To evaluate PCR-generated artifacts (i.e., chimeras, mutations, and heteroduplexes) with the 16S ribosomal DNA (rDNA)-based cloning approach, a model community of four species was constructed from alpha, beta, and gamma subdivisions of the division Proteobacteria as well as gram-positive bacterium, all of which could be distinguished by HhaI restriction digestion patterns. The overall PCR artifacts were significantly different among the three Taq DNA polymerases examined: 20% for Z-Taq, with the highest processitivity; 15% for LA-Taq, with the highest fidelity and intermediate processitivity; and 7% for the conventionally used DNA polymerase, AmpliTaq. In contrast to the theoretical prediction, the frequency of chimeras for both Z-Taq (8.7%) and LA-Taq (6.2%) was higher than that for AmpliTaq (2.5%). The frequencies of chimeras and of heteroduplexes for Z-Taq were almost three times higher than those of AmpliTaq. The total PCR artifacts increased as PCR cycles and template concentrations increased and decreased as elongation time increased. Generally the frequency of chimeras was lower than that of mutations but higher than that of heteroduplexes. The total PCR artifacts as well as the frequency of heteroduplexes increased as the species diversity increased. PCR artifacts were significantly reduced by using AmpliTaq and fewer PCR cycles (fewer than 20 cycles), and the heteroduplexes could be effectively removed from PCR products prior to cloning by polyacrylamide gel purification or T7 endonuclease I digestion. Based upon these results, an optimal approach is proposed to minimize PCR artifacts in 16S rDNA-based microbial community studies.