Normative data for verbal fluency and naming tests in Spanish-speaking adults in the United States

2.1Participants

The initial sample consisted of 253 healthy individuals who were recruited from the U.S. (California, Connecticut, Florida, Indiana, New Jersey, Oregon, Virginia, and Wisconsin). To participate in the study, individuals met the following eligibility criteria: a) be between 18–80 years of age, b) have been living in the U.S. for at least 1 year (12 continuous months), c) self-identify Spanish as their “dominant language”, d) have at least 1 year of formal education, e) be able to read and write in Spanish, f) obtain a score of ≥23 on the Mini-Mental State Examination (MMSE) (Folstein et al., 1975; Villaseñor-Cabrera et al., 2010), g) obtain a score of ≤10 on the Patient Health Questionnaire–9 (PHQ-9) (Kroenke et al., 2001), and h) obtain a score of ≤10 on the Generalized Anxiety Disorder-7 (GAD-7) (Spitzer et al., 2006).

Individuals were ineligible if any of the following exclusion criteria were present: a) history of a neurodevelopmental disorder, b) history of a learning disorder, c) past or present neurological condition, d) past or present chronic medical condition that might affect cognition (e.g., metabolic syndrome, chronic heart failure, sleep apnea, complications associated with SARS-CoV-2), e) past or present use of psychotropic medications that might affect cognition, f) past or present history of substance abuse or dependence, or g) past or present history of a psychiatric disorder.

Eight participants were excluded from data analysis due to incomplete sociodemographic information, resulting in a final sample of 245 participants. The majority of the sample was comprised of women (60.8%, n = 149), the mean age was 41.1 (SD = 14.9), and the mean years of school completed was 15.1 (SD = 4.2). Sociodemographic characteristics are reported by Rivera et al. (2024).

Each of the institutions obtained their own institutional review board protocol approval to cover the ethical conduct of the study at their own site. All participants signed an informed consent form and were offered $25 for their participation in the study.

2.2Measures

For the present study, phonological and semantic verbal fluency tests were administered in Spanish. Participants were asked to provide as many words as possible within 60 seconds in each of the six phonemes (/f/, /a/, /s/, /p/, /m/, and /r/) and two categories (animals and fruits). Olabarrieta-Landa et al.’s (2017) administration and scoring guidelines were followed. For this study, the total score was used.

Participants also completed the standard (60-item) Spanish version of Boston Naming Test (BNT) that differs slightly from the original BNT (Kaplan et al., 1983). For example, item 14 is a carrot instead of a mushroom, item 19 is a muffin instead of a pretzel, item 28 is a crown instead of a wreath, item 40 is a pacifier instead of a knocker, item 48 is a needle instead of a noose, item 51 is a keyhole instead of a latch, and item 57 is a watering can instead of a trellis. The BNT total score is the sum of correct spontaneous answers plus correct answers followed by a semantic clue (Kaplan et al., 2005). The administration of the BNT was conservative, and thus synonyms were not considered correct responses.

It should be noted that although a large portion of the sample was bilingual, participants were required to provide responses only in their dominant language of Spanish. Therefore, credit was not given for any correct English responses.

2.3Statistical analysis

The detailed statistical analyses used to generate the normative data for this test are described in Rivera et al. (2024). In summary, since the verbal fluency total score consisted of the number correct words produced in 60 seconds [Y_i  (i = 1,   … ,  n)], Poisson distribution was assumed [Y_i ∼ Po (μ_i)]. However, BNT total score was considered as 60 trials or attempts, where each trial can be classified as success or failure, assuming a binomial distribution [Y_i ∼ Bin (m_j,  π_j)], where m is the total number of trials (m = 60) and π is the probability of success on each trial.

A Bayesian approach was adopted to determine which variables should be included as predictors, wherein the available prior information was combined with the information from the data. This combination is summarized in a posterior probability distribution for each unknown quantity in the model. In the variable selection procedure, the unknown quantity is the true model and, for each of the 2^p combinations, where p refers to the number of covariates, a candidate is obtained that may be potentially related to the output (Li & Clyde, 2018).

The full regression models included as predictors: age, age², education (log- transformed), sex, the Bidimensional Acculturation Scale for Hispanics (BAS; Marin & Gamba, 1996), the Bilingual Dominance Scale (BDS, Dunn & Fox, 2009), and all two-way interactions between these variables. Age was centered (age in years – X¯Age in the sample) before computing quadratic age to avoid multicollinearity (Kutner et al., 2005). Once the variables were selected, the model estimations were obtained using Poisson and Binomial regression models, respectively. The models can be expressed as follows:

The Bayesian inference procedure was performed using Markov Chain Monte Carlo methods through the software JAGS (Just Another Gibbs Sampler) and its R interface rjags (Plummer, 2022). For the convergence of three chains for each parameter, a burning period was used. A total of 3000 samples of each posterior distribution were left.

