522 JENNINGS— HEREDITY IN PROTOZOA. [April 24, 



chance of avoiding differences due to environment and growth. It 

 is conceivable that congenital hereditary variations exist, but that 

 they are few in number compared with those due to environment 

 and to slight differences in ways of living, so that in our selection 

 we always get the mere environmental variations. There are decided 

 differences between the specimens of the same line beginning fission, 

 as Table XIII. (page 442) well shows; here the length varied from 

 156 to 204 microns. It is possible that selection among specimens 

 beginning fission might have a better chance for success. I have 

 attempted this, but it is extremely difficult; I hope to return to it. 



We must consider, however, that if the non-inheritable differ- 

 ences are so much more numerous and marked than the inheritable 

 ones as to render conscious selection by human beings ineffective, they 

 would apparently have the same effect on selection by the agencies 

 of nature. The same ground for selection offered by heritable varia- 

 tions is offered so much more fully by those not heritable that there 

 would be as little effect in selection by nature as in selection by man. 



Certainly, therefore, until someone can show that selection is 

 effective within pure lines, it is only a statement of fact to say that 

 all the experimental evidence we have is against this. The results 

 set forth in the present paper tend to strengthen that explanation of 

 the observed facts regarding selection, regression, etc., in mixed 

 populations, which is set forth by Johannsen (1903). We need not 

 discuss these in detail here ; they are essentially as follows : 



1. Selection in a mixed population consists in isolating the 

 various different lines already existing. 



2. If selection is made, not of single individuals, but of consid- 

 erable numbers having a certain characteristic, then by repeated 

 selection it will be possible to approach nearer and nearer to a 

 certain end. 



Thus, if we select from such a heterogeneous collection as is rep- 

 resented in Table LXI. all the larger individuals, we shall have taken 

 representatives of many different lines. Our selection will include 

 the larger individuals of lines of median size, as well as the average 

 individuals of lines of large size. The progeny of this selected lot 

 will then consist of various lines, some larger, some smaller, but with 

 the average higher than in the original collection. Another selection 



