The genetic diversity of a vesicular stomatitis virus population was analysed by RT-PCR, cloning and sequencing of two approximately 500 nucleotide regions of the virus genome. PCR amplifications were performed in parallel experiments with both Taq and Pfu DNA polymerases, and important differences were observed. Between 10 and 22 mutations were detected when virus populations were analysed by Taq amplification (20 clones from each region), whereas amplification of the same samples with Pfu revealed between 0 and 5 mutations. PCR fidelity assays, performed under the same PCR conditions as those used in the population analysis, showed that the Taq error-rate estimate of 0.27 x 10(-4) misincorporations per bp per cycle was within the range estimated elsewhere from PCR amplification of recombinant plasmids (0.27-0.85 x 10(-4) errors per bp per cycle) or from functional assays (0.2-2 x 10(-4) errors per bp per cycle). The error rate of Taq was found to be 9.3 times higher than the error rate of Pfu with DNA as a template, and about 10 times higher with cDNAs obtained by reverse transcription of viral RNA templates from natural populations. In the present study, we discuss (i) the implications of Taq errors on the analysis of genetic variability, based on both the frequency and nature (replacement vs synonymous) of the observed substitutions and (ii) the sample size required to assess the genetic variability in a virus population generated by a single infection.