3.1Variable selection

Twenty-nine possible candidate covariates were examined. Table 1 shows the posterior inclusion probability (PIP) values of all covariates in each neuropsychological score. From the PIPs, an elbow plot (x-axis=each covariate, y-axis=PIP values) was made and the substantial change in PIPs was studied to select variables for the regression model for each neuropsychological score (Table 1).

3.2Phonological verbal fluency

Table 2 presents the mean, median, mode, and credibility interval (95%) of the posterior distributions for the for the parameters in each model. The final model for /f/ showed that age by BDS interaction arose, such that people over 50 years old who had higher English competence generated more words than those with higher Spanish proficiency. However, people younger than 50 years old with higher Spanish proficiency produced more words than those with higher English competence (Fig. 1A). Also, an education by BDS interaction was found, such that after ten years of education, people with higher Spanish proficiency produced a greater number of words than those with higher English proficiency. However, people with fewer than ten years of education but higher English proficiency produced more words than those with higher Spanish proficiency (Fig. 1B).

Fig. 1

Predicted mean count as a function of demographic variable included in the phonologic verbal fluency models. Note: BAS = Bidimensional Acculturation Scale for Hispanics; BDS = Bilingual Dominance Scale.

Education was related to the /a/ phoneme, such that scores increased according to education’s natural logarithmic transformation (Fig. 1C). Moreover, a sex by BDS interaction was found, such that women with higher Spanish competence produced a greater number of words whereas men did not (Fig. 1D). Finally, a sex by time in the U.S. interaction was found, such that women with a greater number of years living in the U.S. generated fewer words, whereas men generated a greater number (Fig. 1E).

The /s/ phoneme was also associated with education, such that scores increased according to education’s natural logarithmic transformation. Furthermore, a BAS by time in the U.S. interaction was found, such that people who spent a greater number of months in the U.S. and had lower Hispanic acculturation performed better on this phoneme. However, people with fewer months living in the U.S. but higher Hispanic acculturation achieved higher scores (Fig. 1F). Moreover, a sex by BDS interaction was found, such that women with higher Spanish proficiency produced a greater number of words whereas men did not.

Regarding the /p/ phoneme, an age by education interaction was found, such that people with greater years of education and older age achieved higher scores compared to older people with fewer years of education (Fig. 1G). Furthermore, a sex by BDS interaction was found, such that women with higher Spanish proficiency produced a greater number of words, while men maintained their level of performance regardless of their Spanish or English proficiency. Finally, a BAS by time in the U.S. interaction was found, such that people with higher Hispanic acculturation performed better on this phoneme regardless of the time spent in the U.S. However, people with lower Hispanic acculturation perform worse if they had spent more time in the U.S.

For the /m/ phoneme, scores increased according to the natural logarithmic transformation of education. A sex by BDS interaction was found, such that men generated a greater number of words with higher English competence compared to women, but with higher Spanish competence, women produced a greater number of words (Fig. 1H). Finally, a BAS by time in the U.S. interaction was found, such that people with higher Spanish acculturation and more time living in the U.S. performed worse compared to those living in the U.S. for less time. However, when people had lower Hispanic acculturation, they performed equally well regardless of the time living in the U.S.

Finally, for the /r/ phoneme, an age by education interaction arose, such that people with greater years of education and older age performed much better than those with fewer years of education (Fig. 1I). Lastly, a sex by BDS interaction was found, such that women produced a greater number of words with greater Spanish proficiency, while men produced fewer words.

3.3Summatory scores for FAS and PMR triads

Summatory score for FAS triad showed an education by BDS interaction, such that after 10 years of education, people with higher Spanish competence produced a greater number of words than those with higher English competence. However, people with fewer than 10 years of education but higher English competence produced a greater number of words than those with higher Spanish competence. Also, a sex by time in the U.S. interaction was found, such that women with a greater number of years living in the U.S. generated fewer words, whereas men a greater number. Finally, an age by sex interaction arose, such that while men’s performance maintained stable with increasing age, women’s performance improved with increasing age until they reached age 65 after which a decline in performance was observed.

On the other hand, summatory score for PMR triad showed a sex by BDS interaction, such that men generated a greater number of words with higher English competence compared to women, but with higher Spanish competence, women produced a greater number of words. Moreover, an age by education interaction was found, such that people with greater years of education and older age achieved higher scores compared to older people with fewer years of education. Finally, as with FAS triad, an age by sex interaction arose, such that while men’s performance maintained stable with increasing age, women’s performance improved with increasing age until they reached age 65 after which a decline in performance was observed (see Table 3).

