Association of majority versus minority first language multilingualism and socioeconomic status with cognition among older adults residing in India

Abstract

INTRODUCTION

Multilingualism has been proposed as a protective factor against cognitive aging, but its effects are often intertwined with socioeconomic and sociolinguistic contexts, particularly whether an individual's first language is the societal majority or a minority language.

METHODS

We analyzed data from 3854 older adults in the Longitudinal Aging Study in India—Diagnostic Assessment of Dementia (LASI‐DAD). Participants were classified as monolingual, majority‐language multilingual, or minority‐language multilingual. Cognitive outcomes were examined across SES strata using multiple‐group models.

RESULTS

Higher SES was consistently associated with better cognition. Multilingualism was associated with better cognitive performance at medium and high SES levels, but not at low SES. Among multilinguals, majority‐language speakers outperformed minority‐language speakers at high SES; however, these differences were largely driven by SES advantages among majority‐language speakers.

DISCUSSION

Cognitive benefits of multilingualism are not uniform but vary by SES. These findings underscore the importance of context in understanding multilingualism and cognitive aging.

Highlights

• We examined whether multilingual benefits vary by SES and language background

• Multilingual benefits were seen at medium and high SES, but absent at low SES

• Majority‐multilinguals had higher SES but similar cognition as minority‐multilinguals

• Multilingual cognitive benefits depend on SES and sociolinguistic context

1. BACKGROUND

Multilingualism may be associated with better late‐life cognitive outcomes, but the current evidence remains inconclusive. A systematic review of 24 studies on multilingualism and cognition in older adults found that 38% of studies reported a cognitive benefit for multilingual individuals, while 16% observed the opposite (i.e., a bilingual ‘disadvantage’), and almost half of the studies yielded mixed results, highlighting the complexity of this relationship. ¹

Efforts to explain the variability in findings emphasize the need to consider the heterogeneity among multilingual people. ² Multilingualism arises through diverse experiences and for different reasons, shaped by the (socio) linguistic contexts in which it develops. ³ In monolingual‐dominant countries such as the United States, multilingualism often emerges through migration, where individuals acquire the majority societal language alongside a non‐majority first language. Conversely, non‐majority languages may also be acquired through formal education. These distinct pathways involve different acquisition contexts—such as occupational and socioeconomic opportunities—and are associated with varying sociolinguistic features, including patterns of use, social prestige, and language attitudes. However, despite their diversity, multilingual people are often reduced to a monolithic group. To better understand how the knowledge and usage of more than one language contributes to cognition, it is essential to account for the heterogeneity among multilingual people and the different contexts in which multilingualism arises. This study aims to address this gap by differentiating among multilingual individuals between those whose first language was the majority language (majority‐multilingual) and those whose first language was not the majority language (minority‐multilingual).

RESEARCH IN CONTEXT

1. Systematic review: We reviewed the literature through March 2025 using traditional sources (e.g., Scopus). Findings are mixed on the cognitive consequences of multilingualism. While many studies acknowledge the heterogeneity of multilingual populations and the risk of confounding—particularly by SES—few examine how broader social and linguistic contexts shape cognitive outcomes.

2. Interpretation: Using data from 3854 older adults in India, we found that multilingualism was associated with better cognitive performance only among individuals with medium or high SES, but not at lower levels of SES. These findings suggest that multilingualism should not be treated as a uniform protective factor; instead, its cognitive effects may depend on early‐life educational and socioeconomic opportunities and the broader sociolinguistic environment.

3. Future directions: Our study underscores the need to account for both socioeconomic conditions and multilingual context in research and policy aimed at promoting cognitive health and healthy aging—particularly in low‐ and middle‐income countries. Future studies should examine how intersecting structural factors—such as SES, education, and language background—influence the cognitive consequences of multilingualism.

