502 Further Evidence of Natural Selection in Man 
We shall of course work with the sixth difference correlations in order to free 
ourselves substantially from the time-factor. 
Here again the judgment based on the partial correlation of the crude 
deathrates is in all six cases reversed. For every one of the partial coefficients 
of crude deathrates shows that for intervening year with a constant deathrate, 
an increase of deathrate in the earlier year means an increase, not a decrease in 
the later year. Actually an increase in the one year is shown in Table X in all 
cases to be followed by a decrease at two years' interval. 
TABLE X. 
Influence of Natural Selection at Interval of Two Years. 
Result of an increase of 10 deaths per mille in the second following year. 
Increase of 10 in Deathrate of 
For constant death- 
rate in 
? 
1st Year on that of 3rd Year 
2nd Year on that of 4th Year 
3rd Year on that of 5th Year 
2nd Year 
3rd Year 
4th Year 
Decrease '81 
Decrease '61 
Decrease '99 
Decrease 1-28 
Decrease '52 
Decrease 1'4 
It will be seen that these values are appreciable although far less important 
than the decreases produced in a following year by an increase in the immediately 
preceding year. Thus we judge that a selection of the weakly children in one 
year is largely influential on the deathrate of the immediately following year, and 
diminishes, as we might anticipate, with increase of time. 
Some objection might, however, be taken to the sixth difference correlations, 
when we consider deathi'ates of the same group two years apart. They are 
Male. Female. 
rs,rn,.s,,n, + -227 ± -159 + -200 ±-161 
+-339 ±-149 +-377 + -144 
'\,„3.5o'»5 +-397 ±-142 +-393 ±-142 
It will be seen that while they are all of the same sign and fairly accordant for 
both sexes the probable errors are becoming very substantial relative to the 
coefficients. We have indeed too limited a range of years. 
(7) If now we take out the correlation coefficients of the sixth differences for 
three years' interval, and again for four years' interval we find great irregularities. 
Male. 
+ •205 + -161 
-•030+ 168 
Female. 
+ -035 ± -168 
+ •072 + -167 
r^ 
-•181 + 163 
-•251 + -158 
5y J?(i . 5o"'6 
The correlations now do not agree in sign, they are insignificant having regard 
to their probable errors, and there is no close correspondence for the two sexes. 
