Oral streptococci formerly classified as Streptococcus sanguis or Streptococcus mitis have recently been divided into four species. Two additional species have also been proposed for this group. Each species is genetically distinct, but they have many traits in common, which makes it difficult for clinical isolates to be identified by phenotypic tests. Genotypic comparison may provide an alternative approach. This study used DNA fingerprint analysis for comparison of genotypes of 21 reference strains-classified as Streptococcus gordonii, Streptococcus sanguis, Streptococcus oralis, “Streptococcus parasanguis”, or “Streptococcus crista” in previous DNA hybridization studies-and 17 clinical and laboratory strains placed in those groups on the basis of phenotypic tests. HinDIII and PuuII digests were run in conventional horizontal agarose gels. SfiI digests of reference strains and two laboratory strains were run in field-inversion gels. Fingerprint patterns were compared by visual examination, cluster analysis of densitometric traces, and lane-matching software. Only two “S. crista” strains and two parent mutant lineages showed fingerprint patterns that were identical by visual examination. Fingerprint patterns of all other strains were unique. Cluster analysis results could not be considered valid, since replicate lanes in different gels were not grouped together. HinDIII and PvuII digests contained too many bands for correct matching by lane-matching software. SfiI digests were correctly matched by computer, with the same results as visual examination. Results indicate that the diversity of strains within these streptococcal species was too great to permit species identification by DNA fingerprint patterns. This genotypic diversity suggests that isolates from unrelated hosts may have been separate for long periods of time. Fingerprint patterns were useful for strain identification, and this may facilitate studies of the ecology and transmission of oral streptococci. Field-inversion electrophoresis of SfiI digests may be particularly suitable for that purpose, since computers can be used for comparison of large numbers of strains.