Multilingual contexts, however, not only shape language backgrounds, but are often also associated with divergent sociodemographic backgrounds. Within the context of migration, multilingualism can co‐occur with socioeconomic adversity faced by (ethnic) minority groups. ⁴ Conversely, when learning an additional language at school, multilingualism may be more closely connected to education and, thereby, enhanced socioeconomic opportunities. Socioeconomic status (SES) —defined as representing an individuals’ possession of normatively valued social and economic resources ⁵ — in and of itself is a strong determinant of late‐life cognitive health, with studies showing that both childhood SES (e.g., parental education, childhood financial rating) and adulthood SES (e.g., education, occupation, income) are associated with better cognitive outcomes in late‐life. ⁶ Not appropriately accounting for socioeconomic backgrounds limits our ability to infer whether differences in cognition between multilingual and monolingual adults are due to multilingualism or socioeconomic factors. ⁷ , ⁸

The reasons to become multilingual, such as whether your first language is majority or minority, are associated with socioeconomic factors (i.e., migration, education) in most western and high‐income countries, but this association may be weaker or more varied in contexts with a higher prevalence and more diversity in multilingualism, such as India. India is one of the most linguistically diverse countries in the world, with 122 languages spoken by over 10,000 people and an additional 1599 documented languages across its regions. ⁹ Moreover, multilingualism is a common occurrence in Indian society, with around 26% of the population estimated to be multilingual, which may be an underestimate due to a lack of consensus on differentiating languages from dialects. ¹⁰ Given that multilingualism is inherent to Indian society, it may be more equally distributed across levels of SES than in most Western and high‐income countries. Additionally, India has experienced unprecedented economic growth in recent decades, increasing socioeconomic variability and making it particularly suitable for studying the interaction between SES, and multilingualism on cognitive outcomes.

In this study, we evaluated the independent and synergistic contributions of SES and majority‐/minority‐multilingualism on cognitive functioning among older Indian adults. We examined two multilingual contexts: Multilingual people who reported the state majority language as their first language (majority‐multilingual), and multilingual people who did not report their states’ majority language as their first language (minority‐multilingual), where minority denotes the relative status of this group's first language. We hypothesized that higher SES and multilingualism would be independently associated with better cognitive functioning. However, we further hypothesized that SES would modify the magnitude of multilingual effects, such that multilingual benefits would be more evident among higher‐SES individuals. Additionally, we examined whether the cognitive benefits of multilingualism differed between majority‐ and minority‐multilingual individuals, and to what extent such differences might be attributable to SES. Specifically, we hypothesized that any advantages observed in the majority‐multilingual group would be amplified at higher SES levels, consistent with the idea that majority‐multilingualism may more often reflect socioeconomic pathways of additional language acquisition, whereas minority‐multilingualism may be more indicative of majority‐language immersion.

2. METHODS

2.1. Participants

Participants were from the Longitudinal Aging Study in India—Diagnostic Assessment of Dementia (LASI‐DAD) which includes community‐residing older adults from eighteen states and union territories in India (n = 4096). LASI‐DAD is a subsample of adults aged 60 and older from the LASI; a nationally representative study on aging in India. The study visit was carried out in the participant's preferred language which was available across twelve languages: Hindi, Kannada, Malayalam, Gujarati, Tamil, Punjabi, Urdu, Bengali, Assamese, Odiya, Marathi, and Telugu. To facilitate accurate assessment due to the linguistic diversity inherent in India, all LASI‐DAD examiners were recruited from the states in which testing was conducted such that they were multilingual in the local languages. See Lee and colleagues for additional details regarding the LASI‐DAD sampling procedure. ¹¹

Analyses were conducted on the harmonized Wave 1 data (Version A.3), publicly available from the Gateway to Global Aging Website (https://g2aging.org/).

2.2. Measures

2.2.1. Language status

Multilingual status was assigned based on self‐report of any language(s) in addition to the first language, as well as evidence of multilingualism when the testing language differed from the first reported language. We determined the majority language of each participant's residential state based on the 2011 Indian Census (see Supplementary Table 1 for majority languages by state). Participants who spoke multiple languages and whose first language matched the majority language of their state were classified as multilingual‐majority speakers. In contrast, multilingual‐minority speakers reported knowing multiple languages, but their first language did not align with the majority language of their state. Moreover, around 25% of our multilingual participants reported speaking three or more languages. However, due to sample size limitations, bilingual vs multilingual comparisons were not investigated, and participants were classified into one of our multilingual groups. Finally, monolingual‐majority speakers were those who reported only one language, which was also the majority language in their residential state.

