A METHOD OF CORRECTING FOR SOIL HETERO¬ 
GENEITY IN VARIETY TESTS 1 
By Frank M. Surface and Raymond Pearl, 
Biologists, Maine Agricultural Experiment Station 
Men with practical experience in conducting variety tests and fertilizer 
experiments are free to admit that in many cases the results of ordinary 
field trials are of little or no value. The reason for this lies in the large 
number of factors which are beyond the control of the experimenter. 
In many instances variation in any one of these uncontrollable factors 
may influence the final results to a greater extent than the one controlled 
variable for which the experiment was undertaken. 
On the other hand, field trials and variety tests play an important 
part in agricultural investigations. Such tests are an indispensable 
adjunct to plant-breeding work. The final test of new varieties or new 
strains must be made under field conditions. It is therefore of the 
greatest importance that methods should be devised which will in some 
measure at least take account of these uncontrollable factors. 
No one of these factors is of more importance than the variation in 
the soil in different plots. It is practically impossible to secure for such 
field trials a tract of land that is absolutely uniform. The literature of 
variety tests abounds in illustrations of this fact. 
In 1897 Larsen (8), 2 on the basis of results with timothy, reached the 
conclusion that more exact results were obtained where a given area 
was divided into a large number of plots than when it was divided into 
a few larger ones. 
Holtsmark and Larsen (7) extended this idea and supplied additional 
evidence. Hall (1) in 1909 and Mercer and Hall (9) and Hall and Russell 
(2) in 1911 laid great emphasis upon soil heterogeneity in field tests. 
Among other things they did much to determine the most suitable sizes 
for experimental plots. 
Montgomery (10, 11) has produced evidence showing that systematic 
repetition of plots over a given area reduces the variability in proportion 
to the number of repetitions; further, that while increase in the size of 
a plot decreases the variability up to a certain limit, a further increase 
in size is not attended by a corresponding decrease in variability. 
As a result of these several investigations, it has become evident that 
much more reliable results are obtained by using several systematically 
repeated small plots than by using a single large one. This method is 
rapidly coming into more general use in field tests of all kinds. Never- 
1 Papers from the Biological Laboratory of the Maine Agricultural Experiment Station, No. 93. 
2 Reference is made by number to " Literature cited,” p. 1050. 
Journal of Agricultural Research, 
Dept, of Agriculture, Washington, D. C. 
cl 
(1039) 
Vol.V.No. 2a 
Feb. 28, 1916 
Maine—7 
