A program combining 70,000 purchased and 3,500 random selection-no charge measurements with extensive questionnaires is described. Methods for reducing biases in purchased measurements have been reasonably successful. Distributions are roughly lognormal with σg ≈ 3.0 but with an excess of very high measurements. Winter and spring-fall measurements average about 60 percent and 40 percent, respectively, higher than summer measurements. Basements average 2.5 times higher radon levels than upper floors. Tightening of homes since 1974 has increased radon levels by less than 10 percent, but well-weatherized houses have ∼40 percent higher levels than poorly weatherized houses. Open windows reduce levels by a factor of about 2.5. Low-cost newer houses have much lower levels than more expensive newer houses, but among older houses this is reversed. Low income families have substantially lower radon levels than average. There is little variation with education level. Nonsmokers have higher radon levels than smokers, but this seems to apply mainly to livingrooms and dining rooms. Suburban houses have higher levels than urban, and rural houses have still higher levels. Owner-occupied houses have much higher levels than rented houses. Variations of radon levels with weatherization do not seem to be due to cost factors. A strategy for locating areas with high mean radon levels is described.