The early origin of four vertebrate Hox gene clusters during the evolution of gnathostomes was likely caused by two consecutive duplications of the entire genome and the subsequent loss of individual genes. The presumed conserved and important roles of these genes in tetrapods during development led to the general assumption that Hox cluster architecture had remained unchanged since the last common ancestor of all jawed vertebrates. But recent data from teleost fishes reveals that this is not the case. Here, we present an analysis of the evolution of vertebrate Hox genes and clusters, with emphasis on the differences between the Hox A clusters of fish (actinopterygian) and tetrapod (sarcopterygian) lineages. In contrast to the general conservation of genomic architecture and gene sequence observed in sarcopterygians, the evolutionary history of actinopterygian Hox clusters likely includes an additional (third) genome duplication that initially increased the number of clusters from four to eight. We document, for the first time, higher rates of gene loss and gene sequence evolution in the Hox genes of fishes compared to those of land vertebrates. These two observations might suggest that two different molecular evolutionary strategies exist in the two major vertebrate lineages. Preliminary data from the African cichlid fish Oreochromis niloticus compared to those of the pufferfish and zebrafish reveal important differences in Hox cluster architecture among fishes and, together with genetic mapping data from Medaka, indicate that the third genome duplication was not zebrafish-specific, but probably occurred early in the history of fishes. Each descending fish lineage that has been characterized so far, distinctively modified its Hox cluster architecture through independent secondary losses. This variation is related to the large body plan differences observed among fishes, such as the loss of entire sets of appendages and ribs in some lineages.