Some of the problems in determining the cancer risk of low-level radiation from studies of exposed groups are reviewed and applied to the study of Hanford workers by Mancuso, Stewart, and Kneale. Problems considered are statistical limitations, variation of cancer rates with geography and race, the "healthy worker effect," calendar year and age variation of cancer mortality, choosing from long lists, use of proportional mortality rates, cigarette smoking-cancer correlations, use of averages to represent data distributions, ignoring other data, and correlations between radiation exposure and other factors that may cause cancer. The current status of studies of the Hanford workers is reviewed.