We only included monolingual‐majority participants (n = 2437) for the monolingual sample and excluded the subset of monolingual participants who did not report the majority language as their language since they were a relatively small group (n = 170; 4%) and we wanted to minimize potential differences in socioeconomic opportunities due to not knowing the majority language. Similarly, we only included multilingual‐minority participants (n = 396; 10%) who reported the majority language of the state as (one of) their additional language(s) and excluded the subset of multilingual‐minority participants who did not report the majority language of the state as (one of) their additional language(s) (n = 64; 2%). The total analytic sample for this study included 3854 participants.

2.2.2. SES

Years of education and parental education were self‐reported in terms of completed years of formal schooling. Caste was self‐reported and grouped into Scheduled Caste/Scheduled Tribe, Other Backward Class, and No/Other Caste, applying the categorization from the Indian Census. Household consumption was used as a proxy for economic well‐being, offering a more stable reflection of standard of living than income which can fluctuate considerably. ¹² Per capita household consumption, including spending on food (over the past week), non‐food items (over the past 30 days), outpatient health expenses (over the past 30 days), and inpatient health care costs (over the past year), was categorized into quartiles for analysis.

A summary indicator of SES was derived using an ordered confirmatory factor analysis on education, parental education, caste, and consumption. The socioeconomic factor model demonstrated good fit (CFI = 0.98, RMSEA = 0.07, SRMR = 0.04). Using the factor score, participants were classified into tertiles representing Low‐, Medium‐, and High‐SES. This categorization preserves the underlying continuous factor information while providing clear comparisons across socioeconomic strata, with higher values reflecting greater education, parental education, consumption, and caste‐related social advantage.

2.2.3. Cognition

LASI‐DAD administers a comprehensive neuropsychological battery from the Health and Retirement Study's (HRS) Harmonized Cognitive Assessment Protocol (HCAP) adapted to be linguistically and culturally appropriate for India. ¹³ The battery evaluates executive functioning, language, memory, and visuospatial abilities. Cognitive domain factor scores were derived using a confirmatory factor analysis which demonstrated good fit and supports the psychometric distinction of the domains. ¹³ The assessments used within the LASI‐DAD battery are in Table 1. Previous work has supported the cognitive factor structure of the battery and demonstrated comparable measurement across the study languages. ¹⁴

TABLE 1.

LASI‐DAD cognitive test assessments by domain factor.

Domain	Assessment
Memory	10‐Word recall, immediate, delayed, and recognition
	3‐Word recall, immediate and delayed
	Brave Man, immediate and delayed recall
	Logical memory, immediate and delayed recall, and recognition
	Constructional praxis, delayed recall
Executive functioning	Raven progressive matrices
	Go/no‐go trial 1 and 2
	Serial 7s
	Backward day naming
	Symbol cancelation
	Digit span, backwards and forwards score
Language	Naming common objects
	Animal naming
	Writing or saying a sentence
	Reading or repeating a phrase
	Close your eyes
	Paper‐folding 3‐stage task
	Naming described objects
Visuospatial	Interlocking pentagons
	Constructional praxis, immediate

2.2.4. Covariates

Education, parental education, caste, and consumption were derived as described under SES. Participants’ age was derived using the LASI‐DAD interview month and year relative to the participants’ birth month and year. Sex/gender was self‐reported as male versus female. Rurality was classified as urban or rural given the participants’ residency at the time of the interview using the 2011 Indian Census categorization.

3. ANALYSES

Descriptive statistics were calculated within each socioeconomic stratum comparing language status. One‐way analyses of variance and chi‐square tests evaluated differences amongst groups in covariates. The associations of SES and language status with cognition were evaluated using two approaches: First, separate multiple‐group comparisons compared language status within each socioeconomic stratum per cognitive domain while adjusting for sex/gender, age, and rurality, and the relative strength of these associations was statistically compared across strata using group contrasts. Second, given that differences in SES can remain even between groups within the same socioeconomic strata, statistically significant associations were then evaluated in separate linear regression analyses stratified by SES examining the main effect of language status, adjusting for all individual socioeconomic covariates: education, parental education, caste, consumption—in addition to sex/gender, age, and rurality. Tukey pairwise post hoc comparisons evaluated differences in language status stratified by SES. All analyses were conducted in R 4.3.1.

