sanitsky, Reference Abu-Rabia and Sanitsky2010; Chachashvili-Bolotin & Kreiner, Reference Chachashvili-Bolotin and Kreiner2022; Geiss et al., Reference Geiss, Gumbsheimer, Lloyd-Smith, Schmid and Kupisch2022; Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Kopečková, Reference Kopečková2016; Maluch et al., Reference Maluch, Kempert, Neumann and Stanat2015, Reference Maluch, Neumann and Kempert2016; Maluch & Kempert, Reference Maluch and Kempert2017; Nguyen & Winsler, Reference Nguyen and Winsler2021; Rauch et al., Reference Rauch, Naumann and Jude2011; Schwartz et al., Reference Schwartz, Geva, Share and Leikin2007). For instance, Cenoz and Valencia (Reference Cenoz and Valencia1994) conducted one of the earliest studies on this topic, examining secondary school bilingual students learning a third language (L3), English, in the Basque Country, where both Basque and Spanish (L1 and L2) hold social significance and are integrated into the education system. They assessed English proficiency using a composite score that included vocabulary, grammar, listening and reading comprehension, as well as speaking and writing skills. The study compared the performance of Spanish-Basque bilinguals to that of their monolingual Spanish-speaking peers of the same age. The findings indicated that bilingual students outperformed monolinguals, and bilingual effect remained significant even after accounting for other factors such as intelligence, age, socioeconomic status, motivation and instructional exposure. In a similar vein, Maluch et al. (Reference Maluch, Kempert, Neumann and Stanat2015) examined the EFL achievement of German-speaking monolingual and bilingual sixth graders who speak a language other than German at home. Their findings indicated that bilingual students outperformed their monolingual counterparts on English proficiency measures including reading proficiency, vocabulary, grammar and spelling, even after controlling for cognitive abilities, age, gender and socioeconomic status. However, despite the generally positive trend linking bilingualism to enhanced FL learning, some studies report no such effects (Edele et al., Reference Edele, Kempert and Schotte2018; Lorenz et al., Reference Lorenz, Rahbari, Schackow and Siemund2020, Reference Lorenz, Toprak-Yildiz and Siemund2023, Reference Lorenz, Toprak-Yildiz and Siemund2024; Sanders & Meijers, Reference Sanders and Meijers1995; Stolvoort et al., Reference Stolvoort, Mackaaij and Tribushinina2024; Zoutenbier & Zwitserlood, Reference Zoutenbier and Zwitserlood2019). For example, Lorenz et al. (Reference Lorenz, Toprak-Yildiz and Siemund2024) conducted a longitudinal study investigating whether Russian- or Turkish-German bilingual students differed from their monolingual German-speaking peers in FL learning. Over three measurement points spanning 1.5 years, their findings revealed no consistent differences between the language groups. Notably, by the third measurement point, students in the Turkish-German bilingual group demonstrated lower English scores compared to both their Russian-German bilingual and monolingual peers.
The inconsistent findings in this area can be attributed to several factors. First, the type of bilingualism may influence the results. Research suggests that the bilingual advantage is more pronounced for additional language acquisition in balanced bilinguals compared to bilinguals who are dominant in one language (Cummins, Reference Cummins1979; Edele et al., Reference Edele, Kempert and Schotte2018; Lambert, Reference Lambert, Aboud and Mead1974). Second, the sociolinguistic environment plays a critical role. It has been argued that bilinguals are more likely to develop cognitive and linguistic advantages when both languages are actively supported within the education system (Jessner, Reference Jessner1999). For language-minority students, the majority language is typically fostered through formal schooling, whereas the minority language often receives little to no institutional support (Cenoz, Reference Cenoz2003). This disparity can result in disproportional language competencies, contrasting sharply with children enrolled in bilingual education programs. Third, the bilingual advantage appears to be more pronounced in younger age cohorts and tends to diminish as children grow older (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Maluch et al., Reference Maluch, Neumann and Kempert2016; Siemund & Lechner, Reference Siemund and Lechner2015). One possible explanation for this age-related difference is that older students often develop multilingual profiles through FL learning at school, which may narrow the gap between early bilinguals and monolinguals over time. Finally, the consideration of confounding variables is essential. Factors such as parental education (Bellocchi & Bonifacci, Reference Bellocchi and Bonifacci2023), non-verbal IQ (Lorenz et al., Reference Lorenz, Toprak-Yildiz and Siemund2023), proficiency in the majority language (Maluch et al., Reference Maluch, Kempert, Neumann and Stanat2015; Siemund et al., Reference Siemund, Lorenz and Toprak-Yildiz2024) and extramural exposure to foreign languages (Tribushinina, Boz, et al., Reference Tribushinina, Boz, Aalbers and Blom2024; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023) can significantly influence FL learning outcomes. Therefore, these variables may affect the comparability of results and should be carefully controlled in empirical studies. Yet, none of the previous research included these factors all together as control variables.
1.2. Potential sources of the bilingual effect in FL learning
Although the effects of bilingualism have been extensively studied, the mechanisms underlying the bilingual advantage remain underexplored. Researchers suggest that bilingualism influences FL acquisition through two pathways: direct transfer of knowledge and indirect mechanisms (Hirosh & Degani, Reference Hirosh and Degani2018). The positive transfer of vocabulary and grammar knowledge to a new language in bilinguals is well-documented (e.g., Kolb et al., Reference Kolb, Mitrofanova and Westergaard2022; Tribushinina, Boz, et al., Reference Tribushinina, Boz, Aalbers and Blom2024; Westergaard et al., Reference Westergaard, Mitrofanova, Mykhaylyk and Rodina2017). Indirectly, bilingualism may enhance FL learning by fostering metalinguistic awareness, improving cognitive skills and promoting positive attitudes toward language learning (Cenoz, Reference Cenoz2003; Hirosh & Degani, Reference Hirosh and Degani2018; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). However, these factors have not been systematically examined in sufficient depth. This section explores these three potential mechanisms as mediators of the relationship between bilingualism and FL learning performance.
The first individual factor that may explain the bilingual effect in FL learning is metalinguistic awareness. Broadly defined, metalinguistic awareness refers to “the ability to think about and reflect upon the nature and functions of language” (Pratt & Grieve, Reference Pratt, Grieve, Tunmer, Pratt and Herriman1984, p. 2). Research has consistently shown that bilinguals exhibit higher levels of metalinguistic awareness compared to monolinguals (Dolas et al., Reference Dolas, Jessner and Cedden2022; Hoffmann, Reference Hoffmann2001; Hofer & Jessner, Reference Hofer and Jessner2019). To date, only one study has explicitly examined metalinguistic awareness as a mediator in the relationship between bilingualism and FL learning. Rauch et al. (Reference Rauch, Naumann and Jude2011) investigated L1, L2 and L3 reading proficiency, as well as metalinguistic awareness, in German and Turkish-German secondary school students. Their findings revealed that fully biliterate students outperformed both monolingual and partially biliterate students in L3/EFL reading proficiency and metalinguistic awareness. Furthermore, within the group of biliterate students, the individual degree of biliteracy positively influenced L3 reading proficiency, and this effect was mediated by metalinguistic awareness. The mediating role of metalinguistic awareness in the link between bilingualism and FL learning may stem from bilinguals’ exposure to and management of two language systems. This experience compels them to compare and analyze structural aspects of language earlier and more deeply than monolinguals (Bruck & Genesee, Reference Bruck and Genesee1995). Furthermore, bilingualism requires individuals to coordinate two linguistic systems, focusing on relevant features of language input and output (Bialystok, Reference Bialystok, Kroll and De Groot2009; Sanz, Reference Sanz and Chappelle2012). Through these extended experiences and heightened metalinguistic awareness, bilingualism creates a favorable foundation for FL learning.
