vary independently of geographic location. 

 Rounsefell did not find this character of use in 

 showing the distinctness of any population. 



Anal -ray counts. --The range in counts of 

 herring anal rays found by Rounsefell was from 

 14 to 20; mean values in fish from different 

 localities ranged from 16.61 (Puget Sound) to 

 17.23 (Dogfish Bay) . Rounsefell considered that 

 the evidence from anal -ray counts was inconclu- 

 sive . As with dorsal -ray counts, there is no 

 general change associated with latitude. There 

 is as great a tendency for large differences be 

 tween adjacent localities as between distant ones: 

 For example, Dogfish Bay, with a mean count of 

 17.23, is adjacent to Shuyak Strait with a mean 

 count of 16 . 71 . The difference between two 

 samples from one locality (Elrington Passage) 

 was 0.35 ray, a variation as large as between 

 most of the localities sampled by Rounsefell. 



Difference in head length. --The only mor- 

 phometric character used by Rounsefell (1930) 

 was head length. He found that head lengths were 

 smaller proportionately in large fish than in 

 small, and he concluded that the length of head 

 "does not increase as rapidly as the length of the 

 fish, " (ibid: 267) . He expressed head lengths as 

 percentages of body length and compared curves 

 of these percentages plotted against standard 

 length; he was not able to express mathematical- 

 ly the differences between the regression curves. 

 Subsequently, a covariance method of comparing 

 morphometric measurements has been developed 

 (Mottley 1941), that permits an assessment of the 

 significance of the differences between regres- 

 sion curves. As a matter of interest, regressions 

 were made of head length on standard length for 

 two localities, "Southeastern Alaska total" and 



Prince William Sound , using data given in 

 Rounsefell (ibid: table 25) . Only average values 

 based on five or more observations were used. 



Both regressions are rectilinear with little 

 variation about the fitted lines (sy.x for both lines 

 is only 0.3 mm) . Statistics describing the two 

 regressions follow: (below) 



A linear relationship exists between the 

 rate of growth of the head and the standard length . 

 (See in this regard, Marr (1955:29) . A difference 

 in growth of herring as compared to trout may be 

 noted; Mottley (1941) found it necessary to use 

 logarithms of the measurements when dealing 

 with regressions of head length on standard length 

 in trout. A simple linear relationship between 

 head length and standard length has also been 

 found for yellowfin tuna (Schaefer 1948), Pacific 

 mackerel (Roedel, 1952), and other fishes. 



Herring "stocks" in southeastern Alaska. -- 

 Rounsefell and Dahlgren (1932) pointed out that a 

 high negative correlation exists between the aver- 

 age temperature during the developmental period 

 and the average number of vertebrae in different 

 year classes . Hence, it is important when com 

 paring samples of herring to deal with individual 

 year classes rather than composite samples. 

 Rounsefell and Dahlgren (1935) analyzed the "races" 

 of herring in southeastern Alaska in greater detail 

 than was possible in Rounsefell' s earlier paper. 



They studied material from 32 localities in 

 southeastern Alaska. The chief character used 

 was counts of vertebrae, determined by year 

 classes for each locality. Other characters em- 

 ployed were rates of growth and relative abundance 

 of year classes . Homogeneity of the material was 

 shown for individual localities . A comparison 

 was made of counts from each locality with those 

 of adjacent localities, using fish of the same year 

 class. The basic data of vertebral distribution 

 by year class and locality are given in table 2 . 

 The authors list 71 comparisons of which 53 were 

 considered not to show differences, 5 were ap- 

 proaching statistical significance and 13 were 

 considered by the investigator to be statistically 

 significant. Five of the statistically significant 

 comparisons are between the 1926 class in Peters- 

 burg and the following localities: Vicinity of Juneau, 

 Point Gardner, Meade Pt . , Warren Island and 

 Wrangell; three are between Noyes Island and 

 localities just north of Cape Lynch (i.e., Corona- 

 tion Island and Warren Island) . Two of the statistic- 

 ally significant comparisons are between Cape 



48 



