1010.] Classification of Mammalian Orders, etc. 463 



some of them may not withstand the destructive criticism which will be 

 raised by new (Uscoveries or a more extended analysis. 



One advantage of the use of the sui)erordcr is that it permits the bn^akino- 

 uj) of unnatural sequences. Juxtaposition and scMpience are inseparable 

 features of all classification, but since it is impossible to represent branchino- 

 kinship in a linear sequence of names, some authors have gone to the other 

 extreme and failed to adopt definite rules of se(}uence. For instance, in a 

 linear arrangement of the orders the Edentates must come in somewhere, 

 but wherever we put them they are bound to stand next to son)e order to 

 which they are but very remotely related. On the other hand, if juxtaposi- 

 tion is to mean nothing, if we are to understand a " Da Capo " sign in front of 

 every order, we then lose sight of the fact that some orders (e. g., Dermoptera, 

 Chiroptera) do not run back independently to the beginning but seem to be 

 rather closely related. 



The classification given below usually jiasses from the older and more 

 generalized to the newer and more specialized. It is intended to express 

 first degrees of homological resemblances, secondly, degrees of kinship, and 

 thirdly, successive grades of specialization. 



A difficult (juestion for the taxonomist is how far mori)hological differ- 

 ences are to be neglected in favor of expressing linear derivation, /. r. of 

 grouping ancestral and derived divisions together. For example, if the 

 Ancylopoda are to be grouped in the same suborder with the Perissodactyla, 

 why not unite the Sirenia with the Proboscidea, the Bats with the Insec- 

 tivora ? The obvious answer is that the systematist has generally given due 

 weight to the magnitude of the morphological resemblances and differences, 

 and has tried to express grades of specialization, as in the terms Prototheria, 

 Metatheria, and Eutheria; although, of coiu'se, it is impossible to be thor- 

 oughly consistent, especially in assigning a position to the Edentate orders, 

 and any classification recjuires some arbitrary decisions and artificial sun- 

 derings. 



One other general principle here adopted is that truly convergent resem- 

 blances between members of widely removed orders are accompanied by 

 fundamental differences which may usually be discovered upon close exam- 

 ination of adequate material. If, for example, the resemblances in the 

 dentition, skull and limbs between the Eocene ]MiacidjTp and the Oligocene 

 Canidfe are convergent to the same degree that the resemblances between 

 Noiorijctes and ChrijsoehJoris are convergent, then almost all phylogenetic 

 speculation from existing material is useless. But as a matter of experi- 

 ence the kind of characters separating Chrijsochloris from Notoryctes are 

 found to be of a very different nature from the kind of characters (primitive) 

 separating the Miacidse from the Oligocene Canidje. 



Thus a crucial problem raised by the present work is whether convergent 



