THE EVOLUTION OF LIVING BEINGS. 25 



would generally be detected by the presence of a 

 number of aborted ovules and aborted pollen- 

 grains. 



Forms showing such abortion are however from 

 the very beginning suspect, so that one does well to 

 exclude them from experiments necessitating the 

 use of pure species. 

 3dly. To make the Davis-test as reliable as possible, we 

 must furthermore cross the form to be tested with 

 more than one other form as pure as obtainable. 

 This will f. i. be absolutely necessary in the 

 case of albinotic forms, because in these cases 

 constitutional differences remain invisible in 

 the F a generations, obtained by crossing two al- 

 binos of different constitution, f. i. two of the 

 three kinds of white mice described in Bateson's 

 Principles on p. 75. 

 In such cases crossing with a colored form reveals 

 the difference. 



But from the fact that the Fj generation of crossed 

 albinos does not necessarily reveal the difference in 

 constitution between those albinos, it is safe to con- 

 clude that the possibility exists that the F x generation, 

 obtained by crossing two colored individuals, may also, 

 in some cases, fail to reveal certain differences in con- 

 stitution existing between these individuals. 



It will therefore be safest to test for specific purity in the 

 following way.: 



istly. cross the form to be tested with as many sup- 

 posedly pure species as obtainable, reciprocally. 

 If the F x generations so obtained, are all uni- 



