320 
Breeding Experiments ivith Foxgloves 
families. The number of the families is, however, too few for drawing any absolute 
conclusions. With such a high variability in these families it would appear im- 
probable that the families obtained by self-fertilizing the individuals at the 
extremes of the range would both exhibit the same mean. It may be added that 
in the parthenogenetic aphis {Hyalopterus) it has been shown that the variability 
of the family amounted to 60 % of that of the race. 
C'haracter 
No. of 
individuals 
in genera- 
tion 
Standard 
deviation 
(S.D.) of 
generation 
No. of 
families 
with six 
or more 
individuals 
No. of 
individuals 
Mean 
standard 
deviation 
(S.D.) of 
the families 
S.D.fam. 
S.D. gen. 
xlOO 
Intensity of general 
90 
3-490 
8 
69 
2-349 
67 
colouration {DR in- 
dividuals) 
_ , . Breadth 
Katio ^ 7- corolla 
172 
3-324 
13 
154 
2-414 
73 
Lengtli 
Spotting 
178 
1-816 
13 
160 
1-000 
61 
Mean of percentages 
67 
It must be remembered that Johanessen's experiments have been shown by 
Karl Pearson to be inconclusive with reference to the pure-line theory. The size 
of the different seeds produced by any given individual plant varies greatly, and there 
is no inherent reason to suppose that the offspring obtained from the small seeds 
and large seeds taken from the same plant would exhibit any difference in the 
size of their seeds. It might reasonably be supposed that a small seed and a large 
seed from the same plant would both inherit the average character of the parent 
irrespective of their actual size. One might as well expect that the plants obtained 
by budding from a small begonia leaf would have smaller leaves than those obtained 
from a larger leaf of the same plant. 
IV. Fraternal Cor relation. 
The likeness of brothers to one another has been investigated, and the relation- 
ship was studied after the method adopted by Prof. Pearson. He takes each 
individual of a family and compares its deviation from the mean with that of every 
other member of the family. These combinations can be plotted out in a correla- 
tion table in a similar manner to the relationship of offspring and parent. The 
coefficient of correlation for all possible pairs of brethren will be a measure of their 
resemblance, just as the standard deviation is a measure of their dissimilarity. 
The fraternal correlation was calculated for (1) the intensity of the general 
colouration of the purple flowers, (2) the area of the spots, (3) the ratio of the 
width to the length of the corolla. 
For the intensity of the general colouration only offspring having a similar 
gametic constitution were taken. Heterozygous individuals, or those which are 
