Unless allowances are made, random errors in radiation dose estimates cause underestimation of linear risk estimates and distort the shape of dose-response curves. These errors also result in spurious associations between radiogenic endpoints, exaggerating possible variations in individual sensitivity to radiation. Statistical methods have been developed which reduce these biases, based on assumptions regarding the nature and magnitude of dose-estimation errors. Some understanding of the underlying statistical basis for these methods is necessary to both those interested in interpreting radiogenic effects and those interested in the dosimetry system. This paper discusses the basic statistical issues and their implications, presents some statistical methods to deal with the problem, and indicates the sensitivity of certain results to assumptions about the magnitude of the dose-estimation errors.