346 U.S. BUREAU OF FISHERIES 



analyses of vertebral counts, growth rates, the proportions of various 

 year classes and by the recovery of tagged herring. 



In analyzing the vertebral counts only counts of herring of the 

 same year class were compared as it was shown in a previous report 

 on the herring of Prince William Sound, and is too apparent in 

 these data to need proof, that the mean vertebral count differs 

 between herring of different year classes from the same locality. 

 Segregation of the material by year classes has not been followed 

 in the European racial work on herring, which fact doubtless 

 accounts for many of the inconsistencies in results. 



In order to be certain that grouping the samples by localities 

 would not in itself bring out differences that were really due merely 

 to random sampling two tests were first made to determine if the 

 data as a whole were homogeneous. The first test was to determine 

 wdiether or not any correlation exists between the mean vertebral 

 'Count in the various localities of one particular year class and the 

 temperature during the spawning period of each locality. High 

 negative correlations were found for the 1927 and 1926 year classes, 

 respectively. The second test was to analyze the variances of 158 

 samples of the vertebral count in the manner show^n by R. A. 

 Fisher, after first discarding four of the samples whose variances 

 exceeded the normal range of variances. This test showed very 

 conclusively that the samples are not homogenous, and that the 

 differences between the means are too great to be assigned to chance 

 sampling. 



Application of the same test to the samples from each of seven 

 major localities gave opposite results. In each case all of the dif- 

 ferences between the means of samples could be assigned to random 

 sampling. This also was in accord with the assumption that 

 different localities might possess different populations of herring. 



Comparisons of the means of the vertebral count from the various 

 localities revealed three groups of herring that differ significantly 

 from their neighbors : namely, Petersburg, Noyes Island and vicin- 

 ity, and the localities east of Clarence Strait and south of Sumner 

 Strait including Wrangell. 



Comparisons of the length distributions of herring of the same 

 year class show that herring of four localities : the Noyes Island area, 

 the Douglas Island-Icy Strait area, Affleck Canal and Peril Strait 

 are all much slower growing than those from the other localities. 

 The Peril Strait herring appear to be the slow^est growing of any 

 yet encountered in Alaska, the median of the 4-year olds taken in 

 June 1930 being only 176 millimeters. 



Comparisons of the age distributions of purse-seined material 

 (avoiding the selected distributions derived from gill-netted samples) 

 caught in 1929 and 1930 show (1) the 1926 year class to be over- 

 wdielmingly dominant in most of the localities, (2) the 1926 and 

 1927 year classes to be approximately equal at Noyes Island, (3) the 

 1927 year class to be very dominant in Peril Strait, (4) the 1926 and 

 1923 year classes both dominant at Douglas Island and at Favorite 

 Bay, (5) a large percentage of the catch older than the 1923 year 

 class at Douglas Island. These facts support the evidence given by 

 the vertebrae and the growth rates which separate the Noyes Island 

 area, Peril Strait, and the Douglas Island-Icy Strait area from 

 neiijhborino- localities. 



