Norm Comparisons of the Spanish-language and English-language WAIS-III: Implications for Clinical Assessment and Test Adaptation

Abstract

This study provides a systematic comparison of the norms of three Spanish-language WAIS-III batteries from Mexico, Spain and Puerto Rico, and the US English-language WAIS-III battery. Specifically, we examined the performance of the four normative samples on two identical subtests (Digit Span and Digit Symbol-Coding) and one nearly identical subtest (Block Design). We found that across most age groups the means associated with the Spanish-language versions of the three subtests were lower than the means of the US English-language version. In addition, we found that for most age ranges the Mexican subsamples scored lower than the Spanish subsamples. Lower educational levels of Mexicans and Spaniards compared to US residents are consistent with the general pattern of findings. These results suggest that because of the different norms, applying any of the three Spanish-language versions of the WAIS-III generally risks underestimating deficits, and that applying the English-language WAIS-III norms risks overestimating deficits of Spanish-speaking adults. There were a few exceptions to these general patterns. For example, the Mexican subsample aged 70 and above performed significantly better on the Digit Symbol and Block Design than the US and Spanish subsamples. Implications for the clinical assessment of US Spanish-speaking Latinos and test adaptation are discussed with an eye toward improving the clinical care for this community.

Of the over 50 million Latinos residing in the US, more than 9 million (19%) reported speaking English not well or not at all (US Census Bureau, 2012). As a result, the assessment of Spanish-speakers has received increasing attention from clinical researchers (e.g., Siedlecki et al., 2010) and test developers (e.g., Riverside Publishing and Pearson). Understanding the normative samples of measures developed for Spanish-speakers can improve our ability to assess this linguistic group residing in the United States.

The Wechsler Adult Intelligence Scales (WAIS) have long been the gold standard for the evaluation of cognitive abilities in adults in the US, and several Spanish-language versions have been developed. During the last 15 years, the three most relevant published Spanish-language versions of the WAIS include one from Mexico (WAIS-III-M; Wechsler, 2003), Spain (WAIS-III-S; Wechsler, 2001), and Puerto Rico (EIWA-III-PR; Wechsler, 2008). Of these batteries, the WAIS-III from Spain has garnered the most attention by researchers (e.g., Choca, Krueger, de la Torre, Corral, & Garside, 2009). The WAIS-III-M and EIWA-PR are also important to consider as their normative samples are drawn from Mexico and Puerto Rico, the origin of the greatest number of US Latinos (Ennis, Rios-Vargas & Albert, 2011).

It is not clear what the strengths and weaknesses are of the norms for the three versions. Prior research indicates that the normative samples from the Mexican WAIS-III and the earlier Puerto Rican EIWA scored much lower than the US English language normative sample (e.g., Suen & Greenspan, 2009). As a result some clinical investigators argue against their use (e.g., Melendez, 1994). There is some evidence that these differences are related to differences in the educational level of the normative samples (López & Romero, 1988). In addition, previous research has only examined in any given study the norms of a single Spanish-language WAIS as it relates to the norms of the US WAIS (e.g., López & Taussig, 1991). It is unclear how the different Spanish-language norms compare to each other.

In the current study, we examine the normative data of the three published Spanish-language versions of the WAIS-III from Mexico, Spain and Puerto Rico, as well as the US WAIS-III from which they were adapted. Our purpose was not to assess the differences in language used across the Spanish versions. Instead, we examined differences in the performance of the normative samples by comparing group performance on identical subtests. Based on past research, we first predicted that the Spanish-language normative samples would have lower educational attainment than the US sample. Accordingly, we then hypothesized that the Spanish-language normative samples would score lower than the English-language sample on identical subtests. We also explored whether the Spanish-language versions of the WAIS differed from each other with regard to the normative samples’ level of education and their performance on identical subtests. Together these data have the potential to guide clinicians and researchers in the difficult task of assessing Spanish-speaking immigrants to the United States.

