Cognitive trajectories among English‐ and Spanish‐test‐takers in the NACC

Abstract

BACKGROUND

Cognitive assessments were traditionally developed using English‐speaking populations, creating a potential disadvantage and misrepresentation for non‐English speakers. We aimed to determine whether English‐ and Spanish‐test‐takers have similar or different cognitive trajectories.

METHODS

Participants included 931 Hispanic/Latino adults from the National Alzheimer's Coordinating Center. Using mixed‐effects regression analyses, we examined baseline differences and longitudinal changes in memory, attention/working memory, executive functioning, and language between Spanish‐ and English‐test‐takers. Models controlled for age at baseline, education, sex, Hispanic/Latino heritage, and cognitive status.

RESULTS

English‐test‐takers performed significantly better than Spanish‐test‐takers across all domains at baseline. No differences in cognitive trajectories were detected, except for attention/working memory, in which Spanish‐test‐takers declined at a slower rate than English‐test‐takers.

DISCUSSION

Despite baseline differences, both groups exhibited largely similar cognitive aging trajectories. These findings suggest that cross‐sectional differences may reflect measurement bias rather than differences in cognitive aging and an underestimation of cognitive abilities among Spanish speakers.

Highlights

• Baseline disparities in cognition were observed across all domains, except for language, with Spanish‐test‐takers scoring significantly lower than English‐test‐takers.

• Despite baseline differences, rates of cognitive decline were largely similar across language groups, suggesting potential measurement bias rather than differences in cognitive aging.

• Spanish‐test‐takers showed greater maintenance in attention over time, pointing to possible benefits from repeated testing or cultural factors that warrant further investigation.

1. INTRODUCTION

Hispanic/Latino adults in the United States (U.S.) have a higher risk for neurodegenerative diseases, such as Alzheimer's disease (AD), compared to non‐Hispanic/Latino White adults. ¹ This population is projected to experience the largest increase in AD cases over the next four decades, ² yet remains underrepresented in dementia research. Several culturally relevant factors, such as primary language and educational background, may influence cognitive aging and disease risk in this group. ³ , ⁴ Language of testing has been identified as a key predictor of cognitive performance among Hispanic/Latino adults. ⁵ , ⁶ , ⁷ , ⁸ Cross‐sectional studies consistently show that English‐test‐takers outperform Spanish‐test‐takers across several cognitive domains. ⁹ , ¹⁰ However, longitudinal studies reveal fewer and less consistent differences in rates of decline. ¹¹ , ¹² This distinction is critical: though one‐time assessments may be biased by sociocultural factors, ¹³ longitudinal decline is a more specific indicator of cognitive aging and, possibly, neurodegeneration. ¹⁴

Differences in cognitive testing performance by test language may be partially explained by differences in education. Spanish‐test‐takers are more likely to have lower educational attainment and to be educated outside of the United States. ⁴ , ⁹ , ¹⁵ Educational background has a strong influence on cognitive test performance, even when accounting for variables such as age, acculturation level, and years in the U.S. Education may also improve familiarity with testing contexts which can influence scores. ¹⁶ , ¹⁷ However, the impact of education on cognitive trajectories over time is less well understood. For instance, in a 9‐year longitudinal study of U.S.‐ and Mexico‐based participants, education predicted baseline performance on verbal memory and a global cognitive screener (i.e., the Mini‐Mental State Examination [MMSE]), but only influenced change over time on the global screener, not memory tests. ¹⁸ Interestingly, Spanish‐test‐takers may benefit more from repeated exposure to testing, perhaps due to the increased familiarity. ¹⁶ , ¹⁷ , ¹⁹ , ²⁰

Bias in the assessments themselves also contributes to disparities. Cognitive tests may reflect cultural norms and values embedded in English‐speaking, U.S.‐based populations. As a result, individuals from different linguistic or cultural backgrounds may be mischaracterized as cognitively impaired. ²¹ , ²² Even when tests are translated or co‐developed in Spanish, versions can differ in difficulty levels, ²³ raising concerns about validity and equity. Thus, the development of linguistically and culturally harmonized tools, such as the Spanish and English Neuropsychological Assessment Scales (SENAS), ²⁴ , ²⁵ is critical to reducing measurement bias and ensuring equitable cognitive assessment across diverse populations.

