A multicentre cohort study was carried out to study the possible association between exposure to ethylene oxide and cancer mortality. The cohort consisted of 2658 men from eight chemical plants of six chemical companies in the Federal Republic of Germany who had been exposed to ethylene oxide for at least one year between 1928 and 1981. The number of subjects in the separate plants varied from 98 to 604. By the closing date of the study (31 December 1982) 268 had died, 68 from malignant neoplasms. For 63 employees who had left the plant (2.4%) the vital status remained unknown. The standardised mortality ratio for all causes of death was 0.87 and for all malignancies 0.97 compared with national rates. When local state rates were used the SMRs were slightly lower. Two deaths from leukaemia were observed compared with 2.35 expected (SMR = 0.85). SMRs for carcinoma of the oesophagus (2.0) and carcinoma of the stomach (1.38) were raised but not significantly. In one plant an internal "control group" was selected matched for age, sex, and date of entry into the factory and compared with the exposed group. In both groups a "healthy worker effect" was observed. The total mortality and mortality from malignant neoplasms was higher in the exposed than in the control group; the differences were not statistically significant. There were no deaths from leukaemia in the exposed group and one in the control group.