Method

Four versions of the WAIS-III were used (see Table 1). The Escala Wechsler de Inteligencia para Adultos (WAIS-III-M, Wechsler, 2003) was standardized on a sample of 970 Mexicans with an age range of 16 – 89 years. The Escala de Inteligencia de Wechsler para Adultos (WAIS-III-S, Wechsler, 2001) was standardized on 1,369 Spaniards with 15 – 94 years of age. The Escala de Inteligencia Wechsler para Adultos III (EIWA-III-PR, Wechsler, 2008) was standardized on 330 Puerto Ricans between the ages of 16 and 61. Although the US Wechsler Adult Intelligence Scale IV (WAIS-IV, Wechsler, 2008) has already been published, we opted to use the WAIS-III (Wechsler, 1997) because the Spanish versions of the WAIS, at the time this research was initiated, were all adapted from the third version of the WAIS. The WAIS-III-US was standardized on a national sample of 2,400 persons between the ages of 16 – 89. Note that the reliability estimates based on the samples were nearly identical across all available IQ and index scores and slightly lower for the subtests of the Spanish language versions than the US English language version (see Table 1).

Table 1.

Demographic Summary of WAIS-III Measures

Measure	Sample	Education	Education - Regrouped	Reliability
Escala Wechsler de Inteligencia para Adultos (WAIS-III-M; Wechsler, 2003).	970 adults Ages: 16–89 54% Female	Low: ≤8 years (42%) High: ≥9 years (57%)	Low: ≤8 years (42%) High: ≥9 years (58%)	IQ & Index: .70–.87 Subtests: Not provided
Escala de Inteligencia de Wechsler para Adultos (WAIS-III-S; Wechsler, 2001).	1,369 adults Ages: 15–94 51% Female	None (22%) Primary School (32%) Secondary School (38%) Higher/University (8%)	Low: ≤8 years (33%) High: ≥9 years (67%)	IQ & Index: .87–.97 Subtest: .77–.98
Escala de Inteligencia Wechsler para Adultos III (EIWA-III-PR; Wechsler, 2008).	330 adults Ages: 16–61 48% Female	0–8 years 9–11 years 12 years 13–15 years ≥16 years		IQ & Index: .86–.98 Subtests: .73–.92
Wechsler Adult Intelligence Scale –III (WAIS- III-US; Wechsler, 1997).	2,400 adults Ages: 16–89 80% White 8% Hispanic 50% Female	≤8 years (12%) 9–11 years (12%) 12 years (35%) 13–15 years (24%) ≥16 years (18%)	Low: ≤8 years (12%) High: ≥9 years (88%)	IQ & Index: .88–.97 Subtests: .82–.93

Note. The EIWA-III-PR technical manual did not provide percentages for education levels.

Two bilingual raters, a graduate student and an undergraduate student, independently examined the administrative and technical manuals, stimulus booklets, and scoring procedures of the four tests. Differences in administration (e.g., the starting item number in a subtest), instructions, procedures (e.g., time constraints), and scoring (e.g., item value and maximum scores) were documented by both parties. For subtests that were identical or nearly identical in content across measures, we examined whether the means of the samples differed across measures. To carry out these analyses we either used the published means from the technical manuals or we estimated the means and standard deviations by following the procedures outlined in Lopez and Romero (1988).

We carried out this calculation with all the age groups listed in the technical manuals. The EIWA-III-PR was excluded from comparisons because the manual did not provide sufficient information to determine the composition of the age groups. Since the composition of some age groups differed across the tests, we collapsed some of the age groups so that they were comparable to the other instruments for comparison. For example, the WAIS-III-M only has one older adult age group (i.e., 70+) with no further specification. In contrast, the WAIS-III-US delineates 4 specific age groups above the age of 70. To compare these groups, we collapsed the 4 age groups from the WAIS-III-US and took an average of their combined mean scores.

A similar procedure was used to assess the comparability of the educational background of each of the WAIS versions’ samples. To compare levels of education across groups, two education groups were created: a low education group (≤ 8 years) and a high education group (≥ 9 years). The WAIS-III-S required additional attention given that their four educational levels could not simply be collapsed into the same categories of the other WAIS versions: 0 – 8 and 9+ years of age. For the WAIS-III-S, these two educational levels were extrapolated as their actual categories were 0 – 7 years and then three other categories--the highest educational degree that the person obtained from either primary school, secondary school, or higher education/university. The second educational level then ranges from 8 – 10 years since those students would have received a degree from primary school (8 years) but they would not have received a degree from secondary school, defined as 11 years (Choca et al., 2009). To estimate the n of those completing 0 – 8 years of schooling, we divided the n associated with the 8 – 10 years group by 3 and then added one-third to the 0 – 7 years category. The remaining sample comprised the high educational level. See Table 1 for further description of the education groups.

