Research highlights
-
• SES differences between groups substantially shaped cognitive performance patterns.
-
• Sequential bilinguals outperformed others on nonverbal tasks before SES adjustment.
-
• Heritage bilinguals matched monolinguals on verbal tasks, contradicting lexical-deficit views.
-
• Findings clarify mixed bilingual advantage results by highlighting SES effects.
1. Introduction
Researchers have investigated the impact of bilingualism on specific cognitive domains. One such domain is nonverbal skills, which are associated with various aspects of executive function (EF) (Bialystok, Reference Bialystok2009). These skills involve cognitive processes such as planning, inhibitory control and cognitive flexibility, which support goal-directed behavior and adaptive problem-solving in novel situations. Prior research suggested a bilingual advantage, where bilinguals outperformed monolinguals in a variety of cognitive tasks (Bialystok, Reference Bialystok1999; Bialystok & Shapero, Reference Bialystok and Shapero2005; Grundy & Timmer, Reference Grundy and Timmer2017; Lee Salvatierra & Rosselli, Reference Lee Salvatierra and Rosselli2011; Prior & Macwhinney, Reference Prior and Macwhinney2010). However, other researchers argue that bilingualism does not confer cognitive advantages, asserting that studies indicating otherwise are likely to be methodologically flawed or biased (Antón et al., Reference Antón, Duñabeitia, Estévez, Hernández, Castillo, Fuentes, Davidson and Carreiras2014; Duñabeitia et al., Reference Duñabeitia, Hernández, Antón, Macizo, Estévez, Fuentes and Carreiras2014; Hilchey & Klein, Reference Hilchey and Klein2011; Morton & Harper, Reference Morton and Harper2007; Paap, Reference Paap2014; Paap et al., Reference Paap, Johnson and Sawi2015). For instance, Paap and Greenberg (Reference Paap and Greenberg2013) argue that the cognitive advantages of bilingualism may not be as robust as previously claimed, emphasizing the importance of considering individual differences in language experience and other contextual factors. Furthermore, Bialystok (Reference Bialystok2024) also suggest that the bilingual advantage is more likely to be observed in complex tasks that require higher levels of cognitive control, rather than in simpler tasks where both monolinguals and bilinguals tend to perform similarly.
While nonverbal skills have been a primary focus in bilingual research, it is equally important to investigate how bilingualism affects verbal skills (including receptive and productive language skills). Verbal skills tend to improve with age as one gain more experience and learning. The relationship between bilingualism and verbal skills remains a contentious issue in both research and educational practice. While some may argue that bilingual children exhibit lower language skills than their monolingual peers, especially in vocabulary, some research suggests that when the total vocabulary across both languages or quality of language input and exposure are considered, bilinguals often match or even exceed monolinguals in linguistic capability (Bialystok, Reference Bialystok2015; De Houwer et al., Reference De Houwer, Bornstein and Putnick2012; Hoff et al., Reference Hoff, Core, Place, Rumiche, Señor and Parra2012; Pearson et al., Reference Pearson, Fernández and Oller1993; Peets et al., Reference Peets, Yim and Bialystok2022; Thordardottir, Reference Thordardottir2019; Unsworth et al., Reference Unsworth, Brouwer, De Bree and Verhagen2019). Kroll et al. (Reference Kroll, Dussias, Bice and Perrotti2015) further highlight that while bilinguals may have smaller vocabularies in each language, their combined linguistic knowledge across both languages is often broader, especially in receptive tasks.
In sum, bilingualism has been associated with cognitive abilities; however, while some studies report positive cognitive effects for bilinguals, these findings are often nuanced, varying according to the age of participants, the type of task and the specific language experiences of the bilingual individuals involved. These effects, especially in children and adolescents, may be moderated by factors such as socioeconomic status (SES) and the complexity and variability of bilingual language experiences (Duñabeitia et al., Reference Duñabeitia, Hernández, Antón, Macizo, Estévez, Fuentes and Carreiras2014; Hilchey & Klein, Reference Hilchey and Klein2011; Morton & Harper, Reference Morton and Harper2007; Paap et al., Reference Paap, Johnson and Sawi2015). Thus, the goal of this article is to explore the role of language experiences and SES in evaluating cognitive outcomes among bilingual and monolingual children, by explicitly testing group differences in cognition both before and after adjusting for SES and other demographic factors.
Given that the cognitive effects of bilingualism are influenced by developmental stage, this article focuses on adolescence – a critical yet relatively understudied period in this research. Beginning around age 10 in girls and 12 in boys, adolescence is marked by substantial cognitive, social and neurological development, including brain changes driven by puberty that influence cognitive flexibility and executive functions (Dahl et al., Reference Dahl, Allen, Wilbrecht and Suleiman2018; Giedd et al., Reference Giedd, Clasen, Lenroot, Greenstein, Wallace, Ordaz, Molloy, Blumenthal, Tossell, Stayer, Samango-Sprouse, Shen, Davatzikos, Merke and Chrousos2006; Luna et al., Reference Luna, Garver, Urban, Lazar and Sweeney2004). Meta-analytic evidence suggests that age may moderate the cognitive effects of bilingualism, with advantages more likely to be observed in younger populations—including children and, in some cases, adolescents, highlighting developmental windows of heightened sensitivity to bilingual experience (Yurtsever et al., Reference Yurtsever, Anderson and Grundy2023; but also see Lowe et al., Reference Lowe, Cho, Goldsmith and Morton2021). Adolescence may thus represent a period in which bilingual experience could exert a measurable influence on cognition, as the brain remains highly plastic and cognitive abilities continue to develop. Some research suggests that bilingual adolescents sometimes outperform monolinguals on flanker tasks; however, evidence is mixed, and effect sizes are generally small (Paap, Reference Paap and Schwieter2019). For example, Chung-Fat-Yim et al. (Reference Chung-Fat-Yim, Himel and Bialystok2019) found that bilingualism was associated with faster overall reaction times, but not reduced interference, indicating that performance differences may not reflect true improvements in inhibitory control. Additionally, the traditional flanker effect exhibits poor test–retest reliability and limited convergent validity with other nonverbal interference measures; therefore, any observed advantages should be interpreted with caution.