The second factor that may mediate the relationship between bilingualism and FL learning is executive function skills, which have been extensively studied in the context of multilingualism. Executive functions are higher-order cognitive processes that facilitate goal-directed actions, emotions and thoughts. These processes comprise three core components: inhibition, working memory and cognitive flexibility (Diamond, Reference Diamond2013; Miyake et al., Reference Miyake, Friedman, Emerson, Witzki, Howerter and Wager2000). First, inhibition involves the ability to control attention, behavior, thoughts and emotions by suppressing strong internal impulses or external distractions. This component includes selective attention – the capacity to focus on relevant information while disregarding irrelevant distractors – and sustained attention, which entails maintaining focus on a target over time (Diamond, Reference Diamond, Griffin, McCardle and Freund2016). Second, working memory refers to the ability to hold information in the mind and manipulate it mentally (Diamond, Reference Diamond2013). Third, cognitive flexibility encompasses the ability to shift perspectives or approaches to a problem, adapt to new rules or priorities and adjust flexibly to changing demands (Diamond, Reference Diamond, Griffin, McCardle and Freund2016). Research demonstrates that bilingual children tend to outperform their monolingual peers on tasks assessing executive functions, though findings are not always consistent (for reviews, see Lowe et al., Reference Lowe, Cho, Goldsmith and Morton2021; Yurtsever et al., Reference Yurtsever, Anderson and Grundy2023). For example, empirical evidence suggests that bilingual children exhibit advantages in inhibitory control (Morales et al., Reference Morales, Calvo and Bialystok2013; Nguyen et al., Reference Nguyen, Hutchison, Norvell, Mead and Winsler2024), working memory (Blom et al., Reference Blom, Küntay, Messer, Verhagen and Leseman2014; Morales et al., Reference Morales, Calvo and Bialystok2013), cognitive flexibility (Morales et al., Reference Morales, Calvo and Bialystok2013; Nguyen et al., Reference Nguyen, Hutchison, Norvell, Mead and Winsler2024) and selective attention (Blom et al., Reference Blom, Boerma, Bosma, Cornips and Everaert2017; Comishen et al., Reference Comishen, Bialystok and Adler2019) but not in sustained attention (Boerma et al., Reference Boerma, Leseman, Wijnen and Blom2017; Ebert et al., Reference Ebert, Rak, Slawny and Fogg2019). It has been argued that the regular cognitive demands of managing two languages serve as a form of intensive cognitive training. This training involves processes such as selecting the relevant language, inhibiting the non-target language, holding dual linguistic representations in mind and switching flexibly between languages (Antoniou, Reference Antoniou2019; Bialystok, Reference Bialystok2017). The lack of differences between bilinguals and monolinguals in sustained attention may be attributed to the nature of many sustained attention tasks, which often do not involve conflict resolution – a skill that is frequently exercised in bilingual contexts (Bialystok, Reference Bialystok, Kroll and De Groot2009). Furthermore, the continuous cognitive engagement required for bilingual language management may lead to enhanced executive function skills. Given that executive functions play a pivotal role in language learning (e.g., Gooch et al., Reference Gooch, Thompson, Nash, Snowling and Hulme2015; Shokrkon & Nicoladis, Reference Shokrkon and Nicoladis2022; White et al., Reference White, Alexander and Greenfield2017), it is plausible to hypothesize that executive function skills mediate the relationship between bilingualism and FL learning performance.
The third factor that may explain the bilingual advantage in FL learning relates to social factors, particularly motivation to learn a new language. Motivation is widely regarded as one of the most critical determinants of individual differences in FL learning (Dörnyei & Ryan, Reference Dörnyei and Ryan2015; Gardner, Reference Gardner2006; Kormos & Csizér, Reference Kormos and Csizér2008; Lamb, Reference Lamb2012). Leona et al. (Reference Leona, van Koert, van der Molen, Rispens, Tijms and Snellings2021) conceptualized motivation to learn English by integrating two influential theoretical frameworks. The first is the Socio-Educational Model of Second Language Acquisition, which encompasses motivation itself as well as attitudinal precursors, such as the learner’s attitude toward the learning environment, integrativeness (the learner’s desire to communicate with or integrate into the target language community), instrumental orientation (practical reasons for learning the language, such as securing better job opportunities, earning a higher salary, or passing an exam) and language anxiety (e.g., Gardner & Smythe, Reference Gardner and Smythe1975). The second framework is the Second Language Motivational Self-System, which highlights self-related beliefs and learners’ perceptions of themselves as successful second-language users. Key components of this model include the ideal L2 self (the learner’s idealized self-image as a competent L2 speaker), the ought-to L2 self (attributes learners feel they should possess), L2 learning experience and linguistic self-confidence (Boo et al., Reference Boo, Dörnyei and Ryan2015; Dörnyei, Reference Dörnyei2005, Reference Dörnyei, Dörnyei and Ushioda2009). Motivational factors, such as willingness to communicate, have been proposed as potential indirect pathways through which bilingualism influences FL learning (Hirosh & Degani, Reference Hirosh and Degani2018). Supporting this idea, studies involving adolescents and adults have shown that bilinguals experience greater enjoyment and lower anxiety when learning an FL compared to their monolingual peers (Botes et al., Reference Botes, Dewaele and Greiff2020; Dewaele, Reference Dewaele2007; Dewaele & MacIntyre, Reference Dewaele and MacIntyre2014). However, contrasting evidence comes from Lorenz et al. (Reference Lorenz, Toprak-Yildiz and Siemund2023), who found no significant differences in motivation to learn English between German monolingual secondary school students and their bilingual peers speaking German and Russian or Turkish. A possible explanation for this null finding could be related to the measure of motivation used in the study, as the questionnaire consisted of only four items, which may have limited its ability to capture the complexity of motivational factors. Additionally, Leona et al. (Reference Leona, van Koert, van der Molen, Rispens, Tijms and Snellings2021) demonstrated that motivational factors, such as the desire to learn English and linguistic self-confidence, positively predicted English performance in primary school children. Thus, it is plausible that bilingual children may exhibit higher motivation to learn an FL, which could, in turn, enhance their FL performance.
1.3. The bilingual effect at young ages
The effect of bilingualism on FL learning has been studied primarily in adults (e.g., Aguasvivas et al., Reference Aguasvivas, Cespón and Carreiras2024; Kaushanskaya & Marian, Reference Kaushanskaya and Marian2009b; Keshavarz & Astaneh, Reference Keshavarz and Astaneh2004) and secondary school students (e.g., Edele et al., Reference Edele, Kempert and Schotte2018; Lorenz et al., Reference Lorenz, Rahbari, Schackow and Siemund2020, Reference Lorenz, Toprak-Yildiz and Siemund2023, Reference Lorenz, Toprak-Yildiz and Siemund2024; Nguyen & Winsler, Reference Nguyen, Winsler, Kersten and Winsler2023; Rauch et al., Reference Rauch, Naumann and Jude2011), with fewer studies focusing on primary school children (e.g., Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Maluch et al., Reference Maluch, Kempert, Neumann and Stanat2015; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). Limited evidence suggests that the impact of bilingualism on FL learning may be stronger at younger ages. For example, Hopp et al. (Reference Hopp, Vogelbacher, Kieseier and Thoma2019) found that bilingual primary school children initially outperformed their monolingual peers in FL vocabulary and grammar learning, but this advantage diminished between grades 3 and 4. This pattern suggests that while bilingual children may have an early advantage in FL learning, monolingual children can catch up as they gain more exposure and proficiency over time. Given this potential age-related variability, it is crucial to examine whether the bilingual advantage in FL learning emerges even earlier, before primary school. However, research in preschool-aged children remains scarce. To our knowledge, Festman (Reference Festman2018) is one of the few studies investigating this question. In a language intervention program for preschoolers, which included instruction in the majority language (German), the home language for bilingual children, and English as a FL, bilingual children showed greater gains in German than their monolingual peers, but no such advantage was observed for English. These findings suggest that studying the bilingual effect in younger age groups has important implications for early childhood education policies and practices. It can provide insights into the potential benefits of introducing FL learning at an earlier stage, particularly for bilingual children, and inform the development of tailored teaching strategies that capitalize on young learners’ linguistic and cognitive strengths. Understanding these effects in kindergarten could ultimately help optimize early FL instruction and support long-term language learning outcomes.