The present study leverages the National Alzheimer's Coordinating Center (NACC) dataset which includes a diverse sample of Hispanic/Latino adults assessed using a comprehensive neuropsychological battery to examine whether cognitive trajectories differ between Spanish‐ and English‐test‐takers. In line with the findings of Mungas et al. (2018) ¹⁸ and Early et al. (2013), ¹⁹ we hypothesized that English‐test‐takers would perform better at baseline, but that rates of cognitive decline would not differ, suggesting that observed performance gaps are due to sociocultural factors and measurement bias rather than underlying pathology. We also explored whether adjusting for education and other sociodemographic factors attenuates group differences.

2. METHODS

2.1. Data source

Data for this study were drawn from the NACC, collected between 2005 and 2022 (extracted June 2022). ²⁶ , ²⁷ The Uniform Data Set (UDS) includes standardized clinical and neuropsychological data collected across 40 Alzheimer's Disease Centers (ADCs) in the United States, each with unique recruitment strategies (e.g., clinician referrals, community outreach, self‐referral). Assessments were conducted approximately annually in participants’ preferred language by trained interviewers. Institutional review board approval and informed consent were obtained at each ADC site.

RESEARCH IN CONTEXT

1. Systematic review: We reviewed prior studies using PubMed and other online databases, focusing on cognitive aging, language of test administration, and neuropsychological assessment among Hispanic/Latino older adults. Relevant articles are cited.

2. Interpretation: In this large and diverse sample of Hispanic/Latino older adults from the National Alzheimer's Coordinating Center, Spanish‐test‐takers scored lower than English‐test‐takers across several cognitive domains at baseline. However, rates of cognitive decline were largely similar across groups, except for attention, in which Spanish‐test‐takers declined at a slower rate. These findings suggest that language‐based cross‐sectional cognitive performance differences may reflect differences in sociocultural factors and/or measurement biases, rather than differential risk for cognitive decline.

3. Future directions: Further research is needed to examine how language of test administration, education, and acculturative factors (e.g., bilingualism, migration history) interact to influence cognitive aging trajectories among Hispanic/Latino adults. Culturally and linguistically valid neuropsychological tools are necessary to reduce diagnostic bias in diverse populations. Extending work by Mungas and colleagues on the Spanish and English Neuropsychological Assessment Scales (SENAS), offers a promising pathway for improving diagnostic accuracy and equity in cognitive aging research.

2.2. The UDS Neuropsychological Battery

The UDS Neuropsychological Battery assesses attention/working memory, processing speed, executive functioning, episodic memory, and language. ²⁷ The UDS neuropsychological battery consists of tests intended to characterize healthy cognitive aging, mild cognitive impairment, and AD. The battery is administered in a standardized manner at all ADCs. Spanish versions were developed through consensus by the Spanish Translation and Adaptation Work Group (STWAG) and became available online in April 2007. ⁵ The following tests were included in this study:

1. Craft Story 21: Immediate and Delayed Recall

   The Craft Story 21 is a test of episodic memory that assesses the ability to recall a short story twice. First, immediately after the story is read (Immediate recall) and then again 20 min later after an interference period (Delayed Recall).

2. Number Span: Forward and Backward

   Number Span is a test of working memory. Numbers for both Forward and Backward span tests are presented with sequences in ascending order of difficulty. Two trials are administered at each sequence length. For this study, we used the total number of correct trials.

3. Category Fluency: Animals and Vegetables

   Category Fluency is a measure of semantic fluency. The participant is asked to name various items of a given semantic category (e.g., animals, vegetables). Participants are given 60 s to generate as many distinct responses as they can, and the number of unique, correct responses is scored.

