450 Growth, Variability and Correlation in Young Trout 
In Table III are given the corrected values for the mean, standard deviation 
and coefficient of variability of the total length in the last four stages on the 
assumption that this 5% (say 40 individuals) should have been uniformly spread 
over these stages. The extra individuals have been placed (a) in the lowest class 
only of the distribution, and equally in the lowest and the highest class. The 
TABLE III. 
Total Lengths. 
Values as given in the 
tables 
Corrected Values 
(a) By the inclusion of 
10 extra individuals in 
the lowest class 
(/3) By the inclusion of 
5 extra individuals in 
the lowest and 5 in the 
highest class 
M 
G. V. 
M 
a 
C. V. 
M 
a 
C. V. 
I 
184-03 
24-13 
13-11 
II 
238-75 
18-25 
7-65 
236-22 
21-29 
9-01 
238-16 
20-20 
8-48 
III 
284-65 
14-55 
5-11 
282-19 
18-92 
6-70 
284-11 
16-88 
5-94 
IV 
319-23 
16-13 
5-05 
316-26 
19-39 
6-13 
319-11 
18-88 
5-92 
V 
358-00 
27-70 
7-74 
351-99 
36-21 
10-29 
356-66 
33-55 
9-41 
If all 40 extra individuals are put in the lowest class in Stage II, then 
J/= 228-36, o- = 26-44, G. F. = 11-58. 
variabilities are of course increased, and more by the former than by the latter 
method. The former perhaps is fairer since Mr Severn says those that died were 
deformed. But even so, the series shows the same decrease up to Stage IV and a 
rise in Stage V, and, what is more important, the variability in no case becomes 
equal to that found in Stage I. Only by including all 40 in the lowest class of 
Stage II can the variability be made even to approximate to that seen in Stage I. 
The values are given at the foot of the tables. In this case of course the values 
for the coefficient of variability in Stages III and IV remain unaltered. 
I admit that the method is unsatisfactory, but in the absence of any actual 
knowledge of the size of those that died, it is the best that can be done. The 
decrease of variability is then perhaps not wholly due to a selective death-rate. 
Is it then to be attributed to decrease in growth-rate ? This seems possible, 
but the absence of any marked increase in growth-rate in the last stage suggests 
that still some other factor may be concerned ; and again there is the apparent 
absence of any connection between the irregularities in the change of growth-rate 
and anomalies in the alteration of variability. 
I had thought at one time that the increased variability in Stage V might be 
due to imperfect sampling in the collection of the material, smaller, weaker 
individuals having been perhaps left at the bottom of the box in earlier stages, and 
