Xo. 641] SHORTER ARTICLES AND DISCUSSION 



559 



correctness of this conclusion will be made sufficiently obvious 

 by consideration of a simple illustrative case : Let us assume 

 that there are involved in a given pair-mating the six duplicate 

 size-factors, AABBCCDDEEFF, and tli,eir absences or corre- 

 sponding recessives, aabbccddeeff. The relation of the F, and 

 variability will be exactly the same no matter which of the four 

 following types of mating is made: {a) aabbccddeeff X 

 AABBCCDDEEFF (ditference between parents, 12 units) ; (h) 

 AAbbccddeeff X aaBBCCDDEEFF (parental difference, 8 

 units) ; (c) AABBccddecff X aabbCCDDEEFF (parental dif- 

 ference, 4 units) or {d) AABBCCddeeff X aabbccDDEEFF 

 (parental difference, zero). Neither Castle's original scheme 

 nor Wright's suggested modification^ of it can be true, at one 

 and the same time, for more than one of these types of mating 

 and they have been specifically designed only for the first-men- 

 tioned type (a) ; but as we have seen in a previous paragraph, if 

 the number of factors involved is large, the number of matings 

 of type (a) is almost infinitesimal in comparison with the num- 

 ber of matings of the other types, (b) {c) + (d) + (e) 



It is not necessary, however, to suppose that all factors which 

 affect a blending character are additive. Indeed, it is quite cer- 

 tain that they are not. and that some factors act in a negative 

 direction and others in a positive direction. Inhibiting and 

 depressing factors have been fully demonstrated, as may be ex- 

 emplified by such classic cases as the inhibitors of horn produc- 

 tion in caltlc and sliccp : llic inhibitor of indeterminate growth 



have made probable in the ease of human statures. The fact that 

 Castle himself was able to make progress in the minus direction 

 in his selection experiments suggests the probable accumulation 



