The paper by J.S. Puskin (2009) on "Perspective on use of Linear-No Threshold Theory (LNT) for radiation protection and risk assessment by the U.S. EPA" begins its scientific section with the statement "Results from laboratory studies of irradiated animals and epidemiological studies of irradiated human cohorts are generally consistent with a linear, nothreshold dose-response, down to the lowest doses". With regard to animal studies, perhaps the clearest statistically indisputable violations of that statement are in the work at Argonne National Lab on injecting radioactive materials into mice (Finkel and Biskis 1962, 1968 and at Oak Ridge on gamma ray exposure of mice (Ullrich and Storer 1979). Another is tumor induction by irradiation of mouse skin throughout life (Tanooka 2001) where weekly irradiation with 1.5 Gy, 2.2 Gy, and 3 Gy resulted in tumors to 0%, 35%, and 100% of the mice respectively-hardly a linear no threshold response.With regard to human studies, a clear and statistically indisputable demonstration of LNT failure was of bone cancers among dial painters and others occupationally exposed to ingested radium (Evans 1974) where there were no tumors in the several categories exposed to less than 10 Gy, but for dose ranges around 18 Gy, 35 Gy, 75 Gy, and 200 Gy 25% to 38% in each category developed tumors. Another case of LNT failure is the observation that lung cancer mortality in U.S. counties decreases dramatically with increasing mean radon levels in homes, with or without corrections for smoking prevalence and with full consideration of over 500 potential confounding factors (Cohen 1995(Cohen , 2006.The Puskin paper then presents a highly over-simplified theoretical basis for supporting LNT down to the lowest doses-the DNA damage is proportional to the number of radiation hits and hence to the dose. This ignores the many microarray studies (e.g. Yin et al 2005) which show that the genes affected by low level radiation are very different from those affected by the cancer-causing high levels from which risk estimates are derived (Tubiana and Aurengo 2005).