cn. viii] DIFFERENTIATION 71 



Up to the present, this theory is the only one which can make 

 any pretence of explaining the hollow curve. The latter is so 

 universal that it is evidentlv a general law which must be ex- 

 plained. But it is, of course, in direct contradiction to the theory 

 of natural selection. With the latter theorv one can make no 

 predictions as to what may be found in the arrangement and 

 characters of families, genera, or species. With differentiation 

 one can make a beginning in this direction, and this alone makes 

 a strong claim in its favour. 



If differentiation be the rule, it is clear that the ultimate result 

 of the growth of a family from one original genus ^4 to a fair 

 number of genera should in general be the formation of groups 

 within groups, like the cat group or the dog group within the 

 larger group of Mammals. By the principles of differentiation 

 and of divergence of variation, each genus thrown will tend to be 

 markedly different from the parent that throws it. If it were 

 thrown very far back in the history of the famil}^ it will have 

 had time to throw more (^enera in its turn. These mav, but so far 

 as one can see, not necessarily must, display the character or 

 characters that made their parental genus, B for example, dif- 

 ferent from its own parent A upon the main line of the family. 

 When these genera upon the second line B had become more or 

 less numerous, or in any case if the characters of their parent had 

 included two or more of the characters w^hich we usually rank as 

 "family" characters, they would form a group Bh, Be, Bd, etc., 

 well marked off from the first group which was being formed by 

 the genera upon the main line. A, B, C, D, etc., and to these two 

 groups we should probably give the rank of sub-families, or 

 tribes, according to our conception of the value of their characters. 

 One of these groups would most probably be larger and perhaps 

 more widely dispersed than the other, and both would continue 

 to grow and to spread. Supposing that the family escapes with- 

 out very great damage all the various accidents that may befall 

 it, and that all its genera behave fairly closely in the same way, 

 as would be the case under differentiation, the original parent A 

 will have the largest number of species (theoretically) and the 

 largest area of occupation, while the other genera, B, C, etc., 

 will be successively smaller in these respects, as we have seen 

 in the ^Monimiaceae for example. The difficulty in defining what 

 is or is not a sub-family or a tribe is the same as that of defining 

 a genus or a species. We have no standard to work by in defining 

 the value of a certain character, other than the way in which it 



