38 
PROFESSOR K. PEARSON ON THE INFLUENCE OF NATURAL 
These give for our case : 
M., = 40‘9 5, Mo = 30‘24. 
Thus a group of males, selected to have the same stature as the females, would 
have a slightly shorter femur and a slightly longer humerus. A slightly longer femur 
in woman and a slightly longer humerus in man would thus appear to be sexual 
characters. 
Turning to the variations, these are given by (xlix.). We find : 
No = 2-36, N3=1-53. 
This shows us that while the selection would give the same variabilitv of humerus 
to the men that women have, it would fail to produce the reduction of variability in 
the femur, which is characteristic of the women. 
From (li.) we deduce 
= '808, = ‘806, 
while from (Ivi.) we have 
1-03 = -840. 
Thus we see that very small changes would be made in the correlations, stature 
and femur, stature and humerus, and femur and humerus, if we selected French men 
to have the same size and variability of stature as French women. The explanation 
Oi this lies in the nearly equal absolute variability of the two sexes with regard to 
stature, for, as we have seen, it is the selection of variability which modifies correlation. 
Looking at the table of values on p. 36, we see that the largest difterence of 
variability in the two sexes lies in the femur, and accordingly it is from a selection of 
femur that we should expect the greatest differences in the variability and correlation 
of the two sexes to have arisen, but even this difference alone would not account for 
the observed sexual differences in the correlation. Indeed, it would be suiq^risiug if it 
did, for the selection of other organs, notably the pelvis, must have played a con¬ 
siderable part in the differentiation of sex."^ 
(9.) I shall now proceed to a series of problems, which will show the application of 
results, such as those obtained in this memoir, to questions which arise in dealing 
with inheritance and selection. If we suppose a general population to have statistical 
“ constants,” wliich remain constant at any rate for a moderate interval, we stiU want 
to knoAV not only the error which may arise from a random sampling, but also the 
sort of effect which results from our sample being too much drawn from one kind of 
environment, from a rather limited class, or from any other practical^ necessary or 
unconsciousl}^ introduced limitation of the random character of our sample. 
* “Primitive man and voman are more nearl}' equal in size, variability, and correlation than highly 
civilized man and woman ” (‘Roy. Soc. Proc.,’ vol. 61, p. 354). 
