Abstract

We review past DNA-hybridisation studies of marsupials and present a reanalysis of the data, utilising results from our and additional studies to formulate and rationalise a new classification of Marsupialia. In the reanalysis, 13 individual DNA-hybridisation matrices, many lacking some pairwise comparisons, were sutured in stages to provide the basis for generating a tree of 101 marsupials plus an outgroup eutherian; a fourteenth matrix provided data for a tree including eight additional eutherians and a monotreme. Validation was achieved by jackknifing on taxa for each matrix as well as on tables combining two or more matrices generated during assembly of the 102-taxon data set. The results are consistent with most conclusions from the individual studies and dramatise the unevenness of hierarchical levels in current classifications of marsupials. In particular, the affinities of the American marsupial Dromiciops gliroides with, and the distinctness of marsupial bandicoots from, Australasian metatherians are reaffirmed, while opossums are shown to be as internally divergent as are most members of the order Diprotodontia. Calibration of the 102-taxon tree and dating of the major dichotomies suggest that no extant marsupial lineage originated before the latest Cretaceous, and that all of them together with most South American and all Australasian fossils should be recognised as a monophyletic group contrasting with a largely Laurasian (if possibly paraphyletic) taxon. These inferences, together with the details of the phylogeny, mandate that the misleading ‘Australian’ v. ‘American’ distinction be abandoned, even as a geographic convenience.