For both dementia and Alzheimer's disease (AD), data regarding incidence rates in the oldest old and time trends in incidence are limited. The authors reanalyzed previously reported data on the incidence of dementia and AD in Rochester, Minnesota, from 1975 through 1984, using three new strategies. First, incidence rates were corrected by removing age-, sex-, and calendar year-specific prevalent cases from the census-derived denominator figures. Second, incidence figures for persons above age 84 years were disaggregated. Third, time trends were investigated graphically using age-specific curves and birth cohort curves. Dementia diagnosis and AD diagnosis followed defined ad hoc criteria. Analyses were conducted for men, women, and both sexes combined, and for dementia and AD separately. The age-specific incidence rates were similar in men and women, continued to increase after age 84 years, and remained stable over time for both dementia and AD. No birth cohort effect was present for either dementia or AD. The similar risks seen in men and women, the continuing increase in incidence after age 84 years, and the stability of incidence over time have important implications for etiologic research on AD.