Abstract
Background:
Better English proficiency and higher frequency of using English among non-native speakers are associated with lower dementia risk.
Objective:
We investigated if Mexican American older adults who use English and Spanish to a more similar degree demonstrate better cognitive function than those who use one language more than the other.
Methods:
We used data from waves one (1992/93) to eight (2012/13) of the Hispanic Established Population for the Epidemiological Study of the Elderly. At baseline, participants were asked what language they usually use across communicative contexts. We based dual language on participants’ use of Spanish and English within and across contexts. We categorized participants as low (n = 1,145), medium (n = 717), and high (n = 702) dual-language users. Linear mixed models were used to estimate the association between dual-language use, baseline Mini-Mental State Examination (MMSE) scores, and change in MMSE.
Results:
Participants in the medium and high dual-language use categories scored 1.91 points and 3.03 points higher at wave one compared to the low dual-language use category. Adjusting for education reduced the association between dual-language use and baseline MMSE (medium B = 0.99 SE = 0.19 p < 0.01; high B = 1.41 SE = 0.21 p < 0.01). The association between dual-language use and decline in the MMSE was not statistically significant.
Conclusions:
Greater dual-language use was associated with higher MMSE scores but not change in MMSE scores among Mexican Americans aged 65 and older. Future work should characterize bilingualism with greater nuance and use more rigorous cognitive measures to identify the components of the bilingual experience that may benefit the cognitive functioning of older adult bilinguals.
INTRODUCTION
Bilingual speakers must switch to or select the appropriate language in each communicative context to communicate efficiently. The processing of words in one language can co-activate phonetically or semantically similar words in the other language [1, 2]. Bilingual speakers must, therefore, reduce interference between languages, thereby avoiding the use of the incorrect language in a given linguistic context (i.e., cognitive control) [1, 2]. Both language selection and switching are mediated by neural mechanisms that support general cognitive functions that regulate attention, memory, and cognitive control [3]. Prior psychological work has proposed that bilingual experience can compensate for some age-related symptoms of cognitive decline by enhancing these cognitive functions [4–6]. However, the clinical literature has reported mixed findings [7].
The evidence that bilingualism may benefit cognitive functioning has motivated epidemiological research on language use and cognitive functioning among older adults. Epidemiologic studies in the United States have primarily used data from cohort studies of Hispanic aging, as Hispanic persons are the largest group of bilingual speakers in the United States [8]. Most of these studies have investigated English proficiency or frequency of using English among older Hispanic adults who were native speakers of Spanish. These studies hypothesized that greater English proficiency and more frequent English use would be associated with better cognitive functioning because of the increased cognitive processing reinforced by dual language experience and mental flexibility that may be gained from using multiple languages. A cross-sectional study of Hispanic adults aged 45–74 born outside the US and whose self-reported first language was Spanish found that higher English proficiency and higher English use were associated with better performance on learning, recall, verbal fluency, and working memory [9]. Longitudinal studies have also found that Hispanic older adults with higher English proficiency and who use English more often have higher cognitive functioning than older adults who are monolingual Spanish speakers [10–13]. However, these longitudinal studies did not find that greater English proficiency was associated with less cognitive decline [10–13].
When studying the relationship between language use and late-life cognitive functioning, it is important to consider the interactional contexts of communication crucial to the bilingual experience. Three interactional contexts have been defined: 1) single-language, 2) dual-language, and 3) dense code-switching [14, 15]. In a single-language context, the speaker uses only one language per social context. In a dual-language context, a bilingual speaker uses two languages but typically between different conversation partners. In a dense code-switching context, a bilingual speaker will frequently switch between two languages within utterances. The three interactional contexts are hypothesized to affect cognitive control processes differently. The dual-language interactional context is considered the most cognitively taxing, with the greatest recruitment of control processes compared to the other contexts [16–18].
