STUDIES ON OAT BREEDING. 



21 



It has already been pointed out that in 19 13 the absolute yield 

 of grain was mucli lower than in any of the other years (Table 

 i). This was caused by environmental conditions which have 

 been discussed (p. 11). Apparently these conditions caused a 

 larger number of rows to fall below the mean than would ordi- 

 narily be the case. It follows that the rows which were above 

 the mean must have shown much larger average deviations than 

 the minus rows. Just why this should occur is not entirely clear. 

 It appears that some rows were not at all or very slightly affected 

 by the adverse conditions, while the majority of rows were 

 much more seriously affected. Perhaps this is due to non- 

 uniformity of the soil. If the decrease was due to soil toxins, 

 these may have been unevenly distributed through the soil. 



Attention has already been directed to the fact that the total 

 deviations in the year immediately following the selection show 

 an excess in the direction of the selection. This can be brought 

 out better by putting the figures in the form of percentages. 

 From previous discussion it is clear that we need to deal only 

 with one of the selections, either the plus or the minus. Table 6 

 shows for the plus selections the percentage of the total devia- 

 tions falling in the plus rows and in the minus rows. 



Table 6. 



Percentage of Total Deviations for the Plus Selections which 



Fall in the Plus and the Minus Rows. 



Selection made in 



Rows grown in 



+Selection. 







+ 





— 



1911 



1912 



67.87 





42.1s 



' ' 



1913 



50.93 





49.07 



' ' 



1914 



47.24 





52.76 



1912 



1913 



64.30 





45.70 



' ' 



1914 



47.65 





52.35 



1913 





62.61 





47.49 



This 'table shows that the only instances in which the deviation 

 can be significantly beyond the expected ratio of 50 per cent, 

 are the rows in the years immediately following the selection 

 (italics) . Unfortunately if we attempt to determine the probable 

 error of 'these ratios we get into trouble from the manner in 

 which the sampling was done. Whatever may be the signifi- 

 cance of the individual ratios, the fact that the excess is in the 

 same direction and fairlv large in each of the three vears makes 