In general, there are few research studies focusing on adolescence. While studies on younger children indicate a bilingual advantage in cognitive flexibility and inhibition (Yurtsever et al., Reference Yurtsever, Anderson and Grundy2023), more adolescent-focused research is needed to determine whether these advantages persist (Giovannoli et al., Reference Giovannoli, Martella, Federico, Pirchio and Casagrande2020). Additionally, variability in early bilingual educational experiences raises questions about how different backgrounds influence cognitive development, with findings suggesting that early bilingual exposure may offer unique cognitive benefits (Grundy & Timmer, Reference Grundy and Timmer2017). A recent meta-analysis challenges the idea of a broad bilingual advantage in executive function, emphasizing the need for further research on specific age-related contexts where bilingualism provides cognitive benefits (Gunnerud et al., Reference Gunnerud, ten Braak, Reikerås, Donolato and Melby-Lervåg2020). Overall, while evidence suggests potential cognitive benefits of bilingualism in adolescence, further research is needed to clarify the extent and conditions under which these advantages emerge. The goal of this article is to examine cognitive abilities in bilingual adolescents, focusing on the influence of bilingual experience on cognitive development.
In addition to age, differences in bilingual experience have been linked to inconsistencies in findings on bilingualism and cognitive abilities. Bilingualism, the ability to proficiently use two languages, arises from various experiences. For example, simultaneous bilinguals may learn two languages from birth in environments where both languages are spoken, while sequential bilinguals may learn the second language (L2) after the first (L1), often resulting from migration or education (Grosjean, Reference Grosjean2010). As a result, bilinguals can be highly proficient in both languages (balanced bilinguals) or be more dominant in their L1 (unbalanced bilinguals). A special case arises with heritage speakers. These individuals acquire their L1 during early childhood at home; however, they receive education in the societal language (L2) that is different from the home language, leading to greater proficiency in L2 than L1 (Valdés, Reference Valdés2005). These variations emphasize the importance of distinguishing between different bilingual subgroups, rather than conceptualizing them as a single category, when investigating language and cognitive development and when attempting to reconcile differences in findings across studies (Luk & Bialystok, Reference Luk and Bialystok2013).
Indeed, language experience shapes the relationship between bilingualism and cognition (Grundy & Timmer, Reference Grundy and Timmer2017). Vaughn et al. (Reference Vaughn, Greene, Ramos Nuñez and Hernandez2015, Reference Vaughn, Nguyen, Ronderos and Hernandez2021) emphasize the need to move beyond simple bilingual–monolingual comparisons to examine how diverse linguistic backgrounds influence cognitive outcomes. Meta-analyses show that while bilinguals show a small working memory (WM) advantage, this effect is stronger in verbal WM tasks performed in the second language (L2), where managing two languages may enhance cognitive flexibility and inhibitory control (Monnier et al., Reference Monnier, Boiché, Armandon, Baudoin and Bellocchi2022). However, language background plays a key role, highlighting the importance of studying bilingual experience as a moderating factor (Bialystok, Reference Bialystok2021). For example, balanced bilingual children showed stronger cognitive flexibility and inhibitory control compared to monolinguals (Nguyen, Hutchison, et al., Reference Nguyen, Hutchison, Norvell, Mead and Winsler2024). Given the varied experiences of bilingual children in the United States, this study compares the following groups: monolinguals, heritage bilinguals, dual language education bilinguals and other bilinguals. Monolinguals are native English speakers. Heritage bilinguals are those who have a home language (L1) that differs from the societal language (English; L2) and learn this L2 from an early age primarily outside of the home. In the United States, these children often come from lower SES backgrounds with greater exposure to a home language, but limited formal education in that language. Dual language education bilinguals are children who learn L2 formally through bilingual or immersion education. Finally, other bilinguals may have acquired L2 through other formal or informal settings. Thus, both dual language education bilinguals and other bilinguals can be considered sequential bilinguals in this context.
Beyond differences in language experience, socioeconomic status (SES) further contributes to variability in bilingual cognitive outcomes. SES shapes access to educational resources, language exposure and cognitive development, making it a crucial factor in understanding bilingualism’s impact on cognition. In fact, bilinguals of different language backgrounds may also come from different SES backgrounds. Socioeconomic status (SES) significantly influences cognitive development, with lower SES children generally having lower scores on cognitive and linguistic tasks due to limited resources, reduced parental involvement and increased stress (Letourneau et al., Reference Letourneau, Duffett-Leger, Levac, Watson and Young-Morris2013; White et al., Reference White, Pasco, Korous and Causadias2020). In bilingual populations, SES interacts with language experience, as heritage speakers from lower SES backgrounds often exhibit weaker performance on verbal short-term memory tasks compared to higher SES bilingual peers, highlighting SES as a key factor in cognitive outcomes (Balilah et al., Reference Balilah, Archibald and Said2023; Gathercole et al., Reference Gathercole, Kennedy and Thomas2016; Haskins et al., Reference Haskins, Greenberg and Fremstad2004; Meir & Armon-Lotem, Reference Meir and Armon-Lotem2017). Higher SES is consistently linked to stronger executive functions, memory and language processing (Chiat & Polišenská, Reference Chiat and Polišenská2016), which may contribute to the diverse cognitive profiles observed among bilinguals.
