1048 
Journal of Agricultural Research 
Vol. V, No. aa 
where the corrected average yield is markedly different from the ob¬ 
served. An instance of this is seen in the Early Pearl variety (Table I). 
The observed average yield of this variety (86.6 bushels) was the highest 
obtained in 1915. The difference between the yield of this and the 
Minnesota No. 26 was nearly 5 bushels. The corrected average yield 
of these two varieties is practically the same, differing only in a fraction 
of a bushel. By referring back to figure 4 it is found that the high 
average yield of the Early Pearl was largely due to the influence of two 
plots, Nos. 945 and 970. These two plots happened to lie in exception¬ 
ally good soil. Their observed yields of 96.5 and 90.9 bushels per acre 
were reduced to the corrected yields of 87 and 75.4 bushels, respectively. 
As is to be expected, the corrected average yields show in nearly all 
cases a much lower variability. This is true of both the absolute and 
relative variability. In one or two instances, as the Imported Scotch 
(Table I), the variability is greater in the case of the corrected yield. 
If all the varieties (Table I) are taken, the corrected yields will show 
an average decrease in the coefficient of variation of about 1% per cent. 
The table shows that with systematically repeated plots the yields 
corrected by this method do not differ radically from the actually ob¬ 
served yields. Such changes in the order of yield as do occur we believe 
more truly express the relative value of these varieties. This statement 
is based on the experience of several years with these same varieties. 
In using this method attention should be called to one or two points. 
In the first place where a field of plots is very large or where it is rela¬ 
tively long and narrow better results will usually be obtained by break¬ 
ing it up into smaller blocks for calculation. For example, our 1914 
test field was 6 plots wide and 28 plots long. More satisfactory results 
were obtained by breaking this up into three blocks, two of which were 
9 plots long, the other 10 plots. Each block was calculated as a separate 
field. In doing this, care should be taken that the blocks are not so 
small as to be unduly affected by a possible preponderance of very good 
or very poor varieties. 
Another point to be remembered in the practical use of this method 
is that it can not be used to take account of uneven seeding, ravages of 
birds, or other irregularities in certain plots. Corrections, if any, for 
these factors should be added before employing the above method. 
SUMMARY 
It is generally admitted that field trials, including variety tests, are 
often of very little value because of the large number of uncontrollable 
factors. Nevertheless, field trials are becoming more and more a neces¬ 
sity in many phases of agricultural investigation. 
Within recent years a number of investigators have shown that the 
experimental error in such trials can be greatly reduced by the use of 
