RACES OF HERRING, SOUTHEASTERN ALASKA 
133 
being the same. This is not sufficiently great odds to be able to state definitely that 
Warren Island and Culebra Island represent distinct populations. Thus while the 
significant differences between Warren Island and Noyes Island and between Corona- 
tion and Noyes Islands tend to indicate the lack of migration across Iphigenia Bay, 
the lack of a statistically significant difference between Warren Island and Culebra 
Island averages does not confirm this view. 
A test of the homogeneity of the Prince of Wales Island samples with the Warren 
Island data included, made by above-described methods, yielded an observed z of 
0.3247 which happens to be exactly the same as the calculated z for a probability of 
0.01, thus indicating that this group of samples is not all drawn from one population. 
The same test made without the Warren Island samples gives an observed 2 of 0.0248 
which is many times less than the calculated z of 0.4420 for a probability of 0.05. 
Therefore, it must be concluded that the data point to the lack of migration between 
localities lying north and those lying south of Cape Ljmch (Iphigenia Bay). 
Samples from Warren Island, at the mouth of Sumner Strait, differ from those of 
Meade Point, near the mouth of Frederick Sound, by 2.44 standard errors which 
yields a probability of 0.014. Even if this difference were significant, it could not be 
assumed to give definite information on migration through Iveku Strait as neither 
locality is very close to the entrance to this channel. It does, however, suggest that 
the herring at the mouth of Frederick Sound do not migrate to the ocean by tills 
route. 
A sample from Kell Bay, in Affleck Canal, differs significantly in vertebral count 
from Coronation Island, but does not show a significant difference from Warren 
Island, which is about the same distance as Coronation Island from the mouth of the 
canal. 
The vertebral count comparisons of table 6 also indicate differences between the 
Cape Ommaney herring and those from Port Herbert and Gut Bay, but not from those 
caught in Big Port Walter. The Big Port Walter herring also do not show differences 
from Port Herbert. Therefore these localities cannot be classified without more 
material. 
ANALYSIS OF GROWTH RATES 
In analyzing the growth rates from the various localities for racial purposes no 
data were used except those from freshly caught purse-seined specimens, all of which 
were obtained during the summer months. 
Many of the body length distributions are slightly skewed, and in addition cover 
a wide range, with a tendency in a few cases for slight modes to form near the upper 
or lower range of the distribution. These disturbing factors are probably caused in 
large measure by errors in age reading whereby a length distribution may contain a 
few fish belonging to younger or older age groups. Since these doubtful measures 
near the extremes of the range exert a large influence in the determination of the 
arithmetic mean, whereas, being of doubtful authenticity they should not carry as 
much weight as the more centrally located items, it was decided not to use the 
arithmetic mean but to employ the median for the measure of central tendency. 
In keeping with the use of the median the interquartile range has been used as the 
measure of dispersion. 
To gain an insight into the growth increments during the summer months the 
data have been grouped by 10-day periods. (See table 8.) For Larch Bay 5-year-olds 
