This case-control study examined the relationship between lung cancer and the work histories of male employees at a large Texas refinery. The study included 112 lung cancer deaths observed between 1946 and 1987 and 490 matched controls. Employment histories were obtained from personnel records, and smoking information was available from medical records. Both stratification methods and conditional logistic regression were used in data analyses. Overall employment in four general job categories (administrative, engineering/laboratory, process, maintenance/mechanical) was not associated with lung cancer mortality. Results by hire period (< 1940, 1940+) showed that workers hired into process jobs before 1940 had a nonsignificantly elevated odds ratio (OR) of 1.71 (95% confidence interval [CI] = 0.85-3.45) compared with nonprocess workers hired before 1940. Among process workers hired before 1940, there was a significant trend toward increasing OR with increasing duration of employment in process jobs, and the association with lung cancer was strongest among smokers in the highest duration category of 30+ years (OR = 2.98, 95% CI = 1.07-8.31). Latency analyses of process workers hired before 1940 indicated that their lung cancer risk had peaked between 30 and 50 years since first employment. Definitive statements about causal factors are limited because results among process workers were based on small numbers of subjects in some exposure categories, and there was no information on specific workplace exposures. The OR for maintenance/mechanical jobs after adjustment for smoking was 1.00 (95% CI = 0.55-1.82). Furthermore, there was no pattern in relation to duration of employment in maintenance/mechanical jobs. The results from this study do not support the hypothesis that work in maintenance/mechanical jobs increases lung cancer risk. On the basis of analyses in this study, it is unlikely that asbestos exposure contributed to excess lung cancer mortality. Additional analyses were conducted for specific maintenance jobs with potential exposure to asbestos and by duration in jobs with occasional or routine asbestos exposure. No significant increase in lung cancer was found in any subgroup. Furthermore, there was no significant trend toward lung cancer risk in relation to duration of employment in jobs with asbestos exposure.