Payne: Variations and Selection 



33 



regression toward the normal, the percentage of normal flies 

 going to 65.58 and the mean of the F. extra bristled flies to 

 5.55. In the fourth and fifth outcrosses, the percentage of 

 normal flies and also the mean of the extras come back to 

 approximately what they were in the first and second out- 

 crosses. The percentage of normal flies in of the fourth 

 outcross is 44.48 and in the fifth outcross 47.36 ; the mean of 

 the extras in these two generations is 6.04 and 5.91 respect- 

 ively. The question might be asked whether the second 

 regression is of any significance or whether it is due to the 

 effect of the environment. It is certain that the environment 

 does affect bristle number, but to what extent and in what 

 way is problematical. I admit that the second regression in 

 the third Fo's might be interpreted as an effect of the 

 environment, altho conditions at all times were as nearly alike 

 as they could be made under room conditions. If this be 

 granted, we would have a regression in the first outcross, and 

 then no further change. But this would not satisfy Castle's 

 demands if multiple factors are present, as he holds that 

 there should be a further regression. On the other hand, the 

 second regression could as well be interpreted differently. It 

 will be noticed that the F^ flies in this outcross showed a high 

 percentage of normals (85.10) and the mean dropped to 4.183. 

 This is a marked change from the preceding and also from the 

 following F^'s. It seems to me very probable that in select- 

 ing the female parents for this outcross, individuals were 

 chosen which were not homozygous for the same factor or 

 factors which were present in the preceding and following 

 outcrosses. If such were the case a regression would be 

 expected in F^ and Fo. Some homozygous flies would appear 

 in F2 and in selecting the parents for the next outcross these 

 homozygous flies were chosen. Hence the percentage of nor- 

 mal flies and the mean bristle number of the fourth F.'s 

 come back to what they were in the first and second Fo's. It 

 is a matter of interpretation then whether we accept the ac- 

 tuality of the second regression. If we do not accept it, then 

 there is no further change after the first outcross. As stated, 

 this does not satisfy Castle's demands of multiple factors. 

 If we do accept the second regression as an actuality, then 

 we have results similar to Castle's, except we have two 

 regressions and a swing back in the direction of selection. 

 I can see no significant difference, however, in the two cases. 



