THE ORIGIN OF VARIATION 33 



several generations sorts out the strains. These may differ from 

 one another in all sorts of characters, both morphological and 

 physiological. If an environmental factor modifies the appear- 

 ance or physiology of an animal, it is always necessary to make 

 sure that similar modifications, if not perhaps of the same 

 degree, do not occur in certain strains in nature. 



There are two ways of guarding against this source of error. 

 The most satisfactory is to use an inbred stock. Ten genera- 

 tions of close inbreeding will probably isolate a reasonably 

 homogeneous strain. In many case:>, however, this pro- 

 cedure would be very lengthy or even impossible. The only 

 method is to employ adequate controls, which indicate that 

 the modification does not occur normally in untreated portions 

 of the same stock. It is impossible to say how many control 

 animals should be maintained ; in a variable species the 

 number necessary for stringent experimental procedure might 

 be so large as to make some preliminary inbreeding almost 

 essential. Even with large numbers of controls a mutation by 

 a coincidence may happen to arise in the experimental animals, 

 but the reduplication of experiments with different stocks 

 reduces the risk of misinterpretation. 



2. The Elimination of Selection. — The experimental treat- 

 ment to which animals are subjected frequently causes con- 

 siderable mortality. If the survivors show some modification, 

 it is always possible that the mortality has been selective 

 and the survivors are that part of the original stock which was 

 genetically fitted to live in the novel environment. The 

 ' modification ' of the survivors may be therefore only the 

 expression of their particular genetic constitution. Such 

 forms will be especially liable to lead the investigator to wrong 

 conclusions, because their characters will of necessity be 

 inherited. 



The safest way of guarding against this error is to bring 

 to maturity every individual of every family throughout the 

 course of the experiment. If the experimental treatment 

 necessarily leads to considerable mortality, it may be almost 

 impossible to arrive at any convincing result, though the use 

 of highly inbred stock would be a great safeguard. In certain 

 cases (many insects) the size of the family is so great that the 

 stock would rapidly become unmanageably large if every 

 specimen was allowed to breed. In these circumstances it is 