4. RESULTS

On average, participants were 69.6 years old (SD = 7.5) with 3.9 years of education (SD = 4.7); 53% were female and 62% resided in rural areas. All descriptive comparisons were conducted within each SES stratum, comparing monolingual, minority multilingual, and majority multilingual participants (Table 2, see Supplementary Table 2 for sample demographics by language status). Within each SES level, we observed systematic demographic differences across language groups. In the low‐SES stratum, monolingual participants were more likely to be female and to reside in rural areas, and they had fewer years of education compared to multilingual participants. Multilingual‐minority participants showed somewhat higher levels of education than monolingual individuals (nonsignificant), whereas multilingual‐majority participants had the highest education levels within this stratum (significant relative to monolingual individuals). A similar pattern was observed in the middle‐SES group: multilingual‐majority participants reported the most years of education and were less often female compared to both monolingual and multilingual‐minority participants, with the latter again falling between the two groups in education. In the high‐SES stratum, multilingual‐majority participants had the highest levels of both personal and parental education, followed by multilingual‐minority participants and then monolingual participants. Monolingual participants in this group were also more often female and from rural areas relative to both multilingual groups.

TABLE 2.

Sample demographics by language status, and SES.

	Monolingual			Multi‐minor			Multi‐major
n = 3854	Low	Med	High	Low	Med	High	Low	Med	High
	n = 983	n = 817	n = 637	n = 109	n = 149	n = 138	n = 192	n = 338	n = 491
Age	70.0 (7.57)	70.4 (7.85)	68.8 (7.28)	70.7 (7.94)	68.9 (6.76)	67.2 (5.95)	68.0 (7.85)	69.7 (7.52)	69.1 (6.78)
Sex/Gender %Female	59%	59%	59%	54%	48%	50%	55%	39%	32%
Urbanicity %Urban	22%	32%	42%	28%	59%	77%	24%	39%	63%
Years of education	1.0 (1.95)	2.3 (3.29)	6.64 (4.31)	1.1 (1.99)	3.0 (3.46)	7.8 (4.72)	1.4 (2.32)	4.1 (4.04)	10.3 (4.38)
Parental education
Mother %No school	100%	100%	64%	100%	100%	78%	100%	100%	63%
Father %No school	100%	93%	17%	100%	95%	28%	100%	95%	22%
Caste
%Scheduled Caste/Tribe	53%	8%	9%	50%	10%	5%	57%	12%	8%
%Other backward class	47%	46%	40%	50%	45%	43%	43%	44%	31%
%No/Other Caste	0%	46%	51%	0%	45%	52%	0%	44%	61%

Note: Multi‐Major = first language was a majority language; Multi‐Minor = first language was a minority language. OBC = Other Backwards Class.

4.1. Multilingualism versus monolingualism

Multilingual‐majority participants of high‐ and medium‐SES outperformed their monolingual socioeconomic counterparts in all cognitive domains: High‐SES; executive function (B = 0.43, [0.34, 0.52]), language (B = 0.25, [0.16, 0.33]), memory (B = 0.42, [0.31, 0.53]), visuospatial abilities (B = 0.35, [0.26, 0.44]); Medium‐SES; executive function (B = 0.25, [0.15, 0.35]), language (B = 0.13, [0.03, 0.24]), memory (B = 0.15, [0.04, 0.27]), visuospatial abilities (B = 0.16, [0.05, 0.26]). Multilingual‐minority participants of high‐ (B = 0.23, [0.09, 0.37]) and medium‐SES (B = 0.23, [0.11, 0.35]) outperformed monolingual participants in the domain of executive functioning. Additionally, multilingual‐minority participants of high‐SES demonstrated better visuospatial abilities than monolingual participants (B = 0.16, [0.02, 0.3]). There was no difference between multilingual and monolingual participants across any cognitive domain among the low‐SES strata.

