Partially confirming our hypothesis, results of this study revealed that after controlling for age, gender, and education, reading ability accounts for a significant proportion of variance on neuropsychological measures of executive function, attention, working memory, and verbal fluency in our sample of African American older adults. In contrast, measures of memory (RAVLT and FOME) were not related to either reading ability or education, but were significantly associated with age. When WRAT-3 reading was entered into a hierarchical regression analysis, the relationship between education and raw scores on five neuropsychological measures (TMT-B, Stroop Word, Animal Naming, COWAT, and Digit Span) became non-significant, whereas WRAT-3 reading was a significant predictor of performance on each of these measures. Additionally, reading ability was significantly associated with performances on TMT-A, Stroop Color/Word, and Stroop Word, though education was not. Impressively, examination of squared semi-partial correlations indicates that reading ability accounted for 7%–41% of the variance in raw test scores and effect size analyses revealed medium to large effects (Cohen, 1992
) when comparing mean scores on neuropsychological measures for groups with low versus high scores on the WRAT-3.
A wide range of variables influences African American elders' performance on neuropsychological measures; however, normative data are generally stratified only by age, education, and, less frequently, race. Previous studies have demonstrated that years of education are often not an accurate proxy for actual educational experiences, putting African American adults at risk for misclassification, and that reading ability may be a better indicator of educational quality among African American adults (Dotson et al., 2008
; Manly et al., 2002
; Schafer Johnson et al., 2006
.). This study sought to provide further evidence of the contribution of reading ability to test performance across a broad range of neuropsychological measures using a community-dwelling sample of African American elders.
The lack of relationship between performances on measures of memory and education or reading ability is counter to previously published work (Dotson et al., 2008
; Morgan et al., 2008
). The current findings are, however, partially consistent with Wall, Deshpande, MacNeill, and Lichtenberg (1998)
who found that Logical Memory scores from the Wechsler Memory Scale-Revised in a sample of urban African Americans were significantly related to reading levels, whereas scores on the FOME were not. In the current study, there was a trend (p
= .02) for RAVLT Total score to be related to reading levels but even trend-level significance was not found for the FOME. The difference in the nature of the stimuli and their frequency of use (i.e., whether language is to be recalled vs. common objects) may account for the consistent findings of significant or near significant relationships between verbal memory tasks and reading ability versus the FOME. However, this is a finding that should be further investigated.
When interpreting these results, it may be argued that reading ability would be expected to contribute to measures that involve verbal skills. However, reading ability was also related to measures that weigh less heavily on verbal skills, such as the TMT-B and Digit Span, and was not related to test performance on measures that require components of verbal skills such as the RAVLT. Additionally, previous studies have reported that the WRAT-3 reading test accounted for significant variance in non-verbal tasks such as Colored Progressive Matrices (Schafer Johnson et al., 2006
) and the Card Rotation Test (Dotson et al., 2008
) in samples of African American adults.
There are several clinical implications to these findings. Results of this study as well as several prior studies indicate that reading ability is not congruent with reported years of education among African American elders. As such, judging an African American patient's performance on years of education alone may result in misclassification, especially on measures that are highly related to reading ability. As neuropsychologists become increasingly aware of the influence of educational experiences on test performance, providing normative data for some measures stratified in part by reading ability may improve diagnostic accuracy. Currently, studies investigating whether consideration of reading ability improves diagnostic accuracy are limited. Lucas and colleagues (2005)
reported that controlling for reading ability in addition to age and education in the MOAANS sample only modestly improved negative predictive power and decreased positive predictive power in differentiation of normal African American elders from those with dementia. The authors indicate that these findings were not expected and posited that the findings may be explained partially by the lack of heterogeneity in the sample's educational background. As such, they assert that the “variability of reading scores within this ‘restricted’ sample may represent differences in true cognitive ability or cognitive reserve” rather than differences in educational quality. Further work is needed to identify whether there is diagnostic utility of controlling for reading ability in groups of African Americans with varying educational backgrounds.
Although WRAT-3 Reading is a screening tool of reading ability and does not replace a more in-depth evaluation or clinical interview regarding educational achievement and experiences, it is brief and offers clinicians a quick method by which to assess the literacy level. This measure could be easily used by researchers in studies for norm development as well as clinicians performing neuropsychological evaluations. One limitation of this approach is that as educational level increases beyond high school, educational quality becomes a less robust covariate (Ostrosky-Solis, Ardila, Rosselli, Lopez-Arango, & Uriel-Mendoza, 1998
). Though reading ability accounted for variance in our sample beyond age and education, this should be examined in samples with greater mean levels of education. It may be that for those individuals who obtained advanced degrees or had educational experiences of high quality, years of education alone may be an accurate predictor of test performance. Additionally, for some clinical questions, comparing the patient to his or her real-world peers or using a specific cut-point may be best clinical practice.
Additionally, though a comparison across racial groups was not conducted as a part of this study, we do not believe that reading ability accounts fully for test performance discrepancies between African American and Caucasian older adults. There are certainly many other factors that contribute to differences in test performance including disparities in health status (Morgan et al., 2008
), preferred language style (Manly et al., 1998
), and cultural equivalency characteristics of tests.
There are several limitations to this study. First, data for health and psychological conditions were based on self-report. Though self-report is commonly used in many studies, it may be a weakness when gathering health-related information. This method is more likely to yield a conservative estimate of disease burden. It should also be noted that our sample size was limited to just 86 participants and was predominantly women (nearly 92%). Due to these factors, this study should be considered preliminary and our results may apply to only a select population of individuals matching the characteristics of our sample. Future studies using larger and more heterogenous samples of African American older adults would validate these findings. Finally, because the goal of this study was to examine community-dwelling African American older adults, participants were not excluded based on their performances on neuropsychological measures. The average educational attainment in this sample was 12 years; however, education ranged from 8 to 18 years of education. As demonstrated in previous work, namely Crum, Anthony, Bassett, and Folstein (1993)
, greater variability in scores on neurocognitive measures have been found in individuals with low versus high educational levels. As such, the range of scores on neuropsychological measures in this study reflects the lower educational attainment of some participants.
Taken together, this study has important implications for older African American adults. Mitrushina, Boone, Razani, and D'Elia (2005)
state, “Issues of ethnic diversity pose difficult and serious challenges for the field of neuropsychology, particularly as clinicians are more frequently being requested to assess functioning in patients from varied ethnic, cultural, socioeconomic and linguistic backgrounds.” A continuing goal for neuropsychology is to improve the diagnostic accuracy among minority populations. Essential to this is an understanding of contributors to neuropsychological test performance across various populations. This study underscores the need to gather normative data that can be used for patients from various minority groups and suggests that stratification based on education alone is not appropriate for some older adults. Further work is needed to better understand the relationships between reading ability, education, and neuropsychological test performance in all minority groups.