We determined the reliability and validity of a cancellation test of symbols (Symbol Cancellation Test [SCT]), designed to assess visual selective attention deficits in the elderly, on 34 Alzheimer's disease (AD) patients from Bari University Hospital Center, Bari, Italy, and 232 nondemented elderly subjects, aged 68 to 87 years, from the second prevalence survey (1995 through 1996) of the Italian Longitudinal Study on Aging (Casamassima, Bari, Italy). To assess convergent and discriminant validity, the Digit Cancellation Test (DCT), Mini-Mental State Examination (MMSE), and Babcock Story Recall Test (BSRT) were administered. Finally, discriminant accuracy of SCT between AD patients and nondemented elderly subjects was assessed. Inter-rater and test-retest reliability for P1 and P2 was excellent for both AD patients and the normal population, with a high degree of internal consistency. The SCT was significantly correlated with the DCT (0.67), MMSE (0.60), and BSRT (0.33). The classification accuracies of overall performance on the SCT for subjects with and without AD were 0.62 and 0.91, respectively. The SCT is a valid and reliable test to assess selective attention in elderly subjects, in whom dementing illness must be diagnosed and clinically distinct from age-related cognitive decline.