184 BOTANICAL GAZETTE [MARCH 
This class size was adopted in accordance with the usual biomet- 
rical procedure, the variations in the small parent (314) being 
included in only 3 classes. But when this is done, the F, dis- 
tribution is decidedly skew. The theoretical mode is along about 
the fourth or fifth class instead of the central class. What is the 
reason for the production of this type of curve? There must bea 
reason, and it seems to me that this reason must be biological and 
not a mathematical transnomination, as have been all the bio- 
metrical analyses of skew curves. The matter appears clear in 
the light of the following interpretation. 
In ordinary statistical work, one produces a frequency dis- 
tribution by throwing his tabular entries into arbitrary classes 
of equal size. By this procedure he has in all probability distorted 
their relationship. This fact is partially recognized by using the 
coefficient of variability instead of the standard deviