The reliability and the level of taxonomic resolution of the amplifiedfragment length polymorphism (AFLP) method were evaluated with species of patho-genic bacteria involved in human, animal and plant diseases. The method was foundvery versatile as it can be adapted to the individual genome constraints of all testedspecies. The calculation of a genetic distance d corresponding to the average dissim-ilarity between actual overall genome sequences was proposed for comparing AFLPdata. Bacterial models showed clearly different patterns between strains belonging todifferent genomic species, while patterns were clearly similar within a given species.The threshold which distinguishes between inter and infra-specific distances indicatesa critical overall genome diversity of about 14% (d = 0.14). AFLP had more resolu-tion power than serology, phage typing, PFGE and restriction analysis of ribosomalintergenic spacers. In the latter case, regression analysis showed that PCR-RFLPof ribosomal intergenic spacers can only be used to differentiate bacteria which haveat least 3.4% (d = 0.034) nucleotide differences between their respective genomes.Finally, an improved procedure using newly developed software was also proposed inorder to standardize the capture of reliable data and their numeric treatment for thefuture development of AFLP data bases.