PRINCE WILLIAM SOUND, COPPER AND BERING RIVER SALMON STATISTICS 207 
other intervals but it is evident by inspection that the correlations at both 3 and 
6 year intervals would be insignificant. 
This high correlation at two different time intervals (four and five years) is natu- 
rally to be interpreted as indicating that the runs are made up of 4 and 5 year fish in 
approximately equal numbers, and it has seemed important to attempt to devise 
some measure of correlation that would take this into consideration. After trying 
various methods the simple scheme was adopted of correlating each catch with the 
average catch of the fourth and fifth preceding years. Thus the catch of 1910 was 
paired with the average for 1905 and 1906, the catch of 1911 with the average for 
1906 and 1907, etc. Such a procedure, of course, gives equal weight to the two parent 
years — which seemed advisable in this case on account of the nearly equal value of 
the correlation coefficients at 4 and 5 year intervals. Any other weighting, of course, 
could have been used if there had been any good reason, biological or other, for so 
doing. The results were interesting since r calculated in this manner proved to be 
0.89 ±0.040, a distinctly higher and more significant correlation than at either 4 or 
5 year intervals. This apparently confirms the interpretation that the Eshamy 
fish are predominantly 4 and 5 years old at maturity and that the two age groups 
are present in about equal numbers, or, more properly, are produced from each 
brood in approximately equal numbers. 
