Thus the fish assigned to one year class had an intrinsic quality in 

 coCTnon in addition to the number of rings on their scales, peculiar to their 

 class, vifhich persisted during three years of this study. It is fortunate 

 that this peculiarity applies to year class 1938, for io is that class on 

 which the most complete collection of material is at hand, going back to 

 first appearance in the bait fisheries, when identification of its age as 

 fish of the year was virtually beyond question. And it is that class from 

 which the material vras drawn on which the validity of age determination of 

 juvenile pilchards was proved (V/alford and Mosher, 19U3). Thus fish iden- 

 tified with. certainty as belonging to year class 1938 have been marked by 

 Nature with a characteristic that may continue to serve for distiiiguishing 

 them as a group for several additional years. 



Since the relation of scale length to body length can vary among year 

 classes, it is necessary, for approximating past grovifth as accurately as 

 possible, to determine a for each year class separately. To do this requires 

 measurements over a sufficient range of sizes - therefore years of growth - 

 to provide a significant measure of the regression. ViTith present material, 

 this vfas possible only for the four year classes mentioned above. Consequently, 

 in the folloviring discussion, averages concerning those year classes alone 

 have been adjusted as of formula (3) (page 13); and with the appropriate 

 a values (page 13 ). For the rest, formula (1) has had to suffice. 



Summarizing: To dei^ect important anomalies in the past growth of the 

 several year classes, the lengths at the past 'ages (f ) were calculated by 

 Lea's formula (1). These were averaged, and averages for year classes 1936 

 to 1939 v;^Gre adjusted to take into account departures from exact proportion.- 

 ality of scale' length to body length; in other words for the failure of the 

 regression of scale length on body length to intersect with coordinates 

 0,0.i/ Valu-J3 of X^ Y/ere obtained by subtracting Tj^ from Tj^-1. 



The results of thuse calculations for the first year's growth increment 

 (^ ), given in Table h, are remarkably variable. Among year classes 1932 

 to I93O collected in 1939-i;0, for example, the average first year's growth 

 increment (lq_) panged from 98 to 115 mm. Such apparent inconsistency could 

 be caused by changes virith age and size in the scale length-body length re- 

 lationship. It could also occur if increasingly larger portions of the size 

 range of each year class' became available to the fishery as the year class 

 became older. Or it might result from systematic errors in marking the lo- 

 cation of the annuli, vihich were somehov; associated •yv'ith age. 



-' These adjustments -v^ere made graphically as follovirs: The average total 

 length of a year class under consideration v/as located on the scale length- 

 body length regression line, and a line dravm from there to the origin of 

 the graph.- Each f viilue for that year class v/as then located on the latter 

 line; and the intersection betv/een, the ordinate passing through that point, 

 and the scale length-body length regression line gave the adjusted -value of 

 that T. 



110 



