In the last few years many attempts have been made to differentiate more than 20 Bijidobacteriurn species. It has been recognized that identification of bifidobacterial species is problematic because of phenetic and genetic heterogeneities. In order to contribute to our understanding of BiJdobacterium taxonomy, we studied Bijidobacterium phylogeny by performing both 16s rRNA and 16s to 23s (16s-23s) internally transcribed spacer (ITS) sequence analyses. In this study, we determined 16s rRNA sequences of five Bifidobacteriurn strains representing four species, and compared them with the sequences available in the GenBank database, and used them to construct a distance tree and for a bootstrap analysis. Moreover, we determined the ITS sequences of 29 bifidobacterial strains representing 18 species and compared these sequences with each other. We constructed a phylogenetic tree based on these sequence data and compared this tree with the tree based on 16s rRNA sequence data. We found that the two trees were similar topologically, suggesting that the two types of molecules provided the same kind of phylogenetic information. However, while 16s rRNA sequences are a good tool to infer interspecific links, the 16s-23s rDNA spacer data allowed us to determine intraspecific relationships. Each of the strains was characterized by its own ITS sequence; hence, 16s-23s rRNA sequences are a good tool for strain identification. Moreover, a comparison of the ITS sequences allowed us to estimate that the maximum level of ITS divergence between strains belonging to the same species was 13%. Our data allowed us to confirm the validity of most of the Bzjidobacterium species which we studied and to identify some classification errors. Finally, our results showed that Bifidobacterium strains have no tRNA genes in the 16s-23s spacer region.Members of the genus Bijidobacterium are widespread in nature, and the habitats of these organisms range from sewage (53, 54) to human and animal intestines (28,31,44,52,56). The ability to catabolize hexose by a particular pathway via fructose phosphate phosphoketolase is important for recognizing members of this genus (8, 51). In the last few years many ways to differentiate Bifidobacterium species have been developed. Numerical taxonomy analysis, which requires data for many characteristics, has been used to circumscribe clusters and to describe strains. Morphological traits, carbohydrate metabolism data, DNA G + C contents (3), electrophoretic patterns (22,48,50), serologic data (44, 58, 59), cell wall compositions (22), and rRNA gene restriction patterns (26) have been used to subdivide Bifidobacterium species (for a review, see reference 3). Major advances in our understanding of the taxonomy of members of the genus Bijidobacterium have come from the DNA-DNA hybridization studies performed by Scardovi et al. (49,54,55,57). However, problems with the identification of Bijidobacterium species are compounded by evidence that there is phenotypic and genetic heterogeneity in these species (6). Moreo...