Survival from Alzheimer's disease (AD) and other dementias into late old age may be a useful phenotype for genetic studies of successful cognitive aging. To support molecular genetics studies for successful cognitive aging, we conducted a two-stage study to determine an optimal age phenotype for successful cognitive aging. First, risk of AD was evaluated, through informant interviews, in 4,794 parents and siblings of 976 elderly nondemented probands who were divided into three different proband age groups: those aged 60–74, 75–89, and 90+. Relatives of probands aged 90+ had a significantly lower risk than the relatives of the other two proband groups. Second, this sample was combined with an earlier sample (combined nondemented elderly probands: n = 2,025; relatives: n = 10,506), and a series of proband age groups (i.e., 75–79, 80–84, 85–89, 90+) were used to determine which optimally identifies a group of relatives with low AD risk. Using the relatives of the nondemented proband aged 60–74 as the reference group, there were reductions in cumulative risk among relatives of probands aged 85–89 and 90+, but only the latter group also showed significant reductions to the relatives of probands aged 75–79, 80–84, and 85–89. This pattern of results varied little by sex. Finally, cumulative AD risk curves were similar between relatives of probands aged 90–94 and 95+. These results suggest that age 90 is an optimal age threshold to use for both men and women in genetic studies seeking to identify genes associated with successful cognitive aging. © 2007 Wiley-Liss, Inc.