
similar types of errors. This table seems to me to be fairly typical 
of deviations encountered in the U. S. banding files, but the number 
of duplicate cards seems somewhat low in my experience with other. 
blocks of cards. The need to check banders' reports, underemphasized 
by the table, is especially important where birds have been hand-reared. 
In one sample that I once gathered together for a life table, about half 
the birds subsequently proved to be of game-farm origin. 
Verification in this study.—-The frequency of minor discrepancies 
and major arrors In the V.8. bandine files imposes a intolerable Sinden 
on any investigator attempting to carry out a scholarly study. In one 
sample of nearly 200 cards, 39 per cent of the letters could not be 
located for the years between 1930 and 1943. Recovery letters are 
filed by banders' names and in an approximate order of receipt. The 
task of verifying cards against these letters proved to be extremely 
time consuming. As I proceded, I found that large blocks of cards 
shrank to a very small sample when one rigidly defined the type of- 
recovery to be studied (as say, juveniles banded before September 1, 
or known adults not banded in the hunting season). To throw out 39 
per cent of the cards because the dates of recovery could not be 
checked reduced most samples to levels of mathematical absurdity. 
This seemingly leaves one pinned on the opposing horns of a dilemma 
provided by the need for sound scholarship and mathematically based 
logic. 
My solution of this problem was to use cards in spite of the 
failure to verify them against recovery letters, and to avoid drawing 
categorical conclusions so refined that they depended on the sex of a 
bird or the exact month it was recovered. The present report therefore 
contains no refined calculations of life expectancy (mean after life- 
time) such as those computed by Farner (1945). In actuarial work, this 
failure might be serious. In animal population studies, it is often 
counterbalanced by satisfactory calculations of annual survival rates 
for subadult and adult classes. When direct computations of mean after 
lifetime consider the month in which each bird is reported to have been 
found dead, this date is slightly biased for large birds with con- 
spicuous skeletal remains. It is frequently biased, too, by returns 
from hunters, and always depends upon the assumption that reported 
mortality occurs in the same monthly proportion as unreported mortality. 
During the course of my work, I systematically tabulated 
recoveries according to the months alleged for each of them. Birds 
found dead could thus be compared with birds shot; those captured 
alive in Texas could be compared with those captured alive in Oklahoma, 
and so on. Thus, samples of equal bias were contrasted with each 
other. These comparisons in some cases approximate what is happening 
in nature; in many they may not. Their chief use in this study arose 
in the search for differences between types of recoveries. Even when 
recovery dates have been verified with extreme care, their monthly 
variation in highly migratory species may be importantly distorted by 
geographic variations in human literacy, language, and cooperativeness. 
21 
