]\gt ; Simple Mendetian Population Mating at Random .
prove of considerable value in bacteriological investigations .
Further experiments are now being made with other orga1lisms which up to the present time have not been proved to be absolutely distinct .
Numerons experiments were made on the resistance of these endotoxins to heat , and every result shows that they are thermostable eve when } ) osed to a temperature of 10 C. for 30 millutes , and are still able to exert their specific action , as is shown by the following experiment:\mdash ; *This extract was exposed to a temperatul'e of C. for 30minutes .
( In the Deterrnination of the Chief Cltions between Colin the Case of Simple endelian P Mating By E. C. SNOW , B.A. , Biometric Laboratory , University College , London .
( Communicated by Prof. Karl Pear.son , F.R.S. Received June 3 , \mdash ; Read , 1910 .
( 1 ) In a paper communicated to the Royal Society in April , 1909 , Prof. Pearson obtained the ganletic correlations between the offspring and the ancestry in each grade in a simple Mendelian population mating at random .
By a ' simple Mendelian population\ldquo ; for a given character , he understood one which started with any definite ratio of dominant individuals to recedents ( aa ) .
These mating at random give rise to a ulation which may be written in the form ( aa ) , and of which , without , the proportions of dominants , recedents , and hybrids are known to } constant , with continued random mating , during successive generations .
Prof. Pearson that , both in the case of gametic and somatic characters , the ancestral correlations diminished in 38 Mr. E. C. Snow .
Correlations between [ June 3 , a geometrical progression , and thus obeyed the fundamental principle of the Law of Ancestral Heredity , as deduced from observations on man and other living orms .
( 2 ) It is difficult to believe that the characters dealt with in the case of Mendelian investigations on animals can be largely affected by environment , but it is easy to allow for this inflnence by the of partial correlation .
If , in an igatiun on any given cter , the subscript 1 denote offspring , 2 the ancestor in any generation , the offspring 's environment , and 4 the ancestor 's environment , then the correlation between 1 and 2 for constant 3 and 4 is given by If suppose the environment of a given stock to remain constant , this reduoes to the simple form where is the correlation between charactPr and environment .
* This should be absolutely true for pairs of brotbers reared in the same home Many investigations into the value of for a great variety of characters in man have been carried out in the Statistical ) oratory at niversity College , and these show that the value of is remarkably small .
Its value is , and it rarely reaches an intensity as high as .
It can only introduce a second order correction into the correJation , and one well within the limits of the probable error of } ation .
If were , the in fraternal CtJt.relation would only be from to .
Thus , until number and accu1acy of obselvations have reac , hed a much stage , the influence of ellvironment need not be considered .
( 3 ) A noteworthy conclusion reached by biometric workers by actual measnrements on man is that the correlation of brothers is not much larger than the correlation of parent and offspring .
A priori , we A German writer has recently given a formula .
correctin fraternal correlation for the iufluence of environmeut .
It differs fronI the above form , and in a factor multiplying .
He assumes that the high values for the parental and fraternal correlations in mau as compared with to be expected for somatic chardcters in the Mendelian hypnthesis are due to neglect of the influence of environment by the biometricians .
He was ably u of the amount of attention which has been given by biometric schooI to the detsrmiuation of ental correlations and to the comparison of the correlat ions for where 1lment is likely and unlikely to produce effect , e.g. ] ) eredity in statnre fo1earm with heredity in eye ] and cephalic index .
See W. Weinsberg , ' Zeitsc.hrift fiir Vererbungslehre ' 1909 , vol. 1 .
1910 .
] in a Population at should expect it to be larger , for brothers have two parents in common , while between father and son comes in the influence of the mother , exerted only on the latter member of the pair .
Thus Sir Francis Galton , in his ' Natural Inheritance ' ( p. 133 ) , makes the resemblance twicc as great as the paternal .
Pearson* , dealing with three characters in 12 tables of over 1000 each , folmd the average correlation for parent and child to be , and for three characters in nine tables the average fraternal correlation to be .
For eye-colour in man he found the parental correlation and the fraternal , showing a slightly reversed relationship .
Both , however , compel us to affirm that the degrees of resemblance between parent and offspring and between brethren are very nearly equal and differ bnt slightly from .
The values which Mendelian theory leads us to expect are shown in Table IX .
( 4 ) Further conclusions can be drawn now that the measurement of resemblance between pairs of cousins and between uncles or aunts and nephews or nieces has been completed .
by a great ietv of characters , the cousin relationship has been found to be as close as , if not closer than , the avuncular relationship .
