284 Bulletin American Museum of Natural History. [Vol. XVI, 



bears, but must be considered as a side branch paralleling 

 them. Cynarctus would seem somewhat more but by no 

 means exactly in the line of descent. The slender jaw ex- 

 cludes it from direct relationship. 



In a previous paper the writer has discussed an alternate 

 hypothesis of the origin of the Ursidae which derives them 

 from the Creodont family Arctocyonidse, instead of from the 

 Canidae. It was then stated that the apparent chain of ex- 

 tinct types connecting the Canidae and Ursidae formed a most 

 serious objection to considering any other hypothesis as pos- 

 sible, but it was pointed out that there was a wide gap in the 

 series between such genera as Amphicyon and Dinocyon — un- 

 mistakable dogs although bear-like — and Hya;narctus and 

 Ursavus, unmistakable bears, although with the primitive 

 carnivore formula lost by the modern bears. Dr. Schlosser 

 has shown that Amphicyon is a side branch of the Canidae and 

 Hycenarctus of the Ursidae and that the wide gap between the 

 primitive Oligocene dogs, such as Cephalogale and the earliest 

 true Ursidae cannot be filled by any genera hitherto known. 

 The genus here described reduces this gap, as its close re- 

 semblance to Ursavus in the composition of the teeth seems 

 hardly explicable except on the ground of a near relationship ; 

 while in most characters it is as unquestionably a dog as Ursa- 

 vus is a bear. It does not seem, however, to point especially to 

 Cephalogale as an ancestor. Its relationship to the direct line 

 of descent is uncertain. 



Nevertheless this additional evidence in favor of the deriva- 

 tion of the bears from early Canidae seems to render untenable 

 any other hypothesis. The Arctocyonidae must then be con- 

 sidered as a case of parallelism not confined to the general 

 characters of teeth and feet, but extending to the detailed 

 structure of both, the rather exceptional cusp composition of 

 carnassial and molar teeth, the relative proportions of the 

 digits, even certain details in the character of the carpals 

 and tarsals being common to both, besides the more 

 general characters of large quadrate, flattened molars, re- 

 duced premolars, slender canines, plantigrade, large-clawed 

 feet. 



