]\gt ; Prof. Ixarl Pearson .
rr Mathematical Contributions to the Theory volution .
the Law of Ancestral Heredity By KARL PEARSO M.A. , F.R.S. , University College , London .
12 , \mdash ; Read Janualy 27 , 1898 .
's Greeting Francis Galton , January ( 1 ) Tntroductory.\mdash ; In Mr. Galton 's ' Natural Inheritance ' we find theoly of regression based upon " " mid-parent.\ldquo ; This formed point of my own theory of biparental inheritance .
* time .
Galton published his eory I venture to think that .
had not clearly in view some of the laws of multiple COl'relation wi which we are now more familiar .
This cel.bainly was my own co .
dition when my memoir on heredity in 1895 , and in that memoir I pretty fully developed the theory of multip corl eation as applied to heredity , it had not then become such familiar tool as two years ' pretty constant occupation with it has made it .
Accordingly misinterpreted a second propounded by Mr. Galton , and reached the paradoxical lclusior ' ' a knowledge of the ancestry beyond the parents in no aIters our judgment as to the size of organ 01 degres of characterist probable in offspring I assumed Mr. Galton to the coefficients of correlation ween offspring and parent , parent , great-grandparent , \amp ; c. , were to be taken \amp ; c conclusions I drew from this result were , had the result been true perfectly sound .
The recent publication of Mr. Galton 's paper Basset hounds has led me back to the subject , because that pap contains facts in obvious contradiction with the principle cited from my memoir of .
At first , I must confess , I inclined to lay less stress on .
Galton 's general law than , and attributed our divergence to the admitted of coloul ' data .
After some Copdence with } .
Galton and a elldeavour on my part to represent his views in my own language , have come to the conclusion that I shall in future term of Ancestral , if properly interpreted , reconciles tb discrepancy in ' Natul.aI Inheritance ' and between it and my memo ) of 1895 .
It indeed enables us to predict a the values of all tb correlation ( joefficients of heredity , and forms , I venture to thin } fundamental principle of eredity from which all the numerlCg data of inheritance can in future be deduced , at ally rate , to pproximation .
' Phil. rans , , vol. 187 , pp. Ibid. , p. 306 .
that this is the meaning to be racted from pp. 132-5 ' Natural Inlt ' JIathematical Contributions to of olntion .
The I put iu the truth of the law is not Sllred by Ir .
Galton 's researches on ture or on in Basset owever strong evidence these may provide , but l.ather on fact the gives a the ] ation betweell ents and ffspring , and that correlntion is practically identical alue I hnve determined from these and otbcr tions .
With to how " " rent \ldquo ; shall be defined , 1 rould state the law of ancestral heredity as follows:\mdash ; If be the deviation of sth from the meall of Ilcestritl generation , and bc the probable devitttion from he men ] ] of .
of individual , the standard deviation the mid-parental tion , of the generation of the off , then This is the somewhat genernlised form of the law , aIton sums up as " " } parent on an oneuarter , one-sixteenth , or so In , the occupier of each ancestral in the Iegree , ) the of , contributes ' of the lcritage .
' .
The form dlows for secular n1odification of he iliti e of the successive generations .
the cocfficients of elation betwecll -grandparent , \amp ; c. , respecivcly .
remains constant during the tions , old bc relation oetween tho parent of and of the ation i be in ) lin , of ascent , ono of tbesc is the parent of unsb be cred t if be bion b his offspring , may theorelically variety of according to the the sexes the of descent .
If all these 's be ullequal , theli be to repressed .nt their mean value .
It ill be uecessnl our find the between tbc andLth of these oive the correlation between luals .
bc the tion between the midthu deviation of any organ of the llltlc mean of thnt for the encration of , ( of his ale ma ; let be alJyall Ilot is , a Ctlt , lIird ) otntion I):dopted .
The nid ) rent o derived froltl t ] sth ' vol. 61 , 102 .
Prof. Pearson .
determined .
Then there will be male and female parents , and the " " mid-parent\ldquo ; will be defined to an individu having a deviation from the mean of the parcntal generation This is a solncwhat more general definition than Mr. Galton 's .
Since , when the summation extends over a individuaIs of the ation w parell , it follows that mean deviation of all possible rents is zero .
