A retrospective cohort study was conducted to examine the mortality experience of 2174 men employed between 1940 and 1978 by a large chemical company and who had been assigned to a chemical production department that used or produced ethylene oxide (EO). Comparisons were made with the general United States population, the regional population, and with a group of 26,965 unexposed men from the same plants. Comparisons with general United States death rates showed fewer deaths than expected in the EO group due to all causes and for total cancers. There was no statistically significant excess of deaths due to any cause. Seven deaths each due to leukaemia and pancreatic cancer were observed with 3.0 and 4.1 deaths expected. Among the subcohort of men who worked where both average and peak exposure levels were probably highest, however, one death due to pancreatic cancer (0.9 expected) and no deaths due to leukaemia were observed. Four of the seven who died from leukaemia and six of the seven who died from pancreatic cancer had been assigned to the chlorohydrin department where the potential for exposure to EO is judged to have been low. The relative risk of death due to each disease was strongly related to duration of assignments to that department. When men who worked in the chlorohydrin department were excluded, there was no evidence for an association of exposure to EO with pancreatic cancer or leukaemia. Together with the failure to show independent EO associations, the chlorohydrin department results suggest that leukaemia and pancreatic cancer may have been associated primarily with production of ethylene chlorohydrin or propylene chlorohydrin, or both. These results emphasise the importance of examining additional concurrent/asynchronous exposures among human populations exposed to EO.