34 THE VARIATION OF ANIMALS IN NATURE 



necessary to kill off part of each family, but the greatest care 

 must be taken to avoid any selection. With adequate statis- 

 tical treatment such material may still lead to a definite 

 conclusion. 



This difficulty arises in its most acute form when only 

 some of the experimental animals show a modification. It 

 has often been the practice to carry on the stock only from 

 these modified individuals, thus introducing a stringent 

 selection in the direction of the modification. Two suggestions 

 may be made in this connection. First, repeated experiment 

 with different strains may show that the modification always 

 tends to arise in the experimental animals and never in the 

 controls. If the experiment stops when the modified indi- 

 viduals first appear, no selection can have been exercised in 

 that particular direction. If repeatable results of this sort can 

 be obtained, the effect of selection in later experiments is 

 relatively unimportant. Secondly, if the modification is an 

 induced mutation and is permanently heritable from the start, 

 selection is evidently only a secondary issue. To prove that 

 there has been an induced mutation is chiefly a matter of 

 reduplicating experiments with different stocks. 



3. Persistence of the Modifications. — It is necessary to 

 distinguish at the offset between induced mutations and any 

 other sort of induced modification. Induced mutations 

 resulting from subjection to high temperature or to X-rays 

 are now well known in Drosophila and in mice. The dis- 

 covery of other equally effective agencies would be a matter 

 of great interest ; but it is evident that experiments of this 

 sort throw no light on the point at issue here. If one admits 

 that it is unlikely that mutations are really ' spontaneous,' the 

 discovery of agencies which raise the mutation-rate need not 

 excite great surprise, even when the mutations tend to be in a 

 particular direction. The question is whether there is any 

 process by which modifications gradually become hereditarily 

 stable. There is a sharp distinction here from mutations 

 which are stable from the start. 



To prove that an induced modification gradually acquires 

 stability is certainly a difficult matter and there is a danger 

 that experiment will lead to a vicious circle in interpretation. 

 If the process alluded to can occur, then the modification 

 induced by experimental conditions must be expected to be 



