469 The Intensity of Natural Selection in .
By Karl Pearson , F.R.S. ( Received June 4 , \#151 ; Read June 27 , 1912 .
) ( 1 ) In a paper* communicated to the Royal Society in 1899 , and later in greater elaboration published in ' Biometrika , 'f 1901 , it has been shown on the basis of the inheritance of longevity that the selective death-rate in man amounted to at least 60 per cent , to 80 per cent , of the total death-rate .
The matter has been recently reconsidered by Prof. Ploetz , J who , dealing with material wholly different from that of Beeton and Pearson came to similar conclusions .
The point is a very vital one , for , combined with : ( i ) the heredity of physical and mental characters in man , S and ( ii ) the demonstration that the longer-lived have more offspring , || we reach a definite knowledge that Darwinism does apply , and very intensely applies , even to man under civilised conditions .
The difficulty of a direct investigation of the problem lies in securing uniformity of environment .
We have to demonstrate that when under the same environment there is a heavier death-rate among a given group of human beings , then among the survivors of this group in a given later period the death-rate will be lessened .
Now each group of individuals we attempt to deal with has its own environment , and if that is a bad environment we should expect to find a heavy death-rate both at the earlier and later periods ; this obviously must obscure the action of natural selection .
For example in districts with a high infant mortality we might expect a high child mortality , say deaths from two to five years of life , because a bad environment sends up the intensity of both .
The correlation between deaths in the first year of life ( 0\#151 ; 1 ) and in the next four years of life ( 1\#151 ; 5 ) for a given district will certainly be positive if no correction be made for varying environment .
Quite recently this matter has been discussed by determining the correlation between the ages 0\#151 ; 1 and 1\#151 ; 5 in the administrative counties of England and Wales .
IT As ( a ) the group 0\#151 ; 1 was not followed to 1\#151 ; 5 , but the * ' Roy .
Soc. Proc. , ' vol. 65 , p. 290 , et seq. + Vol. 1 , p. 50 , et seq.\#151 ; especially pp. 74-5 .
J ' Archiv fur Rassen- u. Gesellsehafts-Biologie , ' 1909 , vol. 6 , p. 33 .
S " On the Laws of Inheritance in Man .
I.\#151 ; Inheritance of Physical Characters , " ' Biometrika , ' vol. 2 , p. 357 ; II.\#151 ; " On the Inheritance of the Mental and Moral Characters in Man , " ' Biometrika , ' vol. 3 , p. 131 .
|| " On the Correlation between Duration of Life and Number of Offspring , " ' Roy .
Soc. Proc. , ' vol. 67 , p. 159 .
IT * Local Government Board Report , ' Cd .
5263 .
Prof. K. Pearson .
[ June 4 , deaths in these age-groups for the sam years were dealt with , and ( 6 ) no allowance whatever was made for the differential environment of the administrative counties , it is difficult to find any real bearing of the data on the problem of natural selection in man .
The only method by which data for different districts can be compared is by endeavouring to fix the nature of the environment .
We want to know whether under a constant environment the correlation between the death-rates of infancy and of childhood is positive or negative .
Dr. E. C. Snow , * working on both English and Prussian data , finds , using a variety of criteria of sameness of environment , that when this factor is allowed for the correlation between infantile and child death-rate is negative , and substantially negative .
In other words , within the same group under the same environment , the greater the infantile death-rate , the less is the death-rate among the survivors ; that is to say , the physically stronger members of the group are those which survive the ordeal .
Dr. Arthur Newsholme in his recent paper , " National Importance of Child Mortality , " f has asked the question ( p. 332):\#151 ; " Is it certain that a lower infant mortality will produce a survival of an increased proportion of physically inferior children ?
" and he reaches the conclusion that:\#151 ; " A high infant death-rate in a given community implies in general a high death-rate in the next four years of life , while low death-rates at both age-periods are similarly associated " ( p. 334 ) .
" So far from any weeding out of the weaklings being manifest , the counties with a high infant mortality have a death-rate which is relatively higher still in the next four years of life .
" As evidence for this he cites the correlation between infantile and child death-rates for the same period , and the life-tables for Healthy Districts and for all England and Wales .
