In this study, the authors examined the relationship between germination percentage and three-dimensional shape factors of soybean seed in population (Variety: Enrei).1) The group of 8.0-9.0mm classified by sieving was a majority in particle mass distribution of the population of 180kg seeds. In analysis on the reduced sample (600 seeds) selected randomly from the population, the data of the shape factors of seeds in population were regarded as normal distribution.2) A particle mass had high positive correlation with length, surface area and volume of seed, and each linear function was shown in the tested 600 seeds. Particle density had negative correlation with other shape factors of seed.3) Grouping classified by sieving did not effect on both germination percentages on field (GF) and on Petri dish (GP). This result was explained by some overlappings with the neighboring groups. Further, the GF data arranged in order of a particle mass were analyzed, but there were no significant differences between germinated group and ungerminated group by the t-test and the F-test.4) In analysis on the 600 data arranged in order of a particle mass, the germination percentage of heavy mass seed group was higer than that of other groups. But the value was changed by grouping method.5) An additional germination test on the Petri dish was practiced with an extended range of a particle mass and without the overlapping. Arranging in order of germination percentage, the result was heavy mass group>middle mass group>light mass group with the significant difference at 5% level.6) To use only heavy mass seeds is not effective judging from the characteristics of particle mumber distribution. In practice, we propose to use seeds which are in the mean range of a particle mass in population at first, and use seeds in the range extended around the mean at second.