GENETICS: E. C. MacDOWELL 
295 
from the correlation coefficients, that the higher grade flies were pro- 
ducing higher offspring at this time, it is entirely within expectation 
to find that the lower grade flies could produce lower grade offspring. 
On the other hand, when flies of similar low grades were selected from 
the 16th generation of the high selected race, and the selection of low 
grade parents continued for 8 generations as a return selection series, 
the means of the offspring were not significantly lower than those of the 
corresponding generations in the high selected race. In fact the suc- 
cessful result of selecting low grade parents of the same grades in the 
second generation was not even approximated. The correlation coeffi- 
cients indicated that flies with different grades were not differentiated 
genetically in the 16th generation; these breeding tests support the sam€ 
conclusion. The Hne from low parents started from the 16th genera- 
tion, was lost after 8 generations; since low grade flies are the smaller 
ones, this selection, besides separating low grade flies, isolated weak ones. 
This accounts for the reduction of vigor and the final loss of the race. A 
second return selection was started from the 27th generation of the high 
race. For three generations the means of this line closely resembled 
those of the high race, but the weakening effect was soon manifest in 
the small numbers of offspring and the lowered means, and shortly, in 
the loss of the race through the failure of the selected flies to reproduce. 
A still further test has been made of the conclusion already stated. 
If the dissimilar genetic behavior of the unselected and the long selected 
flies rests on a sorting out of differences in the germ plasm, it should be 
possible to bring back these differences by crosses with flies that might 
be supposed to have germ plasm still bearing these differences. Flies 
from the 1 7th generation were crossed with wild normals ; from the extra- 
bristled flies that appeared in the second generation, low grade parents 
were selected. These at once estabhshed a low grade race that gave 
means constantly and unquestionably lower than the high selected race; 
no variation in the environment was great enough to confuse them. 
This low race did not show any signs of weakening on account of the 
selection of low grade flies; it was continued for 19 generations and gave 
large families. The reason for this is that all the flies, even the large 
well fed ones, had fewer extra bristles than the flies in the uncrossed high 
selected race; their germ plasm was different. It is very evident that 
the results of this crossing can not be explained on the basis of non- 
genetic physiological causes, as Castle^ has suggested may explain the 
very closely parallel results of his crosses with self-colored and hooded 
rats. In spite of this recent interpretation, it is well to note that Castle 
has given unmistakable evidence that his crosses involved genetic modi- 
