Species pairs of threespine stickleback, Gasterosteus aculeatus, co-exist in several lakes in the Strait of Georgia, southwestern British Columbia. One species, ‘benthics’ is robust-bodied and is morphologically and behaviourally specialized for benthivory. The other species, ‘limnetics’ is specialized for planktivory in open-water habitats of the lakes. We examined mitochondrial DNA restriction site variation in benthic and limnetic sticklebacks as well as in solitary freshwater, anadromous (sea-run), and marine populations to test: (i) if benthic and limnetic pairs have evolved only once or multiple times (parallel evolution) and (ii) if the species have evolved sympatrically, or allopatrically from ‘double invasions’ of lakes by ancestral anadromous/marine sticklebacks. Stickleback mtDNA comprised a single clade with a low (mean = 0.40%) degree of sequence divergence among the 77 haplotypes resolved. Most nucleotide diversity (97%) was found within (rather than among) populations of anadromous/marine sticklebacks whereas most diversity (77%) was found among populations in freshwater sticklebacks. Significant differences in haplotype frequencies were found between benthics and limnetics in three of the four species pair lakes examined, but in all cases the pairs within lakes were characterized by unique assemblages of closely related haplotypes. Hierarchical clustering of divergence estimates suggested that comparable species from different lakes have originated independently in all lakes because in no case did comparable species from different lakes cluster together. Divergent species within lakes tended to be more closely related to one another than to species in other lakes and there were two cases were benthics and limnetics within a particular lake were monophyletic. In two of the four two-species lakes, limnetics were less divergent from putative ancestral anadromous/marine stickleback as predicted by the double invasion hypothesis, but in the two other lakes benthics were less divergent. Our data argue strongly that the species pairs have evolved independently in each lake were they now co-exist. Further, in two lakes our data are consistent with the species having evolved by sympatric divergence, but allopatric divergence followed by introgression of mtDNA that has obscured ancestral relationships cannot be discounted completely. Finally, despite remaining uncertainty about the geography of speciation, the species appear to have evolved in the face of gene flow arguing that natural selection acting on trophic ecology has been a major component of ecological speciation in sticklebacks.