M. Greenwood 
65 
ages of 25 and 55. Such a " random sample " is, of course, also a selection in that 
it consists solelj' of those patients who died in hospital, and upon whom a post- 
viortem examination was held. Thus it is far from being a random sample of the 
" general population " of the country, many classes of which are never found in the 
wards of a general hospital at all. 
Evidently the population of a general hospital will chiefly consist of, (i) persons 
acutely ill, (ii) those suffering from surgical injuries or diseases, (iii) sufferers from 
medical affections requiring special treatment. Chronic maladies of old age, such 
as bronchitis, indeed, any highly chronic disease, will be under-represented in 
comparison with the general death-rate. Similarly the number of cases of valvular 
heart disease and rarer disorders, such as Diabetes Mellitus or Insular Sclerosis and 
other nervous lesions, will be above the general average. 
Now, as pneumonia and bronchitis, particularly the latter, form a considerable 
number of the so-called " terminal affections " responsible for a large majority of 
all adult deaths, a random thousand necropsies will not give us the information we 
require as to the quantitative relations of average viscera, post-mortem. The error 
resulting from too few cases of senile bronchitis will be lessened, if not minimised, 
by the fact that we have confined our attention to cases of less than 55 years of 
age. But even thus we have too many cases of valvular cardiac disease, and as 
this affection tends to produce hypertrophy of the heart, the average weight in 
the first three tables is probably a good deal higher than that of the ordinary 
population at death. 
It is, of course, to be remembered that this " general hospital population " does 
not mean the " normal " or healthy one. The above remarks are merely intended 
to show that a thousand deaths in hospital will not be due to exactly the same 
causes as a thousand deaths taken at random outside, and that therefore when we 
proceed to select sub-groups, such as "Normal Hearts," "Hearts in Pneumonia," etc., 
the material we have to select from is not what it would have been had we been 
able to start with 1000 random deaths in the population at large. And so, if we 
find that the average weight or variability of an organ is diminished when we 
proceed to special classes, we must bear in mind that possibly the change might 
not have been so striking if we had had a more representative sample to start 
from. 
General Hospital Population. 
TABLE I. Hearts with Livers. Number 1382. 
Mean Heart 13-53 ozs. Standard Deviation 4 '680 ozs. 
Mean Liver 63'01 ozs. Standard Deviation 13 314 ozs. 
Correction of Heart and Liver -1931 ±-0175. 
TABLE IL Hearts with Spleens. Number 1303. 
Mean Heart 13-07 ozs. Standard Deviation 4-067 ozs. 
Mean Spleen 6-61 ozs. Standard Deviation 3-345 ozs. 
Correlation of Heart £^nd Spleen -1827 ±-0181. 
Biometrika iii 9 
