About 10% of hospitalized unipolar depressed patients ultimately became bipolar. Using a blind protocol to assess data from patient (N = 485) and first-degree relative (N = 2,803) interviews and hospital records, the authors found that bipolar probands and unipolar probands had more affective illness in their families than did control subjects. Although more relatives of bipolar probands than of unipolar probands had bipolar illness, the difference in risk for affective illness between relatives of bipolar probands and of unipolar probands was not significant. In general there were more female than male depressed relatives, bit the equal numbers of men and women among bipolar relatives suggest heterogeneity of illness within the families of bipolar probands.