294 
GENETICS: E. C. MacDOWELL 
are regularly positive and significant; where the means are irregulars 
and seem to depend mainly upon environment, the correlation coefficient, 
are irregular, plus and minus, and in general not statistically significant. 
It is important to understand that this failure of selection can not be 
explained by physical incompatibility in the soma. There is obviously 
plenty of room for many more bristles than are generally found and, in 
a few cases where very rare conditions happened to be experienced by 
individual flies, such extreme grades as 12, 13 and 16 extra bristles have 
been found, whereas the most frequent high extreme is 9 extra bristles, 
and the characteristic mean of the race seems to be 4.75. It may be 
concluded that at first there were differences in the germ plasm of the 
different flies and these differences were manifest in the somatic ap- 
pearance, even though environment was also causing yariations; later, 
due to the selection, such germinal differences between individuals as 
were clearly observed at the beginning were no longer found. 
As further evidence for this conclusion the following experiment may 
be described, which meets the objection that the narrow range of parents 
selected in the later generations might account for the absence of corre- 
lation. Accordingly, selection was suspended for three generations 
and as far as facifities would permit, all the progeny of one pair of flies 
in the 50th generation were bred in pairs for three generations without 
respect to their grades, the only restriction being that ma tings were made 
of flies of the same or within one of the same bristle grade, and only 
virgin females were used. All the offspring were counted and graded. 
In the second generation of this experiment over 4000 flies were graded, 
in the third generation, 27,000. 
For these generations the correlation coefficients were as follows: 
Males— 2d generation, r-0. 1436 ± 0.0995; 3d generation, r-0. 0271 ± 0.0391 
Females— 2d generation, r-0. 1378 =t 0.0997; 3d generation, r-0. 0221 ± 0.0391 
This shows that the higher grade parents did not have any tendency to 
produce higher grade offspring than the lower grade parents, and there- 
fore, that the failure of selection to produce further changes in the means 
of the race, is due to the absence of such genetic differences, as were 
originally present. 
From quite a different source comes evidence that the early genera- 
tions of selection modified the genetic constitution of the race. Start- 
ing from the second selected generation a race of low grade flies was 
formed in one generation by selecting low grade ffies as parents. The 
means of this race were entirely distinct from those of the high selected 
race; as the high race rose, this race remained low. Since it is known 