The strength of the association of multilingualism and cognition differed by SES status. The association of majority‐multilingualism was stronger among high‐ compared to medium‐SES participants in memory (B = 0.27 [0.11, 0.43]) and visuospatial abilities (B = 0.14 [0.01, 0.28]) but did not differ between high‐ and medium‐SES in language nor executive functioning (p’s > .05). The association of minority‐multilingualism did not differ among high‐SES and medium‐SES participants. Model estimates for majority‐ and minority‐multilingualism and model fit statistics per cognitive domain are provided in Table 3. Figure 1 depicts the association of majority‐ and minority‐multilingualism relative to monolingualism by SES.

TABLE 3.

Multiple‐group results showing language status association by SES strata in terms of estimates with 95% confidence intervals and the fit statistics, including Comparative Fit Index (CFI), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Square Residual (SRMR).

Executive functioning
	Multi‐major versus Mono		Multi‐minor versus Mono		Multi‐minor versus Multi‐major
	B	CI	B	CI	B	CI
Low‐SES	0.039	−0.064, 0.142	0.032	−0.096, 0.161	0.010	−0.155, 0.176
Medium‐SES	0.250***	0.148, 0.353	0.229***	0.105, 0.352	0.037	−0.123, 0.196
High‐SES	0.429***	0.340, 0.517	0.232***	0.093, 0.372	0.211***	0.076, 0.346
CFI	0.987		0.978		0.967
RMSEA	0.040		0.051		0.053
SRMR	0.016		0.020		0.022
Language
Low‐SES	0.079	−0.033, 0.192	0.087	−0.056, 0.229	−0.01	−0.179, 0.160
Medium‐SES	0.131**	0.027, 0.235	0.096	−0.030, 0.221	0.019	−0.122, 0.160
High‐SES	0.245***	0.160, 0.329	0.018	−0.118, 0.155	0.209***	0.093, 0.324
CFI	0.948		0.968		0.918
RMSEA	0.056		0.040		0.056
SRMR	0.020		0.016		0.021
Memory
Low‐SES	−0.051	−0.170, 0.068	−0.109	−0.256, 0.039	0.058	−0.135, 0.251
Medium‐SES	0.150**	0.035, 0.266	0.115	−0.025, 0.256	0.05	−0.135, 0.236
High‐SES	0.421***	0.311, 0.530	0.064	−0.106, 0.234	0.369***	0.199, 0.540
CFI	0.968		0.964		0.960
RMSEA	0.045		0.043		0.047
SRMR	0.017		0.018		0.020
Visuospatial
Low‐SES	0.035	−0.071, 0.141	−0.003	−0.135, 0.130	0.043	−0.126, 0.213
Medium‐SES	0.158**	0.054, 0.263	0.069	−0.057, 0.195	0.09	−0.070, 0.249
High‐SES	0.347***	0.255, 0.438	0.158**	0.015, 0.302	0.184**	0.045, 0.323
CFI	0.976		0.969		0.979
RMSEA	0.042		0.044		0.026
SRMR	0.016		0.018		0.017

Asterisks denote statistical significance: * p < .05, ** p < .01, *** p < .001.

FIGURE 1.

Multiple‐group analyses contrasting multilingual‐majority and multilingual‐minority groups relative to the monolingual group of equal SES (reference) adjusting for age, sex/gender, and rurality.

When additionally adjusting for individual socioeconomic variables, multilingual‐majority participants of high‐SES still outperformed their monolingual counterparts in all cognitive domains: executive functioning (B = 0.22, [0.12, 0.32]), language (B = 0.14, [0.04, 0.23]), memory, (B = 0.25, [0.12, 0.38]), and visuospatial abilities (B = 0.44, [0.35, 0.53]). Multilingual‐minority participants of high‐SES outperformed their monolingual counterparts in executive functioning (B = 0.20, [0.06, 0.35]), but no longer in visuospatial abilities (B = 0.15, [−0.01, 0.32]), although the latter approached statistical significance (p = 0.06). Among medium‐SES, both multilingual groups performed better than the monolingual group in executive functioning (multilingual‐majority: B = 0.18, [0.07, 0.28]; multilingual‐minority: B = 0.28, [0.15, 0.41]), but no other associations remained significant, despite estimates being positive in all cognitive domains for both multilingual‐majority and multilingual‐minority participants. Figure 2 depicts the additionally adjusted association of majority‐ and minority‐multilingualism relative to monolingualism among high and medium SES participants.

