The spread of transgenes into the genome of wild soybean is a concern when transgenic and wild soybeans are planted sympatrically. The objectives of this study were to investigate the origin and fate of morphological intermediates between wild and cultivated soybeans in their natural habitats in Japan. Twenty nuclear microsatellite and two chloroplast dCAPS markers were used to evaluate genetic variation of 468 wild, 17 intermediate, and 12 cultivated soybean samples collected from six sites between 2003 and 2006. Allelic differentiation of microsatellite markers between wild and cultivated soybeans was sufficient to detect their hybrids. Based on levels of observed heterozygosity, intermediate soybean plants were from two generations: either F1 or an early segregating generation. Genetic admixture analysis and parentage assignment analysis revealed that the parents of all intermediate soybean plants could be assigned to a particular wild soybean plant and late-maturing cultivar. The chloroplast DNA haplotypes revealed that all intermediate soybean plants originated from gene flow from cultivated to wild soybeans at all sites. Based on monitoring at both the phenotypic and molecular levels, hybrids quickly disappeared from natural habitats, and secondary gene flow from these plants to wild soybean was not detected. Thus, while gene flow from transgenic soybean into wild soybean can occur, gene introgression appears to be rare in natural habitats in Japan. This is the first report on the detection of gene flow from cultivated to wild soybean at the molecular level.