132 
BULLETIN OF THE BUREAU OF FISHERIES 
Any two means are compared by dividing their difference by the standard error 
estimated by the formula 
I S(x-x) 2 +S(x'-x') 2 ( 1 " 1 \ 
° V n 1 +n 2 VH + l + w 2 + iy 
if Xi, x 2 , a: n2 + l and x' u x' 2 x'n 2 + l be two samples, 
and 
1 
7H+ 1 
Six), x' 
1 
n 2 J rl 
Six') 
Of the 71 comparisons between the mean vertebral counts given in table 6, 36 are 
between localities not over 25 miles apart, and the remaining 35 are from localities 
over 25 miles apart. 
The results of the comparisons are listed in table 7. 
That there is such close agreement between the results of the two group of com- 
parisons is surprising considering the difference in distance. The median distance 
apart in the close group is only 17 miles as contrasted with 52 miles in the group of 
distant comparisons. These results tend to indicate that ordinarily there is probably 
little intermingling between herring of different races, the boundaries between racial 
areas being quite abrupt. 
Table 7. — Summarized comparisons of vertebral counts 
Distance apart in miles 
0-25 
2G and over. 
Probability of means being the same 
0.99-0.05 
0.05-0.01 
0.01 or less 
27 
3 
6 
26 
2 
7 
The mean vertebral counts of herring from Petersburg, in the northern entrance 
to Wrangell Narrows, for instance, differ by over 4 standard errors from those of 
Juneau, Point Gardner, or Meade Point, by 3.89 standard errors from Warren Island, 
and by 3.19 from those of Wrangell. This is rather definite evidence that herring do 
not migrate through Wrangell Narrows or Dry Strait, and that migrations in Freder- 
ick Sound must be largely confined to that body of water. 
The samples taken in the area south of Sumner Strait and east of Clarence 
Strait, including Wrangell, Anita Bay, Santa Ana Inlet, and Frances Cove, do not 
differ amongst themselves, so that, until additional evidence is collected, it can only 
be assumed that the herring of these localities may intermingle. That this group of 
herring does not intermingle with herring of the outer coast through Sumner Strait 
is clearly indicated by a difference between the Warren Island and Wrangell averages 
of 5.87 standard errors. 
On the west coast of Prince of Wales Island the means of the samples from 4 
closely adjacent localities, Klawak, Port Estrella, Noyes Island, and Culebra Island, 
do not differ amongst themselves, but differ from those of the localities to the north. 
The agreement of the herring of the adjacent localities (captured during the summer 
months) with those taken while spawning, on the important spawning grounds near 
Klawak, is quite in keeping with expectations. Noyes Island differs from Coronation 
Island by 4.56 standard errors and from Warren Island by 2.82 and 2.95 standard 
errors in the 2 available comparisons. However, Warren Island and Culebra Island 
differ by only 2.03 standard errors which gives a probability of 0.04 of the populations 
