150 On the Inheritance of the Mental and Moral Characters in Man 
only, that there existed two pigments, black and red. We might by placing red at 
one end of the scale and black at the other, obtain a single scale which would 
really be a double one, i.e., a scale of diminishing amounts of black pigment from one 
end, and of red from the other. In the one case the fairs are classed with red as 
marking an absence of black pigment and in the other case with the darks as marking 
an absence of red. Fourfold divisions of this table would then give the correlation 
between brethren either in the amount of red pigment or in the amount of black 
pigment. Unfortunately the observer comes across — besides a very deep red type of 
hair which seems to be pure red, and which shades, if enough individuals are taken, 
continuously away from " fair reds " — another red, a " dark red," which I found 
frequently described as "brown red" or "dark brown red," and which seems to be a 
blend of the red and dark pigments. The existence of these brown reds seems to me 
the difficulty of the single scale arrangement. It is on this account that some hair 
scale makers have placed the reds alongside the browns, but this appears to misplace 
the " fair reds " and " pure reds." I am at present working on the problem of a 
practical hair scale, and I am not at all certain that something corresponding to the 
artist's conception of "value" is not what we want, if we are to use hair colour as a 
character for investigations about inheritance. I merely refer to this method because 
I consider these hair colour results somewhat unsatisfactory and subject to revision 
and reclassification. There is another point also to which I must refer. I have found 
a distinct growth in children's hair colour with age. This, of course, has been 
recognized in a general way, but our data supply, as soon as we have settled our scale, 
the quantitative measure of it. Hence, exactly as in the case of head measurements, 
we ought really to allow for the growth change in hair before measuring the 
resemblance of brothers. Allowance for this growth, to judge from the effect of 
growth in other cases, might easily change the value of the correlation by 10 to 
15 per cent. I hope to return to the problems of scale and growth in hair colour ; 
meanwhile I would describe what I have done. The hair correlation tables have been 
worked out in four different ways, namely, by forming fourfold tables at each corner 
of the "brown-brown" category. By doing this I have endeavoured to allow for the 
position of the red-browns, which were classified under reds, i.e., whenever a 
division comes for the fourfold table between brown and dark, it is immaterial 
whether the reds are placed beyond the fairs, between fairs and browns, or between 
browns and darks. The results given for hair are the means of the four correlations 
found by working out the tables in four different ways. I believe on any system 
of "value*" my results will be approximately correct, but it would still need 
correction for growth, i.e., a sensible darkening in the fifteen years of life covered 
by our observations. On the whole, I publish the hair colour results with reser- 
vations f. 
* I hope shortly to be able to publish photographic measures of " value " in hair colour. 
t I think since writing the above I have surmounted the difficulty of scale orders by applying the 
new method of contingency, which completely dispenses with any scale order : see Drapers' Company 
Research Memoirs, I. " On the Theory of Contingency and its Kelation to Association and Normal 
Correlation." (Dulau and Co., Soho Sq., London.) 
