adults, particularly adult males, tends to be concentrated in 
wintering areas, whereas a greater proportion of the kill of 
immatures occurs farther north near the breeding grounds early 
in migration. If wings were collected throughout all the harvest 
areas, these differences in distribution of the kill would not affect 
the data seriously. However, when age ratios are based on only a 
portion of the total harvest area, it is necessary to have some 
method for weighting the various banded samples for the size of the 
kill they represent. It is hoped that more pre-season banding in 
the future will permit an evaluation of the feasibility of weighting 
kill data for differential vulnerability. 
The location of the recoveries from flying mallards banded 
before the hunting season in Saskatchewan is shown in Figure 6. Only 
one of 31 recoveries in the Mississippi Flyway occurred in the north- 
ern tier of States. This suggests that birds coming from Saskatchewan, 
the principal drought area, tended to be taken in the southern and 
western portion of the Flyway and this wuld help explain the low 
age ratios among mallards taken in southern States. 
Sampling Error 
As used here, the term sampling error means the extent to which 
chance might cause a measurement based on a sample to differ from 
the true measurement in the population being sampled. It is very 
difficult to estimate the magnitude of sampling error of estimates 
based on the wing-collection data. First, the structure of the 
sample is complex. Second, there are several reasons why each 
wing received may not represent an independent random sample from 
all waterfowl taken in a specific area: (1) the sample of hunters 
was based on the lists obtained for the Mail Questionnaire Survey. 
Hunters' names were obtained from post offices selected at random, 
and as a result, the residences of these hunters tended to be 
clustered around the sample post offices rather than distributed at 
random throughout the State. The effect of this clustering is 
evident in Figures 3 and 5, and explains why the wing-collection 
samples in Arkansas and Louisiana were heavily concentrated in a 
small portion of the State; (2) the wing-collection samples are 
samples of the kill of a specific group of hunters rather than a 
random sample of the ducks taken by all hunters. It seems likely, 
for example, that the kill of a given hunter would not be a random 
sample of the species composition of the entire kill because of the 
particular areas in which a given person hunts; (3) the wings 
submitted by a certain hunter are taken on the day he hunts, rather 
than in a random manner during all days throughout the seasone 
-40- 
