Interview information was obtained on primary and secondary occupations for 261 male lung cancer patients and 444 controls residing on the island of Oahu, HI. With the use of a computerized job-exposure matrix (JEM), exposure levels were assigned to each subject for each of 5 known and 2 likely lung carcinogens. A multiple logistic regression model was used to compute the odds ratio (OR) for exposure to each carcinogen, adjusted for age, ethnicity, and pack-years of smoking. On the basis of the criterion of either a dose-response effect or a statistically significant OR (P less than .05) or both, 5 of the 7 carcinogens (4 of the 5 known carcinogens) were found to be associated with lung cancer risk. However, on the basis of the single criterion of a significant OR, only 3 of the 7 carcinogens (2 of the 5 known carcinogens) were associated with lung cancer risk. Due to its relative insensitivity, this JEM may have limited usefulness in the identification of exposure-disease relationships and should not be applied uncritically.