We describe here one of the oldest known Japanese mammals, a triconodont lower jaw from the Early Cretaceous Kuwajima Formation, Tetori Group, Japan. This mammal is recognized as the type of Hakusanodon archaeus, new genus, new species, and as closely related to Eurasian and North American Jurassic “amphilestid” triconodonts (Simpson 1925b–c; Jenkins and Crompton 1979; Sigogneau-Russell 2003a), a group of poorly understood early mammals. To better understand the relationships of Hakusanodon, we present an inclusive cladistic analysis of most triconodont mammals and relatives. Our analysis highlights the central role of triconodonts for early mammalian history and corroborates recent findings indicating that traditional groups of Mesozoic mammals such as Triconodonta (Jenkins and Crompton 1979), Amphilestidae (sensu Jenkins and Crompton 1979; McKenna and Bell 1997) and Symmetrodonta (sensu Lillegraven et al. 1979; McKenna and Bell 1997) are artificial. A dental pattern similar to that of “amphilestids” is generalized, playing a central role in mammalian dental evolution and being primitive for triconodontids and spalacotheriids, and therefore for the archaic members of the therian lineage. Australosphenidans (Luo et al. 2001a, 2003; Martin and Rauhut 2005), with fully reversed triangles in their lower dentitions (uppers are not known in any of them with the exception of monotremes), also are likely to have evolved from ancestors with an “amphilestid”/sym-metrodont-like dental morphology.

The enigmatic Amphidon aequicrurius Simpson, 1925a, from the Jurassic of North America (Simpson 1925a) has been treated as related to acute angle symmetrodonts or to triconodonts. Re-study of the type and only specimen suggests that Amphidon Simpson, 1925a, is likely an old, worn-down “amphilestid” (Rougier et al. 2001), a conclusion also supported by the results of the cladistic analysis. Jeholodens jenkinsi Ji et al., 1999, postulated as a basal member of the triconodontid lineage, or eutriconodont (Ji et al. 1999), is more likely a generalized “amphilestid,” which makes this taxon more relevant for the understanding of the stem members of the therian lineage by providing a relatively complete out-group.

The cranium of the Barstovian castoroidine beaver Eucastor tortus (Leidy, 1858), and the Clarendonian species Eucastor burgensis Korth, 2002, Eucastor dividerus Stirton, 1935, Eucastor katensis Korth, 2002, and Dipoides tanneri Korth, 1998, are described and compared. The skull of E. tortus has traits that are similar to those of the Clarendonian Nothodipoides Korth, 2002, that are associated with tooth digging. Other species examined previously referred to Eucastor Leidy, 1858, are nearly identical in cranial morphology to D. tanneri and are similar to the skull of Hemphillian Dipoides Schlosser, 1902. Prodipoides, new genus, is named with Eucastor dividerus Stirton, 1935, as type species, to include the Clarendonian species as new combinations previously referred to Eucastor: Prodipoides burgensis (Korth, 2002), Prodipoides dividerus (Stirton, 1935), Prodipoides katensis (Korth, 2002), Prodipoides lecontei (Merriam, 1896), and Prodipoides phillisi (Wilson, 1968). Eucastor is limited to the nominal species E. tortus and is referred to the Tribe Nothodipoidini. Prodipoides is referred to the Tribe Castoroidini. It is evident that the tooth-digging nothodipoidines diverged from the rest of the castoroidines in the Barstovian with the occurrence of Eucastor, and that the traditional sequence of Monosaulax-Eucastor-Dipoides is unlikely.