This study separated the effects of age, birth cohort, and generation (parents, siblings, and offspring) on familial risk of epilepsy. The study population comprised 9,741 parents, siblings, and offspring of 1,957 adult probands with epilepsy ascertained from 10 voluntary organizations in New York, New Jersey, Connecticut, and Massachusetts between 1985 and 1988. Semistructured telephone interviews with probands and selected family members were used to collect data on the history of epilepsy in the relatives. The risk of epilepsy increased in successive generations (cumulative incidence to age 40: parents, 1.8%; siblings, 2.9%; offspring, 5.6%) but, with the exception of one subgroup (offspring of female probands), these differences disappeared after controlling for age and birth year of the relatives. With age and relationship controlled, the risk of epilepsy increased approximately 50% for each 20 advancing years of birth. Population-based data indicate that age-specific incidence rates of epilepsy have not increased during the age and time periods investigated here; thus, the most plausible explanation for the findings is that epilepsy is underreported in persons born in earlier time periods. These results illustrate a general phenomenon of underreporting in studies of lifetime prevalence, and they caution against interpreting apparent cohort effects in such studies as evidence of secular changes in incidence.