Results

We first identified the identical or near identical subtests across the four Wechsler scales, we then assessed the comparability of the samples’ educational background, and then we tested for differences in the performance of the samples on the identified subtests. Two of the subtests, Digit Span and Digit Symbol-Coding, were found to be identical across all four versions of WAIS. Block Design was nearly identical across the test batteries, as only one difference was noted with the Puerto Rican version; items 7 and 8 were inverted such that they were items 8 and 7 of the three other versions. In addition, the direction of the red interior design was angled to the left for the Puerto Rican version and to the right for the other versions. Because the scoring was the same for both items across all tests, and the differences in the stimuli were minimal, we also included the Block Design in our analyses. The other eight subtests differed in some manner and were excluded from analyses. The coders agreed on all but one of the judgments of whether the subtests were identical to the US subtests (K = .94). The one disagreement occurred with the Digit Symbol subtest. After discussion, both coders agreed that the subtest was identical.

With regard to the educational attainment of the subgroups, the Mexican WAIS-III sample (low educational level 42%, high educational level 58%) reflects a lower educational level than the US WAIS-III sample (low 12%, high 88%), χ² (1, N = 3420) = 406.56, p < 0.001. The educational attainment estimates of the Spanish sample (33% low, 67% high) also are lower than the US sample (12% low, 88% high), χ² (1, N = 3819) =1269.14, p < 0.001, but are higher than the Mexican sample (42% low, 58% high), χ² (1, N = 2339) = 23.82, p < 0.01. Such comparisons were not possible with the Puerto Rican sample due to the lack of educational information.

The estimated raw score means for the three subtests and each age group are noted in Table 2. To examine whether these differences are significant and reliable, independent sample t-tests with Bonferroni adjustments (.05/48) were conducted using the estimated means and standard deviations for each age group within the three subtests. Mean comparisons of the WAIS-III-M and the WAIS-III-US reveal significant differences across most of the three subtests and age groups, such that the Mexican subsamples performed at a lower level than the US subsamples. These differences reflected medium to large effect sizes (Cohen’s d): Digit Span −0.61 to −1.24; Digit Symbol-Coding −0.64 to −1.40 and Block Design −0.42 to −0.83. An exception was found in the Block Design subtest for age groups 16–17, 18–19, 20–24, 25–29, 55–64, and 65–69 in which no significant differences were observed. The analysis also revealed that the Mexican 70+ sample performed significantly better than the US 70+ sample on the Block Design (d = 1.34) and Digit Symbol-Coding (d = 1.66) subtests.

Table 2.

