Experimental Error in Crop Tests 
73 
tests were also made in five-row plats relatively free from 
plat competition and subject primarily only to soil variations. 
In Table 1 (1913) the yields of the thick and thin planted 
wheat rows were, respectively, 389 ±5.3 and 264 ±3.8 grams. 
Altho the probable error for each yield is less than 2 per cent, 
the actual error of the relative yields due to competition is 
24.4 per cent. In 1914 the yields of the thick and thin planted 
wheat rows were respectively 327±6.66 and 115±3.6 grams. 
Altho the probable error for each yield is only 2 per cent, the 
actual error of the relative yields, due to competition, is 56.8 
per cent. 
In 1913 (Table 2) the probable errors for the mean yields 
of thick and thin planted oats rows were less than 2 per 
cent, but the actual error in relative yields, due to competi- 
tion, was 20 per cent. In 1914 the probable errors for simi- 
lar yields were also below 2 per cent, while the actual error 
in relative yields, due to competition, was 34.3 per cent. 
Similar examples are seen in variety tests in Tables 3 to 
7. We would have great confidence in these single-row tests 
were we to judge them by their low ''probable errors." How- 
ever, it is evident that this confidence would be badly mis- 
placed. 
Crop tests are subject to such a multitude of local environ- 
mental influences that errors in them cannot be regarded as 
occurring according to the formulas or rules of chance cal- 
culated from purely mechanical observations. The probable 
error calculation may apply, for example, to the chance draw- 
ing of black and white marbles from a bag at a given ratio 
to each other. But variations in crop yields are no such sim- 
ple matter, and the probable error not only may have little 
significance but may be misleading. 
Water Requirements of Corn and Wheat — As further illus- 
tration of the limitation of the probable error, the following 
simple data from our 1916 water requirements of crop studies 
may be cited. 
The object was to make a comparative test of the relative 
water requirements for grain production of a standard variety 
of both corn and winter wheat. Potometers, 16 by 36 inches 
in size and containing 250 pounds of well-manured moist- 
ure-free soil, were used. (The method of testing is de- 
scribed in detail in Nebraska Research Bulletin No. 6.) 
Previous experiments had indicated that these potometers 
would grow one corn plant in a normal manner. The ratio 
