CE fingerprint methods are commonly used in microbial ecology. We have previously noticed that the position and number of peaks in CE-SSCP (single-strand conformation polymorphism) profiles depend on the DNA polymerase used in PCR . Here, we studied the fragments produced by Taq polymerase as well as four commercially available proofreading polymerases, using the V3 region of the Escherichia coli rss gene as a marker. PCR products rendered multiple peaks in denaturing CE; Taq polymerase was observed to produce the longest fragments. Incubation of the fragments with T4 DNA polymerase indicated that the 3'-ends of the proofreading polymerase amplicons were recessed, while the Taq amplicon was partially +A tailed. Treatment of the PCR product with proofreading DNA polymerase rendered trimmed fragments. This was due to the 3'-5' exonuclease activity of these enzymes, which is essential for proofreading. The nuclease activity was reduced by increasing the concentration of dNTP. The Platinum Pfx DNA polymerase generated very few artifacts and could produce 85% of blunted PCR products. Nevertheless, despite the higher error rate, we recommend the use of Taq polymerase rather than proofreading in the framework for molecular fingerprint studies. They are more cost-effective and therefore ideally suited for high-throughput analysis; the +A tail artifact rate can be controlled by modifying the PCR primers and the reaction conditions.