190 



THE AMERICAN NATURALIST [Vol. LI 



question in these words. Thus the issue between Dr. Castle on 

 one side, and Johannsen and us on the other narrows, and there 

 need be no more difficulty as to the exact meaning of the term 

 unit character. As to the effectiveness of selection in genotypi- 

 cally homogeneous material, all the evidence so far adduced 

 shows that selection in such material is absolutely ineffective. 

 It is evident that selection in a population is usually effective, 

 but this only shows that in ordinary populations, even in so-called 

 pure strains of animals, there is a good deal of genotypic varia- 

 tion, or in other words impurity. 



The fact, for instance, that Dr. Castle's selection in hooded 

 rats was effective, shows that his material was not originally 

 pure for all the genes. In all those instances where the guar- 

 antee for genotypic purity of the material was reasonably good, 

 selection has, until now, proved ineffective. We need only point 

 to the fifty years of selection in wheats by the de Vilmorin fam- 

 ily, and to the numerous selection-experiments with clones of 

 Paramoecinm by Jennings and others. 



As to the so-called instances of the effectiveness of selection 

 on the genes themselves in alleged genotypically homogeneous 

 animal material, we repeat that the only way to show such an 

 effect in material which offers no sure proof of purity would be 

 to change a strain of severely inbred animals by selection to a 

 point removed from the rartge of the ordinary modification in 

 the material, continuing the inbreeding, and then, by contraselec- 

 tion, to bring the character under consideration back to its start- 

 ing-point. Since we wrote down this challenge to the believers 

 in the variability of genes, one such a series of selection-experi- 

 ments has been performed, namely, on flies, and in this series it 

 has been proved to be impossible to get the material back to its 

 original quality. 



According to Dr. Castle, Hoshino 's Table 6 shows the effect of 

 selection within a pure line. In the cases taken from Hoshino 's 

 paper, in which the progeny of an early-flowering and a late- 

 flowering individual of the same "pure line" can be compared 

 (in the original table there is one more case in which the earliest 

 parent gives the latest progeny) there are more instances in 

 which an early parent gives an earlier progeny than a late 

 parent, than cases in which an earlier parent gives a later 

 progeny. But if we examine the figures more closely, we observe 

 that the mean deviation of offspring from parents in the case in 

 which the earlier parent gives the earlier progeny is 0.52 day, 



