PROF. K. PEARSON ON THE MATHEMATICAL THEORY OF EVOLUTION. 283 
Table X.—Sisters of Sisters. 
Mean height. 
64-0454 
Probable error of mean height. 
•0655 
S.D. 
2 3668 
Probable error of S.D. 
•0463 
Coefficient of correlation r . 
•4386 
Probable error of r. 
•0205 
Coefficient of regression. 
•4386 
S.D. of array of sisters of selected sister. 
2Y270 
It will be seen from this table that eider and younger sisters of sisters are 
respectively less and more variable than sisters of sisters in general. It will be noted 
also that sisters of brothers are, both in stature and variation, nearer akin to elder 
sisters of sisters than to younger sisters. It deserves accordingly to be investigated 
whether or not sisters are not on the average older than brothers—on this point I have 
no data. As sisters of brothers approximate to elder sisters of sisters, so brothers of 
sisters correspond more closely to younger than to elder brothers of brothers. These 
are points which require fuller investigation, when ampler statistics are forthcoming. 
Turning to correlation we note that the coefficients in the case of collateral 
inheritance are slightly greater than in the case of direct inheritance. It will be 
remarked at once that the values are much less than those given by Mr. Galton, 
“Natural Inheritance,” p. 133, who has himself drawn attention to the considerable 
difference between the constants for collateral inheritance given by his R.F.F. Data 
and by his Special Data. Mr. Galton having kindly allowed me to use his data, I 
have recalculated from the formula r = S (^^/(noqcro) the value of r for the Special 
Data, taking my pairs of brothers precisely as I had done for the Records of Family 
Faculties. I find r = *5990 with a probable error of *0124. This value is not as 
high as Mr. Galton’s, but differs very widely from the value ‘3913 given above. 
In making the calculations, however, I was much struck by the peculiarities 
presented by a certain portion of the data, which I will speak of as the Essex 
contribution. The brothers therein were very short and remarkably close together. 
I therefore went through the calculations again, separating the Essex contribution, 
and with the following results :— 
Mr. Galton’s Special Data. 
Whole population. 
Essex contribution. 
Remainder. 
Mean height. 
68-544 
67-797 
68-797 
Probable error. 
•0402 
■1013 
•0457 
t for brothers. 
•5990 
•7175 
•5574 
Probable error of r ... . 
•0124 
•0200 
•0152 
2 0 2 
