218 BULLETIN 16 9, UNITED STATES NATIONAL MUSEUM 



them in the Insectivora because of their primitive character and 

 because of certain marked differences from the phenacodonts. 



In 1915 Matthew reviewed the lower Eocene hyopsodonts, which, 

 for the first time, gave him a good knowledge of their foot structure. 

 He then considered them to be condylarths, confirming his tentative 

 suggestion of 1909. He carefully redefined the Condylarthra and 

 included these five famiHes, the last provisionally: Mioclaenidae, 

 Hyopsodontidae, Phenacodontidae, Meniscotheriidae, and ?Pleura- 

 spidotheriidae. 



In his last contribution (Pale. Mem.) Matthew retained this ar- 

 rangement, except that the Mioclaenidae and Hyopsodontidae are 

 reduced by further study to two subfamilies of Hyopsodontidae, as 

 already tentatively foreseen in 1909. 



I now propose to return to Cope's arrangement of 1884, with the 

 only change the inclusion of the Hyopsodontidae (with Mioclaeninae), 

 that is, to his classical conception of a group based on both the phen- 

 acodonts and the periptychids. It seems to me, after careful and 

 long consideration with practically all the pertinent original specimens 

 (including a good deal even unknown to Matthew), that comprehen- 

 sion of this group has been retarded and taxonomy has been in a 

 blind alley since the rise in the nineties of the idea of close periptychid- 

 pantolambdid affinities, an idea to which even Matthew finally sub- 

 scribed after som.e years of resistance. This reactionary view, which 

 at this late date wiU rather seem radical, requires an outline defense 

 even though much of the crucial evidence is not drawn from the pres- 

 ent fauna. 



The original suggestion that Periptychus might be an amblypod 

 (Osborn and Earle, 1895, p. 47) was based on the facts that its tarsus 

 is not serial and that "it has the strictly trigonal molar of the Ambly- 

 poda." It may at once be noted that these arguments have since 

 proved to be valueless. It is now known that the primitive condy- 

 larth tarsus was not serial, and the molars of Periptychidae are not, 

 as a rule, strictly trigonal, those of some condylarths are, and the 

 periptychid molars are otherwise decidedly more condylarth- than 

 amblypod-like. 



Cope (1897, p. 335) stated that he had anticipated that the perip- 

 tychids, with their astragalo-cuboid contact (nonserial tarsus), might 

 be the bunodont ancestors of the Amblypoda, but he awaited dis- 

 covery of their carpus and evidence that it, also, was nonserial. The 

 carpus had not been discovered, but the continuing failure to discover 

 any other possible amblypod ancestry led Cope then to assume the 

 presence of a nonserial carpus in the periptychids and to consider 

 them as this ancestry. He was also influenced by the suggestion of 

 Osborn and Earle. This is too theoretical to warrant much "con- 



