Early-Life Bilingualism and Midlife Cognitive Health: Evidence from High School and Beyond

The Debate on the Bilingual Advantage

The relationship between bilingualism and adulthood cognitive health remains unclear. According to Cognitive Reserve Theory (Stern 2002), cognitively stimulating tasks can strengthen the brain’s ability to cope with damage and resist age-related cognitive decline by developing alternative neural pathways available in the event that some pathways are blocked. Learning a second language may contribute to cognitive reserve due to the frequent use of executive function and engagement in mental processes entailed in language acquisition that, in turn, enhance the brain’s efficiency and flexibility in managing cognitive tasks (Bialystok 2021). Supporting this view, many researchers find that bilingual speakers have better cognitive performance and later onset of cognitive diseases than their monolingual counterparts (Abutalebi et al. 2015; Alladi et al. 2013; Ballarini et al. 2023; Bialystok and Craik 2022; Bialystok, Craik, and Freedman 2007; Bialystok et al. 2014, 2016; Perani et al. 2017).

On the other hand, other researchers do not observe a significant positive association between bilingualism and adulthood cognitive functioning (Nichols et al. 2020; von Bastian et al. 2016; Yeung et al. 2014), and some research even finds a negative association (Anderson, Saleemi, and Bialystok 2017; Folke et al. 2016; Lehtonen et al. 2018). Evidence for the cognitive benefits of bilingualism is thus inconsistent. Unobserved confounders, such as SES, immigration status, and cultural differences, may cause a spurious bilingual advantage (Paap, Johnson, and Sawi 2015). Moreover, research finds that studies supporting the bilingual advantage are more likely to be published than those challenging it (de Bruin, Treccani, and Della Sala 2015), so the cognitive benefits of bilingualism suggested by some studies may reflect publication bias rather than true effects. Preference for positive findings and reliance on laboratory and clinical recruitment have contributed to a literature dominated by small, selective samples with limited socioeconomic heterogeneity (Paap et al. 2015), while evidence from large, nationally representative surveys remains scarce.

Prior debates on the bilingual advantage have been largely shaped by psychologists and neuroscientists, who focus on neurocognitive mechanisms and generally treat bilingualism as exogenous in its relation to cognition. Few studies have examined the potential confounding role of other early-life cognitive skills and broader educational and stratification contexts in this relationship. Our study aims to bridge this gap by situating the bilingualism–cognition link within a broader sociological discussion of education, cognitive skills, and cognitive health across the life course and by examining how early-life educational factors may confound the observed association.

Early-Life Bilingualism and Midlife Cognition

Early-life bilingualism may contribute to midlife cognition through three pathways. The first possible mechanism is neurocognitive. Drawing on Cognitive Reserve Theory, research suggests that acquiring and using different languages may enhance the brain’s efficiency and flexibility in utilizing neural resources, thereby providing protection against cognitive diseases (Bialystok 2021). Critical Period Theory further suggests that early life represents a sensitive window of heightened neural plasticity, during which language acquisition plays a pivotal role in shaping the brain’s functional connectivity (Knudsen 2004; Kwon et al. 2021). Research also indicates that such early neural differences can persist into adulthood (Cheng et al. 2023), implying that exposure to bilingualism during this developmental window could have enduring cognitive consequences. If this neurocognitive pathway is operative, then the association between bilingualism and adulthood cognitive health should be most evident when different languages are acquired early in life.

Second, early-life bilingualism may benefit midlife cognition indirectly by enhancing SES and resources. Human Capital Theory holds that knowledge and skills increase employment opportunities and earnings (Deming 2022), and bilingualism constitutes such a skill. Empirical studies show that bilingual individuals have higher educational attainment, occupational status, and earnings compared to their monolingual peers (Gándara 2015; Rumbaut 2014; Santibañez and Zárate 2014). In labor markets, bilingualism expands opportunities in both mainstream and ethnic economies, reduces unemployment risks, and opens access to jobs with wage premiums for language skills (Agirdag 2014). By improving educational attainment, labor market prospects, and socioeconomic standing, bilingualism may indirectly benefit cognitive health. While few studies examine whether SES mediates the relationship between bilingualism and adulthood cognition, prior research finds that SES mediates the relationship between bilingualism and better self-rated health (Schachter, Kimbro, and Gorman 2012), although the measurement of self-rated health might be sensitive to survey language and cultural context (Kandula, Lauderdale, and Baker 2007).

