solved by Rounsefell and Dahlgren, who devised 

 an internal "iron" tag that could be mechanical- 

 ly recovered in meal lines by electromagnets. 

 Since its introduction, the internal tag has been 

 widely used in subpopulation studies of Pacific 

 Coast fishes, especially in studies on Pacific 

 herring, Pacific sardine and Pacific mackerel. 



The earliest paper dealing with "racial" 

 variation in the Pacific herring is by Hubbs(1925) . 

 He summarized data on fin rays (dorsal and anal) 

 and variation in number of vertebrae, based on 

 counts made by Thompson (1917) and himself. 

 All fish dealt with were either sampled in British 

 Columbia or California . Hubbs pointed out that 

 there is a southward decrease in number of 

 vertebrae in the Pacific herring. 



A much larger mass of information is 

 presented by Rounsefell (1930) . He chose four 

 characters to show structural differences: 

 counts of vertebrae, of dorsal rays, of anal rays, 

 and measurements of head length. Rounsefell 

 examined samples from many localities between 

 San Diego Bay in southern California and Golovin 

 Bay in the northern part of the Bering Sea. 



Variation in number of vertebrae . - -The 

 range in number of vertebrae in the Pacific 

 herring is large, amounting to a difference of 

 13 between the lowest and highest counts record- 

 ed in the literature (46 to 58 including the 

 urostyle). There is a marked difference in the 

 average number of vertebrae in different local- 

 ities . The lowest mean value, 51 .68 vertebrae, 

 is in fish from San Diego Bay, at the southern 

 end of the distribution of the Pacific herring. 

 The average number of vertebrae gradually in- 

 creases to the north and westward, reaching 

 highest average values along the Alaskan penin- 

 sula (Rounsefell, 1930). The highest average 

 value recorded, 55.67 vertebrae, is in fish from 

 the Shumagin Islands. The difference in mean 

 vertebral number between fish from San Diego 

 Bay and the Shumagin Islands amounts to nearly 

 four vertebrae . Data on variation in number of 

 vertebrae in different parts of the range of this 

 species along the west coast of North America 

 is summarized in table 1 . Although this table is 

 based primarily on data given in Rounsefell (1930) 

 it includes counts of other workers. The source 

 is indicated for each locality. For several 



localities the counts of two or more investiga- 

 tors are given . 



Rounsefell summarized vertebral data for 

 32 localities in all: 5 localities between San 

 Diego and British Columbia, 7 localities in 

 southeast Alaska, 15 localities in central Alaska, 

 and 5 localities in western Alaska . It is ob- 

 vious that the extreme variations, such as that 

 between San Diego and the Shumagin Islands, 

 are significantly different . The heterogeneity 

 of the Alaskan populations is also evident. The 

 difference between the mean values for Stephen 

 Passage in southeast Alaska and the Shumagin 

 Islands off the Alaskan Peninsula of 2.31 verte- 

 brae is 68 times the probable error of the 

 difference.—' 



When stocks of adjacent localities were 

 compared, Rounsefell found what he considered 

 to be significant differences in some instances . 

 Thus, in southeast Alaska, Stephens Passage 

 herring differed from samples from Chatham 

 Strait localities by about 12 times the probable 

 error of the difference between the vertebral 

 means. On the other hand, the three localities 

 in Chatham Strait did not differ significantly 

 from one another . 



2/ I have used the formulae given by Royce 

 (formula 5 under concept of overlap) to compare 

 three localities: San Diego Bay, Stephens 

 Passage in southeast Alaska, and the Shumagin 

 Islands . The overlap in distributions of the 

 number of vertebrae in the several localities 

 was as follows: 



Locality Percentage of overlap in 

 vertebral counts 

 San Diego Bay - Shumagin 



Tsland 4 



San Diego Bay - Stephens 



Passage 52 



Stephens Passage - Shumagin 



Island 37 



These values can be assumed to represent 

 the maximum amount of intermingling that 

 could take place between the herring groups in 

 any two localities. The actual values of inter- 

 mingling could be as low as 0, or some inter- 

 mediate value between and the above maximal 

 values . 





46 