* Next let us find the standard deviation of the } If their number the ation be Now , if there be assortative mating ) will equal where is the correlation coefficient for assortative mating in tl geueration and the standard deviations of male and fema ] mates in that geuel.ation .
rther , ) and not be .
zero , because assol.tative mating would mean a claf mating into a like class leading to a correlation between relatior in ; but these sums would be of the order , and since , at an for mall , appears to be very small , we may neglect them to first oximation .
Hencc or . . . . . . . . . . . . . . . . . . .
Now lct us take , which is at our choice , equal to , thetl In the next let us find the coefficient of tion betwee ] two mid-parents , say those of the and generations .
have . . . . .
cve selcti hich would Tveight particular parentq , is ectec or , if ust { sured from the weighfed } ) IIere , as \ldquo ; I the effects of bree require tment .
I hope to publish fullel , based it ' ) stcll of mcitsurelcllts thau arc in Math ) ntions to the of ) Now , if be , ally will only ( of order ) )rrelated with its own icular ents , malc and femitJc ) , there will be such male pttrents such female ents .
Hence ( iii ) .
Herc is of all which and end the cent , ) , of female and e , of th osc which with a male nnd with a female , nd of those which and end with cfore , put ) .
We thus ility of ation to bo constant , ' , . . . . . . . . . . . . . . . .
( iv ) , .
for tho mcall of the icicnts o its individual parents .
It as on the of the eneselection to rem , if it ists , sce that depends only ou the differncc of tions , altd } ) lvritten if clccve breeding , , to small .
H WCCn t nlid-parcnts , cncr.tions { , is of and menn of coefficients and for pobsiblo coml ) innti o Wh btlection , by lliss A and nlyself thab tho possible rfttncc of resnlt is it cicnts betw diffcrent to , , ( oclT i only .
Prof. Pearson .
for first parent and offspring are nearly equal ; * the equality of all possible 's for the pareut and offspring , an neglecting we have . . . . . . . . . . . . . . . . . .
or , we conclude that very tely : The standard det.iation the ?
parent may be obtained from the standard of in vidual parents by dividing by , and the ( jorrelation betu , eenmi sth and mid parents may be obtained by multiplyin between an individual and any parent by ) .
Thus the variability of the mid-parent rapidly decreases as increase , i.e. , we get back in ancestry the mid-parent comes mor and mors ] to represent in all cases the mean of the geners populatio ] .
Whether the correlation tends to decrease or increase depend on the relative rates of change of and Since must always be less than 1 , we obtain at once the interesI ing limit that the correlation of an individual and a parent lways less than For example the correlation between : Offspring and parent musb be less than and ndparents , , , , and dparents , , , , and -grandparents , , , , Their acbual values as deduced from .
Galton 's ] are much smaller , as we shall see later .
( 3 ) The reader will remark in order to get these results in ' simple form we have multiplied female deviations from the meal by a constant factor , which has afterwards been taken equal the ratio of male to female variability .
The reasou for this was two fold .
In the first place is cerbainly not equal to , and , conse quently , would not given , but a complex form .
In the next place we note the clos1 equality of when we neglect hence is the only value which appreciably reduce formula ( iii ) as well as formula ( i ) .
I therefore define a mid-paren to be one in which the deviations of the females are reduced to malc standard by first muItiplying them by the ratio of male to * ' Boy .
Soc. Proc vol. 60 , Jfath ematicnl Contribntions to .
391 male variability .
This does not theoretically agree } Mr. -alton finition , for he reduces the to the male starldard by multi- ying them by the sexunl ratio , or the ratio of the male to the male mean for ths under consideration .
In ordel , therefore , at my factor of reduction should agrce with Mr. , it is ledful thttt the ratio of the standard deviations should be equal to ratio of the , or , in other words , that the coefficient of riation should be the same for the two sexes .
Now for the men women , I find for 1000 cases of each sex the coefficients 07 rest ) , or coefficient of variation is sensibly lor ) .
Galton found from his anthropometric returns for somewl , at fewer numbers , and probably for cIass , valucs of and , again sensibly equal .
co the ] , whether defined in my manner or in , would have a sensibly equal value in the case of stature , hich is one Mr. Galton dealt with in his 'Natural Inhel.itance .
