No. 518] 



VARIATION 



81 



The conception is simple and is capable theoretically of 

 bringing in order many complicated facts, although the 

 presence of fluctuating variation will be a great f.-idor in 

 preventing analysis of data. I have thought of only 

 one fact that is difficult to bring into line. If 8AA, SBB 

 and 8CC all represent homozygous twelve-rowed ears— 

 to continue the maize illustration— and none of these 

 factors are allelomorphic to each other, sixteen-rowed ears 

 should sometimes be obtained when crossing two twelve- 

 rowed ears. I am not sure but that this would happen if 

 we were to extract all the homozygous twelve-rowed 

 strains after a cross between sixteen-row and eight-row, 

 and after proving their purity cross them. In some cases 

 the additional four-row units would probably be allelo- 

 morphic to each other and in other cases independent of 

 each other. On the other hand, this is only an hypothesis, 

 and while I have faith in its foundation facts, the details 

 may need change. 



Castle has raised the point that greater variation should 

 be expected in the F x generation than in the F, genera- 

 tions when crossing widely deviating individuals showing 

 variation apparently continuous. If the parents are 

 strictly j3ure for a definite number of units, say for size, 

 a greater variation should certainly be expected in the F 1 

 generation after crossing. But considering the diffi- 

 culties that arise when even five independent units are 

 considered, can it be said that anything has heretofore 

 been known concerning the actual gametic status of 

 parents which it is known do vary in the character in 

 question and in which the variations are inherited, for tin- 

 race can be changed by selection within it. It may be, too, 

 that the correct criterion has not been used in size meas- 

 urements, for, as others have suggested, solids vary as the 

 cube root of their mass, whereas the sum of the weights of 

 the body cells has usually been measured and compared 

 directly with similar sums. 



Attention should be called to one further point. Many 

 characters in all probability are truly blending in their 