This result is , of course , no mole remarkable than that between the parental alId fraternal nces referred to ) .
However , in view of the publication of a number of avuncular correlations which have been worked out in the aboratory at University College , it seems desirable to show that it is at least ( , onsistent with some one or other theory of heredity .
From Table IX it will ) seen that on the Mendelian theory the relationship between ousius is as close as that between uncle and nephew .
The point is one which is of considerable importance in medical diagnosis .
The parents ' brothers sisters have been usually included , and the con , qins excluded , in considering the family history .
The fact , owever , that a 's cousins resemble him as closely as do his uncles and aunts shows cousins , usually more numerous , provide at least as llaterial for an estimate of his tendencies as his parents ' brothers and.sisters .
The paradox that a man 's father 's brother 's son resembles him , on the average , as closely as his father 's brother , although the nfluel ) of the fathet.'S brother 's wife has come in , is only a paradox so as we overlook the fact that somatic and gametic characters are not related but only correlated .
Almost any determinantal theory of heledity brings out the fact that , on the average , a man resembles his as closely as he does his brother , and , further , that he is not more like to his 'Biometrika , ' vol. 2 , pp. 378 , 387 .
'Phil .
Trans , 1900 , vol. 195 , p. 106 .
40 Mr. E. C. Snow .
Correlations between [ June 3 , uncle than to his uncle 's son .
This point is illustrated in the present paper from the investigation of the collateral correlations of a simple Mendelian population mating at random because that determinantal theory is much in vogne to-day .
Ic will be used here to indicate that the facts reached by obsel'vation as to the degrees of resemblance between an individual and his ascendants and , and considered by some to be paradoxical , are iously e legitimate developments of Mendelian theory .
If they are paradoxes of biometric observations , they are also paradoxes of Mendelian theory , and highly probably of other determinantal hypotheses whioh may hereafter be developed .
three degrees of collateral relationship are worked out in the following paragraphs , viz. :\mdash ; ( A ) The correlation of siblings , i.e. of brethren regardless of sex .
( B ) The rrelation of uncles and aunts with nephews and nieces .
* ( C ) The correlation of cousins regardless of sex .
The ebra presents no special difficulties beyond laboriousness .
Starting with the population of the form ( aa ) , after random of pairs of individuals of each class , the types of families produced will be those given in the square brackets in Table I. The lcies o. the families will be the factors which mnltiply the the brackets .
Thus , for example , the mating of the individuals possessing the constituent ( AA ) with tlIe possessing the hybrid ( Aa ) gives rise to families , in each of which individuals have the protogenic constituent .
and the hybrid constituent ( Aa ) .
The word " " siblings\ldquo ; bas been reintroduced and largely accepted as a couvenient term for brethren of both sexes , but no , terms are in to denote ) or aunt and lephew or niece regardless of sex .
Prof. Pearson has suggested eldersib lor uncle or aunt and szbmag for nephew or niece as coIlvenient Anglo-Saxon uames .
The nnmber of matings is taken of the form 16 in order to prevent fractions occurring later on in the paper .
See Table IV .
1910 .
] in a Mendelian Population at Table I. ( A. ) Correlation between Brothers .
( 7 ) From Table I we can pick out all the pairs of brothers and arrange them according to the gametic character exhibited each member of the pair .
The cases which can arise , and the number of pairs of brothers in each case brothers !
, - . .
Both brothers S. .
Both brothers One brother ( AA ) , other brother One brother ( AA ) , other brother brother ( Aa ) , other brother From the above expressions a symmetrical table can be constructed , from which to obtain the gametie correlation between brothers .
This is giveu in Table II .
By above expressions under the " " \ldquo ; and " " not \ldquo ; another table can be formed to give the somatic COlTelation .
This is given in Table III .
In both tables the factor 16 , which multiplies every term occurring , has , for brevity , been omitted .
1910 .
] in a Population Mating at Random .
Hence from which it follows that the regression is linear and the correlation is ( 9 ) Somatic \mdash ; In this case , from Table III , we have , measuring from the " " not \ldquo ; row , The correlation therefore depends upon the ratio of to , i.e. upon the ratio of the nunlber of individuals in the population the dominant aracter to the number the recedent .
For diHerent values of and the above value of ' can range between .
If its value is 10 ) In considering the value to be given to in these formulae in order that they can be compared with statistical results , we must rememl ) that the latter are obtained from families of varying size , from which we selections of fraternal couples .