Mean Comparisons and Effect Sizes of Three WAIS Subtests by Age Group

		WAIS-III-M		WAIS-III-S		WAIS-III-US		tM-US	dM-US	tS-US	dS-US	tM-S	dM-S
Age		N	M(SD)	N	M (SD)	N	M (SD)
Digit Span
16–17		99	15.00 (4.00)	-	-	200	17.50 (4.25)	−4.88*	−0.61	-	-	-	-
18–19		110	14.50 (3.50)	-	-	200	17.50 (4.25)	−6.08*	−0.72	-	-	-	-
20–24		95	14.50 (3.75)	-	-	200	17.50 (4.50)	−5.63*	−0.70	-	-	-	-
25–29		97	14.50 (3.50)	-	-	200	17.50 (4.50)	−5.77*	−0.72	-	-	-	-
30–34		96	13.50 (3.75)	-	-	200	17.00 (4.25)	−6.88*	−0.86	-	-	-	-
35–44		97	13.00 (3.50)	-	-	200	17.00 (4.25)	−8.57*	−1.06	-	-	-	-
45–54		94	13.00 (3.50)	-	-	200	17.00 (4.25)	−7.99*	−1.00	-	-	-	-
55–64		98	12.50 (4.00)	-	-	200	16.00 (4.50)	−6.54*	−0.81	-	-	-	-
65–69		77	13.00 (3.50)	-	-	200	16.00 (4.25)	−5.51*	−0.74	-	-	-	-
70+		107	10.50 (4.25)	-	-	650	14.63 (3.75)	−10.34*	−1.24	-	-	-	-
Digit Span†
16–19		209	14.75 (3.75)	163	16.00 (4.00)	400	17.50 (4.25)	-	-	−3.86*	−0.36	−3.10	−0.32
20–24		95	14.50 (3.75)	153	16.00 (4.25)	200	17.50 (4.25)	-	-	−3.18*	−0.34	−2.82	−0.38
25–34		193	14.00 (3.63)	272	16.00 (4.25)	400	17.25 (4.38)	-	-	−3.68*	−0.29	−5.30*	−0.50
35–54		191	13.00 (3.50)	408	14.00 (4.00)	400	17.00 (4.25)	-	-	−8.51*	−0.60	−2.96	−0.27
55–69		175	12.75 (3.75)	237	11.00 (3.50)	400	16.00 (4.38)	-	-	−14.98*	−1.23	4.87*	0.49
70+		107	10.50 (4.25)	136	9.00 (3.50)	800	14.63 (3.75)	-	-	−16.83*	−1.56	−3.02	−0.39
Digit Symbol
16–17		99	65.00 (16.00)	-	-	200	77.00 (15.75)	−6.17*	−0.76	-	-	-	-
18–19		110	69.00 (15.00)	-	-	200	81.00 (16.00)	−6.46*	−0.77	-	-	-	-
20–24		95	61.50 (23.00)	-	-	200	80.00 (16.25)	−7.95*	−0.99	-	-	-	-
25–29		97	63.50 (18.00)	-	-	200	78.00 (15.50)	−7.16*	−0.89	-	-	-	-
30–34		96	62.50 (18.50)	-	-	200	77.00 (16.00)	−6.91*	−0.86	-	-	-	-
35–44		97	54.00 (19.50)	-	-	200	75.00 (16.50)	−9.68*	−1.20	-	-	-	-
45–54		94	47.50 (18.00)	-	-	200	70.00 (15.25)	−11.12*	−1.40	-	-	-	-
55–64		98	44.00 (22.00)	-	-	200	61.00 (15.00)	−7.83*	−0.97	-	-	-	-
65–69		77	42.50 (24.00)	-	-	200	54.00 (15.00)	−4.78*	−0.64	-	-	-	-
70+		107	65.00 (16.00)	-	-	200	43.75 (14.63)	13.79*	1.66	-	-	-	-
Digit Symbol†
16–19		209	67.00 (15.50)	163	77.00 (16.50)	400	79.00 (15.88)	-	-	−1.34	−0.13	−6.00*	−0.63
20–24		95	61.50 (23.00)	153	77.50 (15.00)	200	80.00 (16.25)	-	-	−1.48	−0.16	−6.63*	−0.87
25–34		193	63.00 (18.25)	272	76.00 (16.75)	400	77.50 (15.75)	-	-	−1.18	−0.09	−7.94*	−0.75
35–54		191	50.75 (18.75)	408	65.00 (23.00)	400	72.50 (15.88)	-	-	−5.49*	−0.39	−7.48*	−0.66
55–69		175	43.25 (23.00)	237	39.50 (21.50)	400	57.50 (15.00)	-	-	−12.40*	−1.02	1.70	0.17
70+		107	65.00 (16.00)	136	20.00 (14.00)	800	43.75 (14.63)	-	-	−17.35*	−1.61	23.35*	3.03
Block Design
16–17		99	36.50 (10.75)	-	-	200	41.50 (13.50)	−3.21	−0.40	-	-	-	-
18–19		110	37.50 (11.25)	-	-	200	41.50 (13.50)	−2.64	−0.31	-	-	-	-
20–24		95	38.50 (12.50)	-	-	200	41.50 (13.50)	−1.83	−0.23	-	-	-	-
25–29		97	40.00 (10.75)	-	-	200	41.50 (13.50)	−0.10	−0.12	-	-	-	-
30–34		96	35.50 (12.75)	-	-	200	41.00 (13.50)	−3.38*	−0.42	-	-	-	-
35–44		97	29.50 (14.00)	-	-	200	40.50 (13.00)	−6.67*	−0.83	-	-	-	-
45–54		94	30.50 (12.25)	-	-	200	36.50 (12.00)	−3.97*	−0.50	-	-	-	-
55–64		98	29.00 (11.00)	-	-	200	33.50 (12.00)	−3.12	−0.39	-	-	-	-
65–69		77	25.50 (11.00)	-	-	200	29.50 (11.75)	−2.58	−0.35	-	-	-	-
70+		107	36.50 (10.75)	-	-	200	26.00 (8.75)	11.11*	1.34	-	-	-	-
Block Design†
16–19		209	37.00 (11.00)	163	46.50 (11.00)	400	41.50 (13.50)	-	-	4.19*	0.39	−8.26*	−0.87
20–24		95	38.50 (12.50)	153	46.50 (11.00)	200	41.50 (13.50)	-	-	3.73*	0.40	−5.28*	−0.69
25–34		193	37.75 (11.75)	272	45.50 (11.25)	400	41.25 (13.38)	-	-	4.31*	0.34	−7.19*	−0.68
35–54		191	30.00 (13.13)	408	37.00 (12.75)	400	38.50 (12.50)	-	-	−3.16*	−0.22	−13.73*	−1.21
55–69		175	27.25 (11.00)	237	25.50 (13.50)	400	31.50 (11.88)	-	-	−5.85*	−0.48	1.40	0.14
70+		107	36.50 (10.75)	136	15.00 (18.75)	800	26.00 (8.75)	-	-	−13.33*	−1.24	17.19*	2.23