1.4. Current study
The present study addresses two significant gaps in the field of bilingualism. First, it aims to explore potential mediators in the relationship between bilingualism and FL learning. This is the first study to examine the indirect effects of metalinguistic awareness, executive functions, and motivation to learn English on the link between bilingualism and FL learning outcomes. Second, the study investigates whether the bilingual advantage is evident at a young age, specifically during the kindergarten years. To address these aims, we focus on a sample of Dutch-speaking monolingual and bilingual kindergarten-aged children (4–6 years old), for whom English was part of the school curriculum from age 4. We assessed their English proficiency in both lexical and grammatical domains. It is known that monolingual and bilingual children benefit from cognateness to different extents, possibly because bilinguals may have a smaller vocabulary in the majority language, which could explain why the bilingual advantage tends to be smaller for cognate words (Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). Therefore, we included both Dutch-English cognate and noncognate words in the vocabulary assessment to examine whether the bilingual effect is stronger in one of these domains. Furthermore, we measured the metalinguistic awareness, executive functions, and motivation to learn English as potential mediators. Additionally, we control for several confounding variables, including age, parental education, non-verbal IQ, intensity of English instruction at school and exposure to English outside of school.
Research questions of the current study are twofold: (1) Does bilingualism in kindergarten-aged children confer an advantage in FL learning, specifically in English vocabulary and grammar? (2) Do metalinguistic awareness, executive functions and motivation to learn English mediate the relationship between bilingualism and FL learning outcomes? We hypothesize that bilingual children will exhibit greater metalinguistic awareness, enhanced executive functions and higher motivation to learn English compared to their monolingual peers. Furthermore, we expect that these enhanced linguistic, cognitive and motivational factors will predict superior performance in English vocabulary and grammar, thereby mediating the relationship between bilingualism and FL learning outcomes.
2. Method
This study was approved by the ethics assessment committee of the Faculty of Humanities at Utrecht University.
2.1. Participants
A total of 122 children aged 4 to 6 years were recruited from six preschools in the central and southern regions of the Netherlands. Three children were excluded because their home language was English and thus, disqualifying them as EFL learners. The final sample consisted of 85 Dutch-speaking monolinguals (Mage = 63.16 months, SD = 7.83; 39 female) and 34 bilinguals (Mage = 61.60 months, SD = 6.61; 15 female). The bilingual participants spoke Dutch in addition to the following languages: Arabic (n = 4), Polish (n = 3), Romanian (n = 3), German (n = 2), Croatian (n = 2), Spanish (n = 2), Turkish (n = 2) or one of the following: Amharic, Chinese, French, Hindi, Japanese, Moluccas, Papiamento, Persian, Portuguese, Surinamese, Telugu, Tigrinya and Vietnamese (each n = 1). Three children had two home languages: Hungarian and Spanish, Italian and Spanish and Czech and Hungarian. The average age of Dutch acquisition of bilingual children was 16.02 (from 0 to 48) months. Parents of bilingual children were asked about how many hours their children spend in home language/Dutch-related activities during the day. On average, the bilinguals were exposed to the home language 39% (range was 14 to 91%) of their time and to Dutch 61% of their time (see Exposure to English Activities below for more details).
In the Netherlands, children begin kindergarten at age 4, which is integrated into the primary school system and transition to grade 1 in September after turning 6. Children are eligible to enroll in kindergarten immediately after their fourth birthday, although attendance becomes compulsory at age 5. Children in kindergarten at age 4 and 5 usually share the same classroom. Kindergartens in the Netherlands generally do not offer English instruction as part of their curriculum, though they may choose to introduce English lessons. All kindergartens participating in this study received English lessons starting at age 4. A 4-year-old may have had only a few months of English instruction, whereas a 6-year-old could have received 2 years or more. Thus, age was taken as a proxy for the length of EFL instruction.
2.2. Materials
2.2.1. Outcome variables
2.2.1.1. Receptive vocabulary in English
To assess English word knowledge, we used the Receptive Vocabulary Test (RVT; Tribushinina et al., Reference Tribushinina, Niemann and Meuwissen2023) that was specifically developed for Dutch-speaking EFL learners at primary school. The task included 40 test items (20 nouns and 20 verbs), preceded by two practice items to familiarize participants with the format. Half of the words (n = 20) were Dutch–English cognates and the other half (n = 20) were noncognates. Cognates and noncognates were matched for word length and frequency in both English and Dutch. Word frequencies were based on lemma occurrences per one million words, retrieved from the COBUILD database for English and the Institute for Dutch Lexicology database for Dutch, via the CELEX lexical database. Test items were presented through PowerPoint slides, each showing four black-and-white images: one target image and three distractors. Participants listened to a target word while viewing the images and were provided with a printed booklet displaying the same items. They were instructed to cross out or circle the image that best represented the spoken word. Participants could take as much time as needed to respond. The position of the target image was counterbalanced across trials. Participants earned 1 point for each correct response, while incorrect, unanswered or multiple responses received 0 points.
2.2.1.2. Receptive grammar in English
We used the Test for Reception of Grammar-2 (TROG-2; Bishop, Reference Bishop2003) to assess participants’ understanding of English grammar. TROG is an 80-item standardized test for English-speaking children. We performed some adaptations to adjust it for children who learn EFL. We did not apply a stopping rule and started the test from the beginning instead of starting with the age-appropriate item. Furthermore, we administered the first 40 items rather than the whole set (80 items), as it was too long for the kindergarten children. Furthermore, the test items were arranged in the order of increasing difficulty, so we reckoned that the second half of the task would be too difficult for children in this age group. The first half of the task included 10 grammatical structures: Two-element sentences, negative, reversible in and on, three elements, reversible SVO, four elements, subject relative clauses, not only X but also Y, reversible above and below, comparative/absolute. Test items were presented via PowerPoint slides, each displaying four black-and-white images, one of which corresponds to the content of the sentence. The three pictures that do not correspond to the sentence are either lexical or grammatical distractors. While viewing the images, participants listened to a target sentence and used a printed booklet that mirrored the slides. They were instructed to cross out or circle the image that best matched the spoken sentence. Participants were allowed as much time as needed to respond, and the position of the target image was counterbalanced across trials. The task was preceded by a practice item to help participants become familiar with the format. Participants received 1 point for each correct answer, while incorrect, unanswered or multiple responses were scored as 0 points.
2.2.2. Mediators
2.2.2.1. Phonological awareness
To measure metalinguistic awareness, a phonological awareness task was deemed most appropriate for this age group. This type of task has been effective in assessing Dutch-speaking children aged 4 to 5 and can detect subtle differences between mono- and bilingual children (Goriot et al., Reference Goriot, Unsworth, van Hout, Broersma and McQueen2021). Similar results have been observed in other studies across various age groups and countries (e.g., Bialystok et al., Reference Bialystok, Majumder and Martin2003; Chen et al., Reference Chen, Anderson, Li, Hao, Wu and Shu2004; Kuo & Anderson, Reference Kuo and Anderson2010; Laurent & Martinot, Reference Laurent and Martinot2010). To account for potential disadvantages in multilingual children due to limited exposure to Dutch, we tested the phonological awareness in non-words. The task consists of two parts:
Part 1: Phoneme identification. To familiarize children with the task, we started the test with practice items including Dutch words. Participants were introduced to a character, “Puk,” who likes words that start with the same sound as her name. They were given an example word (“pen”) that matches the criterion, followed by two practice items (one matching, one non-matching). Then, they heard 12 practice words (balanced for syllable length and matching status) and were asked whether each word starts like “Puk.” For the test items, non-words were adapted from the Quasi-Universal Nonword Repetition Test (Q-U NWRT; Boerma et al., Reference Boerma, Chiat, Leseman, Timmermeister, Wijnen and Blom2015) and contained phonemes common to many languages. Two aliens, “Beb” and “Tum,” were introduced to avoid syllable repetition. Each alien was associated with six test items (with balanced syllable length and match status). The vowels in the alien names did not match the vowels in the test items to keep the task manageable. Children received 1 point if they correctly judged whether the words began with the same phoneme.
