The current investigation compared neuropsychological test performance among nondemented literate and illiterate elders. The sample included participants in an epidemiological study of normal aging and dementia in the Northern Manhattan community. All participants were diagnosed as nondemented by a neurologist, and did not have history of Parkinson's disease, stroke, or head injury. Literacy level was determined by self-report. MANOVAs revealed a significant overall effect for literacy status (literate vs. illiterate) on neuropsychological test performance when groups were matched on years of education. The overall effect of literacy status remained significant after restricting the analyses to elders with no formal education, and after controlling for the effects of language of test administration. Specifically, illiterates obtained lower scores on measures of naming, comprehension, verbal abstraction, orientation, and figure matching and recognition. However tests of verbal list delayed recall, nonverbal abstraction, and category fluency were unaffected by literacy status, suggesting that these measures can be used to accurately detect cognitive decline among illiterate elders in this sample. Differences in organization of visuospatial information, lack of previous exposure to stimuli, and difficulties with interpretation of the logical functions of language are possible factors that contribute to our findings.