Research suggests that bilingualism can enhance executive functions even in lower SES populations, with bilingual children demonstrating improved inhibitory control, multitasking and cognitive flexibility, potentially mitigating SES-related disadvantages (Carlson & Meltzoff, Reference Carlson and Meltzoff2008; Lawson et al., Reference Lawson, Hook and Farah2018; Thomas-Sunesson et al., Reference Thomas-Sunesson, Hakuta and Bialystok2018). These cognitive benefits, observed across SES groups, highlight the potential role of bilingual experience in cognition, but differences in SES, if not considered, may contribute to conflicting findings in bilingualism research (Adesope et al., Reference Adesope, Lavin, Thompson and Ungerleider2010; Engel de Abreu et al., Reference Engel de Abreu, Cruz-Santos, Tourinho, Martin and Bialystok2012; Krizman et al., Reference Krizman, Skoe and Kraus2016; Naeem et al., Reference Naeem, Filippi, Periche-Tomas, Papageorgiou and Bright2018). Recent meta-analytic evidence suggests that many previously reported bilingual advantages in executive function are largely influenced by SES and study quality, with small positive effects primarily observed in studies using self- or parent-reported measures of language (Lowe et al., Reference Lowe, Cho, Goldsmith and Morton2021). This underscores the importance of carefully accounting for SES and methodological rigor when interpreting cognitive differences between bilingual and monolingual children.
Given the complex relationship between bilingualism and cognition, which is shaped by age, language experience and SES, this study aims to compare bilinguals with varying language experience while accounting for SES background to better understand its influence on cognitive abilities in children. Using data from the Adolescent Brain Cognitive Development (ABCD) study, a large, diverse and longitudinal U.S. dataset (ABCD Research Consortium, 2024; Casey et al., Reference Casey, Cannonier, Conley, Cohen, Barch, Heitzeg, Soules, Teslovich, Dellarco, Garavan, Orr, Wager, Banich, Speer, Sutherland, Riedel, Dick, Bjork, Thomas and Dale2018), this study examines how SES and bilingualism interact to shape cognition over time. Findings may offer valuable insights into language proficiency, cognitive skills and their role in academic and social outcomes, informing policies that support bilingual adolescents in diverse communities.
Studies using ABCD data to examine bilingualism in adolescence have varied in their definitions, ranging from simple home language use (Dick et al., Reference Dick, Garcia, Pruden, Thompson, Hawes, Sutherland, Riedel, Laird and Gonzalez2019) to more refined approaches that account for language proficiency, frequency and context (Kwon et al., Reference Kwon, Yoo, Nguyen, Jeong and Chun2021; Nguyen, Xu, et al., Reference Nguyen, Xu, Vaughn and Hernandez2024; Vaughn et al., Reference Vaughn, Nguyen, Ronderos and Hernandez2021). To better capture these complexities, this study expands on prior strategies by categorizing children into four language groups—monolinguals, heritage bilinguals, dual language education bilinguals and other bilinguals—offering a more nuanced framework that considers both language learning context and formal education. Unlike prior studies that focused on broad bilingual labels, this study provides a more detailed classification of bilingual types, integrating language learning context. By explicitly accounting for SES and other demographic factors, we aim to disentangle the effects of bilingualism on cognition from those of SES—an issue that has contributed to inconsistent findings in past research. This approach offers a novel contribution by addressing both the heterogeneity of bilingual experiences and the critical role of SES in shaping cognitive outcomes.
Given the complex relationship between bilingualism and cognition, which can be shaped by age, language experience and SES, this study compares different types of bilinguals, explicitly accounting for SES and demographic background to better understand their influence on cognitive abilities in children. Using data from the Adolescent Brain Cognitive Development (ABCD) study, a large, diverse and nationally representative U.S. dataset (Garavan et al., Reference Garavan, Bartsch, Conway, Decastro, Goldstein, Heeringa, Jernigan, Potter, Thompson and Zahs2018), we examine how SES and bilingualism interact to shape cognition over time. This approach offers a novel contribution by addressing both the heterogeneity of bilingual experiences and the critical role of SES in shaping cognitive outcomes.
In sum, this study aims to provide valuable insights into cognitive performance differences among children in the ABCD study, considering variability in language background and SES. We ask: (1) How do different types of bilinguals differ in their language backgrounds? (2) How do monolinguals and different types of bilinguals differ in cognitive performance, both before and after accounting for demographic and SES factors?
2. Method
2.1. Participants
The sample in this study included children selected from the ABCD study, which follows over 11,000 children across 22 sites in the United States beginning at nine and ten years old. The current analyses utilized data from the ABCD data release 5.1 (doi:10.15154/z563-zd24), which includes baseline, year 1, year 2 and year 3 data, covering ages 9–13. We determined language groups using demographic data from the Parent Longitudinal Demographic Questionnaire (PLDQ; collected during the 1-year follow-up visit; Haines, Reference Hamilton, Strader, Pratt, Maiese, Hendershot, Kwok, Hammond, Huggins, Jackman, Pan, Nettles, Beaty, Farrer, Kraft, Marazita, Ordovas, Pato, Spitz, Wagener and Haines2011) and the child’s Youth Acculturation Survey (YAS; Haines, Reference Hamilton, Strader, Pratt, Maiese, Hendershot, Kwok, Hammond, Huggins, Jackman, Pan, Nettles, Beaty, Farrer, Kraft, Marazita, Ordovas, Pato, Spitz, Wagener and Haines2011). The key items from the PLDQ for this study were questions about the child’s native language (i.e., “What is your child’s native language?”); the home language environment (i.e., The child’s parents or guardians spoke in English more than any other language after birth; English and another language were spoken equally; The child’s parents or guardians spoke a language other than English more than any other language after birth) and whether the child ever attended a dual language program at school. The YAS asked children about their knowledge of another language (“Besides English, do you speak or understand another language or dialect?”). Using the answers to these questions, we identified four language groups: monolinguals, heritage bilinguals, dual language education (DLE) bilinguals and other bilinguals. Figure 1 depicts the data selection process and the distinction between the four language groups (see Figure S1 in the supplementary materials for depiction of the final sample for each language group). Given that both DLE bilinguals and other bilinguals would have acquired their L2 in school or other contexts after L1, they may collectively be referred to as sequential bilinguals where appropriate. See the Supplementary Materials for full demographic information of the four language groups.
Sample selection process.
