The evolution of tribosphenic molars is a key innovation in the history of Mammalia. Tribospheny allows for both shearing and grinding occlusal functions. Marsupials and placentals are advanced tribosphenic mammals (i.e., Theria) that show additional modifications of the tribosphenic dentition including loss of the distal metacristid and development of double-rank postvallum/prevallid shear. The recent discovery of Eomaia [Nature 416 (2002) 816], regarded as the oldest eutherian mammal, implies that the marsupial-placental split is at least 125 million years old. The conventional scenario for the evolution of tribosphenic and therian mammals hypothesizes that each group evolved once, in the northern hemisphere, and is based on a predominantly Laurasian fossil record. With the recent discovery of the oldest tribosphenic mammal (Ambondro) from the Mesozoic of Gondwana, Flynn et al. [Nature 401 (1999) 57] suggested that tribospheny evolved in Gondwana rather than in Laurasia. Luo et al. [Nature 409 (2001) 53; Acta Palaeontol. Pol. 47 (2002) 1] argued for independent origins of tribospheny in northern (Boreosphenida) and southern (Australosphenida) hemisphere clades, with the latter including Ambondro, ausktribosphenids, and monotremes. Here, we present cladistic evidence for a single origin of tribosphenic molars. Further, Ambondro may be a stem eutherian, making the split between marsupials and placentals at least 167 m.y. old. To test this hypothesis, we used the relaxed molecular clock approach of Thorne/Kishino with amino acid data sets for BRCA1 [J. Mammal. Evol. 8 (2001) 239] and the IGF2 receptor [Mammal. Genome 12 (2001) 513]. Point estimates for the marsupial-placental split were 182-190 million years based on BRCA1 and 185-187 million years based on the IGF2 receptor. These estimates are fully compatible with the results of our cladistic analyses.