84 
Journal of Agricultural Research 
Vol. XXXI, No. 1 
The hill-checking method was used in all comparisons—that is, one 
test plant and one check plant were grown in each hill. These two 
plants were 8 to 10 inches apart in the rows, and always in the same 
order so that they could be identified. The seed for the test plants 
in any one row was of one of the varieties or selections being compared. 
The seed for the check plants was composited by taking a definite 
number of kernels from a definite number of ears and was the same 
for all of the rows in the experiment. 
In general, two kernels were planted where one plant was desired, 
the extra seedlings being removed when they were from 4 to 8 inches 
high. This partly eliminated irregularity in stand. The effect of 
the remaining irregularity in stand, as well as the effect of soil varia¬ 
tion, was avoided by harvesting only perfect hills. A perfect hill 
was defined as one containing a check plant and a test plant that 
could be identified definitely and that apparently had had equal 
growing conditions. 
Each variety or selection in an experiment was grown in single 
rows replicated in different parts of the plat. The perfect hills in 
these rows were harvested, and the production of the test plants and 
of the check plants was determined for each row. The weight of ears 
from the test plants in all rows of one variety or strain in the experi¬ 
ment, divided by the weight of ears from' the corresponding check 
plants, expressed as a percentage, constitutes a relative yield as here 
used. It is believed that the reliability of results obtained by the 
hill-checking method is proportional in general to the total number 
of perfect hills grown. Accordingly, the number of perfect hills on 
which the relative yields are based is given in the data. 
In the varietal experiment in 1912 moisture determinations were 
made on the shelled grain, and these are given in the data, but no 
corrections were made in the yields. In most of the experiments the 
ears were uniformly dry at harvest, or, if not, corrections were made 
on the basis of data from drying samples. 
VARIETAL EXPERIMENTS 
EXPERIMENTS IN SOUTH CAROLINA IN 1912 
The varietal experiments in South Carolina in 1912 were con¬ 
ducted on 10 farms in different parts of the State. Ten varieties, each 
of which had been grown previously on one of these farms for three 
or four years, were used throughout the experiments. One hill- 
checked row of each of the 10 varieties constituted a section which 
was replicated 10 times, each of the varieties being used as the check 
in one section. 
The total yield of all the plants of a variety in the experiment was 
divided by the total yield oi all the plants grown in the same perfect 
hills with it to obtain its relative yield. Computed in this way, the 
relative yields are percentages of the average yield of all the varieties 
in the experiment. 
The experiment was destroyed by storm and flood on one farm. 
Two varieties were found to be somewhat earher than the others, and 
their yields are not presented. This leaves the yields obtained on 9 
farms from 8 varieties, 4 prolific and 4 nonprolific, as a basis for com¬ 
parison. It was evident that these widely distributed farms repre- 