Our study extends prior research in epidemiology and population health by using self-reported information for the language used in different family and social settings to categorize older Mexican Americans according to increasing dual-language use instead of increasing English proficiency or frequency of English use. We use this information to describe patterns in participants’ use of English and Spanish in different contexts and investigate the relationship between dual-language use and performance on the Mini-Mental Status Exam (MMSE). Additionally, our use of a large, representative cohort with long follow-up adds to the evidence from clinical research that has collected detailed assessments of language abilities and cognitive functioning from smaller, and sometimes heterogenous, participant samples. We hypothesize that older Mexican American adults who use English and Spanish equally (high dual-language use) or use one language more than the other (medium dual-language use) will have higher MMSE scores than those who predominately use one language (low dual-language use). Bilingualism may also contribute to an increase in cognitive reserve and compensate for age-related changes in the brain [19, 20]. We expect these associations will be greatest for individuals who gain high levels of language proficiency, taken to reflect regular dual-language use over time. Thus, we hypothesize that high dual-language use will also be associated with less decline in MMSE scores.
METHODS
Data and sample selection
We used data from Waves 1 through 8 of the Hispanic Established Population for the Epidemiological Study of the Elderly (H-EPESE). The H-EPESE is a population-based cohort study of Mexican American older adults living in the southwestern United States [21]. Wave 1 was completed in 1993/94 and included 3,050 participants (response rate 83%) who were representative of the Mexican American population aged 65 and older living in Texas, Colorado, New Mexico, Nevada, and California [22]. Follow-up interviews were completed every two to four years. Wave 8 was completed in 2012/13. Survey data for Waves 1–8 are publicly available at the Inter-university Consortium for Political and Social Research [23]. All publicly available data is de-identified and thus meets exemption four for non-human subjects research.
To obtain the final analytic sample, we first excluded 177 participants who did not receive the MMSE and needed a proxy to complete the entire Wave 1 interview. We excluded these 177 participants because the H-EPESE does not include a proxy measure of cognitive function. The reasons for a proxy interview, as documented in the H-EPESE, are that the participant was physically ill, deaf, away indefinitely, had reduced cognitive capacity or made more than 13 errors on the MMSE. Participants with a low MMSE score could still complete a direct interview or have a proxy complete the parts of the interview they could not do. We included participants who received a partial proxy interview if they had responses for all our variables of interest. Next, we excluded 309 participants with missing information for one or more demographic or language variables at the Wave 1 interview. The final analytic sample included 2,564 participants. The participants contributed an average of four observations over an average follow-up of 8.4 years. Because the H-EPESE does not use a proxy measure of cognitive functioning, we removed follow-up observations in which a participant required a proxy interview and did not have an MMSE score. The sample size for each number of observation waves participants contributed to the analysis were 494 (one observation), 373 (two observations), 307 (three observations), 403 (four observations), 259 (five observations), 280 (six observations), 222 (seven observations), and 216 (eight observations).
Measures
Characterization of dual-language use
The baseline observation of the H-EPESE included a modified version of the Hazuda Acculturation Scale [24]. Participants were asked what language they usually use with 1) their spouse, 2) children, 3) during family gatherings, 4) most of their friends, and 5) most of their neighbors. Participants could respond to each question with only English, mostly English, Spanish and English equally, mostly Spanish, or only Spanish. The responses of participants who were not married or had no children were recorded as not applicable for questions on language use with a spouse or children.
We used the responses from the questions above to create a summary variable representing the balance of dual-language use. We assigned a score to each response option so that a higher value indicated a greater balance of speaking English and Spanish. We scored answers of only English or only Spanish as 1-point, mostly English or mostly Spanish as 2-points, and English and Spanish equally as 3-points. The maximum attainable score was 15 for those with spouses and children. The maximum attainable score for those without spouses or children was less than 15. For example, for a respondent with a spouse but no children, the maximum score would be 12. Therefore, we converted all scores to a proportion of the maximum attainable score relevant for that respondent. We grouped the 2,564 participants in the final sample into low (n = 1,145; 44.7%), medium (n = 717; 28.0%), and high (n = 702; 27.3%) dual-language use categories according to natural breaks in the distribution of the ratio score (Fig. 1). In a sensitivity analysis, the percentage of participants in the low (40.8%), medium (29.3%), and high (29.9%) dual-language use categories when we excluded participants with no spouse or living children were similar to the percentages in these categories in the full sample.
