360 
Journal of Agricultural Research 
Vol. XIV, No. 9 
In order to obtain the desired data in a form comparable for the 4-year 
period and easily presented, the biometrical method was used ( Davenport , 
1907)} 
The subject matter is arranged for presentation in two main divisions. 
In the first division variability both of the seed used and of the resulting 
plants is studied. The means, standard deviations, and coefficients of 
variability are used to this end. In the second division (1) degree of 
relation between weight of seed and characters of the resultant plants, 
and (2) degree of interrelation between characters of the resultant plants 
are shown by correlation coefficients. 
REVIEW OF LITERATURE 
Love and Leighty (1914), working with a pure line of oats, found that 
biometrical constants—that is, means, standard deviations, coefficients 
of variability, and correlation coefficients—vary more or less with en¬ 
vironmental conditions, such as degree of crowding of the plants and 
differences in the weather conditions. With conditions not so favorable 
for plant development, less variability was found in number of culms, 
total and average number of spikelets, and average number of kernels 
per spikelet. In average weight of kernels, greater variability was found 
under unfavorable conditions. When development was arrested by 
environmental conditions, yield was lowered, not by reduction in average 
weight per kernels or number of spikelets produced, but by a reduction 
of the number of kernels per plant, per culm, and per spikelet. Corre¬ 
lations were broadly classified as (a) fluctuating, which vary considerably 
with environmental conditions, and (b) stable, which vary less from year 
to year. Between yield of kernels per plant and their average weight 
no correlation was found in one trial and in two others the coefficients 
were low—but five and seven times their probable error, respectively. 
The interpretation is that, for the years when correlation between these 
two characters occurs, selecting the largest seeds would be securing 
them from the heaviest yielding plants. Average height of plants, as 
correlated with average weight per kernel, gave coefficients of 0.219 ± 
0.029, —0.023±0,034, and 0.217±0.032, respectively, for three years. 
For one year there is no correlation. For the two other years the coeffi¬ 
cient of correlation is seven times its probable error, which is significant. 
For the two years the taller plants had a tendency to produce the larger 
kernels. 
Leighty (1914) found practically the same correlation coefficients 
when determinations were made on single culms as when whole plants 
were used. There was a uniform tendency for the coefficients to be 1 
greater when single culms were used. In studying oats grown in hills 
as compared with that growth in drills rather large differences occurred 
1 Bibliographic citations in parentheses refer to “Literature cited,” p. 391-392. 
