A. M. Carr Saunders 
367 
correlation between any of the hair or eye colour classes and disease in general or 
any one disease in particular. 
There is, however, another point of view from which we may look at this table. 
It shows no orderly system from the standpoint of age, sex and disease. But in 
another way the table seems to be improbable. In 48 cases we have enquired 
whether the samples are so divergent that they ought to occur. Therefore there 
ought to be one case that would only occur once on the average in 48 trials, or in 
other words with a probability of "02. Now, there are actually two such cases "023 
and "019. In the same way there should be five cases which occur once in ten 
trials — or with a probability roughly of "1. Actually there are nine cases instead 
of five below "1. There should be between nine and ten cases below "2 ; actually 
there are fifteen. Thus it would seem that the whole table is somewhat improb- 
able as a result of random sampling ; but there are no systematic differences 
to be made out of it. Is the general irregularity a result of personal equation ? 
General Conclusions. 
The data used in this section of my work consist of random samples of boys 
and girls at two different ages. We possess a classification of their pigmentation, 
and figures giving the number of cases of six specified diseases. An analysis of 
this data has shown that no obvious systematic connection can be discovered 
between any one category of pigmentation and disease. In other words, taking 
any child in our population we should not, starting from our knowledge of its 
pigmentation, be able to say that there was any probability of the child having 
had any more or less diseases than the average, or that there was a greater or less 
probability of the child having had one particular disease. Are we justified in 
going a step further and saying that in this population pigmentation is not a 
factor in natural selection ? Suppose, for example, that the fair haired children 
were in process of elimination through selection, owing to their greater suscepti- 
bility to disease, would these figures show it ? Now it is impossible to compare 
the proportions of the different categories of pigmentation, in the two age groups, 
and thus to estimate if these change as the children grow older, because the 
change in pigmentation cannot be corrected for with sufficient accuracy ; and thus 
we cannot say how much of any change in the relative proportions of the classes 
of pigmentation is due to increasing age and how much to selection. We merely 
have the fact that boys and girls in this population, aged 3 — 7 and 13, do not 
show any connection between pigmentation and disease. Now, we argued before 
that if boys were being selected owing to their liability to disease, the girls should 
show a larger average number of diseases per child. In the same way we may 
argue that if the lighter haired children were being selected, the darker haired 
should show a greater average number of diseases per child. But we do not find 
anything of the kind. This argument, however, is by no means conclusive ; it 
depends upon the assumption that, if one of two classes is in process of elimination 
through disease, that class differs from the other class simply in the fact that 
