STATISTICAL STUDY OF HEREDITY 79 



be due entirely to environmental conditions, and the 

 coefficient of heredity of these is zero. In a mixed 

 population of sweet-peas, on the other hand, where 

 self-fertilisation is the rule and each colour type is 

 produced constantly, the coefficient of heredity would 

 be unity. One should naturally expect many inter- 

 mediate values, depending on the extent to which 

 the gross total of variability is due to environmental 

 modification and inborn differences respectively. 



It is necessary to realise clearly the relation of 

 statistical results, such as we have described, to 

 Mendelian and pure line theory. It has frequently 

 been pointed out that the results of the two opposite 

 methods are not really contradictory. The one, the 

 experimental method, applies to individual cases. It 

 distinguishes sharply between inborn characters and 

 modifications, and it enables us to predict, in any par- 

 ticular mating, the numbers of offspring that will show 

 certain definite inborn characters. But before we can 

 make such predictions we must know, not only the 

 actual characters possessed by the parent organisms, 

 but also their germinal constitutions. 



On the other hand, statistical results apply only 

 to populations in the mass. We do not attempt to 

 analyse the gross total of variation, to distinguish be- 

 tween modifications and inborn variations. Moreover, 

 we do not attempt to distinguish between similar 

 individuals of different constitution. We regard all 

 differences as variations, and determine the extent to 

 which, on the whole, in the mass, these variations are 

 inherited. The coefficient of heredity does not enable 

 us to determine what will occur in an individual case. 

 The son of a six-foot father may be anything within the 

 whole range of statures. All that we can say from our 

 statistical investigation is that his height is most likely 

 to be near five foot ten. 



Thus the two methods are different ways of treating 

 the subject of heredity, and their results apply in 

 different ways. But it would be a poor apology for 

 the statistical method to say that its results were 