These life-tables , however , show , as I indicate below , that a .heavier infantile mortality has actually a lower child mortality associated with it .
It appears to me that the non-allowance for differential environment renders Dr. Newsholme 's reasoning insecure .
It is not generally realised that the infantile mortality in England and Wales has not been falling but steadily rising since the restriction in size of families .
On the other hand , the child mortality has been steadily falling .
The numbers provided by the Registrar-General 's Life-Tables* are :\#151 ; * " The Intensity of Natural Selection in Man , " 'Drapers ' Company Research Memoirs , !
Dulau \amp ; ; Co. , 1911 .
t ' Journal of the Royal Sanitary Institute , ' 1910 , vol. 31 , p. 334 .
X ' Supplement to the 65th Annual Report , ' 1891\#151 ; 1900 , Part I , pp. xlviii 1 , London , 1907 .
1912 .
] The Intensity of Natural Selection in Man .
English Life-Tables .
Period .
Males .
Deaths per 1000 individuals .
Females .
Deaths per 1000 individuals .
0\#151 ; 1 years .
1\#151 ; 5 years .
0\#151 ; 1 years .
1\#151 ; 5 years .
1838\#151 ; 1854 163*50 134 -73 134 -71 132 -60 1871\#151 ; 1880 158 -58 127 -58 128 -73 124 -70 1881\#151 ; 1890 161-04 104-26 131 -13 98 *55 1891\#151 ; 1900 171 -86 94-01 140-66 89*83 Now , these data are the most considerable we can get and they involve all environments in this country .
If we might suppose the environment of the country as a whole to have remained constant , we could only conclude that it is certain that a high infant death-rate in a given community implies in general a low death-rate in the next four years of life .
It is difficult to realise how these data have come to be overlooked .
It is clear from them that the improved environment of the last 30 to 40 years has not effected any improvement in the infantile death-rate .
Either motherhood is less efficient , or the quality of the infants has degenerated , and they can resist worse a better environment .
It seems a priori unreasonable to suppose the improvement in child mortality wholly due to an improved environment which has produced no effect on infant mortality ; it is a priori not unreasonable to assume that some of the improved child mortality is due to increased infantile mortality .
Can we to any extent determine these proportions ?
( 2 ) There are two preliminary points to be considered in this matter .
In the first place the actuarial calculation of a life-table is not based on following a given group of persons through life , and determining how many die at each age .
At first sight , therefore , our objection ( a ) above applies to approaching the problem of natural selection in man from the life-table standpoint .
But a little consideration will show that it is far more justifiable than dealing with local districts in the same year .
We have in the first place an enormous mass of material\#151 ; the whole country\#151 ; and the results are taken for 10 or 16 years .
We thus obtain a stable community representing the average of what is taking place in a community gradually changing throughout the epoch .
The general experience of actuaries who use such tables for all purposes indicates that the death-rates calculated in this manner for various groups of individuals closely represent what occurs , if the same group be followed through life\#151 ; the accidental excesses or defects , scarlet fever or measles epidemics , etc. , are averaged out when a whole Prof. K. Pearson .
[ June 4 ' , country and considerable periods are dealt with .
I think we may say with fair certitude , for example , that of 1000 male children born in the period 1871-1880,158*58 would die in the first year of life , and of their survivors , 841 in number , ( 841/ 1000 ) x 127*58 would die in the second to the fifth years of life .
Further it must be remembered that owing to the period ( 10 and 16 years ) covered , the arrays born in each year but the last one or two have contributed their survivors to die in the whole or part of the child period 1\#151 ; 5 .
... " My next point is the problem of environment .
We are told that the environment has been continuously improving during the last 40 to 50 years , and if we ask for a measure of it we are very rightly referred to the falling death-rate , which remained stationary until 1866 , and then has been falling in a remarkable way ever since .
To correct therefore for continuously improving environment we may take something closely associated with the death-rate ; it occurred to me that expectation of life would bean excellent measure of this change of environment .