FIGURE 2.

Post hoc analyses stratified by SES contrasting multilingual‐majority and multilingual‐minority groups relative to the monolingual group of equal SES (reference) adjusting for caste, parental education, years of education, and consumption in addition to age, sex/gender, and rurality.

4.2. Majority‐ versus minority‐multilingualism

The multiple‐group model indicated that multilingual‐majority participants of high‐SES outperformed their multilingual‐minority counterparts in all cognitive domains: executive function (B = 0.21, [0.08, 0.35]), language (B = 0.21, [0.09, 0.32]), memory (B = 0.37, [0.20, 0.50]), visuospatial abilities (B = 0.18, [0.05, 0.32]). We observed no differences between majority‐ and minority‐multilingual groups at medium‐ and low‐SES levels in any of the cognitive domains. Figure 3 (top‐panel) depicts the association of majority‐multilingualism as compared to minority‐multilingualism.

FIGURE 3.

Multiple‐group analyses [top‐panel] contrasting the multilingual‐majority group relative to the multilingual‐minority group of equal SES (reference) adjusting for age, sex/gender, and rurality and post hoc analyses [bottom panel] stratified by SES contrasting the multilingual‐majority group relative to the multilingual‐minority group of equal SES (reference) adjusting for caste, parental education, years of education, and consumption in addition to age, sex/gender, and rurality.

Akin to the multilingual vs monolingual analyses, we subsequently additionally adjusted for individual socioeconomic variables to evaluate whether that may contribute to high SES multilingual‐majority speakers consistently outperforming their multilingual‐minority counterparts across cognitive domains. When additionally adjusting for caste, parental education, years of education, and consumption, the multilingual‐majority group still outperformed the multilingual‐minority group in memory (B = 0.21, [0.01, 0.41]), but no longer differed in executive functioning, language, and visuospatial abilities (all p’s > 0.05). Figure 3 (bottom‐panel) depicts the additionally adjusted association of majority‐multilingualism relative to minority‐multilingualism stratified by high and medium SES.

5. DISCUSSION

Among older adults in India, we found that while multilingualism was positively associated with cognition, this relationship depended on their SES level, such that multilingual people outperformed monolingual people only among Medium‐ and High‐SES strata but not among the Low‐SES stratum. Moreover, among multilingual people, speaking the societal majority or a minority language as their first language was not associated with differences in cognitive test performance, after fully accounting for socioeconomic differences. However, societal majority‐multilingualism when compared to monolingualism was more consistently associated with better cognition in Medium‐, and High‐SES strata, in contrast to multilingual adults whose first language was a minority language.

Previous studies have observed better cognitive performance among high‐ versus low‐SES older adults, predominantly within North American and European samples. ¹⁵ , ¹⁶ , ¹⁷ Emerging literature in low‐ and middle‐income countries observes similar detrimental effects of low‐SES on cognitive aging, e.g., in Brazil, ¹⁸ Malawi, ¹⁹ Indonesia, ²⁰ and Argentina. ²¹ Our study—using a comprehensive SES factor derived from multiple indicators—further demonstrates that increasing SES is associated with better cognitive functioning.

