CLASSIFICATION OF THE THERIODoNTlA. 77 



stage in the advance in structure of the basicranial region, 

 already discussed in P. Z. 8. 1914, p. 1027, figs. 3 & 4; it is in 

 this region more advanced than any other known Endothiodon- 

 zone GTorgonopsid. It retains as primitive features the very large 

 quadrate, large postfrontal and frontal, and a sloping occiput. 



Thus the Endothiodon-zouB Gorgonopsids show definite ad- 

 vances over the Pelycosaurs in the direction of Dlademodon. 

 Each form is advanced in certain features whilst retaining a 

 more primitive structure in others, so that an imaginary animal, 

 built up by throwing together the most advanced features found 

 in all the actual animals, would be far more advanced than any 

 one is on the average ; although in no point would it be more 

 advanced than a known form. In fact, the evidence existing 

 here, small though it is, suggests that there is a limit to the 

 total amount of advance possible to the members of a group in a 

 given time, and that these changes may be distributed either 

 over the whole animal or concentrated on a definite region, which 

 will then present a structure of much more advanced type than 

 is found in allied contemporaneous forms. A somewhat similar- 

 conclusion seems to have been reached by W. D. Matthew from 

 the study of the more abundant material of fossil mammals. 



Discussion of the Gorgonopsids of the Cistecephalus-y,one is 

 rendered difficult by two factors — the incomplete descriptions 

 and insufficient figures of many of the perfect sknlls in S. Africa 

 and New York, and the fact that the Cistecephalus-zone is a 

 long one and that we do not know the relative ages of the 

 Gorgonopsids from it. It will appear from the evidence to be 

 brought forward in this paper that the forms from Dunedin and 

 Nieuweveld localities are early, those from New Bethesda and 

 the Kagaberg which are associated with Dicynodon tlgricejys 

 considerably later in time. There is, however, no stratigraphical 

 evidence that this is so. 



From the Cistecephalas-zoue Broom and Haughton have 

 described several forms as species of Scymnognatlms — S. iigrlceps 

 B. &■ H., S. parvus Br., S. minor Br., S. cmgusticeps Br., S.serra- 

 tideus Han., are all from the Nieuweveld. These forms may 

 very possibly be congeneric ; they agree with Scymnogncdhus in 

 having i. 5, c. 1, in. 4-5, but quite certainly do not belong to 

 that genus. They differ from Scymnognathus whaitsi in the 

 following characters : — 



The snout is very much deeper, its anterior end instead of 

 being rounded is vertical {cf. Broom, P. Z. S. 1913, p. 225, pi. 36), 

 the external nostril is much larger, the septomaxillary foramen 

 smaller. The anterior end of the nasal does not fully overhang 

 the nostril. The top of the snout may be ridged, and the square 

 section with a, preorbital depression overhung by the prefrontal 

 is entirely lost (cf. S. serratidens, Ann. South Afr. Mus. vol. xii. 

 p. 89, pi. xiii.). The snout is much shorter and the prefrontal 

 in consequence smaller. 



