64 BULLETIN OF THE BUREAU OF FISHERIES 



given year will be followed by a poor catch six years later, and vice versa. If, as 

 we have indicated, there has actually been a change in the Niishagak district from 

 an association between catches at five-year intervals to an association at fonr-year 

 intervals it is most interesting, and certain possible explanations may be sug- 

 gested. Most of the fishing in the entire Bristol Bay region has been carried on by 

 gill nets, and it is generally supposed that this type of gear has a tendency to select 

 the larger individuals in a run. With such a selective agency at work it is possible 

 that a race or several races of predominantly large, five-year fish might be so reduced 

 in numbers as to disappear almost entirely from the commercial catch. It may have 

 been that this occurred in the Nushagak district quite suddenly in 1919, when the 

 run at Nushagak was abnormally low. It is possible, though hardly probable, 

 that the selection of the larger and older fish by the giU nets has operated to change 

 the predominant ages of all the races that make up the Nushagak catch from five 

 to four years. In any event it does not seem likely that this high degree of cor- 

 relation between the catches at four-year intervals is due merely to the operation 

 of chance fluctuations. 



In the case of the other three districts in Bristol Bay the highest correlation occurs 

 between catches at five-year intervals. In the Kvichak and Ugashik catches there is 

 also a low but probably significant correlation at four-year intervals but there is no 

 significant correlation at six-year intervals. In the discussion of the cycles at Nusha- 

 gak we pointed out that a significant correlation between catches at any given interval 

 of years was strong indication that the prevailing age of the fish was of the same 

 number of years. If this be true, the data here\\'ith presented would indicate that a 

 considerable percentage of the Kvichak, Egegik, and Ugashik red salmon are in their 

 fifth year. We find a similar degree of correlation between catches at five-year in- 

 tervals at Karluk, where the fish are known to be chiefly 5 years old; and the main- 

 tenance for many years of a four-year cycle on the Frazcr River, where the fish were 

 predominantly 4 years old, is well known. We wish to emphasize again the fact that 

 no correlation of this sort can be expected to be very high, not only on accoimt of 

 inaccuracies in the data but because of the presence of more than one age group in the 

 catches, fluctuations in the relation of escapement to catch, possible fluctuations in 

 the percentage of fish of different ages coming from different broods, fluctuations in 

 the efficiency of spawning and the rate of mortality at different stages in the life 

 history, and various other factors. 



Table 4. — Coefficients o] correlation (r) between catches of red salmon in the four districts of Bristol Bay 



Figure 8 shows clearly that there is a distinct tendency for the catches in the four 

 districts to vary together, although at times, and especiaUy in certain districts, the 

 catches vary independently to a considerable degree. The extent of the correlation 