We must therefol'e look upon our tions as a random sample of all the possible of brethren obtained from an indefinitely large number of matings .
That is , we must make indefinitely large .
Then we get the gameti correlation to be 12 , and the range for the somatic correlation from to , with the when * This particular value of the somatic fraternal correlation was reached Prof. Pearson in his lnemoir in the ' Phil. Traus , 1904 , vol. 203 .
In that memoir he deaIt with cl1aracters only , and he assumed the population to be generated ) series of hybridisations and not from a combination of individuals exhibiting the protogenic , allogenic and hybrid constituents in certain portions .
But he more general in his supposition that the character depended upon Mendelian couplets , not on a single one .
44 Mr. E. C. Snow .
between [ June 3 , B. ) Corrdation between Uncle and Nephew .
( 11 ) Iu this case we require not only the various types of families given in also the pring of these families when mated at random .
But it will not be sufficient to mate the nine types of families given by Table I with ) eral population in which the proportioI ) ) ossessing the protogenic , hybrid , and genic c are ( AA ) , ( Aa ) , and .
( aa ) , any particular matin will be of a member of one type of family with a member of the same or another type .
An avuncular relationship will arise with the nlembers of both families from which the mating couples are selected .
Thus we must discover the pring which arise from the of each type of family in Table I with itself and with each other .
This is given in Table , in which 4 is taken to be the number of matings of each pair .
This allows for a possible change of fertility in consecutive generations .
( 12 ) To explain Table , consider , for example , the mating of the family of type [ Type ] with the family of type [ Type ] .
Of the of Type will mate with of the from Type , giving ofT'spring : will mate with of the ( Aa ) from iving 2 ( Aa ) offspring .
2 will with of the from Type , giving offspring .
Of the ( Aa ) of Type will with of the ( AA ) from Type , giving will mate with of the ( Aa ) from Type , giving spring .
will mate with of the from Type , giving 2 offspt ing .
The complete result will consist of the of the six combinations written out .
Since the frequency of Type family is , and of the Type is , there will be alLogether 16 such gregations as the above .
The is represented in tlIe table , in which the falnilies derived through any particular individual the constituent , or ( aa ) are written in the same horizontal or vertical line as the stituent itself .
The aggregations can be read off in like manner .
( 13 ) From Table we can pick out uncle and nephew in pairs according .
Soc. .
vol. 83 , oxs [ I of eneration .
' -\mdash ; 1910 .
] in Mendelian Population at Random .
to the particular constituent , protogenic , hybrid , or allogenic , each member of the pair possesses .
It will be more convenient , however , to ltlake use of the fact that the number of pairs of uncle and nephew will be equal to the number of pairs obtained by taking any one of the children from given parents with any one of the grandchildren , minus the number of pairs of parents and offspring among such pairs .
This latter can be obtained more simply than by Table .
For Table I enables us to pick out parent and offspring in pairs and arrange them according as each member of the pair is ( AA ) , ( Aa ) , or ( aa ) .
But we require the number of such pairs in the next generation .
Allowing for the change in the fertility from to , it is clear that the number in the second generation will be obtained from the number in the first by writing for and multiplying by .
From Table I we can at once pick out the following terms for the first generation .
Parent , offspring ( AA ) , , , ( Aa ) ( AA ) , , , ( aa ) ( Aa ) , , , ( AA ) ( Aa ) , , , ( Aa ) . . .
( Aa ) , , , ( aa ) ( aa ) , , , ( AA ) ( aa ) , , , ( aa ) In obtaining these expressions , one parent only was considered .
From the symmetry of the table , if both parents are to be taken into account , each term must be multiplied by two .
Since this factor will occur throughout the , it can be omitted .
Collecting the above terms , and applying the modification necessary to the first generation into the second , we have , for this latter generation , the frequencies:\mdash ; Parent ( AA ) , offspring ( AA ) ( AA ) , , , ( Aa ) ( AA ) , , , ( aa ) , , ( AA ) ( Aa ) , , , ( Aa ) ( Aa ) , , , ( aa ) ( aa ) , , , ( AA ) , , ( Aa ) Table we have now to pick out and arrange the various pairs 46 Mr. E. C. Snow .
Correlations between [ June 3 , which can be obtained by taking any one in the first generation with any one of the same stock in the second generation ( i.e. both sons and nephews ) .
To economise space , the details of the terms are given in the case of only of the nine pairs .
The other six can be written out in the same manner , but the results alone are stated .
The three pairs given in detail Member of the first generation ( AA ) , member of the second genera .
