ELIMINATION OF MUTATIONS 91 



will have the best chance of surviving and producing the 

 most numerous offspring. Thus those individuals that vary 

 away from the most favourable characters will be gradually 

 or suddenly eliminated, while those that vary towards more 

 accurately adapted characters will survive and transmit their 

 characters. The subsequent offspring will therefore start 

 from a new racial mean, round which they will again vary. 

 Selection will keep up this mean to a definite standard so 

 long as the same conditions prevail. The mutation theory 

 assumes rigid types to be departed from only per saltum, 

 and this necessarily assumes also an inflexibility in races of 

 living organisms which we know does not generally exist. 

 On the contrary, we see everywhere that races are influenced 

 by changes in the environment, in the most marked and yet 

 in the most delicate and gradual manner. The frequency of 

 the "fluctuating variations" is enormously in favour of their 

 being the ground upon which new characters are built. They 

 occur always and in all directions. Mutations, on the other 

 hand, are comparatively rare, and also occur in all directions, 

 so the chance of their being in the direction of adaptation 

 is proportionately a small one. In favour of the action 

 of the fluctuating variations is their very smallness, for a 

 great number of them may occur at the same time without 

 throwing the organism completely out of harmony with its 

 environment, which would certainly end in its death. Even 

 single mutations are comparatively rare, and yet if adaptation 

 is brought about by mutations, we must imagine several 

 mutations occurring in various parts of the organism simul- 

 taneously in a favourable direction. The argument that 

 small differences are of two kinds, mutations and fluctuating 

 variations, seems to involve a quite unnecessary assumption. 

 It is admitted that the two kinds cannot be distinguished 

 from each other except by breeding. So many possibilities 

 of error must exist in applying such a test, that even were 

 some such assumption necessary to the plausible explanation 

 of evolution, it would hardly meet with general acceptance. 

 As a matter of fact such an assumption seems quite un- 



