Previously reported body mass estimation formulas based on femoral osteometry are not suitable for the physique of Japanese people before the 1960s due to excessive errors. Therefore, based on the measurements of Japanese skeletons in the first part of 20th century, the author attempted to develop body mass estimation formulas with small errors, considering application to ancient human remains. First, the author applied the body mass estimation formulas previously devised based on the somatometric data of temperate Asian groups including traditional society, Japanese people in the 10 years after the Pacific War, and those in the 1990s, to anatomical bleached specimens from Kinai and Kanto regions, and thereby reconstructed antemortem body mass. In this calculation, the head measurements including the skin thickness and the mean values of the estimated statures based on the maximum length of each limb bone were substituted to these formulas. Assuming that the average values of the estimated body mass obtained from these formulas were the actual antemortem body mass of each individual, multiple regression analysis was performed for each part based on the cranial and limb bone measurements so that it can also be used for scattered skeletal parts. Although this analysis was performed in both sex-specific and combined cases, the multiple correlation coefficients were higher in combined cases at all parts, and the standard error of estimation was suppressed to less than 3 kg. For the upper limbs, the estimation formulas using the diaphyseal diameters or circumferences of the radius and humerus were, while for the lower limbs, the formulas for the tibia and fibula in the lower thighs were more accurate. In particular, the author confirmed that the errors became smaller in the distal bones, and the accuracy of the estimation formulas including the knee joint breadth increased in the lower limb bones.