STUDIES ON OAT BREEDING. 29 



From this table it is seen : 



1. That in the majority of cases the mean index is not far 

 from the theoretical value i. From the indices for all the 

 selections in each year (last column) it is seen that in the 

 majority of cases the deviation from i is not at all or very 

 little greater than the probable error. In two cases, viz. the 

 191 1 selection and 1914 rows, and the 1912 selection and 1913 

 rows the difference is very nearly twice the probable error. In 

 these cases the mean index is less than i indicating a possible 

 effect of the selection. 



2. There is considerable difference in the mean index of 

 the different years. Thus the 191 1 selection and 1912 rows 

 show a mean index considerably above i. However, the 

 probable error of the index for all the selections is practicahy 

 as large as the deviation of the index above i. On the other 

 hand the 1912 selections and 1913 rows show a mean index 

 consistently less than i.o. The difference between i.o and the 

 mean index of all the selections is ,0432, a little less than twice 

 the probable error. In none of these cases is the deviation from 

 1.0 statistically significant. 



3. On the whole the mean index of the plus selection is 

 greater than that of the minus selection. In only one case 

 (1912 selection and 1913 rows) is this relation reversed. In 

 three of the other years the difference is only nominal. How- 

 ever, in two cases (1911 selection and rows of 1913 and of 

 1914) the excess of the mean index of the plus over that ot 

 the minus selections is quite marked. If we average the six 

 mean indices of the plus selections it is found to be 1.0770, 

 while the corresponding average for the minus selections is 

 only 0.9803. It is very doubtful whether this difference has 

 any significance. 



4. There is no evidence from these indices that the selec- 

 tion is more effective upon rows grown the year after the 

 selection than upon rows grown in later years. 



Many of the irregularities in these mean indices are attributav 

 ble to the effect of one or two selections. For example, in Line 

 No. 262 a plant was selected in 191 1 whose yield was only 0.05 

 O'f a gram above the mean. It so happened that several of 

 the rows grown from this plant deviated from their mean 

 much farther than 0.05. Consequently they gave indices much 



