in the legal hunting season number only 19h; the 499 
3 parallel Saskatcheran 
recoveries may be biased because so many of them repres j 
at or near Yorkton. : Present birds banded 
In another paper (Hickey 1951) I have shown that, of 15h, 
Manitoba-banded birds which were later reported, 15 per cent (actually 
1.9 per cent) came from Arkansas and Lousiiana, and that 17.6 per 
cent of 499 Saskatchewan recoveries were likewise reported from the 
same region. Now if the mallard populations breeding in these two 
Canadian provinces were exactly equal in numbers, and these two 
banding samples were perfectly randomized and free of sampling 
errors, it would follow that the ratio of Manitoba to Saskatchewan 
birds in Arkansas-Louisiana is as 14.9 is to 17.6. If the Manitoba 
population was twice that of the Saskatchewan one, the ratio would 
be as 29.8 is to 17.6. The actual ratio can be determined if ran- 
domized samples of the Arkansas~Louisiana population are banded and 
the number of breeding-season recoveries are tabulated for the two 
prairie provinces. From Hochbaum's (19) work in Manitoba, I infer 
that the breeding season may be taken as the months of May, June, and 
July; the exact pericd when no significant number of transients are 
present deserves further study. No thoroughly randomized banding 
operations have been carried out in Arkansas-Louisiana, but the work 
of McIlhenny at Avery Island, Gordon at the Rainey Sanctuary, and 
Van Huizen at the White River Refuge may be used to continue the 
illustration. Of the mallards banded in Arkansas-Louisiana, lk 
have been later reported in the "breeding season" in Manitoba and 
17 in Saskatchewan. We therefore have an equation 
I = 0.149 
Tl? O.176S5 
with M and S the two unknowns. Here 0.97M = 1.0S. The result, that 
the mallard population of Saskatchewan is equal to 97 per cent of 
the mallard population of Manitoba must, of course, be subject to 
sampling variations and to potential biases already mentioned. 
: The pertinent details of a series of similar tests are 
given in table 56. Data furnished by the individual reference areas 
are often so few that the different results may well be the product 
of small-sample variations. I have no other satisfactory explana- 
tion for the divergent values given when Oklahoma is used as a 
reference area. G. A. Swanson has pointed out to me that the validity 
of these comparisons lies in the implication that the two migratory 
populations are banded in a reference area in which the migrants 
have become thoroughly mixed together. It would be improper in such 
analyses to consider North Dakota as a reference area in the compari- 
son of birds breeding in two nearby areas like Saskatchewan and 
Manitoba e 
The data and calculations given in table 56 illustrate a 
technique that could not be further explored in the present study. 
The birds banded in Canada (at least during midsummer) cannot be 
133 
