461 
VI. British Returns due to K. Pearson. 
Hair Colour. 
Fair 
Brown 
Dark 
Bed 
Totals 
Light ... 
Medium ... 
Dark 
210 
117 
23 
107 
158 
63 
47 
113 
125 
20 
15 
2 
384 
403 
213 
Totals ... 
350 
.328 
285 
37 
1000 
My data are from my observations on school children, in this case for boys only. I have put 
my categories ' Dark ' and ' Jet Black ' together as ' Dark ' as the nearest equivalent to the 
continental * Schwarz ' possible under the circumstances. My three eye categories cannot be 
very different from the ' Hell,' ' Melist ' and ' Braun ' of Retzius' tables. Their exacter definitions 
will be found in Biometriica, Vol. ill. p. 162. The importance of complete equivalence is, however, 
for our present purposes, not very great. We want to get the same number of reasonably 
definite gi'oupings in each case. The six tables above each supply a 12-fold classification. 
From these tables the correlation between hair and eye colours was determined for me by 
Dr A. Lee using the method of mean square contingency, — a method which gives results inde- 
pendent of any scale order whatever*. We foimd 
Correlation hetiveen Hair and Eye Colours. 
Sweden (Conscripts) 
Prussia (Schoolboys and Schoolgirls) 
Italy (Conscripts) 
German Jews (Schoolboys and Schoolgirls) 
Baden (Conscripts) 
Great Britain (Schoolboys) 
•2495 
•2714 
•3091 
•3381 
•3540 
•4203 
Now these results seem to show that the correlation between hair and eye colours is by no 
means so close as has been hitherto supposed. Further it apx3ears to be less in those districts 
which, whether light or dark, have a majority of one type. It is not improbable that the higher 
values in Baden and Great Britain are due to the effect of a greater mixture of local races, such 
local races not having interbred largely. In my opinion the results for the German Jews show 
that the population so classified is very far from being purely Semitic. If as I suspect the 
lesser valvies of the correlation occur in the relatively more homogeneous groups, then we might 
infer that increasing purity of race would mark a still slighter correlation between hair and eye 
pigment. In other words, if mankind originally consisted of several races each with a definite 
pure eye and pure hair colour, then the right statement is that the correlation between hair 
and eye colours was interracially unity or perfect, and intraracially zero. The point is of course 
one of terminology, provided the decreasing intraracial correlation with increasing racial purity 
be once established as a fact. But it is an important point of terminology, especially when 
pure races are crossed for hybridisation experiments. Is it correct to say that the correlation 
in white mice between eye pigment and coat colour is perfect? Since both characters are 
invariable, algebraically the correlation takes the indeterminate form 0/0. It is not unusual 
* Pearson: "Mathematical Contributions to the Theory of Evolution. Xlll. On the Theory of 
Contingency and its Relation to Association and Normal Correlation." Drapers' Research Memoirs, 
Biometrie Series, I. (Dulau and Co., Soho Square, London.) 
