Most vertebrate species utilize antibody and T-cell receptor (TCR) genes to create a vast repertoire of antigen sensor molecules on their B and T lymphocytes, respectively. While the organization of these genes exhibits substantial variation between species, one common theme is that, in almost every case, there is at least one variable region with a highly diverse CDR3 region and often much less diversity elsewhere in the binding site. Whereas with alphabeta TCRs this skewing of diversity correlates well with the need to recognize diverse peptides bound to MHC molecules, this cannot explain why this same pattern is evident in immunoglobulins (Igs) or gammadelta TCRs. Instead we have postulated that in the primary repertoire, all or most antigen receptors have a bipartite binding site, in which diverse CDR3 loops act as a highly antigen specific 'core' whereas other CDRs bind in a largely opportunistic fashion. In the case of antibodies, somatic hypermutation then acts to improve the complementarity to a given antigen and increase antibody affinity. A test of this model in mice engineered to have a very limited V region repertoire shows that primary antibodies can be generated that are highly specific for distinct antigens, yet identical in sequence except for their V(H) CDR3. Furthermore, very high affinity antibodies can be raised by repeated immunizations, showing that somatic hypermutation can mold these low affinity antibodies into high affinity ones. Thus, the wide variations seen in V region repertoires amongst vertebrates is likely to be of lesser importance than the preservation of one or more diverse CDR3 regions.