Data for the Problem of Evolution Man .
discourse ) roughly corresponding in their uses to partial coefficients of correlation .
In Section ( IV ) , the values of the Probable Errors , and the correlations of the errors in the chief constants , are obtained .
The probable error of Q is In Section ( V ) , a series of miscellaneous illustrations are given ( association of smallpox attack rate and non-vaccination ; association between temper of husband and wife , inheritance of artistic faculty \amp ; c. , from Mr. Francis Galton 's 'Natural Inheritance ' ; association between vigour of offspring and crossing of parentage in plants from Darwin 's ' Cross and Self-fertilisation ) .
^ In Section ( VI ) , the " Association of defects in children and adults , " is treated more at length as an example of the methods advocated , the material being drawn mainly from the Report of the Committee on the Scientific Study of Childhood .
It is shown that the association coefficient is almost uniformly higher for women than men , and for children than adults .
This last effect is however a mixed one , due partly to selection , partly to change in the individual , and the material available does not enable us to separate the partial effects .
These two laws of association appear to correspond to similar ones for correlation ; women being more highly correlated than men , and children than adults .
" Data for the Problem of Evolution in Man .
III.\#151 ; On the ^Magnitude of certain Coefficients of Correlation in Man , \amp ; c. By Karl Pearson , F.R.S. , University College , London .
Received November 20 , \#151 ; Read December 7 , 1899 .
1 .
This paper contains a number of data bearing on the correlation of characters , \amp ; c. , in man which have been worked out by my collaborators during the last few years , and several of which seem of considerable importance for problems relating to the .
evolution of man .
In each case the data were procured or reduced with a view to answering some problem which had directly arisen during our inquiries as to the action of natural selection on man .
Questions as to the alteration of correlation with growth or the influence of homogamy on fertility demand definite answers before the general theory of the influence of natural selection on a growing and reproductive population can be effectively developed .
0-6745 i-g Prof. Karl Pearson .
( A).\#151 ; On a Monthly Period in the Birth-rate .
2 .
While it is well known that there is an annual period in the birthrate , no attempt , as far as I am aware , has been made to ascertain whether a lunar period exists .
Accordingly , I applied to Dr. E. C. Perry , superintendent of Guy 's Hospital , who most kindly placed at my disposal the maternity records of that charity .
As preparatory work Mr. Yule and I extracted upwards of 6,000 births with their dates .
These , with the assistance of Mr. L. Bramley-Moore , we arranged in four groups , male and female , and in twenty-nine and thirty day lunar months as given by the almanac of the year .
In each lunar month the number of births on each successive day following the new moon was tabulated , and each month was then reduced to the same total , so that no month might be weighted by its relation to the annual variation in birth-rate .
Thus four Curves were obtained , each embracing the material for twenty-four months , and giving the daily fluctuation in male and female birth-rate for twenty-nine and thirty day lunar months .
In none of these curves was there any significant deviation from the diurnal average on any day .
The curves were then harmonically analysed by Mr. Yule ; the result gave no approach to agreement between the amplitudes or phases in the four cases .
Had there been any approach we should have gone on to 20,000 births as we originally proposed , but it seemed merely a waste of labour .
I conclude , therefore , that if there be any monthly period in the birth-rate , it is of very small importance .
There is little or no correlation between lunar phase and birth frequency .
The object of this inquiry was the following : The average regular recurrence of the monthly period in woman has been taken to suggest a tidal influence on the primitive ancestry of mankind ; * there is an indisputable correlation between birth and the date at which a monthly period would have taken place had pregnancy not intervened .
Hence a positive result might have confirmed this suggestion of tidal influence , as well as explained a certain amount of folklore connecting birth and lunar influence .
Our negative result merely shows that if lunar or tidal influence ever fixed the period of the menses , sensible correlation between the two has now disappeared .
( B).\#151 ; On the Correlation between Weight and Length of Infants at Birth .
3 .
A table of the length and weight of infants at birth was given in a " Report of the Anthropometic Committee of the British Association " * " In the lunar or weekly recurrent periods of some of our functions we apparently still retain traces of our primordial birthplace , a shore washed by the tides , " Charles Darwin , The Descent of Man , ' p. 161 , 2nd ed. * See " Variation in Sian and Woman , " The Cliances of Death , ' rol .