Third, bilingualism may support cognitive functioning through enhanced social connections. Individuals who speak multiple languages can interact with a wider range of people across different cultural and linguistic groups, which broadens the diversity of their social contacts. Exposure to such heterogeneous social environments provides more varied cognitive stimulation, through novel conversations, perspectives, and problem-solving interactions, which may strengthen cognitive skills (Peng and Perry 2025; Perry et al. 2022). Additionally, expanded social networks can provide greater access to support when needed, buffering stress and consequently promoting cognitive well-being (Moorman and Pai 2024). In this way, bilingualism may indirectly bolster midlife cognitive health by expanding opportunities for social engagement and support across social networks.

Despite these plausible pathways, the observed bilingualism–cognition association may also reflect the influence of other early-life cognitive skills and academic factors. As mentioned before, bilingualism develops alongside other cognitive abilities, such as academic skills of language, reading, and mathematics (Fernandez and Nielsen 1986; Nielsen and Lerner 1986), as they use shared neural networks for attention, working memory, and executive function, and they are cultivated within overlapping educational and developmental contexts. These other cognitive skills themselves are strong predictors of adulthood cognition. For example, using data from High School and Beyond: 1980, researchers demonstrate that individuals’ academic performance in high school strongly predicts midlife cognition, even net of educational attainment (Muller et al. 2025). Similarly, research using the Wisconsin Longitudinal Study (WLS) shows that early-life cognitive ability is a powerful predictor of cognitive functioning later in life (Moorman and Kong 2025; Zhang, Liu, and Choi 2020), despite the lack of national representativeness and racial and ethnic diversity in the WLS sample. These early-life cognitive skills also shape educational attainment, occupational status, earnings, and health behaviors (Farkas 2003; Hall and Farkas 2011; Pokhrel et al. 2013), each of which is, in turn, associated with adulthood cognitive health (Greenfield, Akincigil, and Moorman 2020; Herd and Walsemann 2025; Vik et al. 2004).

Contradicting claims of the benefits of bilingualism, some research finds little evidence that bilingual adults outperform their monolingual counterparts in educational achievement, labor market outcomes, or economic returns (Fry and Lowell 2003; Mouw and Xie 1999; Shin and Alba 2009). For instance, Fry and Lowell (2003) show that bilingualism itself does not significantly predict wages net of educational attainment. Taken together, these findings underscore the need to situate the bilingualism–cognition link within the broader context of early-life cognitive development and educational inequality in order to assess whether bilingualism itself is protective for cognition or whether its apparent cognitive benefits are shaped by underlying educational and social advantages.

We distinguish between becoming bilingual before and after starting formal education, a distinction rarely examined in prior research due in part to data constraints. This step is important because one major challenge in studying the relationship between early-life bilingualism and midlife cognition is the potential confounding of early-life educational experiences and school contexts, which are well-established determinants of cognitive health across the life course (Moorman, Greenfield, and Garcia 2019; Moorman and Kong 2025; Muller et al. 2025; Walsemann and Ailshire 2020). By separating language learning that occurs before versus after school entry, we can better probe potential mechanisms. If becoming bilingual before school entry shows a stronger association with midlife cognition than after school entry, it would strengthen our confidence that early-life bilingualism is associated with midlife cognition independent of educational contexts. Moreover, it can provide further support for Critical Period Theory, which suggests that earlier language learning may more strongly influence brain development due to heightened neuroplasticity in early childhood (Hartshorne 2022). However, if becoming bilingual after school entry is more strongly associated with midlife cognition, it suggests a greater need to be cautious about the potential confounding role of formal educational experiences in the relationship between early-life bilingualism and midlife cognition. Either way, findings derived from this distinction will expand our understanding of the relationship between early-life language and educational experiences and midlife cognition.

