462 Bulletin American Museum of Natural History. [A'ol. XXVII, 



II. A Classification and Provisional Phylogeny of the Orders 



AND Suborders of Mammals. 



In the preceding chapters the writer has endeavored to collate and 

 interpret the work of many authors, to add his quota of new observations 

 and to develoj) therefrom a general hypothesis of interordinal relationships. 

 This is summarized in the present chapter, first in the form of a classifica- 

 tion of the orders and suborders, secondly in the form of a phylogenetie 

 tree (Fig. 32, 33). The classification is founded, so far as possible, upon a 

 consideration of the " totality of characters drawn from various parts of the or- 

 ganism" (Gill), and thus contrasts, for example, with Cope's classification 

 of the Ungulates, which was founded upon practically a single character of the 

 carpus and tarsus. As other principles of classification followed in the pres- 

 ent work were discussed in the writer's review ' of the phylogeny of the Teleo- 

 stomous fishes (1907, pp. 440-444), the only matter here requiring discussion 

 is the use of the superorder. 



The orders of mammals have probably been fairly well determined in the 

 majority of instances. It is not likely, for example, that the Perissodactyla 

 will ever be split up into several orders, at least if the term "order" is to 

 retain its historical meaning. Consequently the orders are capable of exact 

 definition, that is, the ordinal characters apply to all members of the order. 

 The superorder, on the contrary, as here used, stands only for an hypothesis 

 of common origin; its definition describes the group from which two or 

 more parallel or divergent orders were derived but whose ancestral characters 

 are lost in varying degrees in the actual orders. Consequently the super- 

 order is usually a more elastic and indefinite conception than the order. 

 The superordinal grouping here proposed may in several cases prove er- 

 roneous, but the abandonment of the names and the bringing together of 

 new groups should not cause confusion, whereas the abandonment of any 

 of the well established ordinal names and groups would be justifiable only 

 after the discovery of a most convincing accumulation of evidence. The 

 history of classification warns us against taking superordinal groupings too 

 seriously. They are bound to expand and divide and recombine to some 

 extent, because it may be assumed that the pahieontological record will always 

 remain very imperfect and classification must, therefore, remain a com- 

 promise between the "vertical" and "horizontal" factors, ?". e., it must be an 

 expression of the mingled results of heredity on the one hand and of parallel 

 and convergent evolution on the other. In view of the more or less hypo- 

 thetical character of the superordinal groupings here adopted, the writer 

 therefore does not expect to see all of them gain very wide acceptance, and 



1 Ann. N. Y. Acad. Sci., Vol. XVII, pt. ii, no. 3, pp. 437-508. 



