From clinically based series it has been proposed that, in homozygotes for the apolipoprotein E epsilon 4 (apoE epsilon 4) allele, Alzheimer's disease is almost inevitable by the age of 80. A population sample of persons aged 70 years and over was interviewed in 1990-91 to ascertain the presence of dementia or cognitive impairment. The sample was reinterviewed in 1994, when the apoE genotype was also determined. Prevalence data for the 638 persons who completed the second examination revealed a linear association between having an apoE epsilon 4 allele and both dementia and cognitive impairment (for heterozygotes, odds ratio for dementia 1.89, 95% confidence interval 1.04-3.44 and for homozygotes OR 3.58, 95% CI 1.08-11.82; both adjusted for age). However, even in subjects homozygous for epsilon 4 the estimated prevalence of dementia by age 90 was only about 50%. Persons with one or two epsilon 4 alleles were more likely to have a family history of dementia than those with none. This study confirms in a population sample that the epsilon 4 allele is a risk factor for dementia, but refutes the suggestion that homozygosity for the epsilon 4 allele is sufficient for the development of Alzheimer's disease: persons with either one or two epsilon 4 alleles may reach late old age without cognitive impairment.