135 



DEPARTMENT OF THE NAVAL SERVICE 



7 GEORGE V, A. 1917 



Judging from these averages, there might possibly be some slight difference 

 between the most northerly locality and that farthest to the south. 



In order to test the value of the differences thus found between the various samples, 

 all possible differences between growth dimensions of the same character were first 

 ascertained, the nine samples giving thirty-six differences for each dimension, making 

 ISO in all for the five first dimensions. 



For each of these differences (-0), the corresponding standard error (d) was then 

 calculated, and the fraction ^/d formed. This fraction, which expresses the difference 

 in units of its error may serve as a i^ind of indicator for the importance of the differ- 

 ence. Where the value of the fraction is small, there is but little probability that the 

 difference is due to other causes than accidental fluctuations, and vice versa, and in 

 particular, where its value exceeds certain limits, the probability becomes an empiric 

 certainty that the difference is not merely due to fluctuations of sampling. 



Table 16 shows the manner in which the various values of the fraction ^/d 

 arrange themselves, firgt for each separate growth dimension, and finally for all 

 together. It will be noticed that in 91 of the 180 cases, the fraction is less than 1, in 

 154 less than 2, and in 171 less than 3. Only in 9 out of 180 cases is the fraction 

 over 3, i.e., in these cases the difference is more than three times as great as the 

 error. In the last column will be found figures showing the arrangement which 

 should have resulted had all samples been drawn from an entirely uniform stock and 

 subject only to fluctuations of sampling. 



Table 16. — Showing distribution of values of — for all possible comparisons between 

 the nine samples from Newfoundland. (Year class 1904, t^ — t^). 



A comparison of tlu'.-;e theoretic values with the figures actually found, inclines 



us to suppose that the nine cases where the value of — — exceeds three are probably 



significant. And on examining these nine differences, it will be seen that they are 

 due to peculiarities in the sample from White Bay and in that from St. George's bay, 

 only in one instance is there a difference noted where neither of these samples is 

 included, i.e., between samples 7 and 8 for the growth-dimensions ^3. The small t^ and 

 the large t^ in the White Bay sample, with the large t^ the small t. and t.^ in that from 

 St. George's bay, give rise to the differences. It would thus seem that the dissimilarity 

 noticed in table 15, as between these two extreme samples, is significant. It is 

 impossible to say with certainty what importance should be attached to these differ- 

 ences, as several possible explanations may be given. As far as I can see, the most 

 reasonable supposition would seem to be that the sample from St. George's bay is not 

 quite " pure," i.e., that it consists of a majority of individuals, belonging to the same 

 growth type as that of the remaining samples, mixed up with a minority of another 

 type. From the nature of the dissimilarity noted, the growth of this second type 

 should be characterized by a large t^ and smaller i.^ t- t^ and t.,. 



