Feb. a8 t 1916 
Correcting for Soil Heterogeneity 
1049 
systematically repeated plots. Nevertheless, if the number of repeti¬ 
tions is not large, certain experiments may still be unduly influenced by 
irregularities in the field. It would therefore be desirable if some method 
could be devised by which the yields of individual plots could be corrected 
in such a way as to take account of these irregularities. 
Check plots have frequently been used for this purpose. But, aside 
from the extra labor and expense involved, the results from check plots 
have been far from satisfactory in many cases. 
In the present paper a method is proposed for use in correcting for 
differences in the soil of different plots. The method in its present form 
is adapted for use only when the plots are arranged in blocks similar to 
those in figure 4. The method of obtaining this correction factor is as 
follows: In the first place the probable yield of each plot is obtained by 
the contingency method. This *'‘calculated” yield represents the most 
probable yield of each plot on the supposition that they have all been 
planted with a hypothetical variety whose mean yield is the same as the 
observed means of the field. 
This “calculated” yield may then be used as a basis for determining 
a correction factor. If the calculated yield of a given plot is above the 
mean of the field it must be taken that the soil of this plot is better 
than the average of the field and a corresponding amount must be 
deducted from the observed yield. Likewise, if the calculated yield is 
below the average, a proportional amount must be added to the observed 
yield in order to make the plots comparable. 
Still more comparable results will be obtained if the correction factors 
are based upon the percentage of the mean rather than upon the absolute 
figures. 
Tests of the efficiency of this method by means of the measure of soil 
heterogeneity proposed by Harris (6) show in all cases a very marked 
reduction in the amount of heterogeneity when the corrected figures are 
used. When tested on our own experimental plots, this method leads to 
results which from other evidence, we have reason to believe, more nearly 
represent the truth than do the uncorrected yields. 
It is realized that this method is not ideal and does not obviate all the 
difficulties connected with soil differences in plot experiments. It is 
hoped that this method may prove useful in certain kinds of plot experi¬ 
ments and that it may lead to further study of this problem. 
