252 On a New Method of Determining Correlation 
meaiis of the arrays are, however, not so smooth* as in the former case and there 
is some sign of the regression curve being heteroclinal. 
(C) Handwriting and Eye Colour. I have chosen this illustration because it 
admits of using formula (ii) directly. The material is represented in the follow- 
ing table: 
Handwriting. 
Very 
0«od 
Good 
Moderate 
Poor 
Bad 
Totals 
Light 
45 
249 
312-75 
112-5 
28-75 
748 
Medium 
60 
252 
302-75 
99 
27-25 
741 
Dark 
41-5 
137-5 
182 
63 
8 
432 
Totak 
14G-5 
638-5 
797-5 
274-5 
64 
1921 
Let us call the range of medium eye colour h and measure the means first 
from the boundary of light and medium {x) and then from the boundary of 
medium and dark {x). We havef 
•2809, 
•7558, 
1 0368, 
•5039, 
•5731, 
h/a, = 
1 0770, 
•9626, 
-2794, 
•7880, 
1-0674, 
(To/ax = 
•9718, 
^3/0-3 = 
-2737, 
•7448, 
h/a, = 
1-0184, 
0-3/o-x = 
roiso, 
•2280, 
•7405, 
hi a, = 
•9684, 
cr^jax = 
1-0705, 
^,/-.= 
1276, 
1-1504, 
1^2780, 
0-5/0-^ = 
-8112. 
We see that the last column gives us the means of finding the yO-xjcrx of the 
formula (ii). Also we have alternative methods of determining t] according as 
to whether we use the first or second column, i.e. x or x. 
Using the first column we find : 
7;^ =-0835 --0789 = -0046, 
or: 7? = -068. 
Using the second column we find : 
7;'-^ = -5754 --57 13 = -0041, 
or: r?' = ^064. 
The close accordance of these results speaks well for the application of the 
method to the present material. We may test it again, assuming that medium 
and dark eyes are classed together, and using formula (iii), we find in this case 
7]" = ^070, 
* Eeduced to a common unit a^, they are - -O'JSl, 7-Jaj.= - -QiQ^, .Ts/cTj. = + -1064, 
54/<^x= + '0345, .T,>^= + -0821. 
t Reduced to 0-^ we have .Ti/(r^ = -4851, .T2/<r^=:r-2714, .1^/0-^.= -2780, .I-4/<7x = -2441, .fj/(r^ = -1035. 
