544 



THE 



^ BALI ST [Vol. XLVHI 



term wliicli are too large. Moreover, this feature appears 

 in a more complex cross which I have carried out, in- 

 volving four pairs of linked genes (YWVM), and in each 

 separate part of all these experiments, regardless of how 

 the crosses were made. It may, then, he taken as a con- 

 stant relation. It can only mean that there is some rela- 

 tion between A and C besides that resulting from second- 

 ary reduplication. In other words, to use Bailey's terms, 

 Trow's "special" hypothesis is not valid. 



Let us then examine what Bailey calls Trow's "gen- 

 eral" hypothesis. Suppose the primary series to be of 

 the following values : 



AB = l:l, 

 BC = m:l, 

 AC = n:l. 



Trow's general formula for calculating what should be 

 the observed value of the AC series is 



Imn + n 

 AC = f+m ' 



The special formula is derived from this by assuming 

 n = 1, when the formula becomes 



Since this always gives a value which is too large, it 

 follows that n is always less than one. This means that 

 the AC primary series is reversed— that the combinations 

 present in the parents tend to be reproduced in fewer 

 numbers than the new combinations. T have worked this 

 out for the case of BCvSp (see Table IV), and find the 

 primary series there to be 0.6:1, though the observed 

 series is 1.0. The "fundamental" AC 1 series has been 

 obtained for most of the cases in Table III, and has 

 always been found to be of the usual form (/. e., n:l, 