4. Trails Making Test A and B

   Trails Making Test parts A and B provide information on processing speed, mental flexibility, and executive functioning. In Trail A, the participant is asked to connect numbered circles in order as quickly as possible. Trail B includes letters and numbers, and the participants are asked to draw the lines in order while alternating between numbers and letters. For both tests, completion time is the primary outcome measure.

5. Multilingual Naming Test (MINT)

   The MINT is a picture naming test that was developed for use in several languages (e.g., English and Spanish) with roughly equivalent difficulty of items across languages. The test consists of 32 standardized pictures that participants are asked to name, one at a time. The total number of correctly named pictures is the primary outcome of interest.

2.2.1. Analytic Sample

Inclusion criteria were: (1) self‐identified Hispanic/Latino adults, (2) neuropsychological evaluation in either English or Spanish, (3) at least two visits of the neuropsychological evaluations were completed, (4) no diagnosis of dementia at baseline as determined by the NACC; ²⁶ and (5) age 65 or older at baseline. The final sample included 678 (Spanish‐test‐takers = 595, English‐test‐takers = 336) participants.

2.2.2. Composite Cognitive Scores

Composite scores were constructed for four cognitive domains: memory, attention/working memory, executive functioning, and language. These composites followed a factor structure previously validated in the NACC sample. ²⁸ Specifically, the memory composite was derived from Craft Story 21 Immediate and Delayed Recall (two scores). Attention/working memory was derived from the Number Span, Forward, and Backward subtests (two scores). Speed/Executive functioning is a composite score derived from Trails A and B (two scores). Language is a composite score derived from category fluency (animals and vegetables) and MINT total (three scores).

All tests were first standardized by subtracting each individual's score from the global mean score for each test across both language groups and dividing by the global standard deviation. Note, Trails A & B were reversed (by multiplying standardized scores by ‐1) for consistent interpretability. There was variability in the number and types of tests that were completed by each participant. Therefore, composite scores were subsequently generated for each participant by adding the tests that belong to a given cognitive domain and dividing by the number of tests included in a given domain for a given person.

2.2.3. Covariates

Sociodemographic variables included age in years, years of education, sex (male, female), cognitive status (e.g., cognitively normal or mild cognitive impairment [MCI]), and Hispanic/Latino heritage (e.g., Mexican, Chicano or Mexican American, Central American, Cuban, Dominican, Puerto Rican, South American, and other/not reported). These were recorded during the participant's first UDS visit in their preferred language.

2.3. Statistical analysis

Baseline differences in demographic characteristics by test language were evaluated using chi‐square tests and analyses of variance (ANOVAs). Cohen's d values were calculated for all continuous variables. Longitudinal changes in cognitive performance were modeled using linear mixed‐effects regression (LMER) ²⁹ models in R (v4.4.2) with random intercepts for participants. In linear mixed‐effects models, the fixed effects estimate the average associations between predictors (e.g., language, time, covariates) and cognitive outcomes across all participants. The random effects account for individual variability in baseline cognitive performance, allowing each participant to have their own intercept.

We modeled cognitive change as a function of chronological age rather than time since baseline, centering age at 65. This approach allows us to estimate group average trajectories of cognitive aging across the adult lifespan, rather than change relative to study entry. Because participants enrolled at different ages and contributed varying numbers of visits, the data structure is inherently unbalanced. Linear mixed‐effects models are well‐suited to handle this type of design, as they flexibly account for both between‐ and within‐subject variability. ³⁰ , ³¹ Random intercepts and slopes allow individuals to deviate from the average trajectory while still contributing to the estimation of group‐level trends. As a result, the reported estimates reflect model‐derived group averages over age, even when participants vary in the age at which they entered the study and the number of assessments they completed.

Each cognitive domain was modeled separately. Fixed effects included language of test administration, age, and their interaction (language × age), sex, heritage, cognitive status, and education, depending on the model.

