Ernest Warhen 
339 
all together there is obtained a population of 8497 individuals derived from 30 
different nests. 
The constants are: Meau 251'8, Standard Deviation 1986, Coefficient of 
Variation 7-89. 
The distribution of the frequencies is irregular, and the polygon shows a 
well-marked double peak. The cause of this is chiefly attributable to the great 
disproportion of the size of the samples, and the consequent undue influence of 
one or two individual nests. To avoid this disturbing effect, random samples 
of 100 individuals from each of the 30 nests were taken. In this way there 
results a population of 3000 (Table V.), and the constants are : Mean 243'5, 
Standard Deviation 17'08, Coefficient of Variation 7-02. The distribution of the 
frequencies is much less irregular than in the former case, and the polygon is 
consequently much smoother. 
The coefficient of variation of the population of 3000 is thus 7-02 as compared 
with the mean value of 3-02 for the families. In other words the variability of 
the family of small soldiers is about 43 7o of that for the population. The vari- 
ability of the population compared with the fraternal variability is much higher 
than was expected, judging from the value obtained from certain breeding experi- 
ments with a moth, S. clathrata, conducted several years ago at University College, 
London. The cause would appear to be attributable to the great effect of the 
environment on the means and variabilities of the different castes of a nest. 
(8) Correlation of the Means. 
The correlation between the means of the small soldiers and large workers has 
been determined in the case of the 27 nests entered in Table I. The correlation- 
surface is shown in the following table. The calculated constants are: mean of 
means of small soldiers, 244-3 units, standard deviation, 15-29; large workers, 
24r4 units and 13'04 respectively. 
Correlation coefficient = •953, probable error + •012. 
For the calculation of the correlation between the large soldiers and small 
soldiers only 20 nests were available. 
The constants are: Large soldiers, mean of means, 435 '5 units, standard 
deviation, 16"9 ; small soldiers, 247'3 units and 17'14 respectively. 
Coefficient of correlation = "831, probable error — ± '046. 
Since the probable error is so considerable, there is no reason for supposing 
that the correlation between the large and small soldiers differs appreciably from 
that between the small soldiers and the large workers, which we found to be 
•953 + '012. It may be assumed that the correlation of the means between any 
two castes is about 9. It is curious that the correlation between similar castes 
(large and small soldiers) does not appear to be greater than that between dis- 
similar castes (small soldiers and large workers). 
