in the 1926~30 bandings is a mean of conditions occurring from September 1, 
1527 to August 31, 1932. The successive mortality rates of 6h, 56, and 
46 per cent for this cohort thus represent points on a 5-year moving 
average. The birds banded in the cohort from 1931 to 1935 represent 
similar average points for later years (1932-37, etc.). They are, in 
fact, a 5-year moving average that suggests a definite trend. The 
possibility that this might be associated with trends in bag limits 
or length of season is examined in figure 17. It is clear that we 
have here strong presumptive evidence for a correlation between 
mortality rates and hunting pressure throughout this period. While 
the separate effects of season or bag limit require a more detailed 
breakdown of the data, the increase in season and mortality. rates at 
the right-hand end of the curves deserves notice. It is perfectly 
possible that these adult mallards might have become less vulnerable 
to gunning as they grew older, and that the change in mortality rates 
was in part due to this increase in experience and in part to hunting. 
This possibility will be explored in other samples. 
Mortality Rates on the Pacific Coast 
British Columbia and Oregon Birds.--This exploration of 
changing mortality rates was continued in a study of 138), adult male . 
mallards banded on the Pacific coast, chiefly by A. J. Butler in 
British Columbia and by George M. Benson and his colleagues in the 
Fish and Wildlife Service at the Malheur National Wildlife Refuge 
near Voltage, Ore. The raw mortality data are given in table 62, 
where they are converted into survival data, and mortality "rates" 
are calculated for all birds available. Thus for the year 1928-29, 
when 162 were reported shot out of a cohort of 252 banded in 1927, 
9 birds were similarly reported out of a cohort banded in 1926 out 
of which 18 were known to be alive. We can say, then, that 171 were 
shot (162 - 9) out of 270 alive at the start (252 - 18). This 
technique of adding different cohorts to obtain a mortality rate for 
a given year I call the method of cumlative cohorts. It has the 
merit of using all the banding data available and increasing the size 
of samples for study. It has a potential defect in that it cannot 
always be defended as a random sample of the adult population, since 
in some years the proportion of rather old adults in the sample my 
depend on banding work in the past and so vary considerably. If 
increasing age and experience make for a higher survival rate in 
mallards, a long series of cumlative cohorts such as this one would 
tend to show progressively lower mortality rates. It should, however, 
at least betray marked increases in mortality rates when these occur 
in specific years. The rates based on cumulative cohorts (given at 
the bottom of table 62) did not differ by more than 3 percentage 
points from those calculated for single cohorts (shown in the last 
column on the right of this table). 
The. explanation for the changes in mortality rates for 
these Pacific coast mallards is at least partly evident in figure 18. 
The first sharp drop in rates occurred when a presidential proclama- 
tion cut the United States hunting season from.107 days to 30 days 
143 