In order , however , to introduce no spurious correlation by taking the expectation of life at birth , which would include the influence of the very mortalities I am dealing with , I have taken the expectation of life at the age of 6 years as the factor by which to correct for the secular change in environment.* Of course , I very fully realise the audacity of determining correlations from four life-tables only , f but it must be remembered that each one of my figures is based upon a population of millions , and even if we considered our total number of observations four only , the fundamental partial correlations are still immensely significant as compared with their probable errors .
Further , I shall show that calculated and observed results are in agreement .
Tables I and II give the data .
Underneath the tables I have placed the chief statistical constants ; i = infantile death-rate ( 0\#151 ; 1 years ) in deaths per thousand ; c \#151 ; death-rate of children ( 1\#151 ; 5 years ) in deaths per thousand ; e \#151 ; expectation of life in children aged 6 .
* Should it be said that the expectation of life at six years of age is influenced by the mortality which occurs in the first five years , this in itself would be to admit that the death-rate is truly selective , the very point we have set out to prove , as against those who hold that the infantile death-rate is not selective .
t We have to bear in mind the vast amount of work involved in computing a table of this kind , and recognise that in calculating 14 life-tables in four series the General Registry Office has achieved a great task .
1912 .
] The Intensity of Natural Selection in Man .
Table I.\#151 ; Males .
Years of life-table .
i. c. e. 1838\#151 ; 1854 163 -50 134 -73 49-39 1871\#151 ; 1880 158*58 127 -58 50-38 1881\#151 ; 1890 161 -04 104-26 52-19 1891\#151 ; 1900 171*86 94-01 52-88 Means mi \#151 ; 163 *744 mc = 115 -145 me = 51 " 21 S. D. 's \lt ; n = 4 9992 \lt ; tc = 16 -6085 \lt ; re = 1 -3920 From these data we further deduce:\#151 ; = -0-6359 = +0-5620 rce =b \#151 ; 0-9941 These lead to the partial correlation coefficient of infantile and child death-rates for constant environment erd = -0-8605 \#177 ; 0-0875 .
Notwithstanding the small number of life-tables dealt with this is most substantially significant .
With our measure of constant environment , every increase of the infantile death-rate is accompanied by a marked decrease in the death-rate of the survivors in childhood .
The multiple regression equation is c = 115-145\#151 ; 0*3748 ( i\#151 ; 163*744)\#151 ; 11*0038 ( e\#151 ; 51*21 ) = 740-021 -0-3748 i-11*0038 e. ( 1 ) Thus for constant environment , i.e. e constant\#151 ; Increase in child death-rate = \#151 ; 0-3748 ( increase in infantile death-rate ) .
Or , looked at in another way\#151 ; Percentage decrease in child death-rate = 0*533 ( percentage increase in infantile death-rate ) .
In other words if you increase the infantile death-rate by 10 per cent , you decrease the child death-rate by 5*3 per cent. The approximation of the method is indicated by the following table:\#151 ; Period .
Observed c. Calculated c. A. 1838\#151 ; 1854 134 -73 186 -26 + 0-53 1871\#151 ; 1880 127 -58 126 -21 -1-37 1881\#151 ; 1890 104 -26 105-38 + 1-12 1891\#151 ; 1900 94-01 93-73 -0-28 474 Prof. K. Pearson .
[ June 4 , The results are in far better accord than we might have anticipated , and we see from these male data that\#151 ; ( i ) Our environmental measure appears to be justified by its close correlation ( 0*994 ) with the child death-rate .
( ii ) Notwithstanding this the infantile death-rate is fairly closely correlated with the environmental measure in the negative sense .
( iii ) The infantile and child death-rates , correcting for the environmental factor , are very substantially negatively correlated .
We may now turn to the female data\#151 ; Table II.\#151 ; Females .
Years of life-table .
\#187 ; .
c. e. 1838\#151 ; 1854 134 *71 132*60 50'00 1871\#151 ; 1880 128 -73 124 '70 62'56 1881\#151 ; 1890 131 -13 98 '55 54-35 1891\#151 ; 1900 140 *66 89 '83 55 '18 Means mt- = 133 '81 mc = 111 '42 me = 53 '02 S. D. 's ay = 4 *4922 ay = 17*7251 oy = 1 '9853 From these data we further deduce Tic = -0*5080 rie = +0*3042 rce = -0*9507 The partial correlation coefficient between infantile and child death-rates is eriC = \#151 ; 0*7399 \#177 ; 0*1459 .
