CH. IX] DIVERGENCE OF VARIATION 77 



Cyclantheae has only one genus, Cyclanthus, with four species 

 in tropical South America from Peru northwards, and in the 

 West Indies. It is thus the second genus of the family both in 

 number of species and in dispersal, but as it is so much smaller 

 than Carludovica, we must suppose that it was only cut off from 

 that genus rather late. Its dispersal is much smaller than that of 

 Carludovica. 



There is no reason for supposing the small genera to be relics; 

 it is far simpler to imagine them all split off by large mutations 

 from Carludovica in its gradual dispersal over the whole area of 

 the family. The first split probably gave Cyclanthus, which is the 

 second largest genus, and has a division to itself in the family, 

 whilst it would seem extremely probable that the mutation that 

 gave rise to it was an extra "large'" one, for the difference is so 

 great between it and the rest. The other four genera, smaller, and 

 with less distribution, count in the same group with Carludovica, 

 from which they are not so markedly different. The general im- 

 pression that one gains here, as in almost all cases, is that after 

 the big mutation which first gave rise to the family, there follow^ed 

 others which gradually became less and less marked, and which 

 kept more or less closely wathin the boundaries that were indi- 

 cated by the first mutation that occurred after the formation of 

 the family. 



The first impulse of many will be to say that Cyclanthaceae 

 form an exceptional famil}^ and perhaps also to say that the keys 

 are artificial things. But the exceptional families, and the diver- 

 gences that are shown in the keys, have both to be explained by 

 natural selection, or by any other theory of evolution, just as 

 much as have the ordinary families and the smallest divergences. 

 Natural selection would be very hardly pressed to find any ex- 

 planation of the very remarkable differences between Carludovica 

 and Cyclanthus, especially as it is all but impossible even to 

 imagine that there can be any intermediate stages, and no use- 

 value can be put to either of the extremes or to any conceivable 

 intermediate. 



This supposition, that the first mutation, in a family newly 

 formed by a large change from some ancestral form, may be in 

 turn large, is well supported by an examination of the keys to the 

 various families that are given in any general text-book of 

 systematic botany. In the list in Appendix II, I have extracted 

 from the keys in my Dictionary, 6th ed. (which are mostly taken 

 from the Natiirlichen Pflanzenfamilien), the first dichotomy in 



