Historical cohort studies conducted in highly industrialized countries were discussed with regard to feasibility, methodology, and problems contributing to invalidation. A cohort study of the mortality rates of uranium miners in the United States conducted in the 1950s was used to illustrate the use of cohorts in testing hypotheses. The study design, cohort definition, follow up and analysis were described. Parameters involved in selecting an appropriate cohort were discussed in relation to: defining the epidemiological characteristics of the cohort, sources of information about the agent of interest, going into the field, assessment of exposure, the adequacy of personnel records, the adequacy of study size and statistical power, the importance of sufficient latency, and the length of exposure. The methodology of cohort studies, after identification and definition of the cohort, was described. Strategies for vital status follow up, the use of death certificates, and the selection of comparison disease rates were discussed. The calculation of mortality and morbidity ratios was considered, and examples were presented of calculations to determine person years at risk. The use of standardization to avoid confounding, the need to categorize exposure and latency periods, and the importance of selecting appropriate calendar time periods were illustrated. Statistical testing of the null hypothesis and available computer software were described. Potential problems of cohort studies including: misleading comparison of standardized mortality ratios, the healthy worker effect, immortality due to minimum person years at risk, retired worker bias, and complications due to smoking were discussed. Finally, methods other than cohort studies used to examine occupational mortality were listed.