114 DEPARTMENT OF THE NAVAL SERVICE 



each sample containing a relatively large number of specimens. As the work went on, 

 however, it was found desirable to collect the material on a more extensive basis, with 

 more samples, distributed as widely as possible throughout the area and the season 

 embraced by the fishery. This involved an alteration of the methods employed, and 

 a number of the samples thus collected were now subjected to a simpler process of 

 examination (vide p. 133) in addition to which the question as to possibility of reduc- 

 ing the number of specimens in each sample was considered and tested. With this 

 new method, which was tried and adopted in 1914, it was found that for the object 

 in view, and under the conditions then prevailing among the stock, comparatively 

 small samples might well be used. An average of some 200 specimens per sample was 

 seen to be sufficient, and it was also realized that the advantage gained by operating 

 with larger samples would be altogether disproportionate to the extra work involved, 

 as the discrepancies between the samples could not be essentially minimized thereby, 

 and may be due to other causes than fluctuations of sampling. 



It should be emphatically pointed out, however, that the prevailing situation was 

 exceptionally favourable for the work of age investigations, the contrast between the 

 one enonnously rich year-class 1904, and all the others, being so great that the curves 

 for age distribution would maintain their characteristic appearance even with con- 

 siderable errors of sampling. In other words, the age composition was so characteris- 

 tic, that it was advantageous to work with small samples, and consequently greater 

 accidental fluctuations, as this rendered it possible to deal with a greater total number 

 of samples. 



In cases where the age distribution is less characteristic, e. g., where several 

 successive year-classes are more or less equally strong, accidental fluctuations of 

 samples may impair the agreement between the samples, or at any rate, render it less 

 obvious. In such cases, therefore, it may be desirable to reduce the extent of the 

 fluctuations of sampling by increasing the number of specimens in each sample. It 

 was in anticipation of such a possible change in the situation that the number of 

 specimens per sample was increased to about 400 in the case of the Norwegian spring 

 herring investigations in 1916. 



The situation prevailing during the past few years among the Norwegian spring 

 herring has not been quite so favourable for growth investigation; when so great a 

 percentage of the specimens examined is derived from a single year-class, the result 

 is an almost superfluous quantity of material confined to the group in question, with 

 a corresponding reduction in the amount for growth observations for the many other 

 year-classes represented. Under these circumstances, it would require a dispropor- 

 tionate amount of work to procure material in which every single year-class should be 

 represented by the full number of specimens desirable for growth investigations. 



