62 The Intensity of Natural Selection in Man 
on p. 61, o:iO"i+2 denotes the partial standard deviation of the total mortality in the 
sum of the periods considered, 03 e w denotes the expected (partial) correlation if 
there were no selection (see § XXXIV of the memoir), and 03 b 2 ] denotes the (partial) 
regression of the mortality of the second period on that of the first. 
Dealing first with the results from the English data, we notice that the regres- 
sions by the two methods for males are fairly similar, but for females they are, on 
the whole, smaller by the new method. Having regard to the probable errors we 
can draw no inferences concerning the differences. The correlations for females, 
however, are in two of the three cases considerably smaller by the new measure of 
environment, and this difference appears to be significant. The partial standard 
deviations by the two methods occasionally show fairly large differences, but in no 
single case is the disagreement significant. For all the six cases, however, is 
less than but 04 o- 2 is greater than 0:s o- 2 . The mean of the male regressions by 
the first method of measuring environment is —'142 and by the other —'143, the 
corresponding figures for females being —'179 and —'117. The mean regressions 
of the mortality of the 4th and 5th years of life on that of the first three years 
are — '085 and — 172 respectively*. Thus, so far as males are concerned, the 
intensity of selection appears greater when measured by the regression of the 
mortality of the 3rd, 4th and 5th years on that of the first two years of life than 
when measured by the regression of the mortality of the 4th and 5th over the first 
three years, but the same conclusion does not so definitely hold for females. It may 
be, as was suggested in the memoir, that the age division between infant and child 
mortality is not the same for females as for males, and the inference is put forward 
tentatively that the ailments of infancy (as distinct from those of childhood) 
attack females to a rather greater age than they do males. It will be noticed, too, 
that the regressions and correlations for the 1872 cohort are smaller than for the 
other cohorts, and that this is accompanied by the fact that the mortality of that 
cohort was smaller in the first period. On the whole, so far as the data for the 
English rural districts are concerned, the adoption of a new measure of environ- 
ment leads to no alteration of view as regards the existence of selection, nor, 
roughly, of its numerical intensity. 
When we turn to the results from the Prussian data in which the same periods 
(the first two years and the next three) are used as for the English data, the 
most marked feature to be noticed is the considerably larger correlations and 
regressions which are obtained. The mean value of the regressions for males 
* I take here the opportunity of correcting a mistake which occurred in connection with the work of 
the first memoir, through an error in transcribing from the schedules containing the raw data. In 
§ xvni, in the portion of the Table for the 1872 cohort referring to females, the following alterations 
should be made : 
^ = 487, 0^ = 119-7, 
r m = -812328, r 12 = -718137, n 3 =-944379, 3 r 0 i= + -464859, 3^=- -168670, 103 »-i2= - -184068. 
These entail the following corrections in the Table on p. 33 for the same cohort: 030-1 = 34-844 and 
partial regression =- -0591. These alterations reduce the correlation and regression but make the 
results more consistent, and necessitate little modification of the conclusions drawn from them. 
