Objective To establish the specific 16S-23S rRNA gene spacer regions in different bacteria using polymerase chain reaction (PCR), restriction fragment length polymorphism (RFLP), DNA cloning and sequences analysis. Methods A pair of primers were selected from highly conserved sequences adjacent to the 16S-23S rRNA spacer region. Bacterial DNA from sixty-one strains of standard bacteria and corresponding clinical isolates representative of 20 genera and 26 species was amplified by PCR, and further analyzed by RFLP, DNA cloning and sequences analysis. Furthermore, all specimens were examined by bacterial culturing and PCR-RFLP analysis. The evaluation of these assays in practical clinic practice was also discussed.Results Restriction enzyme analysis revealed one, two or three bands or more observed among the 26 different standard strains. The sensitivity of PCR reached 2.5 colony-forming unit (CFU), and there was no cross reaction with human genomic DNA, fungus or virus. Fourteen species could be distinguished immediately by PCR, while another 10 species were further identified by Hinf Ⅰ or Alu Ⅰ digestion. The only difference between K.pneumoniae and E.durans was located at the site of the 779th nucleotide according to the sequence analysis and only XmaⅢ digestion could distinguish one from another. Of 42 specimens from septicemic neonates, 15 were identified as positive by blood culture at a rate of 35.7%. However, 27 specimens identified as positive by PCR, with a rate of 64.2%, a method significantly more effective than blood culture (P＜0.01). Of 6 cerebrospinal fluid (CSF) specimens, one tested positive for S.epidermidis was also positive by PCR, two culture negative were positive by PCR and diagnosed as S.epidermidis according to the DNA pattern. One positive for C.neoformans was negative by PCR. The other two specimens were negative by both PCR and culture.Conclusions The method of detecting bacterial 16S-23S rRNA spacer regions using PCR-RFLP techniques was specific, sensitive, rapid and accurate in providing a new technique for detecting pathogens in clinical bacterial infections.