The genetic structure of humpback whale populations and subpopulation divisions is described by restriction fragment length analysis of the mitochondrial (mt) DNA from samples of 230 whales collected by biopsy darting in 11 seasonal habitats representing six subpopulations, or 'stocks', world-wide. The hierarchical structure of mtDNA haplotype diversity among population subdivisions is described using the analysis of molecular variance (AMOVA) procedure, the analysis of gene identity, and the genealogical relationship of haplotypes as constructed by parsimony analysis and distance clustering. These analyses revealed: (i) significant partitioning of world-wide genetic variation among oceanic populations, among subpopulations or 'stocks' within oceanic populations and among seasonal habitats within stocks; (ii) fixed categorical segregation of haplotypes on the south-eastern Alaska and central California feeding grounds of the North Pacific; (iii) support for the division of the North Pacific population into a central stock which feeds in Alaska and winters in Hawaii, and an eastern or 'American' stock which feeds along the coast of California and winters near Mexico; (iv) evidence of genetic heterogeneity within the Gulf of Maine feeding grounds and among the sampled feeding and breeding grounds of the western North Atlantic; and (v) support for the historical division between the Group IV (Western Australia) and Group V (eastern Australia, New Zealand and Tonga) stocks in the Southern Oceans. Overall, our results demonstrate a striking degree of genetic structure both within and between oceanic populations of humpback whales, despite the nearly unlimited migratory potential of this species. We suggest that the humpback whale is a suitable demographic and genetic model for the management of less tractable species of baleen whales and for the general study of gene flow among long-lived, mobile vertebrates in the marine ecosystem.