Moreover, our study highlights how multilingualism—and the contexts in which it arises—can be confounded with SES. Multilingual older adults who speak the societal majority language as their first language generally showed advantages across most sociodemographic markers, followed by those from multilingual‐minority contexts, while monolingual older adults were most disadvantaged. Some of these sociodemographic differences persisted even after categorizing groups by SES. However, after additionally adjusting for sociodemographic disparities within SES strata, cognitive differences between multilingual and monolingual groups, as well as among the multilingual groups themselves, were greatly diminished and/or no longer significant within the medium‐ and high‐SES strata. Thus, majority‐ and minority‐multilingual individuals did not differ in cognitive outcomes after accounting for SES, suggesting that multilingualism's cognitive benefits at higher SES levels are not contingent on majority/minority status. Instead, SES appears to exert a stronger role than multilingual context in amplifying cognitive benefits, underscoring the importance of adequately accounting for socioeconomic confounds in studies of multilingualism and aging to avoid erroneously attributing cognitive differences to multilingualism.

Other studies have also found that adjusting for SES reduces and/or eliminates multilingual cognitive benefits, ²² , ²³ but not all. ²⁴ Notably, a large‐scale study by Nichols and colleagues featuring more than 11,000 participants indicated no multilingual benefit within executive functioning after adjusting for SES, including when examining bilingualism as a protective factor in age‐related cognitive decline. ²⁵ In our study, executive functioning was robustly associated with multilingual benefits among medium‐ and high‐SES older adults, surviving additional sociodemographic adjustments in both multilingual‐minority and ‐majority analyses. Methodological differences may contribute to our conflicting findings—e.g., Nichols employing an online testing platform versus our comprehensive in‐person neuropsychological evaluation—yet both studies indicate the necessity of adjusting for sociodemographic differences to avoid overestimating multilingual cognitive benefits.

In our study, multilingualism was associated with better cognitive functioning, with benefits observed across multiple domains—including executive functioning, language, memory, and visuospatial domains—particularly among higher‐SES individuals. Since theoretical accounts of multilingual cognitive benefits most strongly hypothesize benefits within executive functioning, these broader effects may seem unexpected. However, prior work has also revealed cognitive benefits in other domains, e.g., including memory. ²⁶ , ²⁷ Moreover, better language domain performance may appear counterintuitive, given evidence of disadvantages for multilingual adults relative to monolingual individuals. Yet such disadvantages are typically modest and task‐specific (e.g., worse naming and fluency performance), and largely when compared to monolingual participants of higher educational and socioeconomic backgrounds (e.g., non‐immigrant English monolingual white participants compared to predominantly immigrant English‐Spanish bilingual Latino participants) with limited accounting for these differences in their analyses. As such, past findings of multilingual cognitive profiles might not have been generalizable, particularly to multilingual speakers from LMICs. Our findings emphasize the importance of carefully accounting for sociodemographic and socioeconomic factors and including understudied multilingual populations to enhance our understanding of multilingual cognitive profiles.

Nevertheless, in our study, cognitive benefits were limited to medium‐ and high‐SES strata only and never occurred among low‐SES multilingual people. One potential explanation is that there is a threshold to multilingual cognitive benefits which could not be reached in the face of the detrimental impact of low child‐/adulthood SES. While there is a large body of literature that has demonstrated the negative impact of reduced life course SES on late‐life cognitive health, ⁵ the potential benefit of multilingualism to cognition has been mixed. ¹ Some studies report that any multilingual cognitive advantage is likely small, ²⁸ and thus may not overcompensate for the large negative effect of reduced life course SES. Second, multilingual adults are heterogeneous, differing in various linguistic characteristics which may be patterned by SES. For instance, language use and proficiency may differ, where multilingualism among higher‐SES strata may extend to formal language abilities (i.e., reading and writing), whereas proficiency in lower‐SES strata may be limited to informal language abilities (i.e., speaking, understanding). Although our study cannot speak to these dynamic language aspects, 75% of our lower‐SES multilingual older adults reported no formal schooling and generally exhibited greater linguistic diversity, speaking a wider variety of languages, whereas higher‐SES multilingual individuals reported a more homogeneous set of languages. As such, future studies should characterize multilingualism in greater detail to determine whether differences in aspects of multilingualism across SES might further elucidate our findings. Lastly, it is possible that, despite accounting for various aspects of life course SES, unmeasured SES confounds (i.e., quality of education, markers of environmental SES) may still drive differences between multilingual and monolingual adults at increasing SES levels.