1 , pp. 296 , 307 , and 308\#151 ; 309 .
Data for the Problem of Evolution in Man .
25 for 1883 , and is published in the 'B .
A. Transactions ' for that year , p. 286 .
The measurements were comparatively few in number ( 450 for each sex ) and were made partly in London and partly in Edinburgh .
Accordingly I thought it better to obtain new material for calculating correlation .
Through the courtesy of Dr. J. D. Rawlings I was able to obtain copies of the measurements made on between 2000 and 3000 new born infants at the Lambeth Lying In Hospital .
From these 1000 male and 1000 female babies born at the normal period were taken , twins being excluded .
The correlation tables and the calcula-tian of the variation and correlation constants are due to Mr. L. Bramley-Moore .
The correlation tables seem of such importance for medical and other purposes that they are given below ( see Tables II and III ) .
The following table contains a summary of results .
I.\#151 ; 'Weight and Length of New-born Infants .
1000 of each Sex .
Females ... Mean .
Standard deviation .
Coefficient of variation .
Coefficient of correlation .
Weight .
Length .
Weight . .
Length .
Weight .
Length .
Weight with Length .
lbs. 7-073 \#177 ; 0-021 ins .
20 -124 \#177 ; 0-025 lbs. 1-006 \#177 ; 0 -015 ins .
1-177 \#177 ; 0-018 14-228 5-849 0-622 \#177 ; 0 013 Males 7-301 \#177 ; 0 -024 20 -503 \#177 ; 0-028 1-144 \#177 ; 0 -017 1-332 \#177 ; 0-020 15 -664 6-500 0-644 \#177 ; 0-012 Table I confirms the view already expressed by me , that the male infant is at birth more variable than the female : that female infants from six to ten years of age are more variable than male in both weight and height , appears to be not a result of selection but of growth.* Both sexes lose not only variability but correlation as they grow older , and this is a fundamental point to be borne in mind , when attempts are made to trace the influence of natural selection in the change of variation or correlation in a group of growing animals .
Even the co-efficient of variation , which is far more stable than the standard deviation ( or absolute variability ) is seen to alter considerably with growth .
So far as weight and height are concerned female and male both lose variation as they grow older , but women .
less rapidly than men .
So far as correlation between weight and height is con* See " Variation in Man and Woman , " The Chances of Death , ' vol , 1 , pp. 296 , 307 , and 308\#151 ; 309 .
Prof. Karl Pearson .
cerned , men start with a scarcely sensible advantage over women as infants , and conclude as adults with an immensely less correlation than women , among whom it appears to have slightly increased , or at any rate not to have decreased .
Until we have accurate numerical determinations of the change in the correlation between organs with growth , it is impossible to attempt to measure quantitatively the influence of a selective death-rate on growing living forms .
We can only deal with the influence of sudden selection on growing organisms , or of long continued selection on adult life .
( C).\#151 ; On the Correlation between Stature , Weight , Strength , and Head Index in the Case of Adults .
4 .
The measurements upon which these results are based were taken from cards in the possession of the Cambridge Anthropometrical Committee , who kindly allowed me to have copies taken for 1000 cases of male students , and for the whole series of female students , which unfortunately were only about 160 in number.* The whole of the lengthy arithmetic involved in the calculation of the constants was undertaken by Miss C. D. Fawcett , B.Sc. The bulk of the students were between nineteen and twenty-five years of age , although some few were older ; they may be taken to represent adults , who in the great majority of cases were in good physical health and training , and were not troubled with the superfluous weight of a later period of life .
I will first put in a separate table the results for weight and height , in order that the constants for adults can be easily compared with those for new-born infants .
IY.\#151 ; Weight and Height of Adults .
1000 Males , 160 Females .
Females ... .
Mean .
[ Standard deviation .
Coefficient of variation .
Coefficient of correlation .
Weight .
Height .
Weight .
Height .
Weight .
Height .
Weight with Height .
lbs. 125605 \#177 ; 0-773 ins .
63 -883 \#177 ; 0-130 lbs. 14 -030 \#177 ; 0 -546 ins .