Note. For comparisons with the US, the negative values of the t and d indicate that the Mexican or Spanish mean is lower than the US test norm. For comparisons between Spain and Mexico, the negative values refer to the Mexican sample doing worse than the Spanish sample.

^†Mean comparisons with the WAIS-III-S have a small age range and require a separate listing.

^M-USWAIS-III-S and WAIS-III-US mean comparisons.

^S-USWAIS-III-M and WAIS-III-US mean comparisons.

^M-SWAIS-III-M and WAIS-III-S mean comparisons.

^*= p < .001.

When comparing the WAIS-III-S to the WAIS-III-US, the pattern of findings resembles a subtest by age group interaction. On the Digit Span subtest, the WAIS-III-S had lower scores than the WAIS-III-US across all age groups (ds ranged from −0.29 to −1.56). On the Digit Symbol-Coding subtest, however, the results varied across age groups. There were no significant differences between the performances of the younger subgroups (16–34 years of age). For subsamples 35 years old and above, however, the Spanish subsamples had lower scores than their US peers (ds ranged from −0.39 to −1.61). For the Block Design subtest, the findings for the two same age groups diverged again. This time the younger Spanish subsamples scored significantly higher than their US counterparts (ds ranged from 0.34 to 0.40), whereas the older Spanish subsamples (35 years and older) scored significantly lower than the older US subsamples (ds ranged from −0.22 to −1.24).

The performance of the Mexican and Spanish normative groups on the Digit Span subtest also varied by age group. The 25–34 age group of the Mexican subsample performed significantly lower than the Spanish sample (d = −0.50) whereas for the 55–69 age group the Mexican subsample performed significantly better than Spanish subsample (d = 0.49). Comparisons across other age groups were not significantly different. On the Digit Symbol-Coding subtest, the Mexican subsample had lower mean scores than the Spanish subsample (ds = −0.63 to −0.87) in every age group with the exception of the 55–69 (d = 0.17) and 70+ age group (d = 3.03). On the Block Design, a similar pattern emerged in which the Mexican subsample performed lower than the Spanish subsample for most age groups (16–54, ds = −.68 to −1.21) except for the 55–69 group (d = 0.14) and the 70+ age group, for which the Mexican subsample of the eldest group again outperformed the Spanish subsample (d = 2.23).

Discussion

Consistent with previous findings (Lopez & Romero, 1988; Melendez, 1994, Suen & Greenspan, 2004), the means associated with the Spanish-language versions of the three identified subtests were lower than the means of the US version across most age groups. Figure 1 illustrates this general pattern. Although we were unable to compare statistically the EIWA-III-PR, nearly all of the subgroup means of the Puerto Rican sample fell below those of the US sample. Comparisons of the Spanish-language samples also revealed differences between the subsamples. For most age ranges from 16 – 54 years, the Mexican subsamples performed lower than the Spanish subsamples across subtests. These findings were robust as 9 of the 12 comparisons were statistically significant after statistical corrections. These observed differences suggest that the WAIS-III-M and the WAIS-III-S are not interchangeable.

Figure 1.

Subtest raw score means by age group.

