In the attempt to explore complex bacterial communities of environmental samples, primers hybridizing to phylogenetically highly conserved regions of 16S rRNA genes are widely used, but differential amplification is a recognized problem. The biases associated with preferential amplification of multitemplate PCR were investigated using 'universal' bacteria-specific primers, focusing on the effect of primer mismatch, annealing temperature and PCR cycle number. The distortion of the template-to-product ratio was measured using predefined template mixtures and environmental samples by terminal restriction fragment length polymorphism analysis. When a 1 : 1 genomic DNA template mixture of two strains was used, primer mismatches inherent in the 63F primer presented a serious bias, showing preferential amplification of the template containing the perfectly matching sequence. The extent of the preferential amplification showed an almost exponential relation with increasing annealing temperature from 47 to 61 degrees C. No negative effect of the various annealing temperatures was observed with the 27F primer, with no mismatches with the target sequences. The number of PCR cycles had little influence on the template-to-product ratios. As a result of additional tests on environmental samples, the use of a low annealing temperature is recommended in order to significantly reduce preferential amplification while maintaining the specificity of PCR.