Three models were estimated for each outcome. Model 1: Language, age, and their interaction; Model 2: Model 1 adding sex, cognitive status at baseline, and Hispanic/Latino heritage; and Model 3: Model 2 adding years of education.

The key term of interest was the language × age interaction, which tested whether cognitive trajectories differed between English‐ and Spanish‐test‐takers.

3. RESULTS

At visit 1, Spanish‐ and English‐test‐takers significantly differed on certain baseline characteristics (Table 1). On average, English‐test‐takers had more years of education than Spanish‐test‐takers (mean difference = 2.47 years, p < 0.001) and completed more follow‐up visits (mean difference = 0.6 visits, p = 0.01). Groups also differed significantly in the distribution of Hispanic/Latino heritage and cognitive status, such that English‐test‐takers were predominantly of Mexican, Chicano, or Mexican‐American heritage and were less likely to be diagnosed with MCI compared to Spanish‐test‐takers. English‐test‐takers also scored higher than Spanish‐test‐takers across all cognitive domains (ps < 0.001; Table 1). Cohen's d values were calculated for all continuous variables to assess the magnitude of these differences. While differences in number of visits were small (d < 0.2), the difference in years of education was medium to large (d = 0.59). The largest differences in cognitive performance were observed on the MoCA (d = 0.71), Number Span Forward (d = 0.71), Number Span Backward (d = 0.65), and the Attention/Working Memory composite (d = 0.76), indicating notably better performance among English‐test‐takers (Table S1). Moderate differences were also seen across memory, executive function, and language composites (ds = 0.50–0.66), highlighting the pervasive association between language of test administration and baseline cognitive scores (Table 1).

TABLE 1.

Demographic and composite cognitive characteristics by language of testing.

Variable	Level	English (n = 336)	Spanish (n = 595)	Cohen's d	p‐value
Age		73.4 (6.0)	73.5 (5.8)	−0.02	0.727
No. of visits		4.6 (3.5)	4.0 (3.1)	0.18	0.01
Years of education		14.4 (3.3)	12.0 (4.6)	0.56	<0.001
Body mass index		28.0 (4.9)	28.2 (4.8)	−0.04	0.560
Hispanic heritage	Central American	11 (3.5)	50 (8.6)		<0.001
	Cuban	17 (5.4)	152 (26.0)
	Dominican	2 (0.6)	34 (5.8)
	Mexican, Chicano, or Mexican‐American	175 (56.1)	127 (21.7)
	Other	23 (7.4)	13 (2.2)
	Puerto Rican	60 (19.2)	78 (13.4)
	South American	24 (7.7)	130 (22.3)
Cognitive status	MCI	88 (26.2)	241 (40.5)		<0.001
	Normal	248 (73.8)	354 (59.5)
Sex	Female	208 (61.9)	403 (67.7)		0.084
	Male	128 (38.1)	192 (32.3)
MoCA total		0.37 (0.78)	−0.25 (0.93)	0.14	<0.001
Memory		0.17 (0.91)	−0.28 (0.89)	−0.12	<0.001
Executive function		0.34 (0.69)	−0.21 (0.93)	0.16	<0.001
Language		0.25 (0.71)	−0.18 (0.73)	0.10	<0.001
Attention/Working memory		0.32 (0.90)	−0.31 (0.76)	0.01	<0.001

Note. Cohen's d is only reported for continuous and composite variables. Categorical comparisons use chi‐squared tests.

Abbreviation: MoCA, Montreal Cognitive Assessment.

In models 1 and 2 (models adjusted for sex, cognitive status at baseline, and Hispanic/Latino heritage, language of testing, and age), Spanish‐test‐takers had significantly lower baseline scores across all four cognitive domains (Table S2). However, baseline differences in the language domain were no longer significant after adjusting for years of education (Model 3; b = ‐0.124, SE = 0.104, p = 0.231). Other baseline differences in memory (b = ‐0.312, SE = 0.120, p = 0.009), attention/working memory (b = ‐0.563, SE = 0.107, p < 0.001), and executive functioning (b = ‐0.220, SE = 0.074, p = 0.003) were attenuated, but remained significant.