This is again substantial and significant .
It means again that by fixing the environment as far as lies in our power , we have increased the negative correlation between infantile and child death-rates .
The multiple regression formula is now c = 111*42\#151 ; 0*9514 ( i\#151 ; 133*81)\#151 ; 7*8335 ( e \#151 ; 53*02 ) = 654*06-0*9514 *-7*8335 e. ( 2 ) Clearly for a constant environment\#151 ; \#166 ; Increase in child death-rate = \#151 ; 0*9514 ( increase in infantile death-rate ) .
Or , in other words\#151 ; Percentage decrease in child death-rate = 1*1424 ( increase in infantile death-rate ) .
1912 .
] The Intensity of Natural Selection in Man .
Thus in the case of females the child death-rate goes down about 1 per cent , for every rise of 1 per cent , in the infantile death-rate .
As we have seen , the correlations are not as high in the case of the females as in that of the males .
This is probably , to some extent , due ( i ) to the ages of the women being less reliable than the ages of the men , and this would affect e , and ( ii ) to the lower infantile death-rate in women , the selection being very likely pushed to a higher age ; the mere difficulties of birth are greater in the case of the boy and selection may thus be more immediate and stringent in his case .
The lower values of the correlations make the agreements between observed and calculated values less close , still there is nothing much to complain of here:\#151 ; Period .
Observed .
Calculated .
A. 1838\#151 ; 1854 132 *60 134 *22 + 1-62 1871\#151 ; 1880 124 *70 119 -86 -4-84 1881\#151 ; 1890 98*55 103 -55 + 5-00 1891\#151 ; 1900 89-93 87-98 -1-85 The remarkable feature of these English life-tables has been the falling child death-rate accompanying the rising infantile death-rate , a phenomenon which should have led those who assert that a high infantile death-rate implies a high death-rate in the next four years of life to pause .
It is noteworthy that the Registrar-General has drawn up life-tables in six cases for " selected material , " i.e. for " selected healthy districts."* These tables give the following results:\#151 ; Period .
Males .
Females .
i. c. e. i. c. e. 1849\#151 ; 1853 ... 112 -80 82 -40 53*94 92*64 79-13 53 '48 1881\#151 ; 1890 ... 119-15 61 -30 56-37 93-50 56 -62 57 -81 1891\#151 ; 1900 ... 121 -50 51 *31 57 *52 95 08 48*32 58-80 Without venturing to work out anything on the basis of two sets of three tables , we can yet recognise from the crude numbers that precisely the same phenomena occur even in the " healthy districts , " improving environment is related to an increasing not a decreasing infantile mortality and while the child mortality decreases with better conditions , it is highly correlated in a negative sense with infantile mortality , a correlation which will not be * ' Supplement to the 65th Annual Report , ' Part I , pp. lviii\#151 ; cxi , 1907 .
476 The Intensity of Natural Selection in Man , reduced but emphasized when correction is made for the secular change in environment.* It is not my purpose to enter into a discussion of the rise in infantile mortality which has gone on in this country since the fall in the birth-rate started , but I think enough has been done to indicate that when allowance is made for environment , a heavy infantile death-rate indicates a reduced child death-rate .
Further , so far from the conclusion that " a high infant death-rate in a given community implies in general a high death-rate in the next four years of life " being true , it is quite incorrect for 14 of the most important English life-tables .
These tables support parallel evidence of other kinds , that the Darwinian theory has application to civilised man , and that a heavy death-rate does mean the elimination of the weaklings .
To recognise this as a scientific law which controls the evolution of man as of other species is one thing ; to assert that the men of science who accept it desire themselves or through the State to play Herod to our modern infants is another thing and a pernicious thing .
It is the duty of science to discover what is happening in the first place , and there appears to me a fair amount of evidence now to show that Darwinism does apply to man , and that , for a constant environment , the higher the infantile death-rate , the more resistant will be the surviving child population .
* Because the correlation between c and e is so nearly perfect that the term ^/ ( l \#151 ; r^2 ) in the denominator of erCT- dominates the relationship .

