There is no evidence concerning human carcinogenic potential of most substances to which workers are exposed. A case-control type method has been developed to discover heretofore unsuspected occupational carcinogens. Incident cases of cancer of several sites are ascertained and the subjects are interviewed to obtain detailed job histories. The job histories are subsequently examined by chemists and engineers who infer for each job a list of potential chemical and physical exposures. These lists of exposures then become the basis of case-control type statistical analyses, wherein each site included and each exposure noted in any history are evaluated for possible association. Controls can be selected in the general population, among non-cancer hospital patients or, as we do, among other cancer patients. Preliminary findings from a pilot study in Montreal show that 1) an efficient case ascertainment system and an acceptable interview procedure can be set up, 2) job histories provided by respondents are quite valid, 3) chemists' translations of job histories into histories of chemical exposures seem to be meaningful and useful, and 4) at least one carcinogenic agent, cigarette smoking, is detectable whether a population control series or a cancer control series is used.