Estimating premorbid intelligence in African-American and white elderly veterans using the American version of the National Adult Reading Test.

Knowledge of earlier intellectual functioning can be very helpful in the diagnosis of neurocognitive disorders. The National Adult Reading Test (Nelson, 1982), based on the observation that reading irregularly spelled words remains intact in mild dementia, may be useful for this purpose. A version for speakers of American English (AMNART) was recently published (Grober & Sliwinski, 1991). The present study evaluated the validity of the AMNART with demented and nondemented African-American and White elderly veterans. The average differences between African-American and White participants on the AMNART and WAIS-R were similar. The differences between demented and nondemented patients were greater on the WAIS-R than on the AMNART, as expected. Paradoxically, the AMNART appears to over-estimate IQ among nondemented individuals with lower WAIS-R scores, but underestimate IQ among more impaired demented patients. Using Grober and Sliwinski's (1991) 10-point discrepancy rule between AMNART and WAIS-R VIQ scores, 70% of demented and nondemented veterans were successfully classified. No discontinuation rule for the AMNART could be established because word difficulty order varies across samples.