Brazilian soils are originally free from soybean bradyrhizobia and the first inoculants were brought to the country in this century, but a search for adapted strains started immediately and still continues. A strain selection program was established at Embrapa based on the reisolation of strains after a long period of adaptation to the soils followed by a search for variant genotypes with higher N2 fixation capacity and competitiveness. A second approach of this program consists of searching for variant colonies of a single strain with higher N2 fixation rates and competitiveness, following a short period of adaptation to the soil. In this study, using both approaches, strains belonging to three serogroups, CB 1809, 532C and SEMIA 5020, were obtained. In general, the variant strains showed differences in colony morphology (mucoidy) but produced similar protein and lipopolysaccharide profiles. Within serogroup CB 1809, containing variants obtained via the second approach, a low level of DNA polymorphism was detected relative to the parental genotype by ERIC and REP-PCR. However, within the two other serogroups, containing variant strains obtained via the first approach, a high level of polymorphism in ERIC and REP-PCR fingerprints was observed relative to the putative serologically related parental genotypes. These results show that a great variability can be detected following adaptation of Bradyrhizobium strains to the soil, although other potential explanations for the DNA polymorphisms observed are discussed. Some of the variant strains obtained by both methodologies were found to have higher rates of N2 fixation and almost all were more competitive than the parental genotypes, suggesting that it is possible to select variant strains which can contribute to an improved plant N nutrition status.