When these terms are collected , and simplification performed , the whole expression reduces to the simple form .
Member of first generation ( AA ) , member of second generation .
This , when simplified , becomes .
1910 .
] in Mendelian Population Mating at Random .
Member of first generation ( AA ) , member of second generation 16 .
t. t .
reduces to The frequencies of the other pairs are ( the member of the first genelation in each case being placed first ) ( Aa ) with ( AA ) 512 , ( Aa ) , , ( Aa ) , ( Aa ) , , ( aa ) , ( aa ) , , ( Aa ) . . .
512 , ( aa ) , , ( aa ) .
from these expressions those found for the combinations in the case of parent and of the second generation given above , we get the following frequencies for combinations of umcle and nephew : Uncle ( AA ) , nephew ( AA ) 64 ( AA ) , , , ( AA ) , , , ( aa ) 512 ( Aa ) , , , ( AA ) ( Aa ) , , , ( Aa ) ( Aa ) , , , ( aa ) ( aa ) , , , ( AA ) ( aa ) , , , ( aa ) Correlation tables can now be formed from which to derive the gametic and the somatic relationships between uncle and nephew .
Iable gives the gametic relationship ; Table the .
In both of these tables the factor , which multiplies every cell , has been omitted .
Hence Therefore the regression is linear and the correlation .
As before , we must take a random sample from the famihes obtained by making indefinitely large .
The correlation then becomes 1910 .
] in a Mating at Random .
( 15 ) Somatic Corrdation.\mdash ; Table gives , measuring from the " " not \ldquo ; row\mdash ; and Again the correlation depends upon the relative values of and .
The limits between which can lie for different values of and are and .
If , its value is .
Putting infinite , the range for is from to , with the value when ( C. ) Correlation betueen First ousins .
( 16 ) From Table we can pick out every pair of first cousins , and arrange them in groups according as each member of the pair exhibits the constituent , ( Aa ) or ( aa ) .
But , again , it is more convenient to obtain these groups indirectly .
For the number of pairs of cousins in any group will be equal to the number of pairs which can be taken from all children the same two grandparents , minus the number of pairs of brothers among such children .
Since the population is stable during succeeding generations , the latter can be obtained at once from the expressions given in ( 7 ) by writing 4 for 16 in that table and multiplying by the factor 16 .
We therefore have the following expressions for the number of pairs of brothers among all grandchildren:\mdash ; First Second .
brother .
( AA ) ( AA ) ( AA ) ( Aa ) 64 ( AA ) ( aa ) ( Aa ) ( Aa ) ( aa ) ( aa ) The number of pairs of all possible grandchildren ( i.e. siblings , together with cousins ) can be easily obtained from inspection of Table .
From symmetry , only six combinations are now required .
Three of these are given in detail , the final expressions only of the other three being written down .
hout the detailed expressions is written for the product VOL. LXXXIII.\mdash ; B. 50 Mr. E. C. Snow .
between [ June 3 , First randchild ( AA ) , second grandchild S. .
This reduces to First grandchild ( AA ) , second grandchild This becomes 512 .
First grandchild ( AA ) , second grandcIJild ( This becomes The other combinations reduce First grandchild ) , second grandchild To obtain the frequencies of the various pairs of cousins , we have to subtract from the above six expressions the corresponding expressions for pairs of brothers .
We then get the following frequencies:\mdash ; Both cousins One cousin ( AA ) , other ( Aa ) ( AA ) , , , ( aa ) .
Both cousins One cousin , other Bobh cousins The correlations tables for cousins can now be formed .
Table gives the ement for the gametic relationship , Table VIII for the somatic .
In both cases the factor 64 has been omitted .
The tables must , of course , be made symmetrical .
Table \mdash ; Gametic Correlation between Cousins .
( Factor 64 omitted .
) 52 Mr. E. C. Snow .
between Collaterals [ June 3 , Table VIII .
Somatic Correlation between Cousins .
( Factor 64 omitted .
) ( 17 ) Gametic Correlation.\mdash ; From Table we find , after simplification\mdash ; From these , .
Hence the regression again is linear , and the correlation .
When is indefinitely large this becomes ( 18 ) Somatic Correlation.\mdash ; In this case , after reduction , we find\mdash ; and Hence in the case of cousins also the somatic correlation depends upon the ratio of to .
The range of for different values of and extends from to .
If its value is .
When is indefinitely large , the range is from to , with the value when ( 19 ) In Table IX , all the correlations which have been worked out from the Mendelian hypothesis of " " unit characters\ldquo ; are collected ether .