The differences observed across samples may be due to several factors, including differences in socioeconomic level, access to medical care, nutrition, and prevalence of vascular disease and other conditions known to affect cognition. Given the available normative data provided in the test manuals, our study was only able to examine the possible role of formal education. The proportion of the three samples with 8 years or less of schooling was 42%, 33%, and 12% respectively for the Mexican, Spanish (estimated), and US samples. These educational levels are congruent with the generally lower performance of the Mexican sample.

Not all comparisons followed the general pattern of the lower Spanish language norms and the higher US English language norms. Mexican older adults performed significantly better on the Digit Symbol and Block Design subtests than the combined US subgroups aged 70 and above, and they also achieved equal or higher means than most of the younger age groups of their own sample (See Figure 1). The fact that the 70+ subsample does as well as or outperforms many of their younger cohorts raises serious concerns about the Mexican older adult norms. Different sampling methods may have played a role. Sánchez Escobedo and Hollingworth (2009) noted that persons of higher educational level were overrepresented in the Mexican normative sample. This is supported by OECD (2004) data that indicate that only 14% of the Mexican population 25 – 64 years of age completed lower secondary education (approximately 8 years). This contrasts with the large percentage of the WAIS-III-M older adult subsamples that completed at least 9 years of schooling--51%.

Education alone, however, does not seem to explain how well the older adult Mexicans did relative to the US sample. Another possible explanation is that the Mexican subsample is significantly younger than the US subsample. To test this possibility, we examined the performance of the older Mexican subsample (70+) to each of the older age groups of the US subsample, 70 – 74, 75 – 79, 80 – 84, 85 – 89. If the Mexican 70+ subsample was indeed younger, then we would expect no differences at the younger age range but differences at the older age range. The t-tests revealed that on both Digit Symbol and Block Design, the Mexican older adult sample outperformed each of the four older adult US groups by more than one standard deviation (p < .001). Thus, the noted advantage of the older Mexican subsample was not likely due to the subsample being younger than the older US subsample.

A related finding was the precipitous drop in performance of Spanish older adults. We considered two possible explanations for this difference—educational level and limited cognitive screening. With regard to education, only one-third of the 70+ subsample of the WAIS-III-S had at least a 9^th grade education. Choca et al. (2009) suggested that older Spaniards had less access to education than younger adults given the Spanish Civil War in 1936 – 1939 and its aftermath. A second possibility is that there may have been limited screening in ruling out mild cognitive impairment among the older Spaniards. Including some older adults with such impairment would significantly lower the means. In any case, the Spanish older adults’ norms also pose significant problems for assessing this age group.

In addition to the discrepant findings for the older adults, Digit Span had a unique pattern of results. The Mexican older adult subsample outperformed the US older adult subsample on the Digit Symbol and Block Design but not on the Digit Span. Similarly, the Spanish younger adults (a) did better than the US sample of younger adults on Block Design, (b) did the same on the Digit Symbol, and (c) did worse than their US counterparts on the Digit Span. Characteristics of the Digit Span may contribute to this discrepant set of findings. The Digit Span requires individuals to manipulate and hold in memory strings of numbers of increasing length. Given that the number of syllables of stimuli are associated with increasing task difficulty (Baddeley, 1992; Lewandowsky & Farrell, 2008), the differences in the Spanish- and English-language representations of numbers (1.7 versus 1.1 syllables per digit) may increase the cognitive load of this test for Spanish-speakers (see also Loewenstein, Arguelles, Barker & Duara, 1993).

Implications

With regard to the clinical assessment of Spanish-speaking individuals in the United States, the critical question is: What set of norms is most appropriate for a given individual? Much of the available clinical research advocates for applying only one set of norms. Some investigators point out how the Mexican, Spanish, and Puerto Rican norms deviate from US norms and argue that the Spanish-language versions of the WAIS should not be used (Melendez, 1994; Suen & Greenspan, 2009) or should be adjusted (Choca et al., 2009). Other researchers, in promoting standardization in the assessment of Spanish-speaking Alzheimer patients, have put together a battery with a single set of norms from the US English-language WAIS-R subtests of Digit Span and Digit Symbol (Acevedo, Krueger, Navarro, Ortiz et al., 2009).

