The 1952-88 cancer mortality records for inhabitants of the Misasa spa area, Japan, which has a high radon background, and a neighboring control area without any radon spa were analyzed (average outdoor Rn concentration: 26 mBq.liter-1 in Misasa vs. 11 mBq.liter-1 in the control area). Standardized mortality ratios (SMRs) for cancers of all sites were significantly lower among the inhabitants of both Misasa (male 0.538; female 0.463) and the control area (male 0.850; female 0.770), than in the whole Japanese population. Poisson regression analysis showed that the relative risks among the inhabitants of Misasa were significantly lower than in the control area for deaths from cancers of all sites (0.67) and stomach cancer (0.59). The relative risk of lung cancer death was also lower (0.55 times) in Misasa than in the control area, although the difference was not statistically significant. These results suggest that the linear no-threshold hypothesis for radiation risk may not be valid for exposure to low doses of radon.