No significant differences were observed in the rate of cognitive change between groups across most domains (ps > 0.07), with the exception of attention/working memory. As shown in Table 2 and Figure 1, the rate of change in attention/working memory differed significantly by language group (language × age interaction: (b = ‐0.022, SE = 0.008). English‐test‐takers declined in attention/working memory over time (b = ‐0.017, SE = 0.006), while the rate of change among Spanish‐test‐takers was not significantly different from zero (b = 0.006, SE = 0.005), suggesting a maintenance of attention/working memory amongst Spanish‐test‐takers.

TABLE 2.

Time × Language interaction coefficients for three alternative models predicting cognitive trajectories.

	Model 1			Model 2			Model 3
Outcome	β	SE	𝑝‐Value	β	SE	𝑝‐Value	β	SE	𝑝‐Value
Memory	−0.001	0.010	0.960	−0.001	0.009	0.882	0.001	0.009	0.938
Attention/ working memory	0.022	0.009	0.015	0.021	0.009	0.015	0.023	0.008	0.006
Language	0.000	0.009	0.969	0.001	0.008	0.908	0.004	0.008	0.614
Executive functioning	−0.003	0.005	0.588	−0.004	0.005	0.437	−0.003	0.005	0.522

Model 1: Language, age, and language ×age interaction.

Model 2: Model 1 + sex, cognitive status at baseline and Hispanic/Latino heritage.

Model 3: Model 2 + years of education.

Abbreviation: SE, standard error.

FIGURE 1.

Cognitive trajectories for Spanish‐ and English‐test‐takers. p < 0.05*, p < 0.01**, p < 0.001***. Age was centered at 65 years old. Predicted cognitive trajectories for each domain are shown by language of testing, adjusted for sex, cognitive status at baseline, Hispanic/Latino heritage, and years of education. Annotations display estimated slope and standard errors (SE) for Spanish and English test‐takers, as well as the language of testing by age interaction term (Int.). WM, working memory.

Across all cognitive domains, the sequential addition of covariates improved model fit and explained variance but did not change the pattern of results. Marginal R² values increased across the four cognitive domains from Model 1 (range = 0.077 to 0.162) to Model 3 (range = 0.222 to 0.370). Akaike's Information Criterion (AIC) values decreased across models, indicating better model fit (Model 1 = 2061.9 to 6685.0; Model 2 = 1903.7 to 6298.8; Model 3 = 1855.5 to 6073.3).

A series of post hoc analyses were conducted to examine whether practice effects may explain the maintenance of attention/working memory performance observed in Spanish‐ but not English‐ test‐takers. Following the methodology of Early et al. (2013), we ran fully adjusted linear mixed‐effects models for each attention/working memory outcome, including an interaction term between visit number and language of testing. Age at visit was included as a time‐varying covariate to account for developmental change. Three models were specified: one for the attention/working memory composite score and two for its components, Number Span Forward and Number Span Backward. Results revealed a modest but significant interaction between visit number and language of testing for the Number Span Forward task, indicating greater gains across visits among Spanish‐test‐takers compared to English‐test‐takers (β = 0.055, p = 0.042). No significant main effect of visit or visit‐by‐language interaction was found for the attention/working memory composite or Number Span Backward, suggesting that practice effects were not consistently observed across all attention/working memory measures (see Table S3).

