Homogenates from soybean nodules, formed by 12 strains of Bradyrhizobium japonicum, were plated into yeast-extract mannitol agar containing 3 or 37 g mannitol 1-1. Viable counts ranged from 8.298 to 11.265 log10 cells-gram nodule-1. When monitored over the life cycle of the symbiosis, the viability of strains USDA 110 and USDA 123 increased with days after planting (DAP), and at 70 DAP was 95% and 81%, respectively. By contrast, the viability of USDA 38 bacteroids decreased with time, and at 70 DAP was only 1.9%. At 49 DAP, nodules induced by USDA 38 had significantly fewer bacteroids per peribacteroid membrane than those formed by USDA 110 or USDA 123, and at 70 DAP, 27% of the USDA 38 bacteroids showed some degree of degeneration. Viable counts of USDA 123 and USDA 110 bacteroids, isolated from the nodules of 12 different cultivars, ranged from 10.963 to 11.463 and from 10.683 to 11.117 log10 viable cells-gram nodule-1, respectively. Varying the osmolarity of the medium had no predictable effect on bacteroid viability. When surface-sterilized nodules of IPAGO 587 (high bacteroid viability) and USDA 38 (low bacteroid viability) were inoculated into a nonsterile silt loam soil, at rates equivalent to 5.0×108 and 5.0×106 viable bacteroids g-1 soil, respectively, and then incubated at 28° C for 60 days, 4.3×104 and 1.5×104 surviving cells g-1 soil, respectively, were recovered. Thus, despite differences due to host and strain variation, bacteroid viability appears to be unrelated to persistence of individual strains following an annual legume crop cycle.