PACIFIC HERRING 
261 
Eliminating the San Francisco Bay samples, the maximum differences that 
approach statistical significance (between 2.7 and 3.1 times their probable errors) 
are those of Elrington Passage, Macleod Harbor, and Shuyak Strait, 0.27, 0.26, and 
0.26, respectively. Although these differences are not of statistical significance, yet 
the fact that they appear in three localities attests to their validity. Hence in com- 
paring samples from adjacent localities the significance of the differences must be 
open to question if they are smaller than 0.27. 
Returning to the differences in vertebral count between distant localities (Table 
7), it is apparent that they are very far in excess of any differences found between 
samples from the same locality, except in three cases (marked in Table 7 by an 
asterisk), in each of which the two populations compared were far apart and sepa- 
rated by others that are very distinct. This lack of difference is probably due entirely 
to accidental similarity. 
In southeastern Alaska (Table 13), Craig and Stephens Passage (Nos. 7 and 12, 
Table 6 and fig. 14), both differ from the three Chatham Strait localities (Nos. 8, 
9, and 10, Table 6 and fig. 14) by more than the maximum variability found between 
samples from one locality, 0.27 (Table 18), except in the case of Craig and Larch 
Bay, where the difference is only 0.26. In this case, however, the difference is 7.9 
times its probable error, while the maximum difference found in the same locality 
(Table 18) is only 3.1 times its probable error, indicating that the difference between 
Craig and Larch Bay may be valid. Craig and Stephens Passage do not vary sig- 
nificantly between themselves, but their geographical position, with the distinct 
Chatham Strait stock between them, would indicate their independence. 
In Prince William Sound (Table 14), as aforementioned, there are three locali- 
ties — Puget Bay, Snug Harbor, and Port Fidalgo — that do not differ among them- 
selves, but each of which shows a statistically significant difference with all of the 
other localities in Prince William Sound. In all three of these localities the range 
of sizes was small, varying chiefly from 150 to 190 millimeters, and so presumably 
the samples were composed very largely of 3-year-olds. As the fish in these samples 
were largely of one year class, the expected variability would presumably be analogous 
to that of the differences found between age classes of which the maximum was 
0.67 ± 0.113, or 5.9 probable errors. The largest difference, that between Puget Bay 
and McClure Bay, is only 0.45 ±0.048, or 9.4 probable errors so, although on the 
assumption that the sampling was truly random the difference is significant, we will 
not regard it as denoting a race because we know that the samples do not represent 
the same populations in age. McClure Bay differs from Elrington Passage by 
0.14 ±0.038, or 3.7 probable errors, and from Macleod Harbor by 0.18 ±0.040, or 
4.5 probable errors. Both of these differences are less than the 0.27 maximum 
variability between samples from the same locality, so they will not be considered as 
valid. Our data do not, therefore, demonstrate the existence of more than one 
stock of herring in Prince William Sound. 
In the Cook Inlet and Kodiak-Afognak districts the maximum differences be- 
tween localities are small. Dogfish Bay shows statistically significant differences 
between both Kachemak Bay and Shuyak Strait, but the actual differences, 0.26 
and 0.22, are slightly less than 0.27 — the maximum difference found between samples 
in one locality — hence Dogfish Bay can not safely be considered as racially distinct, 
especially as only one sample was obtained. Shuyak Strait differs from* both Old 
Harbor and Shearwater Bay by four probable errors, the actual differences being 
