Family history of Alzheimer's disease (AD) was investigated using a telephone reinterview of 99 cases and 116 controls selected from a case-control study of 170 matched pairs. It was found that the family history method used in the initial interview was satisfactory in identifying first-degree relatives and assessing their ages of birth and death, but the number of first-degree relatives suffering from AD was probably under-estimated. Family history of AD was confirmed as a risk factor for AD. Higher estimates of cumulative incidence were obtained among case relatives than among control relatives. No evidence was found to support the hypothesis that a familial form of AD is more common in those with earlier onset AD (before age 75) in those who display early, prominent features of aphasia or apraxia, or that an AD gene may be sex-linked. The curves for cumulative incidence showed no tendency to reach an asymptote, as is implied by the theory that some forms of AD are due to the action of an autosomal dominant gene. Estimates of lifetime risk by age 90 were within the range found by other investigators. Much larger samples of the very old are needed to obtain estimates of total lifetime risk with smaller standard errors.