While we focus on the confounding role of early-life academic factors, other factors, including family SES, race and ethnicity, immigration status, and cultural differences, may all shape selection into bilingualism and influence midlife cognition. For example, individuals raised in racially and ethnically diverse communities are more likely to become bilingual, and members of such communities often exhibit stronger social connectedness (Pearrow, Sander, and Jones 2019), which can benefit cognition (Perry et al. 2022). Many bilingual speakers are also immigrants who maintain social ties both in their countries of origin and in the United States, which promotes continued use of different languages and facilitates access to social support that may benefit cognition. Thus, ignoring race, ethnicity, and immigration status could overstate the role of bilingualism in fostering social connectedness and consequently cognition.

Another selection mechanism worth noting is immigrant health and achievement selection. Healthier individuals are more likely to migrate, which can make immigrants appear to have better health outcomes than native-born populations (Markides and Eschbach 2005). This advantage has also been observed in cognitive performance (Garcia et al. 2020; Hill et al. 2012). In addition, immigrant students may demonstrate stronger early-life cognitive performance in school than their native-born peers as they are often positively selected from families with greater resources or stronger academic preparation in their countries of origin (Feliciano 2020). Consequently, it is essential to account for immigration status in our study since any observed bilingual advantage may partly reflect this health or achievement selection rather than the causal effect of bilingualism itself. Another related health selection is “salmon bias,” in which immigrants with declining health return to their country of origin and thus are not observed in U.S. data (Riosmena, Wong, and Palloni 2013; Turra and Elo 2008). Given our focus on midlife cognition, this bias might not be substantial as selective return migration primarily operates at older ages (Markides and Eschbach 2005). However, given the long interval between early-life bilingualism and midlife cognition, we will still address selection and attrition between exposure and outcome in our analytic sample.

First Language as a Potential Source of Heterogeneity

Beyond differentiating becoming bilingual before versus after school entry, this study also distinguishes among three linguistic groups: English monolingual, English-first bilingual, and non-English-first bilingual speakers. Bourdieu’s (1977) concept of linguistic capital suggests that the value of language skills depends on the social context and can shape individuals’ life chances and experiences. In the United States, where English is the majority language, the timing of English acquisition may be especially consequential for how linguistic resources translate into educational, social, and institutional opportunities that may have important implications for cognitive health over the life course.

Based on this perspective, English-first bilingual speakers may exhibit the most favorable midlife cognitive performance as their bilingualism typically reflects the addition of a second language to an already advantaged linguistic position, combining cognitive stimulation with fewer early-life barriers to institutional resources and social integration. By contrast, non-English-first bilingual individuals may face early-life constraints associated with limited English skills, including restricted access to quality childcare and schooling, health care, and social integration as well as exposure to discrimination (Akresh et al. 2014; Derose et al. 2007; Murillo and Smith 2011; Nores and Barnett 2014), all of which may have lasting implications for cognitive health across the life course (Glymour and Manly 2008). However, it remains unclear whether the hypothesized neurocognitive benefits of early-life bilingualism can offset such potential long-term consequences of early language-based disadvantage, making non-English-first bilingual speakers exhibit higher midlife cognition than English monolingual speakers. Comparing midlife cognition across the three linguistic groups therefore provides an opportunity to probe the joint role of neurocognitive stimulation and language-based social inequality in cognitive aging.

Data and Sample

We used data from High School & Beyond: 1980 (HS&B:80) to address our research questions. HS&B:80 is nationally representative, reflects the experience of a large sample of people who were initially observed in high school in 1980, and includes abundant information on respondents’ early-life language learning, early-life educational factors, and midlife cognition. In 1980, HS&B:80 interviewed 30,030 sophomores and 28,240 seniors in 1,020 randomly selected public and private high schools (about 60 students per school) in the United States. From the initial sample of 58,270 students, random subsamples of 14,830 sophomores and 12,000 seniors were selected to participate in a longitudinal panel from 1982. Respondents were followed over time until 2021. We used data from the 1980, 1982, and 2021 waves of HS&B:80; the first two waves included information on respondents’ early-life experiences, and the last wave measured respondents’ midlife cognitive health. Our analytic sample was restricted to respondents who participated in the 2021 wave ( N ≈ 13,980). All sample sizes were rounded to the nearest 10 in accordance with the requirements of our restricted-use data license.

