364 PROCEEDINGS OF THE ACADEMY OF {[June, 
experimental indications as to what food items are unattractive or 
distasteful to the birds, thus, are proved to be misleading in 42 cases 
out of 108. This makes a percentage of error of 38 (which will 
grow larger), enough to entirely invalidate the data. Furthermore, 
it is not probable that the data from any other series of experiments 
are any more reliable. The conclusion cannot be avoided, therefore, 
that the rejection of various items of food by captive animals does 
not prove that these items are rejected by the same species under 
natural conditions. 
CoNCLUSION. 
It has been demonstrated that behavior of captive animals toward 
food is not a reliable indication of what wild individuals of the same 
species would do in the presence of the same food. In other words, 
since the feeding habits of an animal in captivity may vary widely 
from its known habits in the natural state, there is no avoiding the 
conclusion that the results obtained under experimental conditions, 
do not indicate the part the animal might play in natural selection. 
We must conclude, therefore, since acceptances and rejections 
in experiments bear no close relation to food preferences under natural 
conditions, that the value of experiments to determine the efficiency 
of warning colors, and other protective adaptations of prey, is very 
questionable. Having no certain value in themselves, they must 
be checked up with definite knowledge of the natural food habits. 
This information is obtained by collecting animals with freshly 
captured prey and by examination of pellets, castings, and the 
contents of stomachs or other portions of the alimentary canal. 
There is no possibility of going back of such evidence on the choice 
of food, nor is there any need of so doing. 
Since this evidence is sufficient in itself, and since experimental 
data must be supported by it to be worthy of any consideration, why 
perform the experiments? The same time expended in collecting 
trustworthy data regarding the natural food habits of animals would 
bring much greater returns, and the result would be truth, not 
imaginative inferences from abnormal behavior. 
