The genetic variability of classical swine fever virus was studied by comparative nucleotide sequence analysis of 76 virus isolates, collected during a half century from three continents. Parts of the E2 (gp55) and the polymerase gene coding regions of the viral genome were amplified by RT-PCR and DNA fragments of 254 and 207 bp, respectively, were sequenced. The comparative sequence analysis of the E2 region revealed two main phylogenetic groups of CSFV, indicating that the virus apparently evolved from two ancestor nodes. Group I (represented by Brescia strain) consisted of old and recent American and Asian viruses, as well as old English isolates from the 1950s. This group was subdivided into three subgroups, termed I.A-I.C. Group II (represented by Alfort strain) consisted of relatively recent isolates from Europe, together with strain Osaka, which was isolated in Japan from a pig of European origin. Based on genetic distances the group was divided into subgroups II.A and II.B. Malaysian isolates were branched into both groups, indicating multiple origins for contemporaneous outbreaks in that country. All ten vaccine strains tested were branched in group I, implying a common ancestor. The Japanese Kanagawa strain, isolated in 1974, and the British Congenital Tremor strain from 1964 were the most distinct variants of CSFV in our collection. The comparison of the nucleotide sequences of the polymerase coding region of 32 European strains distinguished subgroups II.A and II.B which were similar to the corresponding subgroups of the E2 phylogenetic tree. Thus, the results revealed that the E2 region and the polymerase coding regions seem to be appropriate for the grouping of CSFV isolates from all over the world, distinguishing two major groups of the virus. The reliability of these regions for phylogenetic analysis is indicated by the similarity of the results obtained from the two separate parts of the CSFV genome.