2.2. Measures
This article accounted for a variety of child demographic and socioeconomic factors as covariates. These included sex, handedness, pubertal status, race/ethnicity, parental education, marital status and household income. These variables were included because bilinguals and monolinguals in the ABCD sample typically differed on these characteristics, and their inclusion is consistent with prior research (Nguyen, Xu, et al., Reference Nguyen, Xu, Vaughn and Hernandez2024; Ronderos et al., Reference Ronderos, Zuk, Hernandez and Vaughn2024; Vaughn et al., Reference Vaughn, Nguyen, Ronderos and Hernandez2021). Descriptions of these variables are provided next.
2.2.1. Questionnaires
Children’s demographic background was acquired using various measures from the ABCD study to be included as covariates in all analyses. Child’s biological sex, age, gender, ethnicity, country of origin, parent education, household marital status and household income came from the Parent Longitudinal Demographic Questionnaire (modified from the PhenX toolkit; Haines, Reference Hamilton, Strader, Pratt, Maiese, Hendershot, Kwok, Hammond, Huggins, Jackman, Pan, Nettles, Beaty, Farrer, Kraft, Marazita, Ordovas, Pato, Spitz, Wagener and Haines2011), which is completed yearly by parents. Information about sex, country of origin and race/ethnicity, however, was only collected at baseline.
Handedness was measured using a brief version of the Edinburgh Handedness Inventory (Oldfield, Reference Oldfield1971; Veale, Reference Veale2014), which contains four items: writing, throwing, using a spoon and using a toothbrush, rated on a 5-point scale from always right hand to always left hand. Based on their responses, participants were identified as right-handed, left-handed or ambidextrous. See Luciana et al. (Reference Luciana, Bjork, Nagel, Barch, Gonzalez, Nixon and Banich2018) for a complete description. Handedness was only measured at baseline.
Pubertal status was measured using the Pubertal Development Scale (Barch et al., Reference Barch, Albaugh, Avenevoli, Chang, Clark, Glantz, Hudziak, Jernigan, Tapert, Yurgelun-Todd, Alia-Klein, Potter, Paulus, Prouty, Zucker and Sher2018; Petersen et al., Reference Petersen, Crockett, Richards and Boxer1988). This questionnaire asks about body hair, skin change (e.g., acne), growth spurt, voice change (males only), facial hair (males only), breast change (females only) and menarche (females only). Children respond to each item on a 4-point scale where 1 = no development; 2 = beginning development; 3 = additional development and 4 = development already past (menarche coded dichotomously as 1 = premenarcheal; 4 = postmenarcheal). For each child, we used the summary scores across each domain as a measure of pubertal status ranging from 1 (no development in any domain) to 4 (development already passed in all domains). Pubertal status was measured yearly.
2.2.2. Matrix reasoning
Fluid intelligence (nonverbal IQ) and visuospatial reasoning were measured using an automated version of the Matrix Reasoning subtest from the Wechsler Intelligence Scale for Children-V (WISC-V; Wechsler, Reference Wechsler2014). Children saw an array of visuospatial stimuli with a missing item, and must select one of four options to complete the array. There are 32 trials, and testing stops if children miss three items in a row. See Luciana et al. (Reference Luciana, Bjork, Nagel, Barch, Gonzalez, Nixon and Banich2018) for a complete description. Matrix reasoning was only measured at baseline.
2.2.3. Language background
Language use includes questions obtained from the Youth Acculturation Survey, which was a subset of questions from the PhenX Acculturation protocol (Haines, Reference Hamilton, Strader, Pratt, Maiese, Hendershot, Kwok, Hammond, Huggins, Jackman, Pan, Nettles, Beaty, Farrer, Kraft, Marazita, Ordovas, Pato, Spitz, Wagener and Haines2011). Children were asked to rank how well they speak English (i.e., poor, fair, good, excellent; convert to a 0 = poor to 4 = excellent scale) and if they speak or understand another language besides English. If children speak or understand a language other than English, they were then asked to identify the other language, rate their language use with family and their language use with friends on a 5-point scale (1 = other language all of the time; 5 = English all of the time), and rank how well they speak the other language. See Zucker et al. (Reference Zucker, Gonzalez, Feldstein Ewing, Paulus, Arroyo, Fuligni, Morris, Sanchez and Wills2018) for more information. We averaged the ratings for language use with family and language use with friends to develop a language use score, also on a 1–5 scale. In this study, 1 = only other language with family and friends and 5 = only English with family and friends. Scores in the middle of this range represent more balanced language use.
2.2.4. NIH toolbox cognition battery
The NIH toolbox consists of the Fluid Composite (includes Dimensional Change Card Sort [DCCS], Flanker Inhibitory Control and Attention, Picture Sequence Memory, List Sorting Working Memory and Pattern Comparison tests) and the Crystallized Composite (includes Picture Vocabulary and Oral Reading Recognition tests) (https://nihtoolbox.org/domain/cognition/). In this study, nonverbal skills were proxied by the fluid composite, while verbal skills were proxied by the crystallized composite. Children completed the tasks using an iPad (Gershon et al., Reference Gershon, Slotkin, Manly, Blitz, Beaumont, Schnipke, Wallner-Allen, Golinkoff, Gleason, Hirsh-Pasek, Adams and Weintraub2013). For this study, we used the age-corrected scores provided by the ABCD research team (M = 100; SD = 15). See Luciana et al. (Reference Luciana, Bjork, Nagel, Barch, Gonzalez, Nixon and Banich2018) for more information. Of these tasks, List Sorting Working Memory and DCCS were administered only at baseline, while the other five tasks were administered both at baseline and year 2. Thus, the fluid composite (nonverbal skills) was only calculated for baseline, while the crystallized composite (verbal skills) was calculated for Baseline and Year 2 data. In this study, these two composites were the main outcomes of interest.