3.4Semantic verbal fluency

For Animal and Fruit categories, an education by BDS interaction was found, such that after 5 years of education, people with higher education and greater Spanish proficiency produced a greater number of animals than those with greater English proficiency. However, in fruits category, the differences between Spanish and English proficiency were maintained regardless of educational attainment, and people with greater Spanish proficiency always produced a greater number of fruits (Fig. 2A, B). See Table 4 for mean, median, mode, and credibility interval (95%) for the posterior distributions for the of the parameters for each model.

Fig. 2

Predicted mean score as a function of demographic variable included in the semantic verbal fluency and BNT models. Note: BAS = Bidimensional Acculturation Scale for Hispanics; z-BDS = Bilingual Dominance Scale – rescaling to z distribution.

3.5Boston Naming Test (BNT)

Education was related to the BNT total score, such that scores increased according to the natural logarithmic transformation of education. In addition, an interaction between education by BAS and education by BDS were found, such that people with higher Hispanic acculturation and greater Spanish proficiency obtained higher scores than those with lower Hispanic acculturation and greater English proficiency. Moreover, performance increased with greater years of education (Table 4 and Fig. 2C, D).

3.6Normative procedure

An expanded explanation of the procedure can be found in Rivera et al. (2024). Briefly, norms (e.g., a percentile score) for the verbal fluency and BNT scores were established using a four-step procedure: 1) The expected test score (μˆi) was computed based on the parameters estimates from the established regression model (βˆp) , using:

Bear in mind that demographic variables were multiplied by 3000 samples of parameters. 2) The probability based on the Poisson/Binomial Cumulative Distribution Function was estimated for each of the 3000 sample parameters, 3) a mean probability was calculated, and 4) this probability was multiplied by 100 to interpret it as a percentile.

3.7User-friendly normative data

To facilitate the understanding of the procedure to obtain the percentile associated with a given score on this test, an example will be given. Suppose you need to find the probability for a woman, who is 50 years old and has 15 years of education. She obtained a BDS score of 10 and a score of 8 on the /f/ phoneme.

Since the method explained above is complex and can be prone to human error due to the number of required computations, we have created an online calculator based on https://www.rstudio.com/products/shiny/. This will facilitate probability calculation as clinicians should only include patient information requested in the calculator (i.e., gross score for the specific test, age, education, and so on). This tool is available for all users at https://neuropsychologylearning.com/datos-normativos-archivos-descargables/. Using the calculator and introducing the information requested, this woman would obtain a mean probability score of.245, that is, at the 24.5th percentile.

4.1Clinical implications

This study has important implications for neuropsychology practice and research. The norms provided may help clinicians and researchers improve accurate identification of neurocognitive impairments relating to phonological and semantic verbal fluency and word naming among Spanish-speaking adults in the U.S., facilitating appropriate treatment selection. The findings have also demonstrated the importance of generating normative data on language measures for dominant Spanish speakers in the U.S., since social, cultural, and linguistic differences have been shown in the current study and past research (e.g. Mungas et al., 2005) to impact neuropsychological test performance. In order to utilize these norms, the BDS and BAS must be administered to capture acculturation and bilingual dominance, two factors that are important when working with the diverse population of dominant Spanish speakers in the U.S. It is also important to note that the use of phonemes differs somewhat from letter fluency because the instructions here were designed to elicit words that begin with a sound rather than a letter. This is a small but important distinction that may be useful for individuals with low education or literacy (Olabarrieta-Landa et al., 2017). The effects of sociocultural variables on traditional demographic variables should be examined in future norming efforts for immigrant populations.

4.2Limitations and future directions

The current study had several limitations that present opportunities for future research. First, the mean years of education in the current sample was more than 12 years, which may limit the generalizability of the norms for individuals with lower levels of education. Additionally, this study only included neurologically healthy adults, highlighting the gap for future research to establish normative data with pediatric populations or adults with neurological and/or psychiatric illnesses. Second, while this study examined the effects of language proficiency on phonemic fluency test scores, the BDS is a subjective measure of language dominance that relies on individuals’ memory and perception of their own linguistic proficiency and language use. Since self-ratings are prone to bias, future research may consider using objective measures to assess bilingualism (Gollan et al., 2024). Finally, the current norms may only be used when the individual is fluent in Spanish and resides in the U.S., whereas the application of these norms to Spanish-speaking individuals outside of the U.S. would increase the risk of misdiagnosis.