Community profiles based on terminal-restriction fragment length polymorphism (T-RFLP) analyses of amplified ribosomal RNA genes are used to monitor changes in microbial community structure and are sometimes employed for semi-quantitative estimates of species richness and abundance in environmental samples. To assess the accuracy of T-RFLP community profiles representing the relative abundance of bacteria in a sample, five species of ruminal bacteria were used to construct simple "communities". Template DNA for PCR amplification was generated either by mixing equal quantities of genomic DNA from pure cultures or by mixing equal numbers of cells prior to DNA extraction. Pairwise mixtures of Fibrobacter succinogenes S85 with Ruminococcus albus 8, Ruminococcus flavefaciens FD-1, Butyrivibrio fibrisolvens 49 and Streptococcus bovis JB1 were created and a 5-member community was constructed. With genomic DNA mixes, relative abundance calculations based on T-RFLP patterns did not reflect input ratios. These discrepancies could not be accounted for by differences in genome size and rRNA operon copy number. In cell mixing experiments, easily lysed cells were overrepresented. To determine if a numerical correction factor could be used to compensate for observed discrepancies, we attempted to quantify biases attributed to DNA extraction and PCR amplification. Biases attributable to these factors led to deviations from expected PCR product ratios by 6% to 38%. We found that interactions were so complex that a suitable factor could not be derived. The unsystematic dependence of T-RFLP peak ratios on variability of DNA extraction and PCR amplification prevents accurate quantification of the relative abundance of microorganisms designed to represent simplified natural populations.