CLASSIFICATION OF THE THERIODONTIA. 65 



or, at most, two drawings of any one skull. Tims it is difficult 

 to carry out any detailed discussion of the mutual relations of the 

 known Gorgonopsids. 



In order to discuss with any satisfaction the classification of a 

 group, it is necessary to know the main outlines of its history, to 

 understand the direction of the advances which make the struc- 

 ture of all late members of it differ from their ancestors, and 

 to work out any adaptive modifications which characterize its 

 different branches. 



It is now generally agreed that the Gorgonopsids include the 

 ancestors of the Cynognathidse, and that the Pelycosaurs are a 

 group which includes self-contained lateral branches springing 

 from the very base of the stem of the mammal-like Reptiles. 

 Thus by comparing the two extreme terms, Varanosaurus and 

 Diademodon, we can gain at once a knowledge of the advances in 

 structure which have occurred in the Anomodonts, and on the 

 assumption that these changes have proceeded regularly we can 

 determine the trend of advance during the evolution of the group. 

 Discussion of intermediate forms will then enable us to decide 

 whether this trend really expresses a true view of the mode of 

 evolution, or whether the actual observed differences between 

 the extremes represent the result of a series of fortuitous changes 

 of indeterminate direction. 



The work of Broili, of Case, of Williston,and the present writer 

 has led to the view that Varanosaurus is the most primitive known 

 Pelycosaur, forming a morphological ancestor to Dimetrodon, 

 through a Deio])eus-\ike form. The view that Diademodon or 

 Trirachodon is the most advanced of known Anomodonts results 

 from the work of Seeley and Broom. 



The differences between the skull of Diademodon and Varano- 

 saurus are : — 



In General Shape. 



In Varanosaurus the snout is very long, square-cut in section, 

 and roomy, compared with the rest of skull, with lateral nostrils 

 and a long straight tooth-row. The large orbits are entirely 

 laterally directed. The small temporal fossa lies entirely on the 

 side of the skull and is almost hidden from above by the very 

 broad parietal region. The occiput slopes forward, but is not 

 deeply cupped. The sides of the skull are nearly straight. The 

 skull is higher than wide. 



In Diademodon the snout is short, small in volume, rounded in 

 section, with nostrils looking more for waixl than laterally. The 

 tooth-row is short and curved The orbit is comparatively small 

 and looks very largely forward. The temporal fossa is very 

 large, lies entirely on the top of the skull, and is not visible from 

 the side, the parietal region being drawn up into a narrow crest. 

 The occiput slopes a little forward and is deeply cupped. The 

 sides of the skull gradually approach one another to the orbits 



Proc. Zool. Soc— 1921, No. V. 5 



