Since 1979 our group has been engaged in a case-referent study of 12 tumour types (oesophagus, stomach, colon, rectum, liver, pancreas, lung, melanoma of skin, prostate, bladder, kidney, lymphoid tissue) and occupational exposure. The study covers the entire Montreal population of males aged 35-70; cases are ascertained in all major hospitals. The interview concerns various sociodemographic variables and a detailed job history. Each job history is evaluated by a team of chemists who infer for each job what the possible physical and chemical exposures were. The list of exposures thus inferred is added to the subject's data file. Each site and each exposure noted in any patient's history are statistically evaluated for possible association. Until June 1982, 1 343 cancer cases and 144 general population subjects had been thus interviewed. For the present purpose, all 1 487 files were reviewed to assign exposure to nickel, and nickel exposure was attributed to 79 subjects. Mantel-Haenszel analyses were carried out comparing each site with a referent group consisting of all other interviewees (other sites and healthy controls). Only the nickel-lung association turned out to be remarkable, with an odds ratio of 3.1, 95% confidence limits 1.9-5.0, and some indication of dose-response effect. The risk was particularly high among stainless-steel welders, though other nickel-exposed workers were also at risk. These findings probably reflect a real risk; however, we cannot be certain that nickel is responsible. Exposure to chromium compounds was so highly correlated with nickel that the observed nickel association may simply have reflected a confounding effect of chromium. Confounding with other occupational exposures is also possible.