iQoi.] Mattheiv^ Additional Observations on the Creodonta. 3 



descendants, are placed in a separate suborder, distinguished by 

 the absence of a specialized character common to all the modern 

 families. Similarly among the Creodont families, the primitive 

 types have been grouped separately from the specialized ones 

 by almost all writers, the family distinctions being on points of 

 specialization or non-specialization. 



The present writer is in a position to appreciate very much the 

 evil effects of unnecessary changes in nomenclature and is in- 

 clined to deprecate most strongly the perpetual changes caused 

 by a too strict adherence to laws of priority which were not de- 

 signed for palaeontology, and do not entirely fit its requirements. 

 With the progress of the science it becomes more and more evi- 

 dent that a very large proportion of the earlier species were 

 founded on indeterminate material ; and the vain attempt to find 

 reasons for identifying these fragments with one or another of the 

 species distinguishable by more perfect material has wasted a vast 

 amount of time, and, however valuable from an antiquarian stand- 

 point, adds not one iota to our knowledge of the fossil animals 

 themselves and their place in nature. Scientific names are not 

 different from other names — they are merely arbitrary terms 

 representing each a certain conception ; and it appears to the 

 writer that custom should be the chief rule to govern their use. 



With classifications the case is different. A classification is 

 not arbitrary, but an expression of the natural relationship as 

 nearly as our knowledge of the species and the limitations of the 

 method used will permit. Conservatism in classification would 

 mean cessation in the increase of our knowledge of a group ;_ 

 uniformity in it would be impossible, unless all observers attached 

 the same relative importance to the same characters. A com- 

 paratively slight divergence of opinion on the last point may in- 

 volve an entire upset in the classification scheme. A graphic 

 method (such as is used on p. 21 in illustrating the relations 

 of three of the Creodont families) would more accurately express 

 the conception ; but in such a method it is seldom possible to 

 include all members of a group — the physical difficulties of space 

 of two dimensions prevents the complete representing of poly- 

 dimensional relationships. 



In venturing, therefore, on very considerable changes in classi- 

 fication from that generally accepted, and embodied especially in 