Part 2: Syllable combination and rhyme identification. In part 2, two tasks adapted from the CELF-Preschool-2-NL (Wiig et al., Reference Wiig, Secord and Semel2012) were administered: the syllable combination task and the rhyme identification task. Both tasks began with practice items to familiarize participants with the procedure. Each task included six practice items using Dutch words and two practice items using non-words, followed by four practice items with non-words. The test items featured non-words adapted from the Q-U NWRT, which were produced without stress to avoid language-specific prosodic patterns. In the syllable combination task, participants were asked to combine two syllables to form a word. During the test, the non-words were bi- and trisyllabic. Participants received 1 point for correctly combining syllables to form a word but did not earn points if there were pauses between the syllables when pronouncing the item. In the rhyme identification task, participants judged whether pairs of words rhymed. The test items for this task consisted of syllables in non-words. Participants earned 1 point for each correct judgment about whether the word pairs rhymed.
2.2.2.2. Motivation to learn English
The Motivation to Learn English Scale was adapted from Leona et al. (Reference Leona, van Koert, van der Molen, Rispens, Tijms and Snellings2021)‘s Young English Language learners’ (YELLs’) Motivation to Learn English Scale. The original scale was developed for primary school children and administered on a paper with 40 items on a 4-point scale. To make the scale more appropriate for kindergarten aged children, we reduced the number of items to 28 with 6 subscales: Interest in Foreign Languages (i.e., It is nice to learn a new language), Integrative Orientation (i.e., When you go to holiday, it is important that you can speak English), Ideal Second Language-Self (i.e., I want to speak English well when I am older), Desire to Learn English (i.e., I think learning English is good), English Use Anxiety (i.e., I dare to speak English with friends) and Linguistic Self-Confidence (i.e., I will write good English stories later).
Previous studies examining young children’s motivation often employed a method involving two stuffed animals with contrasting motivational belief profiles (Mata, Reference Mata2011; Schrodt et al., Reference Schrodt, Elleman, FitzPatrick, Hasty, Kim, Tharp and Rector2019). This approach minimized social desirability bias by presenting both choices neutrally (Baker & Scher, Reference Baker and Scher2002). In the current study, we used two dolls, which were identical except for the color of their necklace and belt. The experimenter explained that one doll represented positive motivational statements (e.g., I will speak English well later), while the other represented negative ones (e.g., I do not have to speak English well later). To prevent bias, the assignment of positive or negative statements to the dolls and the order of their introduction were counterbalanced across items. Children were asked to choose the doll they felt was most like them. Scoring was as follows: 3 points if the child selected the doll with the positive statement, 2 points if the child selected both dolls, 1 point if the child selected the doll with negative statement and 0 point if the child selected neither.
Factor structure of the motivation to learn English scale. We used Confirmatory Factor Analysis to test whether the items in the Motivation to Learn English Scale loaded onto a single global factor. Initially, the 28-item model did not yield a satisfactory fit. After removing two items with the lowest factor loadings (one from the Integrative Orientation subscale, “It is important to learn English because then I can easily interact with people from different countries,” and one from the Linguistic Self-Confidence subscale, “I can understand English movies very well”), the revised 26-item model demonstrated good fit: Comparative Fit Index (CFI) = .97, Tucker–Lewis index (TLI) = .97, Root Mean Square Error of Approximation (RMSEA) = .04, and Standardized Root Mean Square Residual (SRMR) = .09. Factor loadings for these items were significant and ranged from 0.49 to 0.86. The sample size was also suitable, as indicated by a Kaiser-Meyer-Olkin value of 0.85, which exceeds the recommended threshold of 0.70. Bartlett’s test of sphericity was significant (χ 2(325) = 1248.26, p < 0.001), indicating sufficient correlation among items, and the determinant of the R-matrix was 0.000002273, suggesting acceptable multicollinearity. Lastly, the Cronbach’s alpha for the 26-item scale was 0.92, demonstrating good internal reliability. Based on these results, we summed the scores of the 26 items to create a composite motivation to learn English score for each participant for use in further analyses.
2.2.2.3. Executive functions
Verbal working memory. Children’s verbal working memory was assessed using the Monkey Game, a computerized, self-guided task designed to test memory span (Van de Weijer-Bergsma et al., Reference Van de Weijer-Bergsma, Kroesbergen, Jolani and Van Luit2016). In this game, children listened to sequences of monosyllabic Dutch words (i.e., moon, fish). Afterward, they saw a 3x3 matrix of pictures representing the words they had heard and were required to click on the words in reverse order. The task included five levels, each with three sequences, totaling 15 sequences. The memory load increased across levels, starting with two words per sequence in Level 1 and progressing to six words in Level 5. There were no cut-off rules; all children completed all 15 sequences. Performance was scored as the proportion of the correctly recalled positions within a sequence. Cronbach’s α of the Monkey Game was α = .81 indicating an acceptable internal reliability.
Cognitive flexibility. Cognitive flexibility was measured using a Dimensional Change Card Sort task based on Bialystok and Martin (Reference Bialystok and Martin2004), Zelazo (Reference Zelazo2006), and Espinet et al. (Reference Espinet, Anderson and Zelazo2013). In this task, children sorted stimuli (e.g., a red bunny) into one of two categories based on either color or shape. For example, a red bunny could be matched with a blue bunny (same shape) or a red car (same color). Stimuli were presented on a laptop, and children responded using two buttons on a BeexyBox. Participants first completed one sorting game (e.g., the shape game) and then switched to the other game (e.g., the color game) midway through the task. The order of the games was randomized. The task began with a practice round of four items, repeated until participants made no errors, followed by an experimental round with 20 pre-switch and 20 post-switch items. Cognitive flexibility was assessed as the switch cost, calculated by subtracting post-switch accuracy from pre-switch accuracy. Higher switch cost values indicated lower cognitive flexibility.
Inhibition. Cognitive inhibition was assessed using a flanker task adapted from Engel Engel de Abreu et al. (Reference Engel de Abreu, Cruz-Santos, Tourinho, Martin and Bialystok2012) and Boerma and Blom (Reference Boerma and Blom2020). Children were shown a row of five fish and asked to quickly indicate the direction of the middle fish using two buttons on a BeexyBox. In congruent trials, all fish swam in the same direction, while in incongruent trials, the middle fish swam in the opposite direction from the others. The task began with eight practice trials, repeated until children made no more than two errors, followed by 40 experimental trials. Stimuli were displayed for 5000 ms, and failure to respond within this time was recorded as incorrect. Accuracy was measured separately for congruent and incongruent trials. The flanker effect on accuracy was calculated by subtracting the proportion of correct incongruent trials from the proportion of correct congruent trials. Higher flanker effect scores indicated weaker inhibitory control.
Selective attention. Selective attention was assessed using the Visual Sky Search, a subtest of the Test of Everyday Attention for Children (Manly et al., Reference Manly, Robertson, Anderson and Nimmo-Smith1999). The task consisted of two parts. In the first part, children were presented with an A3 sheet displaying 20 identical pairs of spaceships (targets) and 108 non-identical pairs (distractors). Their task was to encircle the targets as quickly as possible while ignoring the distractors. Children were instructed to say “stop” once they believed they had completed the task. To account for motor speed and drawing ability, a second A3 sheet was administered after the first task. This motor-control sheet displayed only the 20 identical targets without distractors, and children were again instructed to encircle the targets as quickly as possible. The selective attention score was calculated by subtracting the average time per target on the motor-control sheet from the average time per target on the distractor sheet. This score reflects the additional time required to focus attention while filtering out distractors.
