
in age distribution to unreported mortality in that sample? This question 
is no easier to answer than the preceding one. It may be rephrased in 
these terms: When a thousand black ducks are banded and only 150 are 
ultimately reported by the general public, are these 150 an unbiased sub- 
sample of the population? How can one be sure that the age distribution 
of the reported dead birds is similar to the age distribution of the 850 
that never are reported? At the present time, we can say that the relia- 
bility of the subsample reported varies according to at least two factors. 
A gradually increasing rate of band loss will obviously mean that the 
reported mortality (or survival) is atypical of what is occurring among 
the birds that are unreported. The seriousness of this possibility in 
the large, salt-water birds has already been mentioned. At the present 
time, there is little published evidence that band loss occurs among the 
smaller passerine birds, many of which have been retrapped over long 
periods when wearing and weakness in the bands could easily have been 
noticed. (My wife and I have, however, seen a blue jay about to lose 
a very old band, and we have good reason to believe that it did so sub- 
sequently. We have also seen a male cardinal remove a band with its 
beak.) A second source of bias occurs when the method of recovery 
does not operate at all ages, or operates unevenly. This appears to 
occur especially where learning and experience coincide with age and 
reduce a bird's vulnerability to some mortality factor like hunting 
which is the source of the reports being studied. If mortality reports 
from other sources are few in number and little affected by this learn- 
ing, reports of juveniles shot will exaggerate the proportion of mortality 
occurring in the first year of life. (This phenomenon will be demonstrated 
in table 13 later.) There are potentially at least two ways to discover 
this type of bias--(a) by productivity analyses and (b) by internal com 
parison of different methods of recovery. 
Checks on bias 

Productivity analyses.--Once a life table is constructed for 
the subadult and adult stages of the life span, ornithologists are some- 
times in a position to calculate the total number of eggs the adults in 
such a table would normally have laid each year, how many young they 
normally would have hatched, and approximately how many fledglings on 
the average should have left their nests. It then becomes possible to 
estimate if the young produced by such a sample of adults will balance 
the mortality of the adults. This type of test enabled Lack (19l3b) to 
recognize the bias introduced by gunning in life tables he set up for 
lesser black-backed and black-headed gulls; it also convinced Kortlandt 
(1942) that a serious loss of bands was occurring in his group of 
European cormorants. Lack (19l43a, 1943b, 1946a) and Farner (195) found 
no important bias in tables they constructed for passerine species. 
Is this test crude or precise? The answer depends in part on 
the investigator's knowledge of the numerical trend in his population, 
the age at which the adults begin to breed, and the availability of 
satisfactory data on nesting success. As table 6 demonstrates, the 
28 
