At low doses of radiation to the whole body, the dose-effect relationship for human must usually be determined by extrapolation from human data at high doses. Reasons for above-linear, linear, or below-linear extrapolation from high doses to low doses are discussed. The linear extrapolation is most common, and probably leads to conservative estimates of risk. Risks from other carcinogens may be compared directly with radiation risk by using cancer incidence as the end point. However, lifeshortening is a more useful index of comparison in many other circumstances, and can be used to compare the time lost due to radiation-induced cancer with that due to accidental deaths and other lost industrial time. Risks for radiation workers exposed at current average occupational dose levels are shown to be comparable with those from other safe industries. Questions of acceptability and public perception of risk are briefly discussed.