Factor structure of executive functions. We tested whether scores from the verbal working memory, inhibition, cognitive flexibility, and selective Attention tasks would load onto a single global Executive Function factor. The results did not indicate a satisfactory model fit, and the task scores did not load significantly onto a common factor. Therefore, we opted to include the executive function task scores as individual variables in subsequent analyses. This approach aligns with common practices in research on bilingualism and executive functions (Bialystok, Barac, et al., Reference Bialystok, Barac, Blaye and Poulin-Dubois2010; Nguyen et al., Reference Nguyen, Hutchison, Norvell, Mead and Winsler2024).
2.2.3. Control variables
2.2.3.1. Dutch proficiency
Since some tasks, such as the Monkey Game and phonological awareness tasks, rely on comprehension of Dutch, children with higher proficiency in Dutch may have an advantage. To account for differences in Dutch proficiency among participants, we assessed their abilities using the Dutch version of the LITMUS Sentence Repetition Task (Marinis & Armon-Lotem, Reference Marinis and Armon-Lotem2015). The Litmus Sentence Repetition Task is considered an appropriate instrument for measuring language proficiency at multiple levels, including lexical knowledge, parsing of spoken sentences, speech production and, particularly, grammatical ability (Klem et al., Reference Klem, Melby-Lervåg, Hagtvet, Lyster, Gustafsson and Hulme2015; Polišenská et al., Reference Polišenská, Chiat and Roy2015). This task evaluates participants’ grammatical proficiency across various structures, including complements, modal verbs and both short and long passives (Jong et al., Reference Jong, Blom and van Dijk2021). The task was designed as a “Treasure Hunt” game and administered via a PowerPoint presentation. It began with two practice sentences, followed by 30 slides, each presenting a pre-recorded sentence that the child was asked to repeat. Children’s responses were recorded and later transcribed by trained research assistants. The task took approximately 8 min. Child responses were scored from 0 to 3; 0 indicating that the child’s response contained 4 or more changes to the target sentence, 1 indicating two or three changes, 2 indicating only one change and 3 indicating no changes to the target sentence.
2.2.3.2. Non-verbal IQ
The Raven Standard Progressive Matrices–2 (SPM) for the age group of 4:0–8:11were used (Raven et al., Reference Raven, Court and Raven1996) as a measure of nonverbal intelligence. The Raven SPM consists of three series (A to C) of 12 diagrams or designs with one part missing. Children are asked to select the correct part to complete the designs from among five answer options printed beneath. Children have to decide which of the alternatives given logically completes the design. The answers were scored as incorrect (0) and correct (1). The total number of correct responses formed the raw score, which corresponded to a standardized score. Subsequently, the standardized scores were converted into an age-appropriate IQ score.
2.2.3.3. Exposure to English activities
Amount of out-of-school exposure to English was collected through parental questionnaires developed for the current study. Parents indicated how much time their children spent on 14 activities (e.g., talking to friends, reading books, gaming) in English, Dutch and home language during weekdays and weekends. An average score of how many minutes children spend on English activities in a day was calculated. Exposure to English activities questionnaire is provided in Supplementary Appendix A.
2.2.3.4. Intensity of instruction
The amount of time schools dedicated to English-related activities each week varied. To gather this information, we asked teachers to report the average number of hours they spent on English-related activities per week. On average, schools provided 1.75 (SD = 1.5) hours of English instruction ranging between 1 and 4.
2.2.3.5. Parental education
Parents were asked to report the highest level of education achieved by each parent and/or caregiver, selecting from six options: primary school (1), secondary school (2), vocational education (3), higher vocational education (4), university (5) and master’s or PhD (6). The numbers in parentheses represent the score assigned to each education level. The scores for both parents were summed to calculate a total. If information was available for only one parent, the reported score was doubled to account for the missing data.
2.3. Procedure
The participants were tested between March and June of 2024. Generally, the tests were completed over two or three sessions, though no specific order was placed on the administration of the tests due to practical considerations. On average, completing the full test battery took about 2 hours. All tasks were conducted individually with the exception of the TROG and RVT, which were completed in small groups of two to five children. All tests took place in a quiet space at the child’s school. The parental questionnaires were taken home by the children with a request for the completed questionnaires to be returned to the teacher. If a questionnaire was not returned, the teacher personally provided parents with a second copy. In cases where the questionnaire remained missing, the experimenters orally asked the parents a selection of questions about the child’s exposure to Dutch, English and their home language.
3. Data analysis strategy
To address the research questions, we conducted two separate path analyses, each with a distinct dependent variable: receptive vocabulary scores and receptive grammar scores. In both models, Language Group (monolingual = 1, bilingual = 2) was included as an independent variable. Phonological awareness, motivation to learn English, verbal working memory, inhibition, cognitive flexibility and selective attention were added as mediators. Additionally, English exposure, intensity of instruction, child age, child IQ and parental education were included as control variables influencing the dependent variables. Child age was also controlled for in relation to verbal working memory and selective attention, given its correlation with these task scores (see Table 2). Similarly, Dutch proficiency scores were controlled in relation to verbal working memory and phonological awareness due to significant correlations (see Table 2). We began with fully saturated models and adopted an iterative model refinement strategy, which involved pruning non-significant paths while adding theoretically or empirically supported relations to improve model fit. This process focused on systematically removing the least significant paths (based on p-values) among control variables, mediators and the dependent variables, continuing until a well-fitting model was achieved (Landis et al., Reference Landis, Beal and Tesluk2000; Molzon et al., Reference Molzon, Mullins, Cushing, Chaney, McNall and Mayes2018; Streiner, Reference Streiner2006). To further enhance model fit, we allowed for meaningful covariations among mediators and control variables where significant inter-variable correlations were observed, for example, between phonological awareness and inhibition. We also specified a covariation between language status and both English exposure and Dutch proficiency, which contributed to improved model performance. For instance, parental education, intensity of instruction, and child IQ were not significant predictors of English receptive vocabulary, and their removal led to a better-fitting model. To assess whether the nested structure of the data (students nested within classrooms) needed to be accounted for, we examined intraclass correlation (ICC) scores. The ICC values ranged from .008 to .11, suggesting a small to moderate clustering effect; thus, we opted not to account for clustering in further analyses.
All models were performed in Mplus version 8.10 (Muthén & Muthén, Reference Muthén and Muthén1998–2017) and estimated with maximum likelihood with robust standard errors (Little et al., Reference Little, Jorgensen, Lang and Moore2014). Full information maximum likelihood was used to handle missing data. This approach allows all information available to be incorporated in the analyses, thus enabling it to handle any missing data on the indicators of the latent constructs (Baraldi & Enders, Reference Baraldi and Enders2010). Model fit was evaluated based on the following criteria: CFI and TLI > 0.90, and RMSEA and SRMR<0.08 (Hu & Bentler, Reference Hu and Bentler1999; Kline & Santor, Reference Kline and Santor1999).
4. Results
4.1. Descriptive statistics
Descriptive statistics for the study variables are presented in Table 1, while correlations among the variables are detailed in Table 2. Group comparison analyses revealed that bilingual children scored higher in receptive vocabulary task compared to monolinguals, but this difference is only present for the noncognate words but not for the cognate words. Moreover, bilingual children had higher scores in exposure to English compared to monolinguals, indicating that bilingual children are exposed to outside-of-school English activities to a greater extent than monolinguals. Bilingual children also had higher levels of parental education compared to monolingual children. Monolingual children, on the other hand, scored significantly higher in phonological awareness, verbal working memory and Dutch proficiency tasks compared to their bilingual peers. As expected, there was a significant positive relationship between receptive vocabulary and grammar scores in English. Additionally, older children showed superior performance across various tasks, including receptive vocabulary in English, receptive grammar in English, phonological awareness, verbal working memory, selective attention and Dutch proficiency. Higher Dutch proficiency scores were also associated with better performance in receptive grammar in English, phonological awareness and verbal working memory tasks.
