112 TEST CASES [ch. xi 



thing that may possibly happen, and again we have no reason 

 why the change should not sometimes be generic as well as 

 specific (3). 



Under the theory of natural selection, the existing plants will 

 only be the lowest row in each family, for evolution, as we have 

 already explained, is at present supposed to work by the forma- 

 tion of slight varieties, which gradually increase to larger, to 

 species, and so on, by the killing out of the less well adapted 

 ancestors, while the ultimate survivors in the way of genera tend 

 to be those that show the greater divergences. And the smaller 

 the family, the greater are the divergences between its genera. 



To return to the diagram, under the supposition of evolution 

 by natural selection, the various genera that occupy the bottom 

 row in each family, whether in line 3, 4 or 5, will simply be genera, 

 or generic stages in the evolution that is continually going on. 

 Under this supposition, there is no reason why the genera in the 

 lowest line, 3, of family A should in any particular way be any 

 different from those in line 4, of family B, or these from those in 

 line 5, of family C. Nor does natural selection offer any test by 

 whose application we may gain any idea as to the relative degree 

 of divergence that there may be between the genera of A, B, 

 and C But upon the theory of differentiation or divergent muta- 

 tion we shall expect that the divergence between the two genera in 

 family A will he about equal to the first divergence in the families B 

 and C, i.e. equal to the divergence between their tribes or subfamilies. 

 At the same level in the diagram, in other words, there will be 

 more or less equal divergences. It has long been known as an 

 axiom in taxonomy that genera in a small family are much better 

 separated than genera in a large one, and here is a simple expla- 

 nation of this. As the family grows in size, new mutations will 

 come in at more and more frequent intervals, but within, or close 

 to, the original divergent mutation. In other words, all the 

 family, sprung from the original genus and its first mutation, 

 will show some at least of the characters shown by these first two 

 genera, which by the hypothesis of divergent mutation will tend 

 to be very divergent. The original family characters will show 

 best in the largest genera, which will be the oldest in the families, 

 and carry the most of the earliest characters. The genera sprung 

 from later mutations will not have, so well marked, many of the 

 characters of the earlier mutations. The generic characters will 

 necessarily become on the whole closer and closer together as the 

 mutations to which they are due are less and less far back in their 



