The mortality experience of a cohort of 3 297 railroad maintenance shopworkers exposed to asbestos was investigated. The study period was 1951-1980, and the vital status was assessed for 99.6% of the men. Individual estimates of cumulative asbestos exposure were based on detailed records on work tasks and divisions. Dust measurements were scanty in earlier decades, and estimates of average fiber levels were therefore based on information on the amount and kind of asbestos used, job descriptions, and interviews with older workers. The overall mortality was lower than expected from the national death rates (standardized mortality ratio = 82). The mortality from lung cancer increased as cumulative exposure increased in consistent dose-response relationships. Employment times of less than 30 years in workplaces with moderate levels of mainly chrysotile asbestos was not associated with any apparent increase in the risk of lung cancer. A subgroup exposed for more than 30 years in workplaces repairing steam engines, where amphiboles were used as well, had a standardized mortality ratio of 192 for lung cancer. This figure may be an underestimation due to healthy worker selection and fewer smokers than normal. The "true" standardized mortality ratio was estimated to be about 300. Five cases of mesothelioma were observed.