Table 1. Descriptive statistics by language group
Abbreviations: RVT = Receptive Vocabulary Test; TROG = Test for Reception of Grammar; DCCS = Dimensional Change Card Sorting; VSS = Visual Sky Search; SRT = Sentence Repetition Task.
Table 2. Correlations among study variables
Note. Pearsons’ Correlations are reported. *p < .05; **p < .01. Eng = English.
4.2. Receptive vocabulary in English
The initial path analysis model assessing the impact of bilingualism on receptive vocabulary scores in English did not indicate a good fit. After refining the model by removing non-significant parameters and adding covariations between mediators and control variables, an acceptable fit was achieved, CFI = .99, TLI = .99, RMSEA = .008, SRMR = .07. Results showed that bilingual children scored significantly higher on receptive vocabulary in English compared to monolingual children. However, the indirect effects of bilingualism on vocabulary scores via phonological awareness (B = .04, SE = .04 p = .30, CI %95 = −.03–.11), motivation to learn English (B = −.002, SE = .01, p = .81, CI %95 = −.02–.01), verbal working memory (B = −.002, SE = .01, p = .85, CI %95 = −.02–.02), cognitive flexibility (B = .004, SE = .02, p = .83, CI %95 = −.07–.05), inhibition (B = −.01, SE = .03, p = .74, CI %95 = −.09–.05) and selective attention (B = −.06, SE = .06, p = .25, CI %95 = −.18–.05) were not significant. These results suggest that the bilingual effect in receptive vocabulary in English is not mediated by phonological awareness, motivation or executive function abilities. Among the control variables, age and exposure to English-related activities were positive and significant predictors of receptive vocabulary in English. A comprehensive depiction of the model with all tested parameters and coefficients is provided in Figure 1.
Figure 1. Path analysis with language group as the predictor of receptive vocabulary in English and phonological awareness, motivation to learn English and executive functions as mediators.
Note. Statistics are standardized regression coefficients. Solid lines indicate significant relations and dotted lines indicate non-significant relations. *p < .05; **p < .01.
4.3. Receptive grammar in English
The full path analysis model evaluating the effect of bilingualism on receptive grammar scores in English initially showed poor model fit. After modifying the model by eliminating non-essential parameters and adding covariations between mediators and control variables, an acceptable fit was obtained, CFI = .91, TLI = .87, RMSEA = .05, SRMR = .09. Findings of this model indicated that there were no significant differences between bilinguals and monolinguals in their grammar scores. The indirect effects of bilingualism on receptive grammar in English scores through phonological awareness (B = .04, SE = .04 p = .31, CI %95 = −.03–.11), motivation to learn English (B = −.01, SE = .02, p = .73, CI %95 = −.05–.04), verbal working memory (B = −.01, SE = .01, p = .64, CI %95 = −.03–.02), cognitive flexibility (B = .005, SE = .02, p = .75, CI %95 = −.03–.04), inhibition (B = −.02, SE = .04, p = .58, CI %95 = −.10–.06) and selective attention (B = −.03, SE = .03, p = .44, CI %95 = −.10–.04) were not significant. This suggests that the lack of a bilingual effect on receptive grammar in English is not attributable to children’s phonological awareness, motivation, or executive function skills. Additionally, age, exposure to English activities, and child IQ were significant positive predictors of receptive grammar scores. The detailed model with all evaluated parameters and coefficients is illustrated in Figure 2.
Figure 2. Path analysis with language group as the predictor of receptive grammar in English and phonological awareness, motivation to learn English and executive functions as mediators.
Note. Statistics are standardized regression coefficients. Solid lines indicate significant relations and dotted lines indicate non-significant relations. *p < .05; **p < .01.
4.4. Follow-up analysis
Previous research suggests that cognate words – those that are similar in form and meaning across two languages – are learned and recalled more easily than noncognates (Mulder et al., Reference Mulder, Van De Ven, Segers and Verhoeven2019; Sheng et al., Reference Sheng, Lam, Cruz and Fulton2016). For example, Tribushinina and Mackaaij (Reference Tribushinina and Mackaaij2023) found that bilingual children with developmental language disorder (DLD) outperformed Dutch-speaking monolinguals on a receptive vocabulary test. However, monolinguals benefited more from Dutch-English cognates than bilinguals. Building on these findings, we conducted a further analysis of English receptive vocabulary scores, distinguishing between cognate and noncognate words in the RVT.
Firstly, we built the same analytic models as described above, and added the cognate word scores as the dependent variable in the analysis. The initial path analysis did not result in acceptable model fit. After pruning the model by removing non-significant parameters and adding covariations among the mediators and control variables, an acceptable model fit was achieved (CFI = .99, TLI = .98, RMSEA = .04, SRMR = .09). Findings indicated that there are no significant differences between bilingual and monolingual children in cognate words of the receptive vocabulary test. Furthermore, monolingual children performed better in verbal working memory task compared to their bilingual peers. Among the mediators, children’s performance in phonological awareness task was positively related to cognate word knowledge in the receptive vocabulary test. However, the indirect effects of bilingualism on cognate vocabulary scores via phonological awareness (B = .08, SE = .06 p = .13, CI %95 = −.03–.19), motivation to learn English (B = −.006, SE = .02, p = .73, CI %95 = −.04–.03), verbal working memory (B = −.09, SE = .11, p = .38, CI %95 = −.37–.11), cognitive flexibility (B = .02, SE = .03, p = .71, CI %95 = −.07–.07), inhibition (B = .006, SE = .02, p = .81, CI %95 = −.06–.06) and selective attention (B = −.01, SE = .04, p = .73, CI %95 = −.09–.06) were not significant. These findings indicate that phonological awareness, motivation or executive function abilities did not mediate the link between bilingualism and cognate vocabulary of the receptive vocabulary test. Among the control variables, only exposure to English-related activities emerged as a significant positive predictor of cognate vocabulary performance in English. A comprehensive depiction of the model, including all tested parameters and coefficients, is provided in Figure 3.
Figure 3. Path analysis with language group as the predictor of cognates in receptive vocabulary in English and phonological awareness, motivation to learn English and executive functions as mediators.
Note. Statistics are standardized regression coefficients. Solid lines indicate significant relations and dotted lines indicate non-significant relations. *p < .05; **p < .01.
The initial path analysis model assessing the impact of bilingualism on receptive noncognate vocabulary scores in English did not indicate a good fit. After refining the model by removing non-significant parameters, an acceptable fit was achieved (CFI = .99, TLI = .98, RMSEA = .04, SRMR = .09). Results revealed that bilingual children scored significantly higher on noncognate words than monolingual children. Among the mediators, cognitive flexibility scores were positively related, while selective attention scores were negatively related to noncognate word performance. Given that lower scores indicate better performance in cognitive flexibility and selective attention tasks, these results suggest that children with lower cognitive flexibility and higher selective attention abilities performed better on noncognate words in English. However, the indirect effects of bilingualism on noncognate vocabulary scores via phonological awareness (B = .04, SE = .04 p = .26, CI%95 = −.03–.12), motivation to learn English (B = −.016, SE = .04, p = .65, CI%95 = −.09–.05), verbal working memory (B = .21, SE = .16, p = .18, CI%95 = −.10–.51), cognitive flexibility (B = .12, SE = .29, p = .69, CI%95 = −.46–.69), inhibition (B = −.02, SE = .04, p = .49, CI%95 = −.09–.04) and selective attention (B = −.10, SE = .07, p = .16, CI %95 = −.25–.04) were not significant. This suggests that the link between bilingualism and noncognate words in the receptive vocabulary in English is not mediated by children’s phonological awareness, motivation or executive function skills. Additionally, exposure to English activities was a significant positive predictor of noncognate word scores. The detailed model with all evaluated parameters and coefficients is illustrated in Figure 4.