The values of the parental and grandparental correlations are from Prof. Pearson 's paper .
* The rest have been obtained in the present paper .
* On the Ancestral Correlations of a Mendelian Population ' Roy .
Soc. Proc 1909 , vol. 81 .
1910 .
] in a Table IX.\mdash ; List of Mendelian Correlations between an Individual and his Relatives .
From this table we notice that , on the Mendelian hypothesis , the parental and sibling gametic correlations are equal , while the grandparental , avuncular , and cousin gametic correlations are also equal and are just one-half of the former .
In Table X the ancestral and collateral correlations in man and certain animals founded on biometric research , together with the correspondgametic correlations found above on the Mendelian assumption of " " unitcharacters are collected .
( 20 ) The following general conclusions can be drawn from the correlations for somatic characters given in Table IX .
( i ) If the cousin correlation is greater than the parental .
This seems remarkable , but it must be borne in mind that the condition is derived from the Mendelian idea of " " unit-characters , \ldquo ; and no prediction can be made as to when the condition is fulfilled .
The only cousin correlations worked out for the biometric school are those by Miss Elderton .
No correlation for a single character in cousins was found to exceed the parental one .
In eye-colour alone did the former approach the htter .
The parental correlation in this case was , while the cousin correlation had the values , and for male and male , male and female , and female and female respectively , the mean value being ( ii ) If is small , i.e. if the recedent constituent is rare , as in the case of albinism in man , the correlations founded on Mendelism have low values , and in certain cases approach zero .
For pathological characters the figures found by Miss Elderton were certainly much smaller than for other characters in 54 Mr. E. C. Snow .
Correlations between Collaterals [ June 8 , Table X.\mdash ; Mean Ancestral and Collateral Correlations in Man and Animals .
* cousins , but this was ascribed to the incompleteness of the records .
No general tendency in the direction of lower correlations for certain characters has yet been noticed in biometric investigations .
( iii ) The correlation between siblings is always greater than that between parent and offspring .
This is in accordance with a priori expectation ( see para .
3 above ) and agrees with statistical results .
( iv ) The grandparental correlation is always the same as the avuncular .
This also agrees fairly well with statistical conclusions .
In his paper of last year Prof. Pearson drew attention to the fact that the * Each biometric correlation given in the table is the mean of a number of values and is compiled from the following sources:\mdash ; ( 1 ) Karl Pearson and Alice Lee , " " On the Laws of Inheritance in Man Parts I and II , ' Biometrika , ' vols .
2 and 3 .
( 2 ) Karl Pearson , ' On Iuheritance of the Mental and Moral Characters in Man ibid. , vol. 3 .
( 3 ) Amy Barrington , Alice Lee.and Karl Pearson , " " On Inheritance of Coat Colour in Greyhoun\amp ; , vol. 3 .
4 ) Amy Barrington and Karl Pearson , " " On Inheritance of Coat Colour in Cattle ibid. , vol. 4 .
( 5 ) Ethel M. Elderton , " " On the Measure of the Resemblance of First Cousins ' Eugenics Laboratory Memoirs , ' , 1907 .
The oharacters correlated for man include stature , span , forearm and eye-colour .
The sibling correlations include also hair-colour , together with those of no less than 16 characters obtained from school records .
The mean fraternal correlation based on this homogeneous maberial was .
The cousin correlations were deduced from eye-colour , hair-colour , health , ability , etc. Amoug the animals , coat-colour was the character correlated throughout .
1910 .
] in a Mendelian values of the ancestral gametic correlations in a Mendelian population obtained from theoretical considerations were in close accordance with those determined for the somatic characters in biometric researches , while the correlation on the Mendelian hypothesis appeared to be too low , if the principle of absolute dominance was maintained .
The results of the present paper for the sibling , avuncular and cousin correlations equally well with the statement made for ancestral correlations .
The general somatic correlation for cousins , , is always greater than the general somatic grandparental and avuncular correlations .
Tloe conclusion of Miss Elderton that " " cousins must be classed as equally important with uncles and aunts , and they may eventually turn out to be as important as andparents paradoxical as it appears , is supported by the theory of Mendelism as well as by biometric experience .
Without making any dogmatic assertion , it seems safe to conclude , both from theory and practice , that the cousin ranks in no way behind aunt or uncle and randparenG in family resemblance .
The importance of this deduction in medical diagnosis has already been indicated .
In conclusion I have to thank Prof. Pearson for considerable adyice while the work .
in connection with this paper has been in progress .

