692 THE AMERICAN NATURALIST [Vol. LI 



generation by the transformation of the dominant allelo- 

 morph. This recessive state of the hereditary substance, 

 however, has a tendency to revert into the original domi- 

 nant state. Such reversion is especially likely to occur m 

 vegetative cells, where each recessive allelomorph seems 

 to be able to revert independently. Consequently, in reces- 

 sive homozygotes the reversion generally will produce 

 heterozygotic cells, either one of the two recessive alle- 

 lomorphs being changed into the dominant. The hetero- 

 zygotic cells thus formed will give rise to partial fertility 

 in otherwise sterile plants. Again, the recessive allelo- 

 morph in heteroz3'gotic cells may be subject to similar 

 reversion, and such reversion may occur both in the 

 heterozygotic cells of sterile plants and in normal hetero- 

 zygotes. Here, however, heterozygotic cells will be 

 transformed into dominant homozygotic cells without 

 visible effect on the plant concerned. The consequence 

 of this reversion in the next generation will be that the 

 proportion of the dominant segregates may exceed the 

 theoretically expected figure. Finally, it may be assumed 

 that between Families A and B there exists a difference 

 in the reverting tendency of the recessive allelomorph, 

 which necessarily will effect the differences in both the 

 intensity of partial fertility of sterile plants and the devi- 

 ations in the segregation ratio. 



In Table III the segregating families derived from the 

 fertile plants of Family B are classified according to the 

 magnitudes of the deviations in terms of probable errors. 

 The true percentage for the recessive is assumed, in the 

 one case as 25 per cent, (the Mendelian ratio), and in the 

 other case as 21 per cent, (an arbitrary number). In 

 comparing the two different frequency distributions made 

 in this manner with the theoretical frequency distribu- 

 tion, it is observed that while the frequency distribution 

 of the deviations from 25 per cent, shows a considerable 

 discrepancy from the theoretical, the latter fits the fre- 

 quency distribution of the deviations from 21 per cent, 

 rather closely, the goodness of fit being P = 0.915. Con- 

 sequently, the ca. 4 per cent, deficiency of recessive segre- 