To further explore the possible presence of practice effects, we estimated attention/working memory trajectories during two distinct segments– the first two visits (when practice effects are expected to be the most pronounced ³² , ³³ ) and excluding visit 1 (i.e., visits 2+). For this analysis, we randomly sampled 300 English‐tested and 300 Spanish‐tested participants. Individual linear slopes of attention/working memory performance (composite score, Number Span Forward, and Number Span Backward) were computed over age within each segment. Separate regression lines were then estimated for each language group and segmented to visually assess change over time. We did not find evidence of practice effects in the first two visits across any measures (attention/working memory composite score, Number Span Forward, and Number Span Backward) as indicated by relatively flat, negative slopes (see Figures S1‐S3).

4. DISCUSSION

In a sample of over 900 Hispanic/Latino adults aged 65–93 years from the NACC dataset, we found that Spanish‐test‐takers performed lower across all cognitive domains at baseline compared to English‐test‐takers, with the exception of language, after accounting for education. However, rates of cognitive decline were broadly similar across groups with the exception of the attention/working memory domain, in which Spanish‐test‐takers showed greater maintenance. These findings are consistent with prior studies demonstrating cross‐sectional differences by testing language but limited evidence of differences in longitudinal trajectories. ¹⁸ , ¹⁹ Mungas et al. (2018) found that individuals tested in Spanish had lower baseline scores on the MMSE and on a verbal memory composite derived from the SENAS battery compared to those tested in English. However, both groups showed similar rates of cognitive change over time. Similarly, Early et al. (2013) reported that, across Hispanic/Latino, White, and Black/African American participants, language of testing was strongly associated with baseline cognitive scores but had little to no effect on longitudinal change. ¹⁹ Collectively, these findings highlight the importance of accounting for the language of test administration on the interpretation of cross‐sectional scores, especially when used in clinical decision‐making.

Our study builds on this literature in several important ways. First, the comprehensive cognitive battery we employed allowed us to model the cognitive trajectories of Spanish‐ and English‐test‐takers across multiple domains (i.e., memory, attention/working memory, language, and executive functioning). Second, earlier studies relied on more regionally specific cohorts (e.g., Davis Alzheimer's Disease Center and the Sacramento Area Latino Study of Aging [SALSA]), and our sample included Hispanic/Latino adults across the U.S., representing a wider range of heritage backgrounds. This is particularly relevant, as cognitive performance and AD risk vary across Hispanic/Latino heritage groups. ¹³ , ²¹ Our study also adopted a novel analytical strategy by modeling time as age rather than years since baseline, allowing for a more biologically relevant depiction of cognitive aging. This approach helps reduce bias introduced by differences in age at study entry, improving comparability across participants. ³⁰

Lower baseline cognitive scores among Spanish test‐takers may be attributed to a range of sociocultural and psychosocial factors, including educational attainment, health literacy, and acculturation, as well as possible measurement bias. ¹³ , ³⁴ Acculturation, particularly in the form of language use, has been consistently linked to cognitive performance in cross‐sectional studies. Higher acculturation—often operationalized as greater English language use—is frequently associated with better scores in episodic memory, working memory, processing speed, and executive functioning, although findings may vary depending on how acculturation is measured. ²¹ , ³⁵ Importantly, though acculturation appears to influence baseline cognitive performance, it is not consistently associated with the rate of decline. Hill et al. (2012) found that higher U.S. acculturation was related to better cognitive performance at baseline but not to slower cognitive deterioration over time, ³⁶ a pattern echoed in our findings. Regarding potential bias, studies have shown that Number/Digit Span tasks are more challenging for Spanish speakers than for English speakers due to the increased phonological and articulatory demands of Spanish number words. Specifically, the Spanish versions of Number Span Forward and Backward require recalling significantly more syllables than English versions, which poses a greater cognitive load and leads to systematically lower scores using traditional scoring methods, even among balanced bilinguals tested in both languages. ³⁷ , ³⁸