' coofficient of variation is , however , not the same for sexes ( 1 the of all organs , hcnce for the purpose of simplifying the ormulae , I th i to think my nlodification Galtoll'S iginal dehnition will p of service .
( 4 ) I shall ) roce , to determinc by the law of ut.edity the rrelatiou b all and any parent rom a of the cssion the and his 1 ] By ion if 11 .
be dard devi tions , and corTelations , then the frequenc.v surface is given by -1R the minor of the consbituent of th row and th * Scc ( nccs of vol. 1 , " " and Womtltl , \ldquo ; p. , p. 311 . .
Gslton 's faluily reorcl d ve 1032 and 1 for tho ntio of cuctIicicnt of iation of solls to hters and of fn to nlothcrs .
Trans , vol. 187 , p. 271 .
in paper on " " in Milt ind Prof. Karl Peal.son .
Putting , . . . .
constants , say , . . . .
, we have for th mean value of the corresponding array of The standard deviation of for this is ... . . . . . . . . . . .
( vii ) These resuIts ( vi ) and ( vii ) the regression formulae .
* Now leb bs the mid-parental of the 1st , 2nd .
3rd , . . . .
order , and the mean value of the in offsprlng .
Then the value of is given by 1 , . . . .
( viii ) , 1 , , / 2 , , 1 , . .
, . . . . . . . . . . . .
1 and the regression formula is : , if we stop at the mid-parent .
Comparing this result with the analytical statement .
Galton 's law of ancesbral heredity given on p. 388 , we see that we must have from ( v ) : .
( ) . . . . . . . . . . . . . . . . . . . . . . . . . .
There will be equations , if we go to the } ) rent , and there are quantities to find .
Thus Mr. statement that the regression coefficients are * See ' .
Tran to the Th volufio .
393 us sufficient equations to find the coefficients of total correl tween the offspring and the successive mid-parents .
Equations ( v ) 11 then enable us to find the coefficients of corl'elation between and any individual ancestor .
But these in turn will ffice to determine all heritance whether direct or ( see low ) .
In short if .
Galton 's law can be firmly established , it is complete ) , at any rate to a.first , of ihcn.hole oblem of .
It throws back the question of inheritance upon constants , which can be once and for all determined ; lies .
I must confess that clement of mplicity first my chief difficulty in ccepting the law as laid the on Basset hounds , and I even yet a certain !
sitation , } to an apparent ence in collateral heredity in fferent ) cial classcs , and ] to the apparent numelical value inh itance of fertility ( 5 ) I obtain a solution of equations ( ix ) .
inors of tlentS of the first row , 11 mely , ( ,2 , and multip ] ying them in succession by the the , 3rd . . . .
row of , we have by the ordinary theory ' nts the system : 1)ivide of by and let us use instead of ( ix ) le more general system which will allow us to consider two ] , and rather more generally Mr. alton dove " " to tax the bequcsts of each ation , \ldquo ; as he : . . .
( ) , and constants ; we then find : Prof. ] Pears . .
Mnltiply the equation by , and subtract from the th . . . . . . .
, . .
Assume , hence : But ce and are both less than uIJity , the term will vanishingly small when is indefinitely large , thus : . . . . . . . . . . . . . .
Substituting in the first of the equations for the ) ' abovt we have : Or , laking as before for very large : ( xiii ) ( xii ) and ( xiii ) contain a complete solution of the equations for the 's given above , so long as we go only to a finit number of mid-parents , i.e. , may be very large , but not ( 6 ) Specied Cases .
4 1 , .
It follows that , and .
Hence 1 jmatical Contribntions to the jory of volution .
the mid-parental correlations would be perfect , and , there , any 0lle } ) arent wouId suffice to fulIy determine other the ) .
individual pal'ental correlations would then be ' ' . . . . .
pareut , ndparent , grent-grandparent , \amp ; c. , } offspring .
* , suppose any of -hich lead , th lencc we find , that is , ; or , all mi elations afect .
Thus , as in case ( i ) , the rental corrclations could be ) resented by , ? . . . . .
are hc valtlGs I took in my memoir of I lues ecaur , they seomed to expl.ess Mr. Galton 's method of Ssing from inilivitlutl parelltal to grand-p total I had not ) thnt there was any antinomy even alton 's theory of ession his of ancestral lity .