Measures

Midlife cognition was the dependent variable. Cognitive tasks were included in both the 2021 HS&B:80 telephone survey and the web survey. The neurocognitive assessments in the telephone survey included immediate and delayed word recall, phonemic and semantic fluency, and forward and backward digit span. The cognitive assessments in the web survey included verbal and visual pairs. We used a measure of general cognitive functioning derived from a hierarchical item response theory model (Huang et al. 2013) that combines information from and accounts for shared variance across all of the aforementioned tasks. To enhance comparability across survey modes and improve measurement precision, we incorporated a broad set of covariates from high school and midlife as an explanatory component of the higher-order latent variable. For additional information, see the online appendix of Muller et al. (2025).

The first independent variable was early-life bilingualism. Respondents were asked whether they spoke another language in addition to English in 1980. We used this information to create a dichotomous variable distinguishing respondents who spoke only one language (i.e., English) from those who spoke more than one language. About 99 percent of respondents who reported more than one language spoke two languages, so we used the term “bilingualism” in this study.

Prior research suggests that language proficiency should also be considered when determining bilingual status (Schachter et al. 2012). Respondents who reported speaking more than one language were also asked to rate their proficiency in understanding, speaking, reading, and writing both English and the other language using the following scale: very well (= 1), pretty well (= 2), not very well (= 3), and not at all (= 4). Over 94 percent of these bilingual respondents rated their English proficiency as very well or pretty well across all four dimensions, while their non-English proficiency showed greater variation. We conducted sensitivity analyses excluding respondents whose non-English proficiency was below pretty well from the bilingual group. Specifically, following prior research (Schachter et al. 2012), we summed the four measures of respondents’ abilities to understand, speak, read, and write the non-English language (α = .9), yielding a total score ranging from 4 to 16. A total score of 8 or lower corresponded to reporting “pretty well” (or better) across all four domains, whereas scores above 8 indicated at least one domain rated below “pretty well.” We reclassified individuals with scores above 8 (about 55.6 percent among bilingual respondents; the comparable proportion for English proficiency was only 3.6 percent) from bilingual to monolingual and reran regression analyses. This threshold represented a conservative definition of bilingualism and was used solely for sensitivity analysis. Results did not change our main conclusion.

The timing of becoming bilingual was the second independent variable. For people who spoke more than one language, HS&B:80 also asked whether they spoke more than one language before school. We distinguished among people who were monolingual, became bilingual before school entry, or became bilingual after beginning school.

The third independent variable was defined by bilingualism and individual first language. Respondents who were able to speak more than one language were also asked what their first language was. HS&B:80 asked respondents two questions: (1) “What was the first language you spoke when you were a child?” and (2) “What other language did you speak when you were a child before you started school?” Responses included “no other language” as well as options to select or report one or more specific language(s). Based on these two questions, we defined respondents as having a non-English first language if they did not report speaking English before starting school. We created a three-category indicator distinguishing English monolingual, English-first bilingual, and non-English-first bilingual respondents.

Analytic Strategy

We used ordinary least squares (OLS) regression because the outcome variable was continuous and approximately normally distributed. We also assessed key OLS assumptions, including linearity, homoscedasticity, the normality of residuals, and no multicollinearity, and found no evidence of major violations. Model 1 estimated the marginal association between early-life bilingualism and midlife cognition without any covariates. Model 2 controlled for respondents’ basic demographic characteristics, including sex, race and ethnicity, childhood SES, religious background, nativity, length of living in the United States by 1980, and early-life disability. Model 3 further controlled for respondents’ academic performance and educational expectations in the senior year of high school. Model 4 further controlled for respondents’ highest educational attainment.