Figure 4. Path analysis with language group as the predictor of noncognates in receptive vocabulary in English and phonological awareness, motivation to learn English and executive functions as mediators.
Note. Statistics are standardized regression coefficients. Solid lines indicate significant relations and dotted lines indicate non-significant relations. *p < .05; **p < .01.
5. Discussion
The current study advances the field of bilingualism by examining the bilingual advantage in FL learning during the kindergarten years and identifying the mechanisms underlying this effect. Specifically, we tested whether the relationship between bilingualism and English as an FL is mediated by metalinguistic awareness, executive functions, and motivation to learn English. Moreover, this study is the first to account for a comprehensive range of potential confounding factors, including child age, majority/school language proficiency, non-verbal IQ, exposure to English outside of school, intensity of EFL instruction and parental education, in assessing the bilingual effect on FL learning. Our findings revealed that bilingual children outperformed their monolingual peers in receptive English vocabulary, specifically in the noncognate words. However, this advantage did not extend to performance on English grammar. Contrary to our hypotheses, metalinguistic awareness, executive functions, and motivation to learn English did not mediate the relationship between bilingualism and FL learning. However, child age and exposure to English outside of school emerged as significant predictors of both English vocabulary and grammar scores, while non-verbal IQ significantly predicted English grammar scores.
One of the most significant findings of the present study is that the bilingual advantage in FL learning is already evident during the kindergarten years. This extends prior research, which has primarily examined the bilingual effect in older age groups, to a younger population. Our findings contribute to the growing body of evidence that the bilingual effect is evident at young ages (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Maluch et al., Reference Maluch, Neumann and Kempert2016; Siemund & Lechner, Reference Siemund and Lechner2015), even during kindergarten years. Notably, the advantage of bilingualism often diminishes as students transition from primary to secondary school (e.g., Maluch et al., Reference Maluch, Neumann and Kempert2016). This highlights the importance of strengthening FL input and instruction during the early years, particularly in kindergarten and primary school. For example, Festman (Reference Festman2018) demonstrated that targeted FL instruction in kindergartens significantly enhanced bilingual children’s vocabulary knowledge. By prioritizing FL instruction in these formative years, educators can leverage the heightened sensitivity of bilingual children to linguistic input.
Our study found that the bilingual advantage was evident in English vocabulary acquisition but not in grammar, with bilingual children outperforming their monolingual peers particularly in noncognate word learning. This pattern aligns with previous findings suggesting that the bilingual effect may be more pronounced for lexical development than for grammatical skills in foreign language learning. While some studies report bilingual advantages in both vocabulary and grammar (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Tribushinina & Boz, Reference Tribushinina and Boz2025; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023), others indicate that this advantage may be limited to vocabulary (Aguasvivas et al., Reference Aguasvivas, Cespón and Carreiras2024). For instance, Aguasvivas et al. (Reference Aguasvivas, Cespón and Carreiras2024) found that bilingual adults outperformed monolinguals in mapping novel words to referents in an artificial language but showed no advantage in tasks involving word segmentation or morphological rule generalization. This suggests that bilinguals may possess more advanced lexical mapping abilities, allowing them to form stronger associations between new word forms and meanings, a skill that may be especially beneficial when learning noncognates, where form similarity offers little support. Moreover, our finding that the bilingual advantage was stronger for noncognate than for cognate words further supports this interpretation. Cognates, due to their phonological and semantic overlap, are easier for all learners to acquire, reducing the scope for a bilingual advantage. In contrast, noncognates present greater learning challenges and may activate bilinguals’ enhanced mapping and inferencing skills. This aligns with the findings of Tribushinina and Mackaaij (Reference Tribushinina and Mackaaij2023), who reported a stronger Dutch–English cognate advantage in monolingual children than in bilinguals with developmental language delays. They attributed this to bilingual children’s smaller Dutch vocabularies, as evidenced by lower Dutch vocabulary scores, providing a weaker basis for transfer. In the case of young children, such as kindergartners, it is also possible that their EFL grammar skills are still in the early stages of development, limiting the potential for observable group differences. Future research should explore the developmental trajectory of grammatical acquisition in bilinguals and investigate under what conditions the bilingual advantage extends beyond vocabulary. Such insights could inform more targeted FL instruction that supports both lexical and grammatical development in early childhood.
Contrary to our expectations, metalinguistic awareness, executive functions, and motivation to learn English did not account for the bilingual effect in FL learning within our sample. One possible explanation for this null effect is the heterogeneity of the bilingual group. Participants varied considerably in their majority language (Dutch) proficiency and age of acquisition. Although we lack data on their home language proficiency, differences in Dutch proficiency, age of onset and exposure to home language suggest that some bilinguals in our sample might have a more balanced linguistic profile (i.e., similar proficiency levels in both their home and majority language), while others have one dominant and one weaker language. Previous research suggests that balanced bilinguals benefit more from the cognitive and linguistic advantages of bilingualism in FL learning compared to bilinguals with one dominant language (Cenoz, Reference Cenoz2003, Reference Cenoz2013; Cummins, Reference Cummins1979; Gallardo, Reference Gallardo2007; Lambert, Reference Lambert, Aboud and Mead1974). Moreover, research has shown that balanced bilinguals demonstrate greater metalinguistic awareness (Cohen, Reference Cohen2013; Dillon, Reference Dillon2009) and stronger executive function skills (Bogulski et al., Reference Bogulski, Rakoczy, Goodman and Bialystok2015; Weber et al., Reference Weber, Johnson, Riccio and Liew2016; Yurtsever et al., Reference Yurtsever, Anderson and Grundy2023) compared to bilinguals with one dominant language and monolinguals. Thus, it is possible that bilingualism influences cognitive and linguistic skills most strongly when individuals have high proficiency in both languages, which in turn facilitates FL learning. In the present study, we did not distinguish between balanced bilinguals and bilinguals with one dominant language. However, it is plausible that bilinguals with more balanced proficiency in their two languages may have advantages in both the tested mediators and FL performance.
Another possible explanation for why the proposed mediators did not account for the bilingual effect in FL scores could be differences in Dutch proficiency between bilingual and monolingual participants. Our results indicated that monolinguals in our sample outperformed bilinguals on the Dutch proficiency task. Research has consistently shown that proficiency in the majority language is a key predictor of FL skills (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Lorenz et al., Reference Lorenz, Rahbari, Schackow and Siemund2020, Reference Lorenz, Toprak-Yildiz and Siemund2024; Maluch et al., Reference Maluch, Kempert, Neumann and Stanat2015). Moreover, monolingual children in our sample also performed better on two tested mediators – phonological awareness and verbal working memory – which measure verbal abilities of children. This difference was observed even though the phonological awareness task used nonwords (i.e., non-existent Dutch words) to minimize the advantage of monolinguals in majority language skills. It is possible that monolinguals benefited from their stronger Dutch proficiency in understanding task instructions, giving them an advantage. Thus, the weaker performance of bilinguals in the tested mediators – potentially due to their more limited Dutch proficiency – may explain why these mediators failed to account for the bilingual effect in English scores. However, it is important to note that our analytical approach rigorously controlled for differences in Dutch proficiency in the modeling, ensuring that the observed effects were not merely due to disparities in majority language skills. These findings highlight the importance of considering differences in majority language proficiency when examining bilingual and monolingual language learning. Future research should account for these disparities and incorporate tasks that do not require advanced majority language skills as potential explanatory factors in FL learning.
