492 THE GRAMMAR OF SCIENCE 



for brethren would denote an average family of something 

 like 2 1, a rather high average for the English middle-class 

 marriages! I have actually found 5.3 for 1000 families 

 taken from the landed gentry when only sterile marriages 

 were excluded, which would denote a fraternal correlation 

 of only .3827. Hence the fraternal correlation offers 

 difficulty ; its solution, I think, lies in the consideration 

 that the parents, while equipotent on the average, are not 

 equipotent within the individual marriage ; any such 

 individual as distinct from sex prepotency raises the 

 fraternal correlation. The avuncular relationship is not 

 so far from half the fraternal as it should be. The grand- 

 parental value is above the theoretical value to be expected, 

 but the divergence is far less improbable than it would be 

 in the case of blended inheritance. Turning to the parental 

 correlation, it differs sensibly from the value to be expected 

 on the latter hypothesis. Both parental and grandparental 

 correlations exclude the notion that we can be dealing 

 with a character which in part blends and in part is 

 exclusive, for they are certainly not, as they then would 

 be, reduced belovj the value required by the exclusive 

 theory. Is there any hypothesis which suits then the 

 facts, namely (i.), that grandparental heritage is increased, 

 and (ii.) that parental heritage is slightly reduced below 

 the theoretical value ? Is not this what we might expect 

 would occur if there were a reversion to the grandparental 

 character in a certain percentage of cases ? Would not 

 reversion strengthen the grandparental heritage, while it 

 weakened the parental ? Will the hypothesis of reversion 

 account for the divergences in the table above from the 

 ideal values given in the previous table for exclusive 

 inheritance ? Unfortunately no hypothesis of reversion 

 as yet propounded appears to account for the observed 

 values. For, while weakening the correlation of the 

 offspring with the parent, they also weaken that of the 

 grandparent. The fact is that some parents now revert 

 to the great-grandparents, and in so doing cost more 

 resemblance to^ the grandparent than the latter gains by 

 direct reversion. 



