608 Bulletin American Museum of Natural History. [Vol. XXXIV, 
the population which lived less than 32.5 days a8 adults had markedly longer 
embryonic periods not only than those with longer adult lives but than the 
general population. 
The results of the starvation experiments were not what I expected them 
to be for I had thought that those larva which fed for a long time would 
have laid up a large supply of reserve material which would enable them to 
withstand starvation in the adult stage better than those which pupated 
early. Perhaps the outcome is a resultant between this factor and the one 
suggested in the preceding paragraph as explaining the negative correlation 
found there. The two physiological conditions might largely neutralize 
each other and the result would be no correlation. At any rate, the fact 
is that no significant correlation was found between the ability of adults to 
withstand starvation and the length of the embryonic periods. In three of — 
the four cases the coefficient is less than the probable errors and in the fourth 
it is less than twice the probable error. The means (Table 2) show the same 
thing. The only case in which there is a possible relation shown is between 
the egg-larval period of the females and their ability to withstand starvation. 
Such as it is, it is in the same direction as that found when considering nor- 
mal adult life. 
It is clear that selective death rate is demonstrated with respect to these 
physiological characters when the adult flies are given all the food they can 
eat. When the adult flies are given no food no selection is demonstrated 
but it may nevertheless exist, being masked by complicating circumstances. 
The explanation, given above, of this masking ‘is not entirely satis- 
factory. There are still other complications as is shown by a study of 
_ the correlations between the durations of the embryonic periods. In the 
American Museum experiments where the temperature was that of the 
laboratory, 7. e. that at which the flies normally live, there was found to 
be a strong positive correlation between the duration of the egg-larval period 
and that of the pupal period. In the Carnegie Institution experiments, 
however, which were conducted at a higher temperature than normal there 
is no significant correlation between the duration of the embryonic periods. 
I have no idea what this difference, which is referred to again below, means. 
I am quite aware that in most of the correlations considered here the re- 
gression is not linear but I do not believe that the correlation ratio would 
alter the significance of the results. ? 
In the starvation experiments we have a pair of anatomical characters 
to consider. They were selected from among the many which might have 
been measured simply because they were easy to measure. They give a 
fair notion of the relative size of the individuals and while not likely to have 
been directly concerned in selection they are no less likely to have been 