Distribution of the ratio for dual-language use score and maximum switching score among H-EPESE participants aged 65 and older.
Cognition
Participants’ cognitive function was measured at each wave using the MMSE [25]. The MMSE is a commonly used screening tool for dementia and cognitive impairment that assesses performance across the following domains: orientation, memory recall, object naming, attention, constructional praxis, and language. The MMSE has a total score of 0 to 30 points, with a higher score indicating better performance.
Covariates
Covariates from the baseline interview were age (centered on the sample mean age of 72), sex, years of completed education (0, 1–5, 6–8, ≥9), marital status (married, not married, widowed), nativity (US-born or foreign-born), and language of interview. Participants can choose to complete each interview in Spanish or English. We used the interview date for each observation wave to calculate the number of years since the baseline interview. We used the interview date instead of age at each observation as our measure of time to avoid inconsistencies in participants reporting their chronological age. Finally, we included a variable that indicated the number of times a participant was observed.
Statistical analysis
We used analysis of variance and χ2 tests to describe the demographic characteristics of the final sample according to dual-language use category. We also described the frequency and percentage of participants who reported using only or mostly English, Spanish and English equally, mostly Spanish, and only Spanish with family, friends, and neighbors according to dual-language use category. For these descriptive analyses, we combined the responses of only English and mostly English because of the low number of participants who responded only English.
We used linear mixed models with random intercepts and slopes to estimate the association between dual-language use category and cognitive function. This is a commonly used method in population-based research on cognitive decline among older adults because it gives valid estimates when the data are highly unbalanced from differences in the number and timing that participants are observed. All models include random effects for time (years since baseline) and intercept to allow for the estimated trajectories and baseline MMSE values to vary for each participant. All models also included an interaction term between time and dual-language use category to determine if there were differences in the trajectory of cognitive functioning by the balance of speaking English and Spanish. We first adjusted for age, sex, marital status, and nativity (model 1). We used additional regression models to adjust for the language of the interview (model 2) and educational attainment (model 3). We treated all covariates except for language of interview as time-invariant. We adjusted for educational attainment in a separate model from the other demographic characteristics to determine how much the association between dual-language use and MMSE scores changed after adjusting for educational attainment. The variance inflation factor value for each variable in model 3 were all below 2.5. We also tested for a statistical interaction between educational attainment, dual-language use category, and time to determine whether education modified the relationship between dual-language use and MMSE scores.
Finally, we conducted several sensitivity analyses. First, we repeated our analyses excluding 116 participants who had a baseline MMSE score of 17 points or lower. Second, we excluded 494 participants who were observed only at the baseline wave. Third, we repeated our analyses excluding participants who did not have a spouse or living children. Finally, we repeated our analysis treating language use as a continuous variable. All analyses were conducted using R version 4.3.1 [26].
RESULTS
Demographic characteristics according to dual-language use category
The average age of participants in the analytic sample was 73.1 years, 58.0% were female, and 56.8% were married. Fifteen percent of participants had no formal education, 45.2% completed 1–5 years, 22.2% completed 6–8 years, and 17.2% completed > 9 years of education. Most participants were born in the US and completed their interviews in Spanish. As shown in Table 1, participants in the low dual-language use category were significantly older, were more likely to be female, and completed fewer years of education than participants in the medium and high dual-language use categories. The average baseline MMSE score ranged from 23.4 points for participants in the low dual-language use category to 26.8 points for participants in the high dual-language use category.
Demographic characteristics of 2,564 H-EPESE participants aged 65 and older according to low, medium, and high dual-language use categories.
H-EPESE, Hispanic Established Population for the Epidemiological Study of the Elderly; SD, standard deviation; MMSE, Mini-Mental State Exam. Sample sizes: a1,145; b717; c702.