JFad ( lon ) so I sIlould c that dnte , have givell preferencc to the , and rejected his law of partial coeffients of in ' of , values .
based numelical ion .
of cocfficients .
Put 1 ; this is .
Galton 's form of the ' at on ] .
we the successive mid.parelltal correlatiollS / ) , , \amp ; c. , ltd for individual tions , , \amp ; c. , \amp ; .
is , I , must follo from of " " of he \ldquo ; and of its exact on cbion .
'Phil .
, vol. , pp. 303-5 .
{ Inheritance , ' .
} .
Mr. Galton puls for for rent , so on .
: 39 .
Karl Pearson .
, , \amp ; .
Further , for the regressions on the mid-parents ( not partial total ) , or , \amp ; c. , we have , on the assumplion that generations ars equally variable , , \amp ; .
Or we may express the law of heredity in .
Galton in the following simple statement:\mdash ; The total regression of rogeny on the tnid-parent of any generation is constant and equal to Let us see how these results with observa } ions .
Mr. tells us that his first estimate of mid-parental regression .
This estimate ctly a with theory .
He afte vards changed the value to , which is less in NIy own calculations , .
on Mr. Galton 's data , give , or .
The probable .
is , however , .
If we do not weigl fel.tility the parental correlationS , a which distinctly too high for Galton 's law .
It must be remembered , hov ever , our deductions from that law based on equalit riation in euch generation , and that this equality is by no the fact .
I hope shortly to get final values for parental heredil from my family ments , which have now reached a total nearly 1,100 families , and thus settle how far GaIton 's law eeds be modified .
On the whole the confirmation obtained from stature dal for the law of ancestral heredity is very striking I am incline 10 think even more convincing than that obtainable from the hounds , and this for a reason to be considered later .
It suffic ' here to observe that we do not need to know the characters of parent -parcnts , great grand-parents to test Mr. single relationship , near or far , direct or collateral ( see below ) , bring its quota of evidence for or against the law .
It will be , seen that table ( p. 397 ) differs in inciple frol Mr. 's on p. 133 of his 'Natural Inheriffince .
' Tn icula supposing equal variability generations , the individual .
rant parcntal is not the of the parental regresslon , the half of it . .
Galton 's law of ancestral hercdity contr * ' Natural Inheritnnce , ' p. 97 .
Ibid. , p. 97 .
'Phil .
Tl.ans .
, ' , vol. 187 , p. 270 .
Proc. of atical ( ntions to the Th of ) .
3 of Hel.editaccol'dino to Galton 's Law .
of fo bc taken can of ions duc to of , if nut , ( llld liks pnrcnt .
Th t ) ( ividttal ) is ) ' the of ring ( same .
In ling with the ]llid-prc ] ] , female to by by ualc it is latter which , filom botb eory and , ] now hold must be ( 7 ) Mr. us the ession coefficients rhcn d the lllid-I){llc1l{s known .
It is desirable to deduce nlultiplo ttbiou the values of the partial r ' efficieni s , :3 , 4 , .
mid-parents only .
YVhen rents aro tnken let the coefficients be , . . . .
we havo for the mean of the : . . . . . .
, I th devintion of the offspring and of epth lion .
this with the regressio ] ( viii ) we have , ( ) .
* I do olumll of Mr. 's table giving the urringlc Crelation tbiliby ( deviation ) of an is the corrolation not the regl.ession .
ility of th grnerations , in case of the idual ill.ent Lo ] ) ced ) to the of ausc triability of the by ( 1 ) is rhat of Prof. Karl Pearson .
Now make use of the general equations for the 's given just equation ( x ) , substituting for the 's in terms of the , and reme .
be that we are to stop at , and that We have after some reductions the system : . . . .
Subbracting the of these equations from the from the , \amp ; c. , and introducing the values and of , we find and so ou , new coefficient being now half the last .
These equ lions give successively the ratios of , \amp ; c. , to the last of the previous set of equations will then glve Thus the partial regression coefficients any limited number can be found .
This last equation also gives us a convenient formula for measuring bow nearly the fsprir of mid-parents , all selected with a peculiar character , has attained that character .
For in this case and ontril)to th of .
30 [ enco the moro nearly unity the more ] the offsprin as the full cter of its cted P. I vellture to of tl , stock .
lt is lueasul.e of the stock reedi truP .
