Men assigned to the chlorohydrin unit of Union Carbide's South Charleston plant in the Kanawha Valley of West Virginia were followed up for mortality from 1940 to the end of 1988. This 10 year update was conducted to verify previous findings of excesses of cancer among the 278 men assigned to the chlorohydrin unit, which primarily produced ethylene chlorohydrin from 1925 to 1957. This process produced ethylene dichloride and bischloroethyl ether as byproducts. Mean duration of assignment was 5.9 years and mean duration of follow up was 36.5 years. Standardised mortality ratios (SMRs) were calculated based on comparisons with the United States white male population. Duration-response trends were assessed by internal comparisons with two different groups of unexposed chemical workers in the Kanawha Valley. The evidence that the earlier finding of an excess of pancreatic cancer was work related is strengthened by the occurrence of two additional cases (0.9 expected). The SMR for pancreatic cancer was 492 (95% CI 158-1140), based on eight observed v 1.6 expected deaths. There were no additional deaths due to leukaemia, but the three to four-fold excess risk for lymphopoietic cancers persisted due to new cases of non-Hodgkin's lymphoma and a death from multiple myeloma. The SMR for lymphatic and haematopoietic cancers was 294 (eight observed v 2.7 expected; 95% CI 127-580). Pronounced increases in risk were seen for total cancer, pancreatic cancer, all lymphatic and haematopoietic cancers, and leukaemia with increasing durations of assignment to the chlorohydrin unit. Most of the cases were first assigned to the unit in the 1930s when chemical manufacturing was in its infancy and exposures were less controlled. These data are insufficient to identify conclusively the causative agent or agents. The weight of evidence, however, based on probable exposure, known toxicity of the chemicals, and animal responses suggest that high exposures to ethylene dichloride, perhaps in combination with other chlorinated hydrocarbons, is the most likely explanation.