HYPOTHESIS OF MULTIPLE FACTORS 193 



(several hundred seeds) in F2. This result was so surprising 

 that one cross which had yielded seventy-eight grains of 

 wheat in F2 was followed into F3, with the following result: 



Expected 



50 plants gave only red seed (being homozygous) 37 



5 " " approximately 63 R : 1 W (being trihybrld) 8 



15 " " " 15 R : 1 W (being dihybrid) 12 



8" " " 3R:1W (being monohybrid) 6 



" " all white 1 



The interpretation given by Nilsson-Ehle is this. The red 

 variety used in this cross bears three independent factors, 

 each of which by itself is able to produce the red character. 

 Their joint action is not different in kind from their action 

 separately, though possibly quantitatively greater. The F2 

 generation should contain one white seed in sixty-four. It 

 happens that none was obtained in this generation. The 

 next generation should contain, in a total of sixty-four indi- 

 viduals, the sorts actually observed as well as a sort which 

 would produce only white seed, the progeny namely of the 

 expected white seed of F2, but as that was not obtained, the 

 all-white plant of F3 could not be obtained either. The ex- 

 pected proportions of the several classes in F3 are given for 

 comparison with those actually obtained. The agreement 

 between expected and observed is so good as to make it seem 

 highly probable that Nilsson-Ehle's explanation is correct. 

 Corroborative evidence in the case of maize has been obtained 

 by East, and in shepherd's-purse by Shull. 



This work introduces us to a new principle which has im- 

 portant theoretical consequences. If a character ordinarily 

 represented by a single unit in the germ-plasm may become 

 represented by two or more such units identical in character, 

 then we may expect it to dominate more persistently in 

 crosses, fewer recessives being formed in F2 and subsequent 

 generations. Further, if duplication of a unit tends to in- 

 crease its intensity, as seems probable, then we have in this 

 process a possible explanation of quantitative variation in 

 characters which are non-Mendelian, or at any rate do not 

 conform with a simple Mendelian system. Consider, for 



