320 
Miscellanea 
It will be seen that the first three, with even grouping, are very close together though 
the number of groups has been reduced from 20 to 5. Similarly the next three are close 
together, and the last is again by itself. 
An examination of the way in which the groups are taken shows that the more the tail 
is bunched together the higher is the value found, and this is what would be expected in this 
particular case, since there is practically no increase of head height with age at the 'old' end of 
the scale, whereas for purposes of calculation we have assumed a constant angle for the regression 
line. But it may be pointed out that r) varies (to the 2nd place of decimals) only from - 29 to - 31 
even if we reduce the twenty groups to two, an extreme proceeding which is never done in 
practice. 
At the same time the ordinary six or eight groups may be expected to give results a little 
too high when, as is usual, the regression line is curved. 
II. On the Hereditary Character of General Health. 
By KARL PEARSON, F.R.S. and ETHEL M. ELDERTON, Galton Laboratory. 
(1) In dealing with the heredity of general health we have to meet at once certain funda- 
mental difficulties. We have first the question of environment and secondly the question of 
variety in health caused by what we may term accident. If we deal with families living in 
widely differentiated environments we shall have, or certainly may have, a spurious correlation 
of health in parents and offspring ; the resemblance in health will be emphasised. On the 
other hand, when a single member of a family is exposed to a specially differentiated environ- 
ment, i.e. goes to the West Coast of Africa, or spends his life in India, or catches enteric at a 
particularly unfavourable moment, the correlation of general health may be decidedly weakened 
in the case of parent and offspring. These difficulties of differentiated environment and what 
we may, perhaps, term accident cannot be wholly overcome, but we may endeavour to meet 
or measure them. In the first place we can confine our observations to one social class and 
thus go a long way to get differentiated environment removed. If, as in the present paper, we 
deal essentially with the professional classes, there is great uniformity of general environment. 
The food supply is sufficient, the doctor is always at command, physical exercise is fairly general 
and markedly insanitary houses or occupations are practically avoided. We do not think 
therefore that, for the data of the present paper, differential environment is a marked factor 
in producing correlation. On the other hand we do consider it possible that "accident" will 
weaken the relationships sought. The reduction in health-correlations below the values for 
other physical characters, might indeed be taken as a measure of random action on health, 
comparable with the random action of death itself in reducing the correlation of duration 
of life, which has already been discussed by one of us*. Indeed heredity of general health 
is almost as significant for the problem of natural selection, as heredity of duration of life. 
A more serious difficulty in the health-inheritance problem is this very question of death. 
If parents are delicate and health or delicacy is hereditary, they will have delicate children, 
and we may anticipate that more of these children will die than in the case of the children 
of robust or normally healthy parents. Thus only the healthier children of delicate parents 
will survive for us to record their state of health, and accordingly the offspring of delicate 
parents will appear healthier than they really should be owing to the selective death-rate. In 
our investigations we have dealt only with offspring who lived to be adult, i.e. to at least 
21 years of age, so that an appreciation of general health could be formed. There is also a 
further difficulty that very delicate parents are themselves likely to die, and we have again 
* 11. S. Proc. Vol. 65, p. 290, and Biometrika, Vol. i. p. 50. 
