This paper summarizes genetic and somatic data on persons exposed to low doses of atomic bomb radiation in Hiroshima and Nagasaki, Japan. Compared with experimental estimates, the new dosimetry system proposed in 1986 underestimates neutron doses, supporting qualitatively the conclusion by the 1965 dosimetry system that Nagasaki A-bomb emitted predominantly gamma rays whereas Hiroshima A-bomb emitted both gamma rays and fast neutrons. A theory based on two recessive mutations in hemopoietic stem cells is proposed to explain radiation leukemogenesis. The theory can explain, at least partly, the acutal dose-response curve for incidences of acute leukemia in Hiroshima but cannot explain chronic leukemia in Nagasaki. Existence of a large threshold dose in the latter''s dose relationship supports the hypothesis that A-bomb radiation at high doses above a threshold value was a promoter and/or progressor of leukemia. Various lines of evidence that support this hypothesis are presented. Hence, it is not warranted to assume that risk of death from cancer at a high dose, say, 1 Gy can be divided by 100 to obtain the risk at 1 cGy. Risk at low doses should be assessed by direct scrutiny of actual data at low doses in spite of their large statistical uncertainty. Actual data show that A-bomb survivors at 1-9 cGy had apparently lower incidences of tumors than unexposed persons.