Hence, it remains an open question why bilingual children tend to outperform their monolingual peers on FL skills. It is possible that an unexamined mediator plays a role in explaining the bilingual advantage in FL learning. It is also possible that direct transfer of language knowledge and/or skills from more than one language is key to explaining the bilingual effect in novel language learning (cf. Hirosh & Degani, Reference Hirosh and Degani2018). Extensive research has shown that bilingual children and adults outperform their monolingual counterparts in novel word learning skills (Aguasvivas et al., Reference Aguasvivas, Cespón and Carreiras2024; Antoniou et al., Reference Antoniou, Liang, Ettlinger and Wong2015; Kaushanskaya et al., Reference Kaushanskaya, Gross and Buac2014; Kaushanskaya & Marian, Reference Kaushanskaya and Marian2009a, Reference Kaushanskaya and Marian2009b; Singh, Reference Singh2018). That is, bilinguals more efficiently link new word forms to their meanings within the mental lexicon. One possible explanation is that their richer lexical networks and greater experience with cross-linguistic associations enhance their ability to resolve lexical ambiguity (e.g., Schröter & Schroeder, Reference Schröter and Schroeder2016). Correspondingly, the superior performance in English vocabulary of bilinguals was present only for noncognate words. It is possible that because bilinguals regularly navigate between two languages, they develop more flexible and interconnected word representations, allowing them to identify plausible meanings of unfamiliar words more quickly. This heightened ability to form and refine word-meaning associations may contribute to their success in FL learning. Additionally, prior experience with multiple languages may facilitate the direct transfer of linguistic knowledge such as vocabulary, syntax, or phonological rules from known to new languages, particularly when there is some structural overlap (Hirosh & Degani, Reference Hirosh and Degani2018). Bilinguals’ early and sustained exposure to diverse phonological systems may shape a more flexible phonological system, enabling them to encode, distinguish and retain novel sounds more effectively, an advantage that is particularly useful in foreign language learning contexts (Hirosh & Degani, Reference Hirosh and Degani2018). Another explanation could be related to bilinguals’ advantage in implicit statistical learning abilities. Evidence from several studies suggests that bilinguals’ advantage in foreign language learning may be partly explained by enhanced implicit statistical learning abilities. For example, De Bree et al. (Reference De Bree, Verhagen and Gerrits2017) showed that bilingual toddlers, but not monolinguals, could extract grammatical regularities from inconsistent input, indicating robustness in detecting patterns even under noisy conditions. In the same vein, Poepsel and Weiss (Reference Poepsel and Weiss2016) found that late bilinguals outperformed monolinguals in a cross-situational statistical word learning task when the input involved multiple mappings, reflecting greater flexibility and reduced reliance on the mutual exclusivity bias. Wang and Saffran (Reference Wang and Saffran2014) also demonstrated that bilingual adults, irrespective of prior tonal language experience, were more successful than monolinguals at segmenting an artificial tone language based solely on distributional cues. Together, these findings indicate that bilinguals may possess a heightened ability to track, integrate and adapt to complex statistical patterns without explicit instruction, which can provide a cognitive foundation for more efficient vocabulary and structure learning in EFL contexts.
A strong contribution of the current study is the inclusion of a comprehensive set of control variables when testing the bilingual effect in FL learning. Among these confounding factors, child age and exposure to English outside of school emerged as significant predictors of children’s English receptive vocabulary and grammar scores. Within the sampled age range (4 to 6 years), older children performed better on English measures, likely due to their more advanced linguistic and cognitive skills, which are reflected in their FL learning performance. Additionally, in the Dutch kindergarten system, children start kindergarten on their 4th birthday, regardless of the time of year, and transition to grade 1 in September after turning 6. As a result, a 4-year old may have had only a few months of English instruction, whereas a 6-year old could have received 2 years or more. Thus, age in this context not only reflects developmental differences but also serves as a proxy for the length of English instruction, highlighting that older children have had substantially more exposure to English in school compared to younger age groups. This study is among the first to include exposure to English outside of school as a control variable when examining the bilingual effect in FL learning (Tribushinina & Boz, Reference Tribushinina and Boz2025; Tribushinina, Boz, et al., Reference Tribushinina, Boz, Aalbers and Blom2024; Tribushinina & Mackaaij, Reference Tribushinina and Mackaaij2023). Consistent with previous research, children who engaged more frequently in extramural English activities demonstrated stronger performance in English tasks, highlighting the importance of accounting for such exposure in future FL learning studies. Finally, child non-verbal IQ significantly predicted English grammar skills, aligning with prior findings (Hopp et al., Reference Hopp, Vogelbacher, Kieseier and Thoma2019; Woumans et al., Reference Woumans, Ameloot, Keuleers and Van Assche2019). However, non-verbal IQ did not significantly predict English vocabulary scores. This discrepancy may stem from the nature of grammar and vocabulary acquisition. Grammar learning requires abstract pattern recognition, rule application and cognitive control – processes closely tied to intelligence. In contrast, vocabulary acquisition is often more dependent on exposure and memory-based learning rather than general cognitive ability. Thus, while children with higher IQs may be better at detecting and applying grammatical rules, vocabulary growth may be more influenced by the quantity and quality of linguistic input rather than cognitive problem-solving skills. This distinction underscores the importance of considering different cognitive and experiential factors in FL learning research.
This study has several limitations, which also provide important directions for future research. One key limitation is the heterogeneity of the bilingual sample in terms of L1 background. Additionally, the lack of information on participants’ home language proficiency prevents a precise determination of their dominant language. This limitation makes it challenging to disentangle whether the lack of mediation effects was driven by variations in L1–L2 dominance or by other factors that we did not take into account in the context of the current study. Moreover, due to the high heterogeneity of the bilingual group with only a few participants sharing similar language backgrounds, it was not possible to conduct meaningful subgroup analyses. The resulting low statistical power would have made any such analyses unreliable. This limits the generalizability of our findings and highlights the need for future research to include more homogeneous bilingual subsamples or sufficiently large and balanced groups to allow for subgroup comparisons based on factors such as heritage language type, age of acquisition and language dominance. Such analyses could provide more nuanced insights into how specific bilingual experiences influence foreign language learning.
Despite these limitations, this study represents an important first step in identifying potential mediators of bilingual advantages in FL learning. Beyond mediation effects, future research should also explore direct effects of bilingualism, particularly in the context of cross-language transfer. Prior research suggests that children can transfer vocabulary and grammar knowledge from their L1 when acquiring an additional language (e.g., Kolb et al., Reference Kolb, Mitrofanova and Westergaard2022; Tribushinina et al., Reference Tribushinina, Dubinkina-Elgart and Mak2022; Westergaard et al., Reference Westergaard, Mitrofanova, Mykhaylyk and Rodina2017), with stronger transfer effects occurring when the languages share structural similarities (Blom & Paradis, Reference Blom and Paradis2015). Investigating how different linguistic backgrounds facilitate or constrain this transfer would deepen our understanding of bilingual advantages in FL acquisition. By addressing these gaps, future research can build a more nuanced and comprehensive understanding of the bilingual effect in FL learning.
6. Conclusion
In the current study, we examined potential mechanisms underlying the bilingual effect in English as a FL learning among kindergarten-aged children. Our findings revealed that bilingual children outperformed their monolingual peers in English vocabulary but not in grammar. The advantages in EFL vocabulary were limited to noncognate words. However, the proposed mediating factors – metalinguistic awareness, executive functions and motivation to learn English – did not account for this advantage. This suggests that other, yet unidentified, mechanisms may explain the link between bilingualism and FL learning outcomes, warranting further investigation.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1366728925100631.
Data availability statement
The data that support the findings will be available in Open Science Framework at https://osf.io/bvndh/?view_only=8dc549b75c704416b0eb9baa7f746b82 following a 3 month embargo from the date of submission to allow for commercialization of research findings.
Acknowledgments
We thank participating reviewers for their helpful comments, participating schools, as well as Daniela Kissová, Louise Brandenburg, Thamar Karstel, Melissa Voorn and Annika Medendorp for their invaluable assistance with data collection and processing. This work was supported by the Netherlands Organization for Scientific Research (NWO) under Grant 406.22.CTW.005 to Elena Tribushinina.
Competing interests
The authors declare none.
Open access