Dual-language use across family and social settings by dual-language use category
Longitudinal association between dual-language use at baseline and cognitive function among H-EPESE participants aged 65 and older
Estimates from linear mixed effects regression models with random intercept and slope. H-EPESE, Hispanic Established Population for the Epidemiological Study of the Elderly; B, beta estimate); SE, standard error; ref, reference category. *p < 0.05; **p < 0.01.
Table 2 describes the use of English and Spanish with family, friends, and neighbors according to dual-language use category. Approximately 98% of participants in the low dual-language use category reported speaking only Spanish in all settings.
Language spoken by H-EPESE participants aged 65 and older with family, friends, and neighbors according to low, medium, and high dual-language use category
The responses of only English and mostly English were combined because of the low number of participants who responded only English. H-EPESE (Hispanic Established Population for the Epidemiological Study of the Elderly). aLow (n = 600); medium (n = 425); high (n = 427). bLow (n = 1069); medium (n = 681); high (n = 659). cLow (n = 1145); medium (n = 717); high (n = 702) *Percent speak only or most English in high versus medium dual-language use category p = 0.04. **Percent speak only or most English in high versus medium dual-language use category p < 0.01.
Nearly half of participants in the medium dual-language use category reported speaking only Spanish with their spouse and 40.2% reported speaking mostly Spanish with their spouse. Most (52.0%) participants in medium dual-language use category reported speaking mostly Spanish with their children and 25.4% reported speaking only Spanish with their children. Additionally, between 50% and 60% of participants in the medium dual-language use category reported speaking mostly Spanish at family gatherings, with friends, and neighbors whereas 13.7%, 27.2%, and 27.3% of participants reported speaking only Spanish in these settings, respectively.
Approximately 56% of participants in the high dual-language use category reported speaking Spanish and English equally with a spouse, and 62.5% spoke Spanish and English equally with their children. Over 70% of participants in the high dual-language use category spoke Spanish and English equally during family gatherings and with friends. The percentage of participants in the high dual-language use category who spoke only or mostly English with a spouse and their children was significantly higher than in the medium dual-language use category. Conversely, the percentage of participants in the medium dual-language use category who spoke only or mostly English with friends and neighbors was significantly higher than in the high dual-language use category.
Dual-language use and cognitive function
As shown in Table 3, participants in the medium and high dual-language categories scored 1.91 points and 3.03 points higher on the MMSE at baseline compared to the low dual-language category (model 1). These findings did not meaningfully change after adjusting for nativity (model 2). When adjusting for educational attainment (model 3), participants in the medium dual-language use category scored 0.99 points higher, and participants in the high dual-language category scored 1.41 points higher than participants in the low dual-language category. This represents a 48.5% and 53.8% decrease in the magnitude of the association for the medium and high dual-language categories, respectively. The differences in change in cognitive function according to dual-language use category were not statistically significant in any model (Table 3). We also did not detect a statistically significant three-way interaction between dual-language use category, time, and educational attainment.
Sensitivity analyses
Supplementary Tables 1–4 present the results from the sensitivity analyses. The baseline differences in MMSE score according to dual-language use category were less but remained statistically significant when excluding participants with an MMSE score of < 18 points (Supplementary Table 1), were not observed at any follow-up waves (Supplementary Table 2), and who were not married or had no living children (Supplementary Table 3). Consistent with the main analysis, we did not detect any statistically significant differences in the change in MMSE score according to dual-language use category when we only included participants with a baseline MMSE score of 18 points or higher (Supplementary Table 1), had at least one follow-up interview (Supplementary Table 2), and who had a spouse and living children (Supplementary Table 3). Finally, a higher total score in dual-language use was associated with higher baseline MMSE scores but no statistically significant differences in the change in MMSE score (Supplementary Table 4).
