K. Pearson 
343 
pairs and covering many types of life — which have to the present date been deter- 
mined for fraternal correlation. 
Now take the 39 cases for homotyposis which have so far been determined. Of 
these 22 are in the vegetable kingdom, and 17, not yet published, in the animal 
kingdom. We get the distribution given below. In both these distributions there is 
no approach to uniform scattei'ing throughout the range of observed values. In the 
Cases 
0—1 
0 
1—2 
4 
2 — 3 
1 
3—li 
5 
7 
r,—G 
11 
6—7 
7 
7—8 
.3 
8—9 
1 
9—10 
0 
Totcal 
39 
Scale of Frequency 
second we see the lump produced by the three cases of Nigella, Malva and Asperula, 
which were doubted before they were dealt with. The mean of the forty-two 
fraternal correlations is now '495 and of the thirty-nine homotypic correlations 
•499. We have still not enough material to reach a typical distribution in either 
case, but we have evidence more than enough to see — notwithstanding the very 
great difficulties of the investigation — that the coefficients tend to cluster about "5. 
Mr Bateson says I "attach importance to the rather close similarity between the 
two average values." I attach just so much as the probable error of 39 or 42 
observations admits. My own words were : 
" Now I do not propose to lay great stress on what at first sight might look 
like a most conclusive equality between the mean values of homotypic and 
fraternal correlations I am quite aware that a few further series added to 
either the homotypic or fraternal results might modify to some extent this 
equality" {Phil. Trans. Vol. 197, A, p. 358). 
