140 



THE AMERICAN NATURALIST [Vol. XLIV 



gression from Amoeba to man, it is evidently necessary 

 that it should give us characters lying beyond the ex- 

 tremes of what already exists. 



5. Our fifth proposition is that in the case of genotypes 

 that cross-breed readily, we may get an indefinite num- 

 ber of combinations of all that lies between the extremes 

 of the existing genotypes— the variety of combinations 

 realized depending on the rules of inheritance. 



Xow, if we test by these propositions the classic cases 

 of effective selection, what is the result? 



Galton's work with peas and with men yields at once 

 to the analysis, giving precisely the results which the 

 genotypic idea requires. A by-product of the analysis 

 is the practical evaporation of the laws of regression and 

 of ancestral inheritance so far as their supposed physio- 

 logical significance is concerned;- they are found to be 

 the product mainly of a lack of distinction between two 

 absolutely diverse things— between non-heritable fluctua- 

 tions on the one hand, and permanent genotypic differen- 

 tiations on the other. 



The experiments of M uller and de Vries on maize yield 

 with equal readiness. In these cases the male parents 

 are unknown; the freest sort of crossing was occurring, 

 and what selection did was pick out the progeny of ex- 

 treme male genotypes, till the result approached the limit 

 of the most extreme existing genotype under the cultural 

 conditions. 



MacCurdy and Castle's experiments in changing by 

 selection the color-patterns of rats and guinea-pigs dealt 