Our findings raise questions about adhering to a given set of norms. When using the norms from the Spanish-language versions of the WAIS, the potential risk of underestimating cognitive deficits of US Latinos has been clearly articulated (Melendez, 1994; Suen & Greenspan, 2009). We agree that this is a valid concern. However, much less attention has been given to the potential risk of applying US norms and even Spanish national norms in evaluating Spanish-speaking Latinos, as the level of cognitive impairment for some will be overestimated. This is suggested by the fact that the US and Spanish national samples outperformed the Mexican and Puerto Rican samples in most age groups and is consistent with prior research (e.g., Lopez and Taussig, 1991). The risk in overestimating deficits is particularly the case with the Digit Span, a subtest commonly used in neuropsychology batteries. Given that many immigrants with very limited educational backgrounds are likely to be poorly represented in the normative samples from the US or even from Spain, using those norms risks ascribing cognitive deficits to low performance on tests when there are no cognitive deficits.

Given that any specific test and their associated norms are at risk to over- or underestimate deficits and functioning, it is difficult to advocate for any one set of norms or any given Spanish-language version of the WAIS. Accordingly, we recommend that clinicians carefully consider the strengths and risks of the given WAIS options. If a clinician prefers using one specific Spanish-language version, it is important that they become familiar with the risks associated with that particular test and take steps to address the potential errors. Consideration should be given to an individual’s education level, their national/ethnic origin, along with other characteristics represented in the respective normative sample. For example, for Spanish-speaking adults who are 70 years and older, both the Mexican and Spanish norms appear problematic. The Mexican norms will likely overestimate deficits and the Spanish norms will likely underestimate deficits. The fact that the type of potential error differs by subtest and age range points out why familiarity with a test is most important. Considering the available test options, being familiar with the risks with each test option allows the examiner to fit the test to the person. Nonverbal tests (e.g., Bracken and McCallum, 1998) are another option although their norms should be carefully reviewed as well given that studies have noted similar challenges in the use of nonverbal measures (Saez et al., 2014).

Having carefully reviewed four versions of the WAIS-III, we also have recommendations for the adaptation of future non-English-language WAIS batteries. Most importantly, if test developers are going to permit test adaptations to be carried out in other countries, then it is important that adequate resources be made available to carry out quality test adaptations. The EIWA-III only sampled 330 native Puerto Ricans across 11 age ranges with the oldest age being 61. There are no reports in the technical manual or relevant publications (e.g., Pons, Matias-Carrelo, Rodriguez, Herrans et al, 2008) specifying the ns within each age level. Limited resources also may have contributed to the divergent norms for the 70 and older age group for the Mexican and Spanish WAIS-III. Every effort should be made to obtain older adult subsamples that represent this segment of the respective countries. Test publishers have the potential to make an important contribution by supporting the strongest test adaptation possible. (See Lopez & Weisman, 2004).

Limitations

First, we were not able to include subtests across the verbal ability domains given that there were no identical subtests across this domain. As such, we are not able to generalize our findings to the verbal domain nor are we able to make inferences as to how our findings relate to the assessment of global intelligence. Second, we could not conduct statistical comparisons with the Puerto Rican version of the WAIS-III due to the limited published data. Third, given that we did not have access to the raw data for the individuals in the samples, we were unable to carry out analyses to determine the specific relationship between the demographic factors and the performance of each sample. The only demographic data that we were able to consider was educational level. That analysis, however, was limited as we reduced educational level to only two levels. As a result, we could only estimate whether education was indeed related to normative performance. Furthermore, there were no data available to consider literacy levels or quality of their formal education, as both factors could affect the appropriateness of a specific set of norms (e.g., Manly, Byrd, Touradji, & Stern, 2004). One final limitation is that the WAIS-IV for Mexico and Spain were recently published and are now available for purchase. Given that the relative educational attainment of these regions has not changed, the general pattern of findings observed with the WAIS-III may still apply.

Conclusion

Having available multiple published Spanish-language assessment tools is a significant advance not only for Spanish-speaking countries but also for the large Spanish-speaking adult population of the United States. The identified normative performance differences and the associated limitations with the specific Spanish-language WAIS-III versions and the US English language version however highlight the importance of psychologists being careful when selecting assessment measures for use with Spanish speaking adults in the US. Persistent efforts by practitioners, researchers, and test developers, and continued communication among these professionals are essential for Spanish-speaking Latinos to receive the very best clinical services.