370 
Pigmentation, Selection and Anthropometric Characters 
between pigmentation and height and weight were established, there would no 
longer be any justification for attributing these characters entirely to the environ- 
ment. In order to elucidate this question the ideal material to examine would be 
data collected from a city where there were known to be different racial elements. 
All the individuals examined should be of the same age, and preferably should 
have been born in the district, or at least have lived a long time there. But such 
material is not easy to get; the examination of school children, however, does 
provide useful data. It is better that the individuals should be of the same age, 
since in that case no correction is required for the change in pigmentation owing 
to increasing age. Although such a correction can be made, it introduces a 
disturbing factor which does not tend to increase the accuracy of the results. 
Birth and not only residence in the same distr ict is a desirable feature for all the 
individuals, since the results are otherwise open to the criticism that the influence 
of an earlier environment has not been taken into account. If, with such material 
as is here indicated, a statistical examination were to show that pigmentation as 
a racial character was correlated with other physical characters, then we should 
have obtained an important result which woidd bear on the correct degree of 
influence to be attributed to social conditions. 
The data which have been employed in the present work are for our immediate 
purposes unfortunately far from ideal. They are derived from two sources. Firstly, 
there is the medical survey of Birmingham School Children, which has been used 
in the first section of this paper. The one advantage which it offers is the fact 
that all the children are the same age ; and therefore there is no need to correct for 
a change of pigmentation. But we do not know how long the children have been 
in Birmingham or where they were born. This, however, is not so serious a 
defect as the fact that the Birmingham population is a singularly homogeneous 
one. Foreign names are extremely rare among the children. This means that, 
if in such a population a connection between pigmentation and stature is found, 
it must be a very important feature indeed. The failure to find it here would not 
be surprising, and it might well exist in a more heterogeneous population. The 
Birmingham data give both height and weight for every child. Now, as has been 
explained in the first part of the paper, if two characters A and B are correlated 
and A is also correlated with another character 0, then there is no reason to 
expect a correlation between B and G unless the correlation between A and B is 
very high. In other words, although we know that height and weight are corre- 
lated, if we find a correlation between pigmentation and height, there is no reason 
to expect a correlation between pigmentation and weight, unless the correlation 
between weight and height is a very high one. Therefore the investigations of 
height and weight and their relation to pigmentation are treated as entirely 
separate. 
Secondly, there are the data collected by Tocher in his "Survey of the Pigmen- 
tation of the Scottish Insane*." In this case we have height and pigmentation 
* Biometrika, Vol. v. p. 298. 
