Karl Pearson and David Heron 
219 
Taking this system of classification we divided up a Gaussian surface of 
"5 correlation into the same groups, and also a Gaussian surface of 3 correlation. 
We pwhlish these surfaces below*. 
TABLE XII. 
Gaussian Surface for r = "5 in Eye-Colour Groupings. 
1 
0 
& 
4 
5 + 6 
7 
8 
Totals 
1 
7-38 
19-85 
4-94 
1-38 
0-26 
0-18 
o-oi 
34 
2 
20-58 
145-47 
78-94 
35-98 
9-72 
9-27 
1-04 
301 
3 
6-01 
93-63 
85-41 
54-34 
18-59 
22-33 
3-69 
284 
4 
1-26 
31-81 
39-49 
31-03 
12-29 
17-36 
3-76 
137 
5 + 6 
0-53 
18-11 
27-79 
25-14 
11-09 
17-62 
4-72 
105 
7 
0-22 
11-02 
21-59 
23-66 
11-86 
21-89 
7'76 
98 
8 
0-02 
2-11 
5-84 
8-47 
5-19 
12-35 
7-02 
41 
Totals 
36 
322 
264 
180 
69 
101 
28 
1000 
TABLE XIII. 
Gaussian Surface for r = 3 in Eye-Golour Groupings. 
1 
2 
3 
4 
5 + 6 
7 
8 
Totals 
1 
4-04 
17-16 
7-55 
3-30 
0-91 
0-92 
0-12 
34 
2 
17-41 
123-59 
79-76 
44-64 
14-61 
17-67 
3 32 
301 
3 
8-86 
93-00 
78-31 
52-04 
19-20 
26-40 
6-19 
284 
4 
2-83 
37 73 
37-24 
27-51 
10-95 
16-31 
4-43 
137 
5 + 6 
1-62 
25-21 
27-75 
22-09 
9-26 
14-64 
4-43 
105 
1-02 
19-50 
24-47 
21-39 
9-58 
16-36 
5-68 
98 
8 
0-22 
5-81 
8-92 
9-03 
4-49 
8-70 
3-83 
41 
Totals 
36 
322 
264 
180 
69 
101 
28 
1000 
In applying the method of contingency to these two tables, no correction for 
mean of (/> 3 in the random sample should be made ; they are actual surfaces and not 
random samples from these surfaces. Further in order to measure what effect 
dealing only with round numbers in the cells would make we replaced the first 
table by the following Table XIV in which the decimals were cut off and a slight 
adjustment made to preserve the total variate frequencies. This table is pub- 
lished so that the reader can judge on what we worked. 
This working to units cannot be expected to give quite as good a result as 
working to two decimal places, but it is more consonant with an actual table. 
Next we took Pearson's Family Data cards and arranged 1000 cases of Father 
and Son, first in order of magnitude of Father's stature, next in order of Son's 
* Mr Yule states that he has divided up the -3 Gaussian surface in a somewhat similar manner, but 
he does not publish his table, and it is therefore impossible to test his results. We should like here 
to enter a protest against this procedure, which recurs in Mr Yule's memoir, and throws an immense 
amount of unnecessary arithmetic on any one traversing Mr Yule's arguments. 
28—2 
