R. C. PUNNETT 91 



which he has hitherto been able to test his hypothesis have been pro- 

 vided by a case among Gregory's primulas, the three factors concerned 

 being M (= magenta), S(= short style), and G(= green stigma). The 

 experimental evidence points to the MS reduplication being on the 

 7 : 1 basis and the MG reduplication on the 2 : 1 basis. The secondary 

 reduplication between S and G should consequently be of the form 

 (7 X 2) 4- 1 : 7 + 2 : 7 4- 2 : (7 X 2) 4- 1, i.e. 5:3:8:5, and the experi- 

 mental data are in fairly close accordance with this expectation (cf 

 Trow (16), p. 316). The data given in this paper from the two sweet 

 pea crosses BEL x be I, and BeL x bEI afford further cases for testing 

 Trow's hypothesis. 



(1) The BEL X bel cross. 



It has been pointed out above (p. 82) that the three reduplication 

 series here are approximately of the following values : 



(a) B and L— 13 : 3 : 3 : 13, 

 (/5) B and E— 63 : 1 : 1 : 63, 

 (7) E and L— 13 : 3 : 3 : 13. 



If we consider (a) and (/3) as the two primary reduplications then 

 the value of (7) should be 



(-yi X 63) + 1 : -'i- + 63 : -'i- + 63 : (-V^- x 63) 4- 1, 



i.e. 4-1 :. 1 : 1 : 4-1, 



a result which is closely in accordance with the experimental numbers. 

 A feature of interest in connection with this case is that it demonstrates 

 that when the value of one primary series is considerably greater than 

 the other, the value of the secondary series will be very close to that of 

 the lower primary series. 



(2) The BeL X bEI cross. 



In this case the relations between B and E, and between L and E, 

 are of the nature of " repulsion." Nevertheless, as Trow has pointed 

 out, even if these be regarded as the two primary series the secondary 

 series (between B and L here) should be of the nature of "coupling." 

 The facts (p. 83) point to the reduplication between L and E being 

 near 1 : 12 : 12 : 1. That between B and E is evidently much higher 

 though how much so we have no direct means of telling. But we have 

 already seen that the "coupling" between B and E is on the 127 : 1 

 basis in families homozygous for L and on the 63 : 1 basis in the 



