6 SCOTT. [Vol. V 



teeth, Wortman says : " There are no cases known to me in 

 which teeth have been added. On the contrary, I am firmly of 

 the opinion that not so much as a single tooth has ever been 

 added to the diphyodont mammalian dentition in the course of 

 development, but that specialization has invariably gone in the 

 other direction, as almost all evidence of palaeontology goes to 

 show" (No. 34, p. 454). 



The bearing of this question upon phylogenetic discussions 

 will be made clear by a single example. Among the Canida, as 

 in other placental mammals, the normal number of true molars 

 is not more than three, and, except as a rare and occasional 

 variation, this number is not exceeded by any known fossil 

 member of the family, or even by any creodont, even as far back 

 as the earliest Eocene ; yet the existing genus Otocyon has four 

 molars, though in other points of structure this genus shows no 

 important differences from other dogs. Nearly all palaeontolo- 

 gists have agreed in regarding this as proof that the ancestors 

 of the Canidcc had four molars, which have been retained only 

 in Otocyon, and that the occasional appearance of a fourth molar 

 in other genera of the family is clearly due to reversion (Hux- 

 ley, No. 14, p. 262), and it is further argued that the clogs must 

 be connected with the carnivorous marsupials. This view may 

 be entirely correct, and yet it involves consequences which have 

 an important bearing upon phylogenetic reasoning. We may 

 suppose : (a) that all the fossil forms with three molars, which 

 are ordinarily regarded as ancestral to the recent dogs, such as 

 Miacis, Daphcejius, Cynodictis, etc., etc., are in reality not con- 

 nected with them at all, but that when the cynoid series is made 

 out, it will be found to consist of forms with four molars down 

 to comparatively recent geological times ; or, (b) that there are 

 two lines, one of which is through Miacis, Dapkeznus, etc., to 

 Cams, and the other through unknown ancestors to Otocyon, 

 both starting from a type with four molars. This alternative, 

 however, involves the assumption of a remarkable degree of con- 

 vergence, for the early forms of the Canidce differ much from 

 the recent members in the structure of the brain, skull, verte- 

 bral column, and limbs, while in all these points Otocyon pre- 

 sents no differences of importance from the other recent genera, 

 and must therefore have reached its present identity with them 

 by an entirely different path, (c) This fourth molar of Otocyon 



