Gene-by-gene phylogenetic analyses of all of the viruses for which sequences are known, as well as analysis of the coding capacities, clearly demonstrated that there are two major groups of viruses in the taxonomic family Geminiviridae. These are of the Subgroup I type, with one genomic component, which mainly infect monocots and are leafhopper-transmitted; and of the Subgroup III type, with one or two genomic components, which infect dicots and are whitefly-transmitted. The existence of "New World" and "Old World" clusters of Subgroup III viruses was confirmed, as well as the possession by the latter of an AV1 ORF not present in New World viruses. A third minor generic group is defined by viruses of the Subgroup II type, which have a single genomic component, infect dicots, and are leafhopper-transmitted. The latter group appear to be the result of an ancient recombination event between a Subgroup III-like and a Subgroup I-like virus. The question of whether one- and two-component Subgroup III viruses should be in the same taxon appears hard to resolve: the only distinguishing feature of the one-component Subgroup III viruses is that they apparently have no second component, as gene-for-gene comparisons of the "A" components of the viruses with other Subgroup III viruses place them within a larger Old World group of viruses, most of which are two component. The possibility exists that these viruses may either have independently lost their B components, or possess a B component that has simply not yet been found. Possible nomenclatural changes to accommodate viruses with the same name which are not closely related to one another, and possible evolutionary scenarios to account for the observed familial, generic and specific diversity of geminiviruses, are discussed.