LnstIy , to find the of the only 10 terms of the miuol.ii of its first .
In the ) hcu has been calculated from these of rce Stock to Galton 's * To it bc ded to , I : 400 Prof. Karl Pearson .
( 8 ) I venture to think this table of considerable S[ggestivene , and will now point out some of tlJe conclusions that may be from it .
With a view of reducing the absofute variability of a is idle to select beyond the grandpareuts , and hardly select beyond parents .
The ratio of the variability of pedigree to the geIleral population decreases 10 per cent. on the selection parents , and only 11 per cent. on the additional selection of grar .
parents .
Beyond this no sensible change is made .
We cannot leduce variability beyond 11 per cent. by the creation of a pedigl stock , i.e. , by breeding from selected parents for 2 , 3 , 4 , . . . .
gene ] tions .
In some cases of course we appear to decrease variability for example , the average size of an organ\mdash ; for absolute variability is then a smaller proportion of the actual size , the relative variability , or coefficient of variation , may thus be decreased .
If Mr. Galton 's law be true , then pedigree stock wou retain only a slightly diminished capacity for variation about new type .
] example , the absolute variabiIity of men of height , inches , being inches , the absolute variability of of 72 inches , obtained by selecting any number of 6-foot aucesto ] would rdly fall short of inches .
* ( ii ) Two different classes of pedigree stock exist .
In the one with the general population , and select special characters , 2 , 3 , . . . .
generations .
In the there we know the pedigree , 2 , 3 , . . . .
generations , bnt have no reason for supposing th before these generations the stock was utely identical with tl general population .
In the former case we put for the mid-parents Hence the regression formula is .
The values of are tabulated in the last column of the above in brackets .
They give the ratio of character in oifspring character in , if ancestors of equal fnll character have selected for generations .
We see that in six generations spring will have been raised to within per cent. of the tral character .
In the latter case we must use the partial regress on coc tion , bock i as a l.ulc .
Contribnfions to th of .
401 .
of the table .
For example , in the case of Mr. Galton 's sset hounds , , and were the coefficients to be , rather than , and , when he proceeded to apply law to three Fations .
These give the proper allowance for ancestry beyond the pcdigree .
Thus the to been given about a fifth more .
If we proceed six in stock of the latter type then the spring will be cent. of the selected , i. e. , eir stability as iven by the last column ( iii ) Now let us apply these results to the all-important ' panmixia and encration .
Suppose a selection made of par- :ular character generations , starting from the general popula .
Then the in the eration will have of the .
Now , selection , Jet us breed th a first neration of mid-pnrents with of the character .
ffspring will have of the will h : ( ) th .
and sn on .
The law is obvious ; the of the character n the selection eased .
stock ond the rlth gelleration , reach the same conclusion .
fions the offspring of pttronl stop belection the ncration ill : fth , and on .
The ( 1 Prof. Pearson .
Thus , on the basis of the law of ancestral heredity the case panmixia is even stronger than it appeared in my memoir heredity.* Assuming .
Galton 's law of regression : I there that panmixia was possible with a stable focus of regression , but the suppol.ters of the consistent of panmixia must place focns of rsgression , in order that degeneration should be continuou in a position inconsistent wibh observed facts ( p. 314 ) .
We now st that with the law of ancestral hel'edity even this is not possible , race with six generations of selection will breed per cen of truth ever unless the focus of regression instead being eteady actually regredes .
Of course there are many ways : this law may be modified .
example , fertility may be maximum with the average .
say , of the unselected original populatio and fter a selection it may remain correlated , having the values of the selected character more fertile than others .
The of course , stock would degenerate with panmixia .
This woul however , be reproductive selection , not panmixia in the ordina ] iigniiiCanCe , reversing natural seleotion .
We are far too ignorant } present of the correlation of fertility with other cters to any sweeping principle like that of degeneration by panmlxla it .
Our attitude present can only be that there are no facts , are there is no workable theory of heredity yet discovered favours in any way degeneration by panmixia .
( 9 ) Taxation of Inheritance.\mdash ; If , assume Mr. Galton 's law ancestral heredity to be a niting statement , we can at once fro our general formulae ascel.tain the influence of " " taxlng .
the ance\ldquo ; in any other than .
