The ratites have stimulated much debate as to how such large flightless birds came to be distributed across the southern continents, and whether they are a monophyletic group or are composed of unrelated lineages that independently lost the power of flight. Hypotheses regarding the relationships among taxa differ for morphological and molecular data sets, thus hindering attempts to test whether plate tectonic events can explain ratite biogeography. Here, we present the complete mitochondrial DNA genomes of two extinct moas from New Zealand, along with those of five extant ratites (the lesser rhea, the ostrich, the great spotted kiwi, the emu and the southern cassowary) and two tinamous from different genera. The non–stationary base composition in these sequences violates the assumptions of most tree–building methods. When this bias is corrected using neighbour–joining with log–determinant distances and nonhomogeneous maximum likelihood, the ratites are found to be monophyletic, with moas basal, as in morphological trees. The avian sequences also violate a molecular clock, so we applied a non–parametric rate smoothing algorithm, which minimizes ancestor–descendant local rate changes, to date nodes in the tree. Using this method, most of the major ratite lineages fit the vicariance biogeography hypothesis, the exceptions being the ostrich and the kiwi, which require dispersal to explain their present distribution.