Most chemical compounds and physical agents recognized as human carcinogens were first identified in the occupational environment by epidemiological studies. Some 2%-8% of all cancers have been attributed to occupational exposures, but the limitations of such estimates should be recognized. The use of existing medical information systems on occupational cancer for hypothesis generation may be improved by comparing the results emerging from different countries. An initiative in this direction is recommended. Clues emerging from such general systems should be submitted to further studies to test specific hypotheses concerning risk factors. Record linkage is necessary for the epidemiological study of occupational cancer. Efforts must be made towards the storage of identifiable records with information on occupation and cancer occurrence.