3. Results
Analyses were conducted using R programming language (https://www.r-project.org/) in RStudio (version 4.4.2; https://posit.co/). All analysis codes and full results are available on the Open Science Framework (https://osf.io/aexrb/). The data from subjects included in the analyses are available on the NIMH Data Archive (doi:10.15154/00fn-m824). Due to the large sample size, the α level is set at 0.01 for all analyses (Gómez-de-Mariscal et al., Reference Gómez-de-Mariscal, Guerrero, Sneider, Jayatilaka, Phillip, Wirtz and Muñoz-Barrutia2021). In general, language groups differed in various ways, but most importantly in SES measures such as household income and parent education (see full results in online Supplementary Materials). Specifically, sequential bilinguals (including DLE and other bilinguals) had significantly higher parent education levels (F(3, 10,102) = 403.63, p < .001, partial η 2 = .11) and household income (F(3, 9881.8) = 206.07, p < .001, partial η 2 = .07) than monolinguals, whose measures were significantly higher than heritage bilinguals. These results indicate the importance of including SES measures as covariates in analyses concerning cognitive outcomes (for more information, review the Supplementary Materials).
3.1. Bilingual language background
Every year, bilingual children self-reported their English and other language skills (Figure 2 top panel, also see Supplementary Materials). We ran linear mixed models to examine differences between the language groups. There were significant differences between groups in self-rated English skills, F(2, 3445) = 223.85, p < .001, partial η 2 = .12. English rating and group differences in English rating did not vary by year. Other bilinguals and DLE bilinguals had similar ratings (b = 0.02, p = .48, 95% CI = [−0.02, 0.06]), and both had higher English self-reported scores than home learners (b = 0.29, p < .001, 95% CI = [0.26, 0.32] and b = 0.29, p < .001, 95% CI = [0.23, 0.32], respectively). For self-report scores of the other language, ratings varied over time, F(3, 3414) = 6.93, p < .001, though the effect is minuscule, partial η 2 = .006. Groups varied in self-rating, F(2, 4952.8) = 103.00, p < .001, partial η 2 = .04, and this effect did not vary by years, F(6, 3426.9) = 2.58, p = .02. Heritage bilinguals had higher self-rated other-language scores compared to DLE bilinguals (b = 0.32, p < .001, 95% CI = [0.14, 0.50]), who had higher self-rated scores than other bilinguals (b = 0.50, p < .001, 95% CI = [0.69, 0.96]).
Bilinguals’ self-report language background.
Children reported their language use with friends and family (Figure 2 middle panel, also see Supplementary Materials). We ran linear mixed models to examine differences between the language groups. Language use with friends varied by year, F(3, 8291.5) = 29.49, p < .001, partial η 2 = .01. There were significant differences between groups in language use with friends, F(2, 3455.5) = 331.64, p < .001, partial η 2 = .16. In addition, the differences in language use with friend between groups varied by year, F(6, 8813.75) = 5.49, p < .001, though this effect was minuscule, partial η 2 = .004. Other bilinguals had greater English use than DLE bilinguals (b = 0.26, p < .001, 95% CI = [0.20, 0.32]), who had greater English use than heritage bilinguals (b = 0.26, p < .001, 95% CI = [0.20, 0.32]).
Language use with family varied by year, F(3, 8048.1) = 13.60, p < .001, though the effect is miniscule, partial η 2 = .005. There were significant differences between groups in language use with family, F(2, 3475.0) = 3356.42, p < .001, partial η 2 = .66. These differences did not vary by year, F(6, 8060.7) = 0.56, p = .76. Other bilinguals and DLE bilinguals did not differ in their language use with family (b = 0.02, p = .85, 95% CI = [−0.06, 0.10]), and both groups had greater English use than heritage bilinguals (b = 2.21, p < .001, 95% CI = [2.15, 2.28], and b = 2.19, p < .001, 95% CI = [2.10, 2.28], respectively).
Language use across friends and family varied by year, F(3, 8075.6) = 31.96, p < .001, partial η 2 = .01. Self-reported English use increased across the four time points. There were significant differences between groups in overall language use, F(2, 3461.0) = 2521.63, p < .001, partial η 2 = .59. These differences did not vary by year, F(6, 8089.8) = 1.73, p = .11. Other bilinguals had significantly higher English use than DLE bilinguals (b = 0.14, p < .001, 95% CI = [0.08, 0.20]), who had higher English use than heritage bilinguals (b = 1.23, p < .001, 95% CI = [1.17, 1.29]).
3.2 Language group across the United States
The ABCD study collected data across 22 sites in the United States. We explored the proportion of language groups within each site, as well as the proportion of children per group that came from any particular data site. We ran chi-square tests of marginal independence to examine differences between language groups and differences between years for sites, and the Cochran–Mantel–Haenszel test of conditional independence to explore whether the differences between groups in sites varied by year. There was a significant difference between language groups in data sites, χ 2 (63, 39,243) = 12,518, p < .001, though sites and the relationship between language group and sites did not vary by year, χ 2 (63, 39,243) = 76.39, p = .11 and χ MH 2 (9, 39,243) = 1.60, p = .996, respectively. Since the pattern of language group distribution did not vary by year, we plotted only the baseline data. Full results can be found at the OSF site (https://osf.io/aexrb/).
We explored the distribution of the four language groups across the 22 sites of the ABCD study (Figure 3). Monolinguals and other bilinguals are distributed relatively evenly across different sites. A greater proportion of DLE bilinguals came from Oregon Health and Science University (OHSU; state: Oregon), University of Utah (state: Utah), University of California San Diego (UCSD; state: California) and Laureate Institute for Brain Research (LIBR; state: Oklahoma). A large proportion of heritage bilinguals came from UCSD and Florida International University (FIU; state: Florida), as well as the Children’s Hospital Los Angeles (CHLA; state: California).
Percent of each language group recruited from each site (baseline data).
