We examined the epidemiology of chronic beryllium disease among a stratified, random sample (n = 895) of nuclear weapons workers using the blood beryllium lymphocyte transformation (BeLT) test and chest radiograph for case identification. Of 18 new cases of beryllium sensitization, 12 had beryllium disease, and three more developed pulmonary granulomas on lung biopsy over the succeeding 2 yr. Beryllium-sensitized cases did not differ from noncases in age, gender, race, ethnicity, smoking, most respiratory symptoms, spirometric or radiographic abnormalities, or job tenure. The six sensitized cases without initial disease differed from beryllium disease cases in having greater pack-years of smoking. Sensitization occurred among workers with inadvertent or bystander exposure, such as a secretary and security guard. However, beryllium sensitization risk was higher for machinists (4.7%) and for persons reporting measured overexposure (7.4%, odds ratio 5.1); exposure beginning before 1970 (3.6%, odds ratio 2.7); consistent beryllium exposure (3.4%); and sawing (4.7%) or band sawing (6.0%) of beryllium metal. We conclude that both individual susceptibility to sensitization and exposure circumstances are important in developing disease.