506 Further Evidence of Natural Selection in Man 
In the light of this result we have found the correlations between deathrates 
of sex A in the third and sex B in the fourth year of life, for constant deathrate 
of sex A in the fourtli year of life. 
We have 
Both of these may be taken as zero, having regard to their probable errors. 
Thus on the vi^hole, while the relation between the deathrate of a group of one 
sex in one year and the deathrate of the remainder in the following year of life 
appears after the annulment of the time-factor to be very considerable and 
negative, there does not appear to be any organic relation between the deathrate 
of sex A in one year and sex B in the following year, if we proceed by the method 
of partial correlation. But at the same time we believe that this method must 
be used with very considerable caution, and that to avoid erroneous conclusions the 
whole problem must be investigated from a variety of standpoints in cases like the 
present where one of the three total correlations is extremely high. The numerator 
N ranges in the cases we have been discussing from about '01 to '05 and with a 
small total frequency like 50, any disturbing cause — apart from random variation — 
may have marked influence*. 
(9) The conclusion which we have formed is tliat in the present problem of 
natural selection it is probably better to annul the environmental factor by 
the variate difference method rather than to proceed by the n^ethod of partial 
correlation as we have hitherto done. 
By the former method we have shown that for both sexes a heavy deathrate 
in one year of life means a markedly lower deathrate in the same group in the 
following year of life, and that this extends in a lessened degree to the year 
following that, but is not by the present method easy to trace further. It is 
difficult to believe that this important fact can be due to any other source than 
the influence of natural selection, i.e. a heavy mortality leaves behind it a stronger 
population. Nature is not concerned with the moral or the immoral, which are 
standards of human conduct, and the duty of the naturalist is to point out what 
goes on in Nature. There can now scarcely be a doubt that even in highly 
organised human communities the deathrate is selective, and physical fitness is 
the criterion for survival. To assert the existence of this selection and measure 
its intensity must be distinguished from advocacy of a high infant mortality as 
a factor of racial efficiency. This reminder is the more needful as there are not 
wanting those who assert that demonstrating the existence of natural selection in 
man is identical with decrying all efforts to reduce the infantile deathrate. 
We have to acknowledge the great assistance we have received from our 
colleague Miss Beatrice M. Cave in the laborious arithmetical work of this paper. 
* If F=NID, where N and D are both small, but F finite, then 8FIF=dNIN - 6Dll> and small 
disturbances produce great results in F. 