To explore the differences between language groups in nonverbal and verbal skillsFootnote 1 (Tables 1 and 2), we ran a series of three-step linear mixed models. All analyses can be found at https://osf.io/aexrb/. We conducted model comparisons for each set of models for each outcome, and reported marginal R 2, AIC, BIC and the log-likelihood of each model, as well as chi-square values of the model comparison.Footnote 2
Fluid (nonverbal) and crystallized (verbal) composite scores by language group
Models predicting cognitive outcomes
Note: *p < .01, **p < .001. AIC = Akaike information criterion; BIC = Bayesian information criterion; LL = log-likelihood. Language group where significant is bold. R 2 reported is the marginal R 2 of the fixed effect.
3.3 Nonverbal skills
The nonverbal skill measures (fluid composite) were only available at baseline, due to some subtests only measured at this time point. For this outcome, we ran a three-step model. Model 0 included the main effect of language group and matrix reasoning; model 1 also included the child’s demographic background as covariates (sex, handedness, pubertal status and race/ethnicity); and model 2 additionally included socioeconomic background as covariates (parent education, household marital status, household income). Continuous covariates were mean-centered, including pubertal status, parent education, household income and matrix reasoning. All models are controlled for the random effects of sites.
$$ \mathrm{Model}\;0:Y\sim \mathrm{language}\ \mathrm{group}+\mathrm{matrix}\ \mathrm{reasoning}+\left(1\vert \mathrm{site}\right) $$
$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;1:Y\sim \mathrm{language}\ \mathrm{group}+\mathrm{matrix}\ \mathrm{reasoning}+\mathrm{sex}\\ {}\hskip9em +\hskip2px \mathrm{handedness}+\mathrm{pubertalstatus}+\mathrm{race}/\mathrm{ethnicity}\\ {}\hskip9em +\left(1\vert \mathrm{site}\right)\end{array}} $$
$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;2:Y\sim \mathrm{language}\ \mathrm{group}+\mathrm{matrix}\ \mathrm{reasoning}+\mathrm{sex}\\ {}\hskip9em +\hskip2px \mathrm{handedness}+\mathrm{pubertal}\ \mathrm{status}+\mathrm{race}/\mathrm{ethnicity}\\ {}\hskip9em +\hskip2px \mathrm{parent}\ \mathrm{education}+\mathrm{marital}\ \mathrm{status}\\ {}\hskip9em +\mathrm{household}\ \mathrm{income}+\hskip2px \left(1\vert \mathrm{site}\right)\end{array}} $$
Language group had a significant effect in modes 0 and 1, but not in model 2 (Table 2, Figure 4). In model 0, both types of sequential bilinguals had higher scores than both monolinguals and heritage bilinguals. There were no differences between other bilinguals and DLE bilinguals, or between monolinguals and heritage bilinguals. In model 1, when demographic variables were added to the model, both types of sequential bilinguals had higher scores than heritage bilinguals, and only DLE bilinguals had higher scores than monolinguals. There were no differences between DLE bilinguals and other bilinguals, between other bilinguals and monolinguals or between monolinguals and heritage learners. Finally, in model 2, when SES variables were added to the model, there were no differences between groups.
Estimated non-verbal skills by group.
3.4 Verbal skills
Verbal skill measures (crystallized composite) were available at baseline and year 2. We ran similar models as described earlier, also exploring the main effect of year, as well as the interaction between language group and year. These models had time nested within children, and children nested within data sites.Footnote 3
$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;0:Y\sim \mathrm{language}\ \mathrm{group}\times \mathrm{year}+\mathrm{matrix}\ \mathrm{reasoning}\\ {}\hskip9em +\left(1\vert ID/\mathrm{site}\right)\end{array}} $$
$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;1:Y\sim \mathrm{language}\ \mathrm{group}\times \mathrm{year}+\mathrm{matrix}\ \mathrm{reasoning}\\ {}\hskip9em +\hskip2px \mathrm{sex}+\mathrm{handedness}+\mathrm{pubertal}\ \mathrm{status}\\ {}\hskip9em +\mathrm{race}/\mathrm{ethnicity}+\hskip2px \left(1\vert ID/\mathrm{site}\right)\end{array}} $$
$$ {\displaystyle \begin{array}{l}\mathrm{Model}\;2:Y\sim \mathrm{language}\ \mathrm{group}\times \mathrm{year}+\mathrm{matrix}\ \mathrm{reasoning}+\mathrm{sex}\\ {}\hskip9em +\hskip2px \mathrm{handedness}+\mathrm{pubertal}\ \mathrm{status}+\mathrm{race}/\mathrm{ethnicity}\\ {}\hskip9em +\hskip2px \mathrm{parent}\ \mathrm{education}+\mathrm{marital}\ \mathrm{status}\\ {}\hskip9em +\mathrm{household}\ \mathrm{income}+\hskip2px \left(1\vert ID/\mathrm{site}\right)\end{array}} $$
There was a significant main effect of year on all models, such that children performed significantly better in baseline than in year 2, for age-corrected scores. There were significant main effects of language groups in all models (Table 2, Figure 5). In model 0, both types of sequential bilinguals had higher scores than monolinguals, who had higher scores than heritage bilinguals. There were no differences between other bilinguals and DLE bilinguals. In model 1, when demographic variables were added to the model, the same trend persisted. In model 2, when SES variables were added to the model, both types of sequential bilinguals had higher scores than both heritage bilinguals and monolinguals. There were no differences between DLE bilinguals and other bilinguals, or between monolinguals and heritage bilinguals. There was a language group-by-year interaction in all models, suggesting that group differences varied by year. We explored the simple effect of language group differences within each year for model 2. At baseline, there were no differences between DLE bilinguals and other bilinguals, or between monolinguals and heritage bilinguals. Both sequential groups scored higher than heritage bilinguals, but only other bilinguals scored higher than monolinguals. In year 2, there were no differences between DLE bilinguals and other bilinguals, or between other bilinguals and monolinguals. DLE bilinguals scored higher than monolinguals, and both other bilinguals and monolinguals scored higher than heritage bilinguals.
Estimated verbal skills by group by year.
