The mortality experience of 7,119 workers who were employed at a Beaumont, Texas, refinery for at least 1 year between 1945 and 1987 was investigated. Mortality analyses based on standardized mortality ratios (SMRs) and 95% confidence intervals (95% CI) showed overall mortality was significantly lower than expected compared with the U.S. general population (SMR = 82, 95% CI = 79-86). Total cancer mortality was also lower than expected (SMR = 92, 95% CI = 84-100). Significant mortality deficits from several malignant and nonmalignant diseases were reported. A significant mortality increase in the broad category of lymphatic and hematopoietic cancers was found (SMR = 133, 95% CI = 103-170). This increase was attributed to a nonsignificant elevation in leukemia of all cell types combined (SMR = 139, 95% CI = 92-201) and a borderline significant increase in other lymphatic tissue cancer (SMR = 158, 95% CI = 101-235). The elevation in leukemia was confined to workers hired before 1950. Furthermore, the leukemia excess was shown to have peaked during the 1960s, with mortality no longer elevated post-1980. Analyses of cell type-specific leukemias showed a similar temporal pattern for acute myeloid leukemia (AML) which was not significantly elevated (SMR = 136, 95% CI = 59-268). Mortality from other leukemia cell types was similar to or lower than expected. Mortality from non-Hodgkin's lymphoma (NHL) (SMR = 140, 95% CI = 88-211) and multiple myeloma (MM) (SMR = 121, 95% CI = 55-230) were increased, but neither was statistically significant nor likely to be related to refinery employment. No death from asbestosis was reported, and mortality from mesothelioma and pulmonary fibrosis was lower than expected. Lung cancer mortality for the overall cohort was similar to expected. For the overall cohort, analyses by duration of employment and time since first employment showed no evidence of any trends for increasing cause-specific mortality. Separate analyses of male workers employed in operator jobs showed mortality patterns that were more favorable than those of the total cohort. Maintenance craftworkers showed statistically significant elevations in mortality for prostate cancer (SMR = 145, 95% CI = 107-194), leukemia (SMR = 179, 95% CI = 111-273), and other lymphatic tissue cancer (SMR = 233, 95% CI = 138-368). Detailed analyses indicated that, among maintenance craftworkers, mortality was elevated for AML, NHL, and MM, but none was significant. Furthermore, no upward trend by duration of maintenance jobs was observed. A small increase of lung cancer was observed among maintenance craftworkers (SMR = 120, 95% CI = 99-145), which was borderline significant. No relationship between lung cancer and duration of maintenance employment was found. In contrast, a deficit of pulmonary fibrosis was reported among maintenance craftworkers (SMR = 62, 95% CI = 17-159). These findings are discussed in conjunction with results from other refinery studies, and the limitations of the study are discussed.