Despite their increasing representation in the population, little is known about the neuropsychological test performance of the oldest old, particularly those who live in residential settings. Limited published data and clinical experience suggest that this group is more likely to perform in the impaired range on standardized tests when cut-offs developed with younger groups are used. We examined the Dementia Rating Scale (DRS) performance of 82 nondemented nursing home residents, aged 80 to 99, with a mean education level of 11 years. Using published norms and cutoffs, a large percentage of this sample performed in the impaired range, particularly on the initiation and conceptualization subtests and on the total score. Education, but not age, was significantly related to performance in this sample. Percentages of patients misclassified were substantial in all groups, but were higher in those with less than 13 years of education. Using a lower total-score cutoff of 110 reduced the percentage of misclassifications markedly. We recommend the development and use of revised cutoff scores for the evaluation of very elderly nursing home residents.