Three regions of the classical swine fever virus (CSFV) genome that have been widely sequenced were compared with respect to their ability to discriminate between isolates and to segregate viruses into genetic groups. Sequence data-sets were assembled for 55 CSFVs comprising 150 nucleotides of the 5' non-translated region, 190 nucleotides of the E2 envelope glycoprotein gene and 409 nucleotides of the NS5B polymerase gene. Phylogenetic analysis of each data-set revealed similar groups and subgroups. For closely related viruses, the more variable or larger data-sets gave better discrimination, and the most reliable classification was obtained with sequence data from the NS5B region. No evidence was found for intertypic recombination between CSFVs. A larger data-set was also analysed comprising 190 nucleotides of E2 sequence from 100 CSFVs from different parts of the world, in order to assess the extent and global distribution of CSFV diversity. Additional groups of CSFV are evident from Asia and the nomenclature of Lowings et al. (1996) [Lowings, P., Ibata, G., Needham, J., Paton, D., 1996. J. Gen. Virol. 77, 1311-1321] needs to be updated to accommodate these. A tentative assignment, adapting rather than overturning the previous nomenclature divides CSF viruses into three groups with three or four subgroups: 1.1, 1.2, 1.3; 2.1, 2.2, 2.3; 3.1, 3.2, 3.3, 3.4. The expanding data-base of CSFV sequences should improve the prospects of disease tracing in the future, and provide a basis for a standardised approach to ensure that results from different laboratories are comparable.