332 BULLETIN OF THE BUREAU OF FISHERIES 



where the winter rings are not so sharply mari^ed, but a gradual transition from sum- 

 mer to winter is seen [p. 90]." As is the case with the circuli, the number of lamella; 

 deposited each year varies as the growth rate of the scale. Lea found further that 

 the outer or "upper covering layer is of almost equal thickness at the edge and near 

 the center of the scale and evidently does not grow thicker; it is thus easy to under- 

 stand that the winter rings, for instance, upon the surface of this layer, continue 

 equally distinct many years after formation * * * [p. 89]." "The scale may thus 

 be considered as a greatly flattened cone composed of fibrillary plates * * * 

 This cone is evidently covered entirely by a nonfibrillary layer, on the upper side of 

 which, however, is found finely marked relief which gives the scale its characteristic 

 appearance [p. 87]." 



Miss Lee's fifth suggestion, then, does not seem tenable. We may look for some 

 other and more plausible factor or factors to account for our paradoxical results. 



It was obvious in our discussion of the scale-diameter measurements of Table 22 

 (p. 331) that there we were, in reality, confronted with Lee's "phenomenon of apparent 

 change in growth rate." This was to be expected but only on the assumption that 

 the scale-diameter measurements and the computed length values based on them 

 are correlated more or less positively, that when the "annidar" scale diameters are 

 large, the lengths calculated from them will be large, and when small, the lengths 

 calculated from them will be small. It is conceivable that such a direct correlation 

 does not exist, as length computations for a particidar year vary as the proportion- 

 ate length of the scale -diameter of that year in the total length of the scale and not 

 as the actual length of that diameter. A length calculated from a large scale diameter 

 may be small, and vice versa. That this is not generally true is evident from the 

 following facts. It has been shown already that the bigger herring of an age group 

 averaged larger at the end of the first year of life than the smaller fish (Table 5). 

 Computations show that the average length of the scale diameters of the first year 

 of hfe is consistently greater in these large fish than in the small. The same 3-year 

 herring and the same size groups employed on page 316 for scale-diameter measure- 

 ments were used here. It was found that the length of the scale diameters of the 

 first year of life averaged 2.85 millimeters in the 40 small 3-year herring of 1921 and 

 2.94 miUimeters in the 26 large fish; 3.01 millimeters in the 48 small 3-year fish 

 taken at Bay City in 1922 and 3.10 millimeters in the 88 large hemng; and 2.78 

 millimeters in the 71 small 3-year fish taken at Oscoda in 1922 and 2.87 millimeters 

 in the 72 large individuals. The grand average for the 159 small herring was 2.87 

 millimeters and for the 186 large fish 2.99 millimeters. As Lee's "phenomenon" 

 appeared in the computed lengths, it must, as the residt of this correlation, appear 

 in the measurements of the scale diameters. Likewise, a more or less perfect direct 

 correlation exists between the total increments of scales and those of the body com- 

 puted from these scales. The paradoxical results shown in the scale increments of 

 Table 22 and discussed on page 330 must, of necessity, then, also occur in the body 

 increments calculated from these scales. 



