statistical analysis » <- As explained in the previous section, the 

 statistical analysis of tagging experiments depends mainly on a line 

 '^or .curve) representing the trend in annual adjusted returnso The inter- 

 cept of the line at year one (an extrapolation) is used to compute catch 

 mortalityo By another computation^ it also results in estimate of total 

 jommercial stock. The rate of decrease in annual returns is taken as 

 the total annual mortality. The remaining statistic, natural mortality, 

 is derived algebraicly from catch mortality and total mortalityo 



Since the trend curve is based on returns during a series of years, 



the statistics resulting therefrom, represent the conditions, not in any 



one year^ but are an average (not necessarily an arithmetic mean) for 

 the series of years. 



It has already been pointed out that the adjustments, particularly 

 that involving the size of the catch, may have affected the data in a way 

 that calls for modification of the above -de scribed computations. Instead 

 of that, determination for a nxxmber of periods of overlapping years might 

 point the way to an interpretation or correction of the statistics for 

 individual periods or even individual years. Thus far, however, records 

 are available for only two periods and these possibilities cannot yet be 

 examined. 



A further aspect of the computations so far made is that when Pacific 

 Northwest catch and Pacifio Northwest returns of fish released in California 

 are excluded different rates are indicated than when they are included. 

 This points either to the non-availability to the California fishery of 

 at least a portion of the stock that migrates seasonally to northern waters 

 or to some discrepancy of experiment or analysis so far unrecognized. 



Fortunately, there is reason to believe that many of the difficulties 

 of interpreting tagging restiits may disappear or be resolvable as more 

 data accumulate. 



The value of the tagging method as a largely independent method of 

 dstermining mortalities cannot be over=emphasized and it is particularly 

 encouraging that the computations have led to determinations practically 

 identical with those resulting from a preliminary application of the 

 method of vital statisticso 



At the same time it should be emphasized that both the tagging, 

 and the vital statistics^ methods depend, in the last analysis, on the 

 representative nature of the sample dealt iwi the At its present stage 

 of development, the method of vital statistics includes a much larger 

 sample, better distributed in time and space and more thorbughly adjusted 

 to exclude the effects of extrinsic influences than does the tagging 

 me thod u 



RECRUnMENT RESEARCH 



Determination of the amount of recruitment would be an outstanding 

 achievement, but would be, of itself, of only limited usefulness so long 



