Age-of-onset data were collected on 142 bipolar probands (61 males, 81 females) and 110 siblings. After correcting for demographic factors the age at illness onset did not distinguish bipolar from unipolar siblings or siblings from probands. This indicates that in a subset of pedigree data, bipolar and unipolar affective disorders may be different phenotypic manifestations of the same underlying disorder. A total of 43% of affected individuals gave evidence of illness before age 30 and 11% were ill as adolescents. Also 2% were affected prior to age 15. The data indicate that affective illness is not uncommon in adolescence and may occur as a childhood disorder. Estimation of the age-of-onset probability distribution was obtained by incorporating demographic factors (survivorship data) into the observed ages at onset of affected individuals. A square-root normal curve fits the age-of-onset probability density function. The implications for morbid risk prediction and genetic analysis of affective disorders were discussed.