DISCUSSION
Using data from a representative cohort of Mexican Americans aged 65 and older, we found that participants who were high dual-language users had higher MMSE scores, on average, at baseline than participants who were low dual-language users. However, we did not detect statistically significant differences in the change in MMSE scores according to low, medium, and high dual-language use. Our findings are consistent with other longitudinal analyses of older Hispanic adults that have found higher English proficiency and greater use of English to be associated with higher initial levels of cognitive functioning but not with less cognitive decline [10–13].
Our description of the use of English and Spanish among participants in the medium and high dual-language use categories in different family and social settings revealed trends that may reflect generational differences in the use of English and Spanish. The percentage of participants in the medium and high dual-language use categories who reported using only Spanish was highest in settings where the communication partner is likely to be a similar age, such as with a spouse. Conversely, settings more likely to involve multiple generations, such as family gatherings, had the lowest percentage of participants who reported using only Spanish and the highest for using English and Spanish equally, particularly for participants in the high dual-language use category. The percentage of participants who reported using only or mostly English was highest for the language spoken with children. A large majority of US Hispanics say it is important for future generations to be able to speak Spanish [27], but the increasing use of English among younger generations has led to a decline in the percentage of Hispanic persons who are bilingual [28]. Additionally, there is a sharp decline among recent generations in the percentage of Hispanic persons who say their parents often encouraged them to speak Spanish [28]. This trend has led to an increase in English proficiency among younger generations of US-born Hispanic adults that has coincided with a decline in Spanish proficiency [27].
Our findings relate baseline performance on the MMSE in older Mexican Americans to the way they use and switch between Spanish and English in family and social settings. The higher baseline MMSE scores for medium and high dual-language users suggest that greater dual-language use may yield cognitive benefits for older adults [6]. Additionally, bilingual speakers’ ability to use multiple languages can contribute to more frequent social interactions than monolinguals [29], which is associated with higher cognitive functioning in old age [30]. The similar rate of change on the MMSE suggests that age-related changes to the brain may have a similar effect on the cognitive functioning of older Mexican American adults who are low, medium, and high dual-language users. Our findings are consistent with a passive model of reserve, which may explain findings that older adults with more education have higher levels of cognitive functioning but similar rates of cognitive decline when compared to older adults with less education [31, 32].
Most participants we characterized as high dual-language use reported using Spanish and English equally across family and social settings, and over 70% reported equal use with friends and at family gatherings. The interactional context (i.e., single-language, dual-language, and dense code-switching) can influence bilingual speakers’ cognitive functioning [16, 33]. A greater degree of dual-language use increases demands on cognitive control as individuals have opportunities to switch between languages, typically with different communication partners, which in turn requires the person to monitor the environment for opportunities to use Spanish while inhibiting interference from the nontarget language. As such, our finding that bilinguals with a greater degree of dual-language use had higher baseline MMSE scores than those with a lesser degree of dual-language use suggests that the frequency of using both languages is a critical aspect of the bilingual experience.
We found that the participants we characterized in the high dual-language category were also more likely to be born in the US and to have completed nine or more years of education than participants in the medium and low dual-language groups. Nativity status and educational attainment are important demographic characteristics associated with cognitive impairment among Mexican American older adults [34, 35]. The differences in baseline MMSE scores according to dual-language use category were unchanged after adjusting for nativity. The association between dual-language use category and baseline MMSE scores remained statistically significant after adjusting for education, but the magnitude of the associations was reduced by more than 60%. This finding indicates that years of education, another life experience related to cognitive reserve, explained a large percentage of the association between high dual-language use and better performance on the MMSE.
When interpreting our results, it is necessary to consider the relationships between nativity, educational attainment, and bilingualism. Over 80% of participants in the high dual-language use category were born in the US. The H-EPESE does not ask participants if they attended school in the US or Mexico, but most US-born H-EPESE participants likely received their education in the US. The number of dual-language education programs is increasing [36]. However, Hispanic children who are English language learners are at a disadvantage in US schools because of the English-only curriculum and the discouragement of speaking Spanish in the classroom [37]. Balanced bilingualism and greater English proficiency are associated with higher educational attainment and better academic performance among Hispanic young adults in the US [38–40]. Future research using longitudinal cohort studies with measures of language use and cognitive functioning over the life course is needed to determine the contributions of bilingualism, language use, and educational attainment on the cognitive functioning of older Hispanic adults.