Our observation of greater maintenance of attention/working memory among Spanish‐test‐takers is somewhat unexpected, but it aligns with some emerging findings in literature related to Hispanic/Latino culture. For instance, Lamar et al. (2023) found that lower acculturation, which included language use, was associated with slower working memory decline in older adults with better cardiovascular health. ¹⁰ Additionally, some, but not all ³⁹ studies suggest that Hispanic/Latino adults who migrate to the U.S. in mid‐life may experience similar or slower cognitive aging relative to their U.S.‐born counterparts after accounting for key sociodemographic factors. ³⁶ , ⁴⁰ , ⁴¹ , ⁴² These findings support the “healthy immigrant effect” or “Hispanic health paradox,” which posits that individuals who migrate to the U.S. are, on average, healthier or more resilient due to the physical, emotional, and logistical demands of the migration process. ³⁶ , ⁴³ , ⁴⁴ Importantly, although language is a strong predictor of acculturation and is associated with age of migration, the migration experience is complex and heterogeneous. More studies with comprehensive data on migration‐related factors are needed to examine the role of this complex experience on the cognitive trajectories of Latino Spanish‐ and English‐test‐takers.

An alternative explanation for the attention/working memory‐related findings involves the role of education and test familiarity. Prior work suggests that individuals with lower educational attainment may demonstrate greater gains or maintenance on repeated cognitive assessments due to an increased familiarity with testing procedures across visits. ¹⁶ , ¹⁷ , ²⁰ In our sample, English‐test‐takers had significantly more years of education. Therefore, we might have expected that Spanish‐test‐takers would benefit more from the test‐retest effects, similar to the findings of Early et al. (2013) that Spanish‐test‐takers with lower educational background had greater improvements in episodic memory over time relative to English‐test‐takers. Despite this, post‐hoc testing in the present study largely did not support that Spanish‐test‐takers had major practice effects from Visit 1 to 2, when the largest practice effects are expected. ³² , ³³

This study has several strengths, including a large and diverse sample of Hispanic/Latino adults and a balanced distribution of English‐ and Spanish‐test‐takers. Our use of validated cognitive composites and age‐based modeling strengthens the reliability of our findings. However, limitations should also be noted. We were unable to account for factors such as age at migration, ³⁶ , ⁴⁵ socioeconomic stability, or healthcare access which may drive language use and access to resources that promote cognitive health. ⁴⁶ , ⁴⁷ Attrition bias is also a concern, as English‐test‐takers completed more follow‐up visits on average than Spanish‐test‐takes, which may have disproportionally affected estimates in the oldest group. Nonetheless, participants were followed, on average, for more than 4 years after enrollment, providing sufficient longitudinal data for modeling cognitive change and supporting reliable estimates of within‐person change. Additionally, we could not assess the potential confounding effects of bilingualism. Prior research has shown that higher degrees of Spanish‐English bilingualism are associated with better neuropsychological performance ³⁶ , ⁴⁵ , ⁴⁸ and that greater English proficiency among those born outside of the U.S. is linked to better cognitive scores and slower decline. ¹⁰ , ⁴⁹ , ⁵⁰

5. CONCLUSION

In this large, diverse sample of Hispanic/Latino older adults from the NACC, we observed significant baseline differences in cognitive performance between English‐ and Spanish‐test‐takers, but broadly similar rates of cognitive decline over time. These findings underscore the importance of interpreting cross‐sectional cognitive test scores with caution, particularly when the language of test administration may introduce sociocultural or educational confounds. Our results also suggest that attention/working memory may follow a different trajectory across language groups, warranting further investigation. Better understanding these differences can inform the development and interpretation of cognitive assessments that are equitable and accurate for individuals from diverse linguistic and cultural backgrounds.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to report and nothing to disclose. Author disclosures are available in the supporting information

CONSENT STATEMENT

NACC was approved by the institutional review boards at all participating institutions, and participants gave written informed consent.

Supporting information

Supporting information

Supporting information

Araujo‐Menendez CEE, Carpio RA, Tarraf W, et al. Cognitive trajectories among English‐ and Spanish‐test‐takers in the NACC. Alzheimer's Dement. 2025;21:e70721. 10.1002/alz.70721