An important strength of our study is our nationally representative cohort including underrepresented populations and diverse linguistic backgrounds, including understudied languages. While the years of schooling seemed low in our sample (M = 3.9) when compared to expectations from high‐income country cohorts, this is not unexpected for India, where census data show around 50% of adults 65 and older have no formal schooling, demonstrating our sample's representativeness. Moreover, our design accounts for both socioeconomic differences and multilingual context. However, limitations of the present study should be considered. First, without detailed language background data, our categorical multilingual classification limits disentangling potential cognitive contributions of specific multilingual aspects, such as age of acquisition, proficiency, or number of languages. Second, our classification of majority‐ and minority‐multilingualism simplifies India's complex linguistic landscape. For example, a Hindi speaker may be a linguistic minority in one state but speak a nationally dominant language. Nonetheless, this binary categorization was necessary to retain analytic power and to examine how multilingual context may interact with SES. Indeed, our assumption that majority‐multilingualism often reflects formal education and minority‐multilingualism reflects immersion was supported by sociodemographic trends. Third, while 95% of minority‐multilinguals were tested in a language other than their first, this does not appear to influence our findings, as minority‐multilinguals did not perform worse than their SES monolingual or majority‐multilingual counterparts. Finally, our SES operationalization combines widely used indicators with contextually specific ones, such as caste, potentially reducing direct comparability with other cohorts outside India. However, despite variation in SES measurement, our approach is grounded in standard operationalizations and contextually relevant considerations. SES was derived from commonly used indicators, including participants’ years of formal education, parental education, and household consumption (covering food, non‐food, and healthcare expenditures), reflecting both childhood and adulthood socioeconomic conditions. While caste does not align strictly with material operationalizations of SES, it reflects broader control over social resources. Although abolished, caste remains a salient determinant of access to education, resources, and opportunities, and we therefore posit that its inclusion captures contextually relevant socioeconomic differences. Moreover, deriving an SES factor score through confirmatory factor analysis is statistically robust as it avoids assuming equal contribution of all indicators and weights indicators proportionally. As such, we believe this approach is statistically robust and strengthens our study by capturing the socioeconomic realities most relevant to cognitive outcomes in India.

6. CONCLUSION

In India's multilingual and socioeconomically diverse context, higher SES was associated with better late‐life cognitive functioning in both mono‐ and multilingual older adults. Multilingualism was associated with better performance only at high SES levels, not at low SES levels. Majority‐multilinguals initially outperformed minority‐multilinguals at high SES, but these differences diminished after adjusting for SES, highlighting how multilingual context intersects with sociodemographic factors. Our findings suggest that SES is a stronger determinant of cognitive functioning than multilingualism, and that the cognitive benefits of multilingualism may depend on specific social and educational conditions.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICAL APPROVAL

Data collection was conducted in accordance with the Helsinki Declaration and reviewed and approved by the University of Southern California's Institutional Review Board, the Indian Council of Medical Research, and all collaborating institutes in India.

Partners in India include: All India Institute of Medical Sciences, New Delhi; All India Institute of Medical Sciences, Bhubaneswar; Dr SN Medical College, Jodhpur; Government Medical College, Thiruvananthapuram; Government Medical College, Chandigarh, Punjab; Grants Medical College & JJ Hospital, Mumbai; Guwahati Medical College, Guwahati; Institute of Medical Sciences, BHU, Varanasi; Indira Gandhi Institute of Medical Sciences, Patna, Bihar; Institute of Medical Sciences, BHU, Varanasi; Madras Medical College, Chennai; Medical College, Kolkata; National Institute of Mental Health and Neurosciences, Bengaluru; Nizam's Institute of Medical Sciences, Hyderabad; Sher‐e‐Kashmir Institute of Medical Sciences, Srinagar; and St. John's Medical College, Bengaluru.

CONSENT STATEMENT

Informed consent was obtained from all participants.

Supporting information

Supporting Information

Supporting Information

Supporting Information

DATA AVAILABILITY STATEMENT

Analyses were conducted on the harmonized Wave 1 data (Version A.3), publicly available from the Gateway to Global Aging Website (https://g2aging.org/).