Strengths and limitations
The current study has several strengths and weaknesses. A strength of our study is that we used data from a large, representative cohort of Mexican Americans aged 65 and older who have been regularly interviewed for up to 20 years. However, a major limitation of our study is that the MMSE is the only measure of cognitive functioning used in all observation waves of the H-EPESE. The MMSE is a commonly used screening tool for dementia and cognitive impairment. However, it has several limitations that affect its ability to assess a person’s overall level of cognitive functioning, especially in older adult populations with low formal educational attainment or who are non-English speakers [41, 42]. The MMSE also has well-documented floor and ceiling effects that make it difficult to detect changes in test performance among older adults with low and high cognitive functioning, respectively [43, 44]. A second limitation of our study is that measures of language use were only collected during the 1993–94 observation wave. Acculturation measures that focus heavily on language have demonstrated change over time, particularly when querying language use, proficiency and/or dominance. With respect to language dominance, well-documented shifts have been observed [45], particularly in immigrant populations [46, 47]. Furthermore, individuals who may be experiencing changes in cognitive status due to an underlying neurodegenerative process may demonstrate regression to the first-learned language and demonstrate susceptibility in their second learned language (though different patterns have been reported in the literature) [48–50]. As such, for those likely to experience cognitive decline, second learned language loss may occur and manifest as diminished language use of the second learned language. We also could not investigate the duration or changes in the use of English and Spanish over the life course because participants were not asked to recall their use of English and Spanish during childhood, young adulthood, and middle age.
While we adjusted for socio-demographic characteristics, we did not include other variables that may confound the association between dual-language use and cognitive functioning. The H-EPESE does not include information on early-life experiences important to acquiring a second language that may also be associated with late-life cognitive functioning [51]. Also, while it is a relative strength that the H-EPESE contains information about dual-language use, it does not ask about other nuances of the bilingual experience that have been associated with bilingual effects. More specifically, the age of second language acquisition and the number/types of languages should be considered in future studies to determine the aspects of the bilingual experience related to cognition.
Conclusions
We found that greater dual-language use was associated with higher MMSE scores at baseline but not less decline in the MMSE among older Mexican American adults. Years of education explained a large percentage of the differences in baseline MMSE scores according to dual-language use. Future work should characterize bilingualism with greater nuance by collecting information regarding the age of acquisition, language proficiency, language dominance, and aspects of language use over the life course. Our findings must also be replicated using cohort studies that include more rigorous cognitive measures. Such research can contribute to identifying the components of the bilingual experience associated with higher late-life cognitive functioning and investigate if these factors are associated with cognitive decline.
AUTHOR CONTRIBUTIONS
Brian Downer (Conceptualization; Formal analysis; Funding acquisition; Investigation; Methodology; Writing – original draft); Sadaf Milani (Conceptualization; Funding acquisition; Methodology; Writing – review & editing); Stephanie Grasso (Conceptualization; Funding acquisition; Methodology; Writing – review & editing); Fernando Llanos Lucas (Conceptualization; Funding acquisition; Methodology; Writing – review & editing); Neil Mehta (Conceptualization; Funding acquisition; Methodology; Supervision; Writing – review & editing).
Footnotes
ACKNOWLEDGMENTS
The authors have no acknowledgments to report.
FUNDING
This work was supported by the National Institutes of Health (RF1 AG068988; R01 AG010939; P30 AG059301; P30 AG024832; P30 AG066614, R01 AG080470).
CONFLICT OF INTEREST
Brian Downer is an Editorial Board Member of this journal but was not involved in the peer-review process of this article nor had access to any information regarding its peer-review.
No other authors have other conflicts of interest to report.
