There is little doubt that underground miners exposed to radon and its progeny have increased rates of lung cancer. Residential radon exposures should carry a possibly smaller risk of increased cancer. When it became possible to collect radon data in a large number of U.S. homes and the data were aggregated by counties, the apparent association with lung cancer was a negative one, even when many other variables were taken into account. Residential radon levels are higher in suburban residences leading to a negative association with population density. Population density is strongly positively associated with lung cancer. It follows that aggregate residential radon and lung cancer rates should be negatively associated for reasons having nothing to do with the possibility of radon being carcinogenic to the lung. A second problem presented by the data is the one of sampling bias since the county lung cancer data are from the whole county population, but only a few residences are tested. Examples of other inherent associations in environmental epidemiology are cited. One strategy is to study areas of the same population density but with radon exposure gradients. This is approximated by choice of rural high radon states. Counties in such states have weak and inconsistent associations between radon and lung cancer, some of which are positive. I conclude that counties are generally inappropriate units for study of radon and lung cancer associations.