MODES OF MAMMALIAN EVOLUTION 655 



the developmental history of each group shows that the smaller 

 and Ughter ancestors were less similar than the larger and more 

 massive descendants. Such subsequently acquired likenesses 

 are thus obvious examples of convergence and were caused 

 by adaptation to similar needs. 



Fiirbringer has shown that among birds size and weight 

 of body determine many resemblances between unrelated 

 famihes, the largest forms displa3dng a more advanced grade 

 of specialization. 



It is thus extremely probable that evolution is a highly 

 complex process, in which divergent, parallel and convergent 

 modes of development are normally concerned. This com- 

 plexity greatly increases the difficulty of determining phyloge- 

 nies, which would be very much easier could every notable 

 resemblance be at once accepted as proof of relationship. 

 It often renders impossible the classification of some isolated 

 group, which seems to have several incompatible affinities. 

 It emphasizes the necessity of foimding schemes of classification 

 upon the totahty of structure and the importance of determin- 

 ing the value of characters, whether they are primitive or 

 advanced, divergent, parallel or convergent, before attempting 

 to use them in classification. 



In looking over the field of maromalian evolution, so far 

 as that is recorded by the fossils, the general impression re- 

 ceived is that the most important process is divergent develop- 

 ment, one line branching out into several. This process became 

 especially vigorous and rapid at times of important change in 

 the character of the environment, what Osborn has called 

 "adaptive radiation." As we have repeatedly observed in the 

 history of particular groups, e.g. the rhinoceroses, horses and 

 camels, numerous parallel phyla of the same family existed 

 together in certain geological stages, but as these phyla were 

 traced back in time, they were found to draw together and dis- 

 play themselves as branches of a single stem. This favours the 

 inference that the mammalian orders, so far as they are truly 



