No. 588] EXPERIMENTS IN MASS SELECTION 723 



tant ' ' race, a highly stable plus variation ; but much oftener 

 gradually, as has occurred continuously in both the plus 

 and the minus selection series. The permanency of these 

 cumulative changes we have tested by repeated crossing 

 of both selected races with the same wild race. The first 

 cross seems to undo to a slight extent the work of selection, 

 causing regression in both plus and minus selected races, 

 but a second back cross with the wild race causes no fur- 

 ther regression. Thus, plus-selected rats of mean grade 

 3.45 were crossed with wild rats and the recessive char- 

 acter was recovered in F 2 in 75 individuals, 24 per cent, 

 of the entire generation. These 75 extracted hooded rats 

 were of mean grade 2.89, a regression of .56 on the mean 

 grade of their hooded grandparents, which is about double 

 the regression shown by the plus selected race when not 

 crossed with wild rats. It seems proper therefore to at- 

 tribute to the wild cross a part of the regression observed 

 in this case and this I have expressed by saying that cross- 

 ing the selected race with wild rats tends to undo the work 

 of selection. The suggestion was tentatively adopted by 

 Phillips and myself that this undoing consisted in the re- 

 moval of "modifiers" of some sort, possibly independent 

 Mendelizing factors. If this explanation were correct, 

 further crossing with wild rats should tend still further 

 to "undo" the work of selection, so that ultimately the 

 extracted hooded race should return completely to its orig- 

 inal modal state, the zero grade. To test this matter, 

 extracted hooded rats ranging from grade +2 to +4 

 (mean grade 3.01) were crossed back a second time with 

 pure wild rats. The theory of independent modifiers 

 would lead one to expect further regression as a result of 

 this cross, but no regression was this time observed. In- 

 stead an advance of .32 took place bringing the mean of 

 the twice extracted hooded recessives back to about the 

 grade of the uncrossed race. The mean grade of the once- 



ber of their twice-extracted hooded grandchildren, was 

 3.01; the mean of the 263 hooded grandchildren was 3.33. 