4. Discussion
To our knowledge, this is the first study to investigate how different types of bilingual experience relate to cognitive outcomes, in comparison to monolingual adolescents within the United States. Using data from the ABCD Study, we examined whether cognitive differences across bilingual and monolingual adolescents remain after accounting for variability in language background and socioeconomic status (SES). Prior research has primarily focused on one of these types of bilinguals (i.e., heritage bilinguals or sequential bilinguals, which consists of DLE bilinguals and other bilinguals). This is likely because communities where research is conducted may have more support for one type or group of bilinguals than others. Research conducted in Florida or Southern California seemed to recruit more heritage bilinguals, whereas research conducted in Utah and Oregon had larger samples of dual language education bilinguals, and research conducted in the northeast and mid-west included more other bilinguals. By making use of the ABCD study, we were able to, for the first time, compare adolescents from each of these groups from across the country.
We found that language groups differed by SES (see Supplementary Materials), with sequential bilinguals coming from families with higher SES than monolinguals, who, in turn, had higher SES than heritage bilinguals. This finding is crucial in explaining variability in research on bilingualism and cognitive abilities, as discussed further next. The language profiles of bilingual adolescents in the ABCD study reflect notable variability, with monolinguals, dual language education (DLE) bilinguals and other bilinguals making up 88.8% of the sample, a higher percentage than the 78.1% reported in the U.S. Census data (Table S1603; 2021). Heritage bilinguals—the largest bilingual group in the United States—primarily spoke Spanish (65–70%) in this sample, aligning with national estimates (70.49%). Self-reported English proficiency was high across all groups, with sequential bilinguals rating their skills similarly to monolinguals, while heritage bilinguals reported slightly lower ratings. In their other language, heritage bilinguals reported the highest proficiency (2.8/4), followed by DLE (2.6/4) and other bilinguals (2.07/4). Language use patterns varied, with heritage bilinguals primarily speaking their home language with family, while sequential bilinguals predominantly used English. These findings suggest the ABCD study provides a representative sample of U.S. children and highlight how variability in bilingual experiences—and crucially, SES—must be accounted for when examining cognitive and developmental outcomes (for a more detailed discussion, see Supplementary Materials).
Most importantly, in this study, we explored group differences in cognition as measured by the NIH Toolbox Cognition Battery. In general, the findings align with previous research suggesting that SES variables, including parental education and household income, can confound the relationship between bilingualism and cognitive abilities (Grosjean, Reference Grosjean2010). For instance, DLE bilinguals, who have higher SES than other groups, initially showed better cognitive performance, but these differences disappeared after controlling for SES and demographics, indicating that their apparent advantage reflected broader access to educational resources rather than language status. Interestingly, other bilinguals also demonstrated cognitive advantages over heritage bilinguals and monolinguals in some tasks, which may reflect their similarity in demographic and SES background with DLE bilinguals. This also suggests that dual language immersion may not be solely responsible for group differences, as sequential bilinguals who learned a language outside of these programs showed similar better cognitive performances compared to the other groups. However, these advantages were not uniformly observed across all cognitive measures (as discussed next), indicating the need for more nuanced interpretations of the bilingual advantage.
The nonverbal skills measures, proxied by the fluid composite, consist of tasks measuring various aspects of executive function or nonverbal skills, such as cognitive flexibility (DCCS; Zelazo, Reference Zelazo2006; Zelazo et al., Reference Zelazo, Anderson, Richler, Wallner-Allen, Beaumont and Weintraub2013), inhibitory control (Flanker; Eriksen, Reference Eriksen1995; Zelazo et al., Reference Zelazo, Anderson, Richler, Wallner-Allen, Beaumont and Weintraub2013), episodic and working memory (Picture Sequence, Dikmen et al., Reference Dikmen, Bauer, Weintraub, Mungas, Slotkin, Beaumont, Gershon, Temkin and Heaton2014; and List Sorting, Tulsky et al., Reference Tulsky, Carlozzi, Chiaravalloti, Beaumont, Kisala, Mungas, Conway and Gershon2014) and general processing speed (Pattern Comparison; Carlozzi et al., Reference Carlozzi, Tulsky, Chiaravalloti, Beaumont, Weintraub, Conway and Gershon2014). Initial comparisons showed sequential bilinguals scoring higher than monolinguals, who outperformed heritage bilinguals, but these differences disappeared after controlling for SES and demographics. This finding highlights the role of SES in shaping cognitive outcomes and may explain inconsistencies in prior bilingualism research.
This pattern aligns with meta-analytic evidence showing that apparent bilingual advantages in executive function are largely influenced by SES and study quality, with small positive effects primarily observed in studies using self- or parent-reported measures of language (Lowe et al., Reference Lowe, Cho, Goldsmith and Morton2021). Notably, immersion or other education programs did not moderate these effects in the meta-analysis, suggesting that the type of bilingual education alone does not drive cognitive differences. Our findings further reinforce that controlling for SES is critical to accurately interpreting cognitive differences between bilingual and monolingual adolescents. This study extends prior work using the ABCD study by Dick et al. (Reference Dick, Garcia, Pruden, Thompson, Hawes, Sutherland, Riedel, Laird and Gonzalez2019) by explicitly comparing bilinguals with different language experience and SES background to monolinguals and showing that the observed group differences are primarily explained by SES rather than bilingual experience itself.
