A systematic evaluation of the value and potential of terminal-restriction fragment length polymorphism (T-RFLP) analysis for the study of microbial community structure has been undertaken. The reproducibility and robustness of the method has been assessed using environmental DNA samples isolated directly from PCB-polluted or pristine soil, and subsequent polymerase chain reaction (PCR) amplification of total community 16S rDNA. An initial investigation to assess the variability both within and between different polyacrylamide gel electrophoresis (PAGE) runs showed that almost identical community profiles were consistently produced from the same sample. Similarly, very little variability was observed as a result of variation between replicate restriction digestions, PCR amplifications or between replicate DNA isolations. Decreasing concentrations of template DNA produced a decline in both the complexity and the intensity of fragments present in the community profile, with no additional fragments detected in the higher dilutions that were not already present when more original template DNA was used. Reducing the number of cycles of PCR produced similar results. The greatest variation between profiles generated from the same DNA sample was produced using different Taq DNA polymerases, while lower levels of variability were found between PCR products that had been produced using different annealing temperatures. Incomplete digestion by the restriction enzyme may, as a result of the generation of partially digested fragments, lead to an overestimation of the overall diversity within a community. The results obtained indicate that, once standardized, T-RFLP analysis is a highly reproducible and robust technique that yields high-quality fingerprints consisting of fragments of precise sizes, which, in principle, could be phylogenetically assigned, once an appropriate database is constructed.