Epidemiologic research has played an important role in identifying and confirming the carcinogenicity of chemicals. This was facilitated in the past by the high levels of exposure and by the simple correspondence between one exposure and one or few job titles. Today we are faced with a different picture: workers are exposed to complex mixtures, exposure levels have decreased while mobility of the work-force has increased. In this changing context, epidemiology needs new methodological tools to improve the validity of risk assessment. The most promising of these seem to be the new techniques of biochemical epidemiology and the development of job-exposure matrices. Among the more traditional tools, record-linkage studies may be relevant in identifying long-term occupational hazards. The methodological problems and the ethical and operative implications of these trends in research are discussed.