Randomly amplified polymorphic DNA (RAPD) fingerprints were generated with M13 and DAF4 primers for 25 isolates of Acinetobacter spp. obtained from 16 different hospitals situated in 12 countries. The overall robustness of the algorithms and the reproducibility of the cluster analysis results generated by two commercially available computer programs (GelCompar and DENDRON) for analysing DNA fingerprinting gels were tested by examining the same set of fingerprinting data independently in two laboratories with the different software packages. Both programs were efficient at recognising and grouping strains with closely similar RAPD fingerprints, i.e., strains which might be expected to have a close epidemiological or evolutionary relationship. However, the relationships suggested for less closely related strains showed considerable variation in terms of the overall similarity or percentage correlation values suggested by the programs. It was concluded that both programs were useful tools for indicating close genotypic relationships between individual strains, but that epidemiological conclusions based on the similarity or correlation values (or the dendrograms derived from them) obtained for less closely related strains should be treated with considerable caution.