490 Farther Evidence of Natural Selection iti Man 
time factor involved in the variate difference correlation method that a selective 
deathrate plays even in highly civilised states a marked part in the natural history 
of man. 
(2) The material dealt with in this investigation consists of the Registrar- 
General's returns for births in England and Wales and of deaths in the first five 
years of life from 1859 to 1908 with the addition of as many years before 1859 
as were requisite to make our highest differences fifty in number, and with the 
addition of as many years after 1908 as were requisite for following up the births 
of that year to the fifth year of life. Thus actually our data extended from 1850 
to 1912. The reason for this procedure lies in the desirability of using a constant 
population, and not reducing by one a relatively small number like 50 on each 
differencing. As a result of this process we had to modify Dr Anderson's values 
for the probable errors for the steady values of the difference correlations because 
in our case the size of the population does not change as we proceed to higher 
differences*. The second cause which requires extension of the data is a very 
important one, and must be illustrated numerically. Consider the table : 
Deaths of those horn in a given year. 
Year 
Female 
Births 
0—1 
1—2 
2—3 
3—4 
4-5 
1908 
478,410 
63,594 
1909 
14,14f! 
1910 
5,020 
1911 
3,449 
1912 
2,341 
Now the deaths of infants 0 — 1 in 1908 are not necessarily of infants all born 
in 1908, but the total deaths 63,594 must represent closely the deaths in the 
478,410 infants born in that year. Disregarding immigration and emigration, this 
gives a deathrate per 1000 of 107'495 and leaves 414,816 children alive. Of this 
group 14,146 may be taken to die in the second year of life, giving a deathrate of 
31'990 per mille. There remain 400,670 children who reach the third year of life 
in 1910, of whom 5,020 die, giving a deathrate of 11*939, and 395,650 survivors. 
These survivors are followed into 1911 and 1912 in the same manner, and thus 
we obtain approximately the deathrate up to the fifth year of the male children 
born in 1908. We thus in bulk follow the same group of children through the 
first five years of life. Tables I and II give the deathrates for males and females 
respectively under the heading of the birth year of each group. These death- 
rates have been taken to three decimals places for the purpose of determining the 
higher differences correctly to one decimal place. The successive differences of 
* All the probable errors of the difference correlations given in this memoir are these modified 
Andersonian values, i.e. they are the probable errors on the assumption that the diiJerence correlations 
have reached steady values. 
