606 | Bulletin American Museum of Natural History. [Vol. XXXIV, 
in which selection acts on the basis of correlation in such a way that the 
surviving population has a lesser positive or a greater negative correlation 
than those which perished. Furthermore Pearson ! has shown that selec- 
tion of the mean and variability of a character influences very markedly the 
correlations between this and other characters. 
The result of these considerations is that we are forced to take a wider 
view of selection than has ordinarily been done. If that portion of the popu- 
lation which perishes differs significantly in either mean, variability or cor- 
relations from that portion which survives we must admit that natural 
selection has been effective. Furthermore, although it is relatively easy to 
demonstrate by statistical methods whether or not selection has influenced 
a given character, it is impossible in the present stage of science to determine 
just what the basis was upon which selection worked. If characters A, B 
and C are correlated in their variabilities, selection acting directly upon the 
mean of character A would change not only the means and variabilities of 
characters B and C but the correlations among the three characters. If we 
studied only B and C we would find that selection had acted but might be 
at a loss to explain its action. The only thing to do, in a case as compli- 
cated as is the problem of selection, is to accumulate facts bearing on the 
subject, keeping the various hypotheses in mind and leave it to suture 
generations to find out the right. 
The present paper concerns the Pomace Fly, Drosophila ampelophila. 
There are two sets of experiments. In one, carried on at the Carnegie 
Institution’s Station for Experimental Evolution, the flies were reared at a 
temperature kept rather close to 20° C. and the adults were given water but 
no food. In the other, carried on at the American Museum of Natural 
History, the flies were reared under normal, 7. e. uncontrolled, temperature 
conditions and the adults were carefully fed. The only unnatural condition 
in the second set of experiments, as far as could be determined, was that the 
adults were not allowed to mate. In both sets of experiments the relation 
of physiological characters, the duration of the embryonic periods, to the 
duration of adult life was studied and in one of them two anatomical char- 
acters also were studied. On account of the practical difficulty of determin- 
ing the exact time of hatching, the egg and the larval periods were combined 
in the records. 
There are two ways of determining whether or not selection acts (directly 
or indirectly) upon the actual size of a character: we may compare the mean 
of the character among those which perished with that among those which 
survived or we may calculate the correlation between the size of the charac- 
ter and the ability to survive. Both methods are used here. 
1 Pearson, K. 1902. ‘On the Influence of Natural Selection on the Variability and 
Correlation of Organs.’ Phil. Trans. Royal Soc. London, Series A, Vol. 200, pp. 1-66. 
