As part of a case-control study conducted to test the association between indoor air radon levels and lung and other cancers, 539 air and 547 water radon samples were measured from homes with drilled wells throughout Maine. The factors tested for an association with air radon levels were season, house construction, draftiness of house, heating source, basement construction, radon levels in the water supply, and the type of overburden and bedrock underlying the house. Logistic regression was used to determine which variables were significantly associated with indoor air radon levels of 4 pCi/L or above and to assess their relative importance. House construction, water radon levels, and overburden permeability showed the strongest association with air radon levels. Woodframe homes were 10 times more likely than mobile homes to have an air radon level of 4 pCi/L or more (95 percent confidence interval [CI] = 1.1 to 103), and brick or stone houses were 108 times more likely to have high air radon levels (CI = 6.7 to 1,760). Homes with water radon between 10,001 and 100,000 pCi/L were 6 times more likely to have high air radon (CI = 1.6 to 22.7), and homes with water radon levels above 100,000 pCi/L were 54 times more likely to have air levels of 4 pCi/L or above than were homes with water radon at or below 1,000 pCi/L (CI = 1.9 to 1,529). Homes built over sand and gravel were 12 times more likely to have high air radon than homes built over clay (CI = 4.1 to 37.1), and homes built over two-mica granite were 4 times more likely to have high air radon than homes built over low-grade metamorphic rock (CI = 1.4 to 11.5). In addition, high air radon levels were more likely in the winter than in the summer (odds ratio [OR] = 2.9, CI = 1.3 to 6.2) and in homes heated with electricity than with other sources of heat (OR = 3.5, CI = 1.4 to 8.8).