Galton 's form .
He has , the ( where by " " inheritance\ldquo ; I de from the mean of general population , not actual of t1 chal.acter ) , .
cent. in each transmission .
There may , be two types of taxation , a general taxation on the receipts and a special tax on each transmls51on--correspon ding , speak , to a duty paid by an individual on com ng receipt of entire ancestral property , and a stamp duty on eac convey ance of ancestor 's contribution .
The is represen ted by tl of onr equation ( x ) , and the second by the .
ted Mr. Gaiton , his memoir on Basset hounds , has ( jonditions of the law of ancestral heredity , and concludes that his tions are only fulfilled by the series * ' Phil. , vol. , p. 308 et seq. is I should ) cscnt account for the degeneration of Pedigr me ence of kind is possibly scnsible in ] ighly , civilised .
Sce ( Reproductive Seleotiou in my ' Chances of JIathematical Contributicns to Th eor ?
y of .
403 It seems to me th they are equally well fulfilled by the ovided the sum of this series is equal to nity , , that is / / , of our previous notation .
Hellcc lilious 1 down are fulfilled by our soluovided .
( xvii ) .
lo not that such a is more ] indeed himplc .
But it back the tlJeoly of inheritance at least one ftry constant , and tbcrcfore while c of heredity ) , allows a greater for riety of in species .
It seel ) worth while to the that result in ancestral tion 1 on " " tax " " As a numerical illns , taL-e at , then .
We ] , hlld by ( xii ) and ( xiii ) : cftn a table exactly like that on p. 397 .
ex ) it , the of a " " \ldquo ; is Il the } tions .
In palticular the more disant they would do under of .
Tax 10 per cent. lIid .
Prof. Karl Pearson .
Galton 's unmodified law .
Now the direct correlations as given law certainly appear somewhat small for both stature and cephal index in man .
Hence it is quite possible that when more extensi data are forbhcoming , it will be found necessary to modify Mr. Galton form and take less than unity .
The above table will suffice bo indica the eneral direction of the changes which resuIt when is aried .
One point should be noticed , the total regression on individual mid-parent ( note , not the parhal regression ) continual eases as we go further back , and will ultimately be greater nnity ; in our case this will happen at the generation .
Now in a generation an individual has 102410th great-grandparents , an they independent , mean of these could hardly differ wide .
the population mean .
Hence the total regression coefficie ] greatel than unity is not so significant as it might at fir sight seem .
What it amounts to is this : that if we only knew of a individual that his rent in a very distant generation had } .
a character than the then population mean , and knew nothir tbout his other mid-parents , then the individual would probably more of that character than the mid-parent .
The apparent paradc rises from the very small variability of a distant mid-parent , are the extreme improbablility of a mid-parent differing from the lation mean .
Of course with close in-andeeding the modification introduced by assol.tative mating cou not be neglected , and our whole investigation would need modific tion .
* Until , however , we have more measurements to deal with , is idle to develop at length all the consequences which flow from tl eneralised form of Galton 's law .
( 10 ) Collateral Heredity.\mdash ; There is another point on which tl of ancestral heredity gives us fnll information , namely , the co elation between brothers , cousins , and all other collateral fn my memoir of 1895 , I felt bound to reject Mr. Galton 's fiSlt coefficient for bl.others , because its value seemed to me in contradi tion with expeI .
I wrote ( p. 285):\mdash ; ' ' There is not , I sufficient g.round at present for definite sion as to the manner in which lineal is relatet tl It does not seem to me tha 'oefficient for the former should be half that for the ]atter , posed by Mr. Galton And again : \mdash ; esto c of sixth parcnts , allil instead of to the Theory ' of olution .
I .
the tion coefficients for collateral hered\mdash ; at any rate in the middJe be reater t\ldquo ; of ancestral hercdity rejected , accepting part of what I uo consider is rolleollS teory of regression , gives just the etwecll lineal ud which 1 was bhen seekino !
Let .
the tion of two brothers part of or its me ) eing measured from the g populntiou diffcr from the meau of the array by and .
' of others in a arlay , thelt incc corresl ) to of ch deviations he .
Hence ' the , is the numbelpaiI s of th serics of ents , \amp ; c. vri te tlu ) there bc t'n .
