Terminal restriction fragment length polymorphism (T-RFLP) analysis is commonly used to analyze microbial communities, including oral microflora. However, accurate identification of terminal restriction fragment (T-RF) origins is prevented by unpredictable errors in sizing, thus necessitating the clone library analysis. To minimize sizing errors, we proposed optimizing the size definition of internal standards. GeneScan-1000 ROX was regenerated as an internal standard by redefining the fragment sizes in terms of molecular weight (MW) based on their mobility relative to 6-carboxyfluorescein (FAM) -labeled restriction fragments derived from the 16S recombinant RNA gene of Porphyromonas gingivalis. Using the new size definition, the average sizing error among eight oral bacteria from six phyla was estimated and compared with that of the conventional method. Microbial communities isolated from saliva were analyzed using the new MW size definition. Bacterial species were assigned to peaks using TRFMA, a Web-based tool for T-RFLP analysis, and compared with those identified in a clone library analysis. Using the new size definition, the average sizing error for 40 T-RFs was drastically reduced from 2.42 to 0.62 bases, and large sizing errors (more than two bases) were eliminated. More than 90% of the total bacterial clones detected by the clone library analysis were assigned by T-RFLP. The size definition of the newly constructed internal standards reduced fragment sizing errors and allowed for accurate assignment of bacteria to peaks by the T-RFLP analysis. This provided a more effective means for studying microbial communities, including the oral microflora.