F. L. Engledow 113 



parental values. Biickhouse (2) records extensive crosses between Polish- 

 Kubanka, and the following values are from his results : 



F^.Mp = 29-0 mm. ; F,.Mk = 9-0 mm. ; F^.M= 18-19-0 mm. 

 Similarly from the records of Biffen [(1), p. 36] for reciprocal Polish x 

 Rivet crosses, the numerical intermediateness of the F^ is quite definite. 

 In the cross to be described the F^ value was thus probably 

 = (30-84 + ll-70)/2 = 21-27 mm., 

 but no further use is made of this speculative value. 



F,. 



Although 530 F^ plants were raised, only 487 of them furnished 

 grain for the raising of an ^3. The causes and distributions of the 

 failures are explained in the tables which follow. The frequency dis- 

 tributions of Fo glume-lengths were : 



Table III (p. 133). The complete F^ of 530 plants. 



Table IV. The 487 F^ plants which matured and whose grain was 

 sown to raise the F^. 



Table V. The VI F^ plants which, although they contributed no 

 grain for the ^3 sowings, did set one or more imperfect grains. They 

 were unripe at harvest. 



Table VI. The 7 F^ plants which matured but were sterile. 



Table VII. The 19 ^2 plants which were unripe at harvest, and 

 which gave no evidence of being able to set grain. 



It is with the plants of Table IV that further observation is con- 

 cerned, for they alone gave F^ progeny, from the examination of which 

 they themselves could be sorted into genetic classes. Table IV (like 

 Table III) exhibits a definite tri-modality. 



Two conclusions may be drawn from the F^ glume-length frequency 

 distributions (Table III or Table IV) viz. : 



(1) Three glume-length types appear in F2. 



(2) None of these types corresponds exactly to the Polish parent. 

 It appears that this parental type is represented in the F^ by a group 

 of plants whose mean and fluctuation ranges for glume-length have been 

 " shifted " down towards the lower end of the scale. 



In the first place an attempt was made to " dissect " the tri-modal 

 F2 distribution into its three constituent types by purely mathematical 

 methods. It was assumed that the distribution of the heterozygote 

 type of F^ did not extend below the lowest "peak" of the total F^ 