1-1 if ( hc tion o \fnof ; g to twcen J I ' li : to ' ' ' 7 of ' hle bc Prof. Karl Pcarson .
and note that is the standard deviation of offspring , is the of pairs of brothers or arents of each order , and as before the rd deviation of the group of qth mid-paren Noicing that havo if be the correlation brothers the sum now referring to all values of and from 1 to unequal to , and taken positive .
Thus : . .
Hcnce summing parallel to the diagonal : .
( xviii by , ( xiii ) , and ( xvii ) .
Let us evaluate this on Mr. Galton 's law and on the hypothesis per cent. tax .
On the first hypothesis , an , hence .
On the second hypothesis ( p. 403 ) , and ; hence We can also o less accurate values of ternal correlation ( bber ways .
Suppose two brothers be considered as sons of or lllid-parent only .
In this case we must takc for the ( see the table , p. 403 ) , or as beforc : , * [ The ave valuc for ternal collelation shows musb bc must bc , only ] Math emafical Contributions to the .
407 If we supposc two parents no assortative have , here and the nale and ] ntions .
With , and .
Assuming the by Mr. Galton in is Inheritance ' p. 133 ) for ression , the fraternal ression deduced rom bo been , and not as obtained Mr. The mcan of the sister-sister , -brother , tions that I fouud 1895 , duly hted for he of in ench cnse , is ctly 0 .
The value as it Jighb h been a } from Galton 's law , with risc to , if we tax\ldquo ; up per cent. I therefore that this law of eredity is at a first tion ireement as complcte as could possibly expected -ith wo as yct kllow as to collateral heredity .
tho took- fraternal heredity cannot be nken ( thtilt value ( about ) obtained data\ldquo ; must bc explained by my : .
, : , cious selcctiou of approximately equal hcights olunteer rncnts ; for the is 's law without a modified to itance of cousins , we noticc that bc rclJrescnted by . . . .
ildrcll of the s parcnts and tlon , and heir other pnrents , and ] a loul , no bion with each , or let .
I or is the 7 , cfore of ' ; the mid-parental tenlS of ' , , if there be no and-in with h for the rray cesponding to , ltol took of , I i ) I in the il .
, 21,1 .
il .
, p. 28 S . . .
[ .
408 Pearson .
where is the number of pairs of cousins corresponding to The factor 2 ( see footnote , p. 404 ) does not occur here , as the cousi : form parts of separate , and not identical , arrays .
Now Iet us sum all possible mid-parental systems , then if be the correIation cousillS and the total number of cousin pairs : But product moment for pairs of others , by what precedes .
is exactly the same as the sum of all offsprin with the mid-parental system of ancestry beyond , since , is not to be equal to for all values of , , Thus , and .
Galton 's valne is ( ' Natural Tnheritance , ' p. Had we , however , applied his method correctly , considering cousin as the offspring of brotbers , and adopted the value given by hi law ancestral heredity for parent and offspring , should hav found , instead of our present .
Considering cousins having two grandparents the same , we shouId have found Cousins.\mdash ; The correlatod parts of heir mid-parental system are where and are cotlsins , and hren , and is the mid-parental system of and In order to work out the relation , we shall vant that of and , or of nephew and uncle .
Here give related parts of the mid-parental systems .
Mathematical Contributions to the of volution .
Hencs if be the uncle-nephew correlation and the lotal mbel .
of nce Thus , or is double correlation of first cotlsins .
, as , the variations of all genelations , in case of unclcs , have been treated as equal .
the correlation of second cousins we have valua n terlns have\mdash ; ; monent of uncles nephcws of offspring and the midsystem of their for all of Iy , .
linlly : } 1 ) fo 1111 liko anncr , may needed for the of : 410 Prof. Karl Pearson .
from single pairs , are given in the table below .
This case , offers some striking applications of GaIton 's law , I postpone for present .
Collatcral Heredity , to Galton 's Law .
... Had we arded secoud cousins as grandchildren of brethren , , should have found instead of for example , showi : the degree of approximation of the incomplet beory .
( 12 ) On Cross Heredity.\mdash ; In my memoir on heredity cf have cross heredity as the correlation between different orga in any two relations .
* If we consider Galton 's law of ancest ] heredity to be applicable to the inheritance of any character quality whatsoever , then can obtain from it a solution of the oblem of cross heredity .
