STUDIES ON OAT BREEDING. 39 



9. The mean indices showiing the effect of two or three 

 successive selections in the same or in different directions are 

 shown in Table lo and ii. These indices are also significantly 

 equal to i with the possible exception of the successive minus 

 selections. These later indices are consistently less than i. 

 Whether they really indicate an eft'ect of the selections or not, 

 cannot be determined from the present data. The values are 

 not very far below i and in view of the other evidence we are 

 inclined to regard them as random fluctuations. 



10. It is pointed out that where selections are made which 

 are only slightly above or below the mean of their pure line, 

 spurious values of this Index are sometimes obtained. This is 

 especially true if a relatively small number of rov/s are grown 

 from such a selection. The reason for this is that the means 

 of some rows mav deviate much farther from the mean of the 

 line than did the mother plant. This will produce indices very 

 large either positive or negative. A priori there is no reason 

 for excluding such small selections. They ought, on the selec- 

 tion theory, to result in rows which would deviate less from 

 the mean of the line than rows grown from large selections. 

 Only one such "small" selection is included in the present data. 

 The eft'ect of omitting the line (No. 262) to which it belongs 

 is shown in tables 9 and 10. 



11. The analysis of the selections for height of plant shows 

 essentially the same results as found for yield of grain. Analy- 

 sis by the method of deviations shows that in only one instance 

 is the excess in the direction of the selection large enough to 

 have any possible significance. For reasons discussed in the 

 text it is probable that this one large deviation is not signifi- 

 cant in connection with the selection. 



12. Indices of selection for height of plant have been 

 calculated for several of the selections. Since they show 

 nothing essentially diff'erent from those for yield they have 

 not been included in the present paper. 



