THE ORIGIN BY MUTATION OF THE ENDEMIC PLANTS OF CEYLON. 81 



invisible steps, and for the theory of their production by mutations. 

 In the rare cases of rapid dispersal of new species a better adap- 

 tation may of course be assumed as one of the chief factors, but on 

 the average the dispersal is very slow in the beginning, giving no 

 argument in favor of this view. 



Furthermore these consideiations lead to the view that wide 

 distribution and commonness are chiefly dependent on age, and 

 only rarely on adaptation. In every family the genera with the 

 widest distribution may be considered as the oldest, those with a 

 smaller domain as younger, and the local endemics as the youngest 

 of all. These principles will be used in subsequent studies to draw 

 pedigrees of families. But the studies made by the author up to 

 this time go to show that nearly all families have the same general 

 type of distribution, that evolution of forms is on the average 

 indifferent, and that most of the so-called adaptations are of no 

 special advantage to their possessors. 



Another argument relates to the possible size of mutations. It 

 is often assumed that mutations must of necessity be small, con- 

 sidering that it seems probable that only one unitfactor will be 

 changed at a time. This conception seems to the author to be an 

 unnecessary handicap to the theory of mutation and he proposes 

 that it should be replaced by the hypothesis that no specific change 

 is too great to appear in one mutation. The difference between 

 endemic species of Ceylon and their nearest allies is often very large, 

 as may be deduced from the fact that they are accepted as well- 

 marked Linnean species by such authorities as Trimen and Hooker. 

 But in many cases they are even larger. For instance, Coleus elon 

 gatus, which occurs only on the top of Ritigala and here only in 

 about a dozen of individuals, differs so much from all other Colei, 

 that it may well be regarded as subgenerically distinct. And for 

 the 17 endemic genera, which have only one species each, it seems 

 at least very probable that the whole genus has arisen at a single 

 step. 



In concluding I might state that my own studies on the production 

 of new forms among the Oenotheras have of late led me to the con- 

 clusion that mutations are in many cases of a far more complicated 

 nature than has been assumed until now. Many of them, as for in- 

 stance the production of 0. rubrinervis 0. nanella and 0. gigas, 

 involve the simultaneous change of two or more characters, in some 

 cases of quite a large number of unit-factors. Why these changes 

 should so regularly go together, we do not, as yet, know, but the 



6 