2-361 \#177 ; 0-092 11 -170 3-696 0-721 \#177 ; 0-026 Males ... ... 152-784 \#177 ; 0-353 68 -863 \#177 ; 0-054 16 -547 \#177 ; 0-250 2-522 \#177 ; 0*048 10 -830 3-662 0-486 \#177 ; 0 016 * The establishment of an anthropometric laboratory at Newnham College will soon increase this total .
Data for the Problem of Evolution Man .
This table shows us the decrease of variability with age in both weight and height .
But the female is now more variable than the male .
Were this change to be attributed to natural selection , with a stronger incidence on the male than the female , then we have the anomaly that the correlation has been reduced in the male , but increased in the female , while theoretical considerations would lead us to the conclusion that it ought to be reduced in both .
We are compelled to-consider the changes in variation and correlation as due to growth or nurture ; or , if there be selection , it is to a large extent screened by these causes .
V.\#151 ; Correlation between Height , Weight , and Strength of Pull .
( Data for 1000 Males and 160 Females .
) Females * , Males ... .
Organs .
Coefficient of correlation .
1 Strength of pull / J and height [ 0-216 \#177 ; 0-052 0-303 \#177 ; 0-019 Females . .
Males ... .
1 Strength of pull f J and weight \ i 0-338 dh 0-049 0-545 \#177 ; 0-015 1 1 expect the greater correlation of the male in these two cases is due to the fact that a better physical training has taught him how to^ make use of his height and weight , especially the latter , in exerting his .
strength .
VI.\#151 ; Correlation between Height , Weight , Strength of Pull , and Head Index .
( Data for 1000 Males only .
) Organs .
Coefficient of correlation .
Heiglit and head index .
i -0-082 \#177 ; 0-021 Weight and head index 0-011 \#177 ; 0-021 !
Strength of pull and head index 0*041 \#177 ; 0-021 Thus it is only in the first case that the correlation with head index can be considered as significant , and in this case it is negative .
There is no reason , then , for supposing brachycephalic persons stronger or heavier than dolichocephalic , but they do appear to be slightly shorter .
We conclude , therefore , that dolichocephalic persons ( 1 races ) ' Prof. Karl Pearson .
also will be found to be taller than the braehycephalic .
Here we have dealt with the correlation between shape of head and physique , the correlation between absolute size of head and physique will be given later .
It would be of the greatest value to obtain the intensity of correlation between shape of head and intellectual capacity .
We hope to return to this on another occasion.* 5 .
We may place here the variation results for pull and head index .
VII .
Sex .
Organ .
Mean .
Standard deviation .
Coefficient of variation .
Female . . .
Male 1 pull in f J lbs. 1 49 *220 \#177 ; 0-452 84 -016 \#177 ; 0-270 8-217 \#177 ; 0-320 12 -676 \#177 ; 0-191 16-69 15-09 Male head index 79:572 \#177 ; 0-064 2 -999 \#177 ; 0-044 3-77 ( D).\#151 ; On the Correlation of Fertility with 6 .
In the reduction of my family measurements , I have been much struck by the very high values obtained for the correlation in characters between husband and wife .
For 1000 cases in which the stature of husband and wife were determined , the correlation was nearly 0*3 !
This is almost as close a resemblance as we have found for some characters between father and daughter .
Now there is little doubt that there is a certain amount of conscious assortative mating in this respect ; a short man does not , as a rule , like a very tall wife .
There is further an unconscious mating arising from neighbours marrying ; neighbours in England often mean persons of the same local race , and such local races differ considerably in their mean statures.f But my data were very largely drawn from the professional classes , and in large towns like London to marry " in the set " hardly means to marry into the same local race .
The parents , however , may in some cases have come to London from the same locality .
The question whether husband and wife spring from the same rural district is one of considerable interest , and deserves special investigation .
I hardly believe , however , that it will be found a source contributing much to the intensity of assortative mating in my own data .
A large proportion of the data cards were filled in by members of the London professional # Measurements are now being made on brothers and sisters in schools , the apparatus being provided by aid from the Government Grant .
It will , however , be some years before sufficient data have been collected .
t Mr. Francis Galton has pointed out this source of indirect assortative mating to me as worthy of consideration , , Data for the Problem of Evolution in Man .
29 classes , and marrying within the district or set cannot , I think , introduce much local race influence .
