We evaluated epidemiologic evidence pertaining to the human carcinogenic potential of triazine herbicides in general and of atrazine, the most common triazine. Cancers for which data are available included non-Hodgkin's lymphoma, Hodgkin's disease, leukemia, multiple myeloma, soft tissue sarcoma, colon cancer, and ovarian cancer. The investigations had methodologic limitations, including lack of in-depth exposure measurements and small numbers of subjects with heavy exposure and/or with many years since starting exposure, possibly required for the induction of cancer. The relation between triazines and non-Hodgkin's lymphoma has been assessed in four independent population-based case-control studies, reporting odds ratio ranging from 1.2 to 2.5. However, chance and/or confounding by other agricultural exposures may have produced these weak statistical associations. Furthermore, a pooled analysis of three of the case-control studies and the combined analysis of two retrospective follow-up studies did not demonstrate the types of dose-response or induction time patterns that would be expected if triazines were causal factors. The epidemiologic data pertaining to Hodgkin's disease, leukemia, multiple myeloma, soft tissue sarcoma, colon cancer, and ovarian cancer were inadequate for determining whether associations with atrazine or triazines exist in humans. For each of these cancers, only one or two studies evaluating the relationship were available, and the results of the studies typically were imprecise.