We compared death certificates for asbestos-associated diseases (mesothelioma, lung cancer, asbestosis) in two asbestos workers' cohorts. One (insulation workers) had current or recent employment and a strong, continuing union support system which gave them much information about the effects of asbestos exposure. The second cohort, asbestos factory workers, had no such advantage. The factory had closed almost 30 years before, and its workers had dispersed into many areas of the state and nation. Accuracy of medical diagnosis was comparable in the two groups, but occupational listings were not. Three-quarters of the insulators' death certificates told of asbestos work, while virtually none of the factory workers' certificates provided such information, even for deaths of mesothelioma and asbestosis. The data indicate that disease categories, based on medical and pathological diagnoses, at least for asbestos-associated disease, tend to be accurate. Attempts to identify groups at risk by sorting occupational categories can give variable results, good for those with current exposures, much less satisfactory for those with long-past occupational exposures.