I accordingly turned the problem round , and asked if there appeared any reason why , in collecting my material , I should have selected unconsciously homogamous marriages .
Now , clearly , I was much more likely to get a return from a large than a small family ; one member of a family of eight was more likely to take interest in the matter than one member of a family of two or three .
It seemed to me that out of 10,000 families I was clearly more likely to get returns from the larger families than the smaller ones , remembering that 26 per cent , of the families correspond to 50 per cent , of the offspring .
Hence arose the important problem is fertility associated with homogamy ?
When like mates with like , is the number of progeny greater than when like and unlike mate ?
Put in this way the problem appears to be of first class importance for the theory of evolution .
If homogamy rather than heterogamy results in fertility , then we get a first gleam of light on what may be ultimately of vital significance for the differentiation of species .
When any form of life breaks up into two groups under the influence of natural selection , what is to prevent them intercrossing , and so destroying the differentiation at each fresh reproductive stage Various hypotheses\#151 ; isolation , recognition marks , physiological selection\#151 ; have been propounded .
But if like mating with like connotes greater fertility , the answer to the problem of differentiation would be simply summed up in differential fertility .
We should have merely a case of genetic selection arising from the correlation of fertility and homogamy .
We must be careful , however , not to rush to any conclusions without ample data .
In particular we must not confuse homogamy with endogamy .
Nor must we argue that relatives being closely alike , kin-marriages would mean increased fertility .
Darwin has shown statistically that , as a rule , self-fertilised flowers are more sterile than cross-fertilised ; kinship , sameness of stock , means likeness of characters , but likeness of characters does not necessarily indicate sameness of stock .
It is quite possible for like individuals of different stocks to be fertile and like individuals of the same stock to be in part or wholly sterile inter In fact , that homogamy means fertility may in all or certain forms of life be dominated by a more potent rule , namely , that endogamy means sterility .
The two statements are not contradictory if we interpret homogamy to mean the mating of two like individuals , not in the first place like because they come of the same stock .
In fact , if a man seeks a wife of stature corresponding to his own , he will as a rule have a larger field of suitable mates in the general population than within his own limited kin .
Bearing this point in mind , I now turn to the somewhat narrow data available at present for the influence of homogamy in the matter of stature on fertility in man .
Prof. Karl Pearson .
7 .
Taking 205 marriages in which I had details of the stature of husbands and wives and the size of their families , two correlation tables were formed ( i ) for husbands and wives , as such , ( ii ) for fathers and mothers , i.e. , for husbands and wives weighted with their fertility .
The tabulation and calculations for ( ii ) are due to Mr. L. Bramley-Moore , with the assistance of Mr. K. Tressler .
For ( i ) I had worked out the data some years ago myself .
The following results were obtained:\#151 ; VIII.\#151 ; Correlation between between Statures of Husband and Wife .
No. of cases .
Yalue of correlation .
Husband and wife ... .
Father and mother ... 205 965 0*0931 \#177 ; 0*0467 0*1783 \#177 ; 0*0210 Now these results seem at first sight significant .
We have practically doubled the intensity of assortative mating by weighting the observations with fertility .
Fertility would thus not be distributed at random , but would increase with the amount of homogamy .
The process of collecting the original data here conceived was totally different from that of my own family data cards , the influence of size of family on chance of procuring data being I consider nothing like as marked.* This is , I think , the source of the difference in correlation of stature between husband and wife being so reduced .
8 .
In order to further investigate the matter directly , a correlation table was prepared for me by Mr. L. Bramley-Moore , in which the variables were ( i ) difference in stature of husbands and wives , and ( ii ) size of family .
In this case the statures of the wives were reduced to male equivalents before the difference was taken , f Thus the difference is zero , when the wife has the female stature which corresponds to her husband 's .
The calculations on this table were made by Miss Alice Lee , D.Sc .
The correlation was found to be , and its value -0T201 \#177 ; 0-0464 Thus it would seem that large difference in stature means small fertility .
Bur there is danger of a fallacy here , which requires careful investigation .
The regression equation for size of family ( / ) in terms of difference of stature of husband and wife in inches ( d ) is * ' Phil. Trans. , ' A , vol. 187 , p. 269 .
