28 



Size of the samples. 



On the basis of these average percentages, the question of variation is 

 investigated in the following manner : The samples are divided into 5 groups, 

 according to number of individuals, (Gr. I, the 5 smallest samples, Gr. V the 5 

 largest) and the standard deviation a calculated for each group. The value of 

 this factor for the different groups, and for all groups together, will be found 

 in Table II. 



Table II. Standard deviation {a) in the 5 groups of samples. 



The table exhibits no distinct relation between a and the size of the 

 samples, from which it would seem that this latter factor does not play any 

 important part in the variation found, as expressed by a. Group I however, 

 which contains samples of less than 150 individuals, seems to have a greater 

 standard deviation than the remainder. 



Economical considerations necessitating the employment of as small 

 samples as possible, the conclusion to be drawn from Table 11 is that samples 

 of the size of those in Group 2 (200 fish) should be preferred, at any rate in 

 cases where no other special object is in view beyond that of ascertaining the 

 numerical value of the richest year-classes. 



Technique of the determination of age. 



As the size of the samples therefore, does not appear to be a factor of 

 any very great importance, we must consider the question as to how far the 

 inevitable errors attaching to the examination of the scales may be regarded 

 äs of serious effect. The material here employed is useless for the purpose 

 of such investigation, as many other possibilities may be imagined. Seven 

 samples from Fuglö and Brandesund, on the other hand, are admirably adapted 

 for this purpose. These samples were examined by seven different persons, each 

 taking one : they were taken on the same day and at practically the same 

 place, from two catches, they are of approximately equal size, and so small, 

 that the variations found may be taken as maximal values of the technical 

 error attaching to the investigation, the more so, as some of the investigators 

 were not practised in the work of examining herring scales. Table III shows 

 the standard deviation of the same four year-classes in these 7 samples. 