In this sample, sequential bilinguals had higher SES than monolinguals, who in turn had higher SES than heritage bilinguals, aligning with national trends (Table S1603; US Census, 2021). The perception of bilingualism also varies by SES, with higher SES families viewing it as an asset, while lower SES families often see it as a barrier to integration (Agirdag, Reference Agirdag2014; Garcia, Reference Garcia2002; Nieto, Reference Nieto2010; Portes & Hao, Reference Portes and Hao2002). These socioeconomic disparities influence educational trajectories and may account for past findings on bilingual cognitive advantages or disadvantages. Consistent with recent meta-analyses, our results suggest no inherent cognitive benefits or deficits tied to bilingualism (Antón et al., Reference Antón, Duñabeitia, Estévez, Hernández, Castillo, Fuentes, Davidson and Carreiras2014; de Bruin et al., Reference de Bruin, Treccani and Della Sala2015; Paap et al., Reference Paap, Johnson and Sawi2015; Paap & Greenberg, Reference Paap and Greenberg2013), though it bears emphasizing that the bilingual adolescents’ ability to communicate in two languages remains a distinct and valuable form of expertise. In the present adolescent sample, differences among bilingual subgroups were largely explained by variation in SES rather than by bilingual experience itself. While we have focused on major factors that differentiate bilingual experiences, we acknowledge that other unmeasured variables may also contribute to cognitive outcomes. Nevertheless, these findings highlight the importance of accounting for co-occurring factors, such as SES, when examining cognitive performance in bilingual populations.
In terms of the verbal skills measures, proxied by the crystallized composite, this measure consists of tasks measuring verbal skills, specifically of vocabulary comprehension and reading decoding (Picture Vocabulary and Oral Reading Recognition; Gershon et al., Reference Gershon, Slotkin, Manly, Blitz, Beaumont, Schnipke, Wallner-Allen, Golinkoff, Gleason, Hirsh-Pasek, Adams and Weintraub2013). Similar to the nonverbal skills findings, sequential bilinguals scored higher than monolinguals, who outperformed heritage bilinguals. However, unlike the nonverbal skills, these differences remained even after controlling for SES and demographics, with sequential bilinguals maintaining an advantage. This challenges earlier theories suggesting bilingualism benefits nonverbal executive function, but hinders language skills (Bialystok & Craik, Reference Bialystok and Craik2010). While bilinguals have been thought to experience a “lexical deficit,” recent research, including a meta-analysis (Bylund et al., Reference Bylund, Antfolk, Abrahamsson, Olstad, Norrman and Lehtonen2023), found no such deficit in either simultaneous or sequential bilinguals. In this study, sequential bilinguals showed gains in language skills, and heritage bilinguals did not exhibit disadvantages when SES was controlled for. Heritage bilinguals in this sample may have initially lagged in English, but seemingly caught up by ages 9–12, contradicting past research that found language-minority children struggle to close the gap (Mancilla-Martinez & Lesaux, Reference Mancilla-Martinez and Lesaux2011). However, prior studies often involved low-SES samples, suggesting that observed deficits may stem from socioeconomic disadvantage rather than bilingualism itself. Since SES influences various aspects of language development (Hoff, Reference Hoff2013), it is crucial to account for it in linguistic research in childhood and adolescence.
Notably, sequential bilinguals consistently outperformed monolinguals, even when SES was accounted for. These children likely acquired their second language through structured school programs, such as dual language immersion. While private school enrollment rates were similar across monolinguals, DLE bilinguals and heritage bilinguals, other bilinguals—who had higher SES—enrolled at much higher rates, suggesting access to quality language instruction may contribute to verbal skill differences. However, it is also possible that children with stronger verbal skills were more likely to receive additional language instruction, perhaps through self-selection or because their parents or teachers identified their potential for enhanced language instruction, making the relationship between bilingualism and cognitive abilities bidirectional. These findings highlight the complexity of bilingualism’s impact and the need to consider diverse bilingual experiences rather than treating bilingualism as a uniform category.
In sum, this study emphasize the need to consider demographic and SES factors when examining bilingualism’s cognitive effects. The absence of consistent cognitive benefits suggests that the “bilingual advantage” is more complex than previously thought and is likely influenced by external factors such as education and socioeconomic background. Expanding research beyond Western contexts can further clarify how bilingualism interacts with cultural and economic diversity. A key strength of this study is the use of ABCD data, which provide a large representative sample and detailed background information. Given that language minority status is often confounded with SES in the United States, isolating these effects is challenging (Hoff, Reference Hoff2013). By analyzing both factors together, this study identified which effects were SES-related and which might be linked to bilingualism itself. However, a limitation of the dataset is that all cognitive and language tasks were conducted in English, making it unclear how bilinguals would perform in their other language. Additionally, there were no objective measures of bilinguals’ non-English language proficiency, particularly for heritage bilinguals, for whom it is their first language. Prior research suggests that literacy and language experience in a child’s home language can enhance their school language learning (Hoff, Reference Hoff2013). Future research should incorporate objective assessments of bilinguals’ skills in both languages to better understand bilingualism’s cognitive and linguistic effects.
5. Conclusion
Overall, this study highlights the complexity of assessing bilingual advantage (or the lack thereof), as differences in cognitive scores were often influenced by demographic and socioeconomic factors rather than language background alone. The initial analysis of cognitive task performance indicated that sequential bilinguals (including dual language education bilinguals and other bilinguals) outperformed monolinguals and heritage bilinguals. However, when SES factors were included as covariates, these cognitive advantages disappeared, confirming that SES rather than bilingualism itself accounted for these differences. This suggests that the apparent cognitive advantages seen in sequential bilinguals were not solely due to their language learning experiences. Instead, factors such as higher levels of parental education and greater household income, which are often associated with better access to educational resources and supportive learning environments, played a significant role in influencing these cognitive outcomes. Thus, socioeconomic status and related demographic factors are crucial variables to consider when assessing the effects of bilingualism on cognitive and educational outcomes, and future research should rigorously control for these variables—through matching or covariate inclusion—to avoid confounding and improve interpretability.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S1366728926101023.
Data availability statement
Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development (ABCD) study (https://abcdstudy.org), held in the NIMH Data Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 children aged 9–10 and follow them over 10 years into early adulthood. The ABCD Study® is supported by the National Institutes of Health and additional federal partners under award numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, U01DA041148, U01DA041093, U01DA041089, U24DA041123, U24DA041147. A full list of supporters is available at https://abcdstudy.org/federal-partners.html. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/consortium_members/. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in the analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time. The ABCD data used in this report came from Release 5.1, DOI: 10.15154/z563-zd24.
Competing interests
The authors declare none.
Open access
