A retrospective cohort study of 14,457 workers at an aircraft maintenance facility was undertaken to evaluate mortality associated with exposures in their workplace. The purpose was to determine whether working with solvents, particularly trichloroethylene, posed any excess risk of mortality. The study group consisted of all civilian employees who worked for at least one year at Hill Air Force Base, Utah, between 1 January 1952 and 31 December 1956. Work histories were obtained from records at the National Personnel Records Centre, St. Louis, Missouri, and the cohort was followed up for ascertainment of vital state until 31 December 1982. Observed deaths among white people were compared with the expected number of deaths, based on the Utah white population, and adjusted for age, sex, and calendar period. Significant deficits occurred for mortality from all causes (SMR 92, 95% confidence interval (95% CI) 90-95), all malignant neoplasms (SMR 90, 95% CI 83-97), ischaemic heart disease (SMR 93, 95% CI 88-98), non-malignant respiratory disease (SMR 87, 95% CI 76-98), and accidents (SMR 61, 95% CI 52-70). Mortality was raised for multiple myeloma (MM) in white women (SMR 236, 95% CI 87-514), non-Hodgkin's lymphoma (NHL) in white women (SMR 212, 95% CI 102-390), and cancer of the biliary passages and liver in white men dying after 1980 (SMR 358, 95% CI 116-836). Detailed analysis of the 6929 employees occupationally exposed to trichloroethylene, the most widely used solvent at the base during the 1950s and 1960s, did not show any significant or persuasive association between several measures of exposure to trichloroethylene and any excess of cancer. Women employed in departments in which fabric cleaning and parachute repair operations were performed had more deaths than expected from MM and NHL. The inconsistent mortality patterns by sex, multiple and overlapping exposures, and small numbers made it difficult to ascribe these excesses to any particular substance. Hypothesis generating results are presented by a variety of exposures for causes of death not showing excesses in the overall cohort.