Only parental occupational prestige had more than 20 percent of missing values (25 percent) due to nonresponse and responses to the open-ended question of specific occupations that were not codable. Among remaining variables, the share of observations with missing values ranged from .1 percent (the highest level of education) to 11.4 percent (religious background). We used multiple imputation with chained equations to address these missing values. All regression analyses were conducted across 50 imputed data sets. We represented cognitive functioning in 2021 using plausible values (Mislevy 1991) to account for measurement error and uncertainty in estimating individuals’ true cognitive ability. These plausible values are drawn from a posterior distribution based on item response theory models, incorporating both observed responses and background variables. This approach yields more valid population-level inferences while preserving the psychometric properties of the cognitive assessments. To facilitate result interpretation, we z-standardized the plausible values of cognition scores so that coefficients reflected changes in standard deviation units (see Figure 1). In multiple imputation and regression analyses, we used the cross-sectional weight provided by HS&B:80 for the 2021 wave, which adjusted for sample attrition and could produce estimates that generalized to the full population of 1980 high school sophomores and seniors for variables collected in the 2021 wave (see Appendix A in the online version of the article), adjusting for school-level clustering.

OPEN IN VIEWER

Figure 1.

The weighted distribution of z-standardized plausible values of cognition scores in 2021 across 50 imputed data sets (N ≈ 13,980).

Supplementary Analysis

Why is the significant association between early-life bilingualism and midlife cognition attenuated to nonsignificance after controlling for high school educational factors? This pattern may be explained in two ways. First, if these educational factors were confounders, the observed association between early-life bilingualism and midlife cognition may reflect underlying academic advantages rather than a direct effect of bilingualism itself. This possibility raises questions about the extent to which early-life bilingual exposure independently contributes to cognitive health and casts doubt on the effectiveness of early-life bilingual education as a preventive strategy for adulthood cognitive impairment. Alternatively, if these educational factors were mediators, then bilingualism may still have played an indirect role in promoting cognitive health through its influence on academic achievement.

To further assess whether early-life bilingualism might enhance midlife cognition through academic performance, we conducted supplementary analyses using second-language course-taking in high school as a proxy for structured bilingual exposure. Specifically, HS&B:80 collected sophomores’ transcript data in 1982 (i.e., their senior year), which included information on individual second-language course-taking in high school. Additionally, HS&B:80 measured sophomores’ academic performance and expectations of educational attainment in 1980. Restricting our analytic sample to sophomores who were monolingual in 1980, we examined whether having a second-language course by 1982 was linked to midlife cognition, controlling for academic performance (including their achievement scores and self-reported grades) and educational expectations measured in 1980.

Table 4 shows the results. While having taken a second-language course by the senior year was initially associated with better midlife cognitive performance, this association was largely attenuated after controlling for academic performance and educational expectations in the sophomore year. This finding supports the interpretation that the association between early-life bilingualism and midlife cognition is more likely driven by confounding than mediation.

OPEN IN VIEWER

Table 4.

Ordinary Least Squares Regression Estimating the Association between Second-Language Course-Taking in 1982 and Midlife Cognition among Sophomores Who Were Monolingual in 1980 (N ≈ 5,810).

	Model 1	Model 2	Model 3	Model 4
Any second language course(s) in 1982	.74*** (.04)	.49*** (.04)	.05(.04)	.02(.04)
Basic demographics (sex, race and ethnicity, childhood socioeconomic status, religion, nativity, U.S. residence, early-life disability)		√	√	√
Academic performance and educational expectations in 1980			√	√
Educational attainment				√

Source. U.S. Department of Education, National Center for Education Statistics, High School and Beyond: 1980/2021.

Note. The analytic sample included sophomores who participated in the 1982 survey, were monolingual in 1980, and had cognition scores in 2021. Sample size rounded to nearest 10 per data-use requirements. Academic performance and educational expectations in 1980 (not shown) were significant in Model 3; academic performance remained significant in Model 4 after controlling for educational attainment. See Appendix Table 3 in the online version of the article. Standard errors are in parentheses.

***

p < .001.