This solution seems so simple plausible that it deserves careful consideration , and I hope short to be to test it by the measurements in my possession .
Let A and be any two relatives ; 1 and 3 represent any ol.gans in , 2 and 4 the same organs in B. Now suppose we investigate the manner in which the index 1 tc is inhcrited by , i.e. , leb us find correlation between the indic 1 to 3 and 2 to 4 .
Let be the coetficient of heredity between of blood A and , and suppose it by Galton 's to ta the sanlc for all qualities and characters , then will be correlation not only between 1 and 2 , and 3 and 4 , but also betwe indices 1 to 3 and 2 to 4 .
The value of this was by me in 'Roy .
Soc. Proc vol. 60 , p. 493 , Equation , and whel .
the coeificients of variation of the four orgar and the their coefficients of correlation .
Now if there be no secular sclection , and , vol. 187 , p. ) to the ThjortJ of ) .
411 , by Galton 's law , , for both are coefficients of cect hedity .
Henco bero is the anic correlation between the two ans in tLlc me individual .
Thus it follows at once that the of ho coefficients of heredity is tho oduct of ecoetfficieltt of direct heredity iuto the correlatiou of the two ans the same .
Now in all cases of relationand bl'other , cousin and cousin , , and is probable that is also true where the ionShip not eable , rent and .
Thus we reach for cross edity .
iJIultiply direcf of , and have , the tion betwecll mertlS rench malcs .
Honcc we should expect to the of parent and humel.us of .
Thus Galton 's law , even if not ) ill serve as a useful standard to .
of hel.cdity .
( tious of Galton 's will ce to extent of its plications .
If eithcr that its suggested modilication , be ntially Crect , they of horcdity .
Tlley bring one simple an of facts , thus I of .
Tt is true thnt thero diificulties to be met , wJlich I wouid llote two in parular : ( i ) the amotltlt of inheritance an onstan ) ves .
It would thus not to be , or one capable of odification by soenls to me ) .
I of or , and we old expect to principlc of ition .
This difiiccll arm of one bo confir- of of of ff 412 ematicat Contributions to the Theor'J of 'volution .
cuIty would be to some extent met by introducing the coefficient which I would pose to call the coefficient of heredity , and cider as of being modified with regaled to bobh character a .
As such a law would cover Mr. Galton 's case , there does seem any objection to using the more general formuJa , until it that the strength of heredity is the same for all characters a : aces .
Of course it may well be argued that heredity is somethi ] prio to evolution , itself evolution , and not determine .
by it .
If this be so , its absolute fixity for all ol.gans and races to be capable of observational proof .
( ii ) For the inheritance of ] in man from parent to offsprin Miss Alice Lee has recently worked out 6,000 male , and 4,000 fema cases .
The result shows thab fertility is probably a heritable ch racter , but the correlation between parent and offspring is scarce one-tenth of that by Galton 's law .
The difficulties of exact determination amount of fertility inherited in under the present artiiicial conditions are very great , but evell allov ing for these , I think we must assert that fertility is inherited man , but in a gree very much less than Galton 's law wou l.equire .
I hold , then , that , as far as our knowledge goes ab present , must be cautious about treating as exactly equal to unity .
That alimiting value which certainly gives strikingly good results for a gre deal of what is yet known , but we must wait at present for determinations of hereditary influence , before the actual degree approximation between law and nature can be appreciated .
wibh regard to such minations , there must be no haste to ass that they actually do contradict Galton 's law .
That law states tl value of certain partial coefficients , the total regressl( coefficients that we have deduced from them are only correct on certa limiting hypotheses , the most important of which the absence eproductive selection , i. e. , the negligible correlation of fertility the inherited character , and the absence of sexual selection .
I pose to deal wibh the resuIts of Galton 's law , when mabing aken into , especially in the case of -an breeding , in another paper .
At present I would stats that , with all due reservations , it seems to me thst the of ancestral heredity is likely to one of the mosb brilliant 1Ir .
Galton 's discoveries ; it is highly probable that it is tho slmp descriptive statement which blings into a single all comp le lines of hereditary influence .
If evoIution be turi selcction combined with , then the statemenb aces the whole field of hel'edity must almost as ep0C llaking to the biologist as law of to the astronomer

