This study presents evidence from mitochondrial and nuclear loci that there is genetic divergence among and within geographic populations of Lagenorhynchus obscurus. The effect of seasonal variation on the genetic structure within New Zealand was examined with mitochondrial DNA control region sequences from 4 localities. Analysis of nested haplotype clades indicated genetic fragmentation and at least 1 historical population expansion within New Zealand. AMOVA and Fst values from nuclear and mitochondrial DNA sequences suggested significant divergence between New Zealand, South Africa, Argentina, and Peru. Dispersal via the west-wind drift was not supported by patterns of population structure among regions. Alternatively, these data support reciprocal exchange among all four regions with 100% posterior probability for a root of origin in the Indian/Atlantic Oceans. The degree of divergence between Peru and other regions indicates the isolation of Peruvian stock is temporally correlated with the constriction of Drake’s passage in the Plio-Pleistocene. There is evidence that the Plio-Pliestocene paleoceanography of the Indian and Southern Atlantic Oceans influenced phylogeography with shifts of temperate sea surface temperatures northward ~5º of latitude, disrupting the dispersal corridor between New Zealand and Atlantic populations. A preference for temperate waters along continental shelves is proposed as an explanation for lack of contemporary genetic exchange among regions. This study supports the polyphyly of the genus Lagenorhynchus. North Atlantic species form a monophyletic Lagenorhynchus. In the Southern Hemisphere, L. australis/L. cruciger and L. obliquidens/L. obscurus do not form a monophyletic group. I discuss the taxonomic implications and propose taxonomic revision of the genus based on these results. Measures of character interaction indicate that combined evidence from nuclear and mitochondrial genes provide better phylogenetic resolution among delphinid lineages than any data partition independently, despite some indications of conflict among mitochondrial and nuclear data.