60 
The intensity of Natural Selection in Man 
method the mortality of the cohort corresponding to the cohort whose history is 
being traced but of the opposite sex, is fixed. In the other, the number of deaths 
of the same sex and between the same age limits in the period under notice, 
apart from the deaths occurring within the cohort, is rendered constant. Neither 
of these, of course, allows for the fact that under a perfectly uniform environment 
the mortality which is taken to indicate that environment should be distributed in 
some way due to random causes. But the general similarity which will be shown 
to exist between the results reached by using the two distinct methods is some 
justification for the claim that each is a satisfactory approximation to the theo- 
retically best method. 
The investigation which is now being described was directed throughout to 
ascertaining the extent of selection in the mortality of the first two years of life. 
In the earlier work of which an account is given in the memoir other periods were 
taken, but later some evidence was adduced to indicate that the first two years 
of life was a natural interval to adopt, as embracing roughly the whole of the 
mortality of infancy and overlapping but very little that of childhood. For the 
second period, on the mortality of which the selective character of that of the 
earlier period is indicated, the next three years of life are taken both for the 
English and Prussian data ; in the case of the latter, also, the next eight years are 
employed as a second period. We thus reach results obtained from English and 
Prussian data by working at the same periods in each case, and the comparison of 
these results is of interest. The notation employed throughout is : 
x 0 = Births of the male or female cohort considered, in, say, year t. 
ajj = Deaths in the cohort in the two years, t and t + 1. 
x 2 — Deaths in the cohort in the next three years, or next eight years. 
x A = Remaining deaths of same sex as cohort in the five years or ten years (see 
Memoir, § vill). 
x 4 = Deaths in the corresponding cohort of opposite sex in the five years or ten 
years considered. 
x s = Births of the corresponding cohort of opposite sex. 
Previous experience suggested that the correction for a constant value of x s in 
addition to constant values of x 0 and x x would have little effect on the correlation 
between x 1 and x 2 . The first case worked out supported this view ; in that (the 
Prussian male cohort of 1881, dealing with the first ten years of life) the correlation 
was only altered in the sixth figure by the extra correction, viz. from — '944206 to 
— '944209. Thus the considerable labour involved in making this further correc- 
tion is not justified by the extra value it gives to the results, and in all the other 
cases it was omitted. 
The partial standard deviations, correlations and regressions with their probable 
errors for the various sets of data are given below, the other standard deviations 
and correlations on which they are based being shown at the end. In the table 
