122 
Journal of Agricultural Research 
Vol. XI, No. 4 
Stated in another form, the differences of standard deviations of the 
four strains when taken as a whole are not large enough to separate the 
strains nearly so well as the differences of means separate them, although 
the standard deviations separate the strains well with regard to some of 
the four characters. For example, we distinguish the high-protein and 
the high-oil strains very effectively from the low-protein and the low-oil 
strains by differences of standard deviations in weight of ears. The 
fact that differences in standard deviations do not enable us to distin¬ 
guish the strains so well as the differences of means does not mean that 
the standard deviations are less stable from season to season for a given 
strain than are the means. In fact, taken as a whole, the standard 
deviations are considerably less variable from season to season than the 
corresponding means. This is in accord with the fact that the probable 
errors of standard deviations are only about 0.7 those of corresponding 
means. We are justified in putting the above statement a little stronger 
by saying that the ratio of fluctuations of standard deviations actually 
experienced to the probable errors of standard deviations are less for these 
four strains than the corresponding ratios of fluctuations of means to the 
probable errors in the means. 
COEFFICIENTS OF VARIABILITY 
The coefficients of variability do not in any marked way maintain a 
definite order of values from year to year. We prepared graphic repre¬ 
sentations of the coefficients, but hesitate to take the space for their 
publication, since the conclusions drawn consist largely in the statement 
of the existence of insignificant differences that may be obtained by 
inspection of the tables. For length of ears these graphs gave a fair 
illustration of as many crossings as we should expect under pure chance. 
On the other hand, in circumference of ears the low-protein and low-oil 
strains have in general somewhat higher coefficients of variability than 
the high-oil strain. 
When we compare differences of coefficients of variability with their 
probable errors, the differences in a fair number of cases are significant 
when judged by the usual requirements in regard to random sampling. 
But these differences are generally insufficient to withstand changes of 
season and soil. Further, we find that the ratio of the average of the 
differences of coefficients of variability in a season to the average of the 
probable errors in such differences varies from 2.7 in the case of rows of 
kernels to 3.8 in the case of weight of ears. The corresponding ratios of 
12 and 23 given above for means show a striking contrast. This accounts 
for the maintenance of a pretty regular order for mean values, while no 
considerable degree of regularity is maintained among coefficients of 
variability. It seems that to maintain any considerable regularity from 
one season to another the differences must be larger, compared to prob- 
