Lung cancer risk in relation to indoor radon was examined in three case-control studies in Stockholm (Sweden), New Jersey (United States), and Shenyang (People's Republic of China). Year-long measurements of radon gas were made in current and past homes of 966 women who developed lung cancer and of 1,158 control women, included in the combined analysis. Nearly 14 percent of the participants were estimated to have a time-weighted, mean, radon concentration in their homes of more than 4 pCi/l (150 Bq/m3) during the period from five to 35 years prior to the date of lung cancer diagnosis (or comparable date for controls). There was a tendency for risk to increase with increasing levels of radon in NJ and Stockholm, but the trends for individual studies and overall were not statistically significant. The estimates of the excess relative risk for indoor exposure per pCi/l were 0.18 (95 percent [CI] = -0.04-0.70) in NJ, 0.06 (CI = -0.05-0.34) in Stockholm, and -0.02 (CI = -infinity-0.03) for Shenyang; these estimates did not differ significantly from each other. The overall excess RR per pCi/l was 0.00 (CI = -0.05-0.07); the confidence limits were sufficiently broad, however, that the overall estimate was still compatible with extrapolations of risks from miners. Cigarette smoking was the predominant cause of lung cancer with the RR significantly elevated in all studies. Within smoking categories, the trend in risk with increasing mean radon concentration was inconsistent. Analyses of data from several studies are complicated by the possibility that there may exist important differences in study bases which might affect results, and which may be controlled only partially through adjustment procedures. Future efforts to combine various residential studies will need to be attentive to the intrinsic limitations of studies to detect low levels of risk as well as the unique uncertainties associated with estimating, accurately, cumulative exposure to indoor radon.