EFFECTIVE FACTORS OR UNITS 



Within human groups mating is not quite at random, but it is 

 near enough to being so for the parental and fraternal correlations 

 to show the relations we should expect. It has been found by Pearson 

 and Lee, in respect of the human cubit measurement, that: — 



rp/o = <^*4i8o and 13/^ = 0-4619 



The slight departure from random mating, and the likelihood that 

 members of one family will develop in environments more alike 

 than those of unrelated people, make detailed estimates of D, H and E 

 somewhat untrustworthy. We can, however, draw some general 

 conclusions from these correlations. They are both so near to their 

 maximum value of o • 5 that the non-heritable component of varia- 

 tion must be small relative to the heritable. And since the fraternal 

 correlation exceeds the parental we have evidence of dominance of 

 the genes controlling variation in human cubit measurement. 



Effective Factors or Units 



We can now return to crosses between true-breeding lines. Where 

 these lines differ in such a way that the increasing allelomorphs 

 {A, B, C, etc.) of all the genes by which they differ are assembled 

 in one of them, and all the decreasing allelomorphs {a, b, c, etc.) 

 in the other, the mean measurement of each will depart from the 

 mid-parent by S(d). Now if we care to assume that all the genes 

 have equally large effects, i.e. d^ = d^^ ^ d^ . . . = d, S(d) = kd 

 where k is the number of genes. Now we can estimate D which, 

 when the genes have equal effects and there is no linkage, will have 

 the value kd^. Then the square of the departure of each line from 

 the mid-parent, when divided by D, will give us an estimate of the 

 number of genes, for it will be 



S^(d) _ (kd)-- _ 

 D ~ kd2 ""^ 



The shortcomings of this estimate are obvious from the assump- 

 tions which nmst be made to obtain it. In particular it will be an 

 underestimate if the increasing and the decreasing allelomorphs 

 are not concentrated in the opposite parents, or if the effects of the 

 genes are not all equal. A second method of estimating k is based 



91 



