Karl Pearson and David Heron 
239 
There seems no doubt that from the Mendelian standpoint, if anterior pigment 
be taken as a " unit," then pure grey eyes should be included with the blues and 
the correlation then comes out the true Mendelian third. Thus the original 
division of the tables between grey and the hazel groups, which appeared at the 
time the most reasonable physiologically and statistically, is amply justified by the 
theory of posterior and anterior pigment. 
Now Mr Yule tells us on the basis of his erroneous theory of pseudo-ranks 
that " the average estimated correlation " of these eye-colour " tables is something 
like A not \" (loc. cit. p. 620). We have indicated earlier in this paper (p. 232), 
that if a fourfold table for continuous variates — skew or Gaussian — give by </> a 
Diagram VI. Regression of eye-colour with eye-colour in brothers, the grade intervals being 
assumed to give a normal distribution. Inset, the same, colours 4 to 8 only, the grade 
intervals being of same value as before, to show resemblance within the darker grades only. 
5&6 
5&6 
4 
1 
(Light) 
3 4 5X6 
First Brother. 
(Dark) 
correlation of "33 this must be increased by something like 40 %> if we wish to 
find the true correlation of the variates. Had Mr Yule asserted that eye-pigmen- 
tation was not a continuous variate, that all brown and hazel eyes were alike and 
differed from blue and grey by the possession of some mysterious unit character, 
then he would have been justified by the mere fourfold tables in asserting that the 
correlation was In doing this he would have taken up the simple Mendelian 
position. That he did not do so was probably owing to the fact that he recognised, 
as the tables indeed show, that within the two groups, 1 + 2 + 3 the blue-grey 
group and 4 + 5 + 0 + 7 + 8 the hazel-brown group, there was distinct heredity of 
sub-divisions. The father with hazel eyes has an offspring less pigmented than the 
father with dark brown eyes. That such heredity exists must dispose at once of a 