They were not a selection of families by size , but rather by the existence of fairly complete ancestral records .
t The ratio of mean statures == 1*08 about , and was added to the female stature to convert it into its male equivalent .
Data for the Problem of Evolution in Man .
31 f = 4*7 - \#151 ; cl. J 10 Thus since the range of difference is from about \#151 ; 10 to +10 inches , we have a fertility varying from 5-7 to about 3*7 , or about 42 to 43 per cent , variation in fertility as we pass from wives relatively 10 inches taller than their husbands to wives relatively 10 inches shorter !
In other words , our homogamous influence is really cloaked by the fact that big husbands and small wives have for extreme cases some 42 per cent , less offspring than small husbands and big wives , a result of considerable interest from the standpoint of genetic selection , and possibly capable of easy physiological explanation , if pelvic measurements are closely correlated with stature .
In order to disentangle the two factors , I divided up my 205 pairs into the quartile groups , and found the following results IX .
Quartiles .
Range of difference .
Total offspring .
Mean size of family .
1st -9*5 " to - 2*25 " 258 *5 5*04+0-26 2nd ... ... -2*25 " to + 0*417 " 260*333 5-08+0-26 3rd. .
+ 0 417 " to + 2*583 " 229 *083 4*47+0-26 4th ... ... .
+ 2*583 " to +11*0 " j 215*083 4-20+0-26 1 The total number of offspring was 963 , or the number to be expected in each quartile 240*75 , while the average size of the family is 4*6976 , with a standard deviation of 2*7826 .
These results show us that very nearly half the marriages occur with the wife relatively taller than the husband , and that such marriages give 54 per cent , of the total offspring as against 46 per cent , produced wnen the husband is relatively taller than the wife .
The mean family with mother relatively taller than father is 5*06 \#177 ; 0*18 , and that with father relatively taller than mother , 4*33 \#177 ; 0*18 , a difference which may be taken as significant .
Grouping the 1st and 4th and the 2nd and 3rd quartiles together , we have for the mean family when husband and wife differ considerably 4*62 \#177 ; 0*18 , and for the mean family when they differ but little 4*77 \#177 ; 0*18 .
This difference in itself , however , unlike that recorded by the previous process of investigation by weighting with fertility , would hardly be sufficient to demonstrate a correlation between fertility and homogamy .
I accordingly made out a fourth correlation table , in which the variables tabulated were difference of relative statures of husband and wife , without regard to its sign and size of family .
The mean relative Mr. W. Steadman Aldis .
the Numerical difference in stature was found to be 2-751 inches , and the standard deviation of its distribution 2-070 inches .
The correlation between difference in stature and size of family was - 0-0236 , or greater fertility appears associated with small differences .
The observations , however , are so few ( 205 ) that the probable error of the correlation is 0-0471 , and thus no stress can be laid on this result .
If the reader asks why is not the result in S 7 conclusive , the answer must be , it would be conclusive , if the means of the husbands and wives weighted with their fertility were the same as when they were unweighted ; increased correlation would then necessarily connote that fertility was associated with homogamy .
Actually the fact that absolutely taller mothers are the more fertile alters the centre of the correlation table , and somewhat obscures the issue as to whether the whole increase of correlation is really due to homogamy being correlated with fertility .
That in man , whether from conscious or unconscious sexual selection , there is far more homogamy than has hitherto been supposed , my family data cards amply demonstrate .
If in man , then with great probability we can consider it to exist in other forms of life .
But the existence of such homogamy is of immense importance for the problem of differentiation .
The present statistics do not enable us to say whether homogamy in man is definitely correlated with fertility ; they do show that fertility is not a random character , but depends upon the relative size of husband and wife , and thus bring evidence in favour of genetic selection .
I can conceive no more valuable investigation than a series of experiments or measurements directed to ascertaining whether homogamy is or is not correlated with fertility , but such investigation , bearing in mind Darwin 's conclusions , should carefully distinguish between exogamous and endogamous homogamy .
" On the Numerical Computation of the Functions G0(\#171 ; ) , G^x ) , and J n(%\/ i)"By W. Steadman Aldis , M.A. Communicated by Professor J. J. Thomson , F.B.S. Deceived and Eeacl June 15 , 1899 .
1 .
The complete solution of the equation may be written y = AIn ( x ) + BK " where ( i ) ;

