346 Doiihleness in Stocks 



the case ?io-c?-non-cream x d-cream), XYW and xyiv are the more 

 frequent, XYw and xyW the rarer terms in the gametic series. If, 

 on the other hand, the XYxyWw heterozygote has been built up 

 from XYw and xyW (as in the case no-c?-cream x d-non-cream), 

 then XYv) and xyW gametes are chiefly formed, those of XYW 

 and xyw composition being comparatively rare. The same scheme 

 of coupling, as already shown, holds in regard to the female germs 

 when the eversporting single is se^-fertilised, but here the symmetry 

 of the gametic series is disturbed by the fact that the male germs are 

 unable to carry either of the dominant factors X or Y. Every ever- 

 sporting single is an XxYy heterozygote and is built up from the 

 combinations XY and xy. In gametogenesis XY and xy ovules are 

 chiefly formed, only comparatively few, we may conclude, are Xy and 

 xY in composition, though direct proof in this case is not as yet possible. 

 For since all doubles are sterile we cannot apply the breeding test, and 

 at present therefore we are unable to demonstrate differences of com- 

 position between the doubles derived from Xy, xY and xy ovules 

 respectively. 



We may now consider the results of the present experiment in 

 detail. If we accept the results as they stand, with the reserve 

 mentioned above, and compare the totals obtained from the 14 mixed 

 families with the results which would follow from gametogenesis on 

 the lines suggested above, we get : 



This latter result agrees very closely with that obtained experi- 

 mentally. If we take the recorded result to be an average sample of F^, 

 it appears that in F^ n may have the value 32, whereas in the eversporting 

 parent, as previously stated, n probably = 16. In several of the smaller 

 families the two rarer forms were not recorded, but, on the whole, it 

 seems probable that their absence in these cases is accidental and is to 

 be accounted for by the small size of the F.i family. For the largest 

 family in which these two forms were absent numbered only 32, and on 

 the present supposition the expectation for both forms is less than 1 in 

 64, and even with the lower value for n would not be quite as high as 

 1 in 32. It may be worth while to note that the occurrence of even 



