Atomic bomb doses in Hiroshima and Nagasaki have been evaluated by many groups. In the 'Hiroshima Atomic Bomb Dosimetry Group' used in this study, radioactivities of l52Eu and 60Co in exposed rock samples have been measured to evaluate neutron doses both in Hiroshima and Nagasaki. These radioactivities were induced by neutron capture reactions. Thermoluminescence dosimetry has also been applied to estimate gammaray doses using exposed tile, roof tile and brick samples. After DS86, our group found systematic discrepancies in measured specific radioactivities and calculation based on DS86. The difference, for example, was that measured data were 5 to 10 times larger than the calculation in Hiroshima. The measured gammaray doses are almost the same as DS86, but they have a similar trend as neutrons, and at 2 km ground range data, are 50-70% larger than the calculation. To find the reason for those problems, some experiments were then carried out by our group. The conclusion of these experiments is that everything seems reasonable except the estimation of neutron and gamma ray spectra at the burst point in Hiroshima. A possible explanation is proposed for the neutron and gamma ray discrepancy in Hiroshima. For this three concepts are assumed as follows: (1) Increase the yield of the atomic bomb about 20%. (2) Leakage of bare fission neutrons from the atomic bomb about 5%, in this model, to the side. That means that for the direction downward, neutrons are shielded with the Hiroshima atomic bomb body. (3) Elevation of burst height about 90 m. By these assumptions all neutron activation data within 1 km ground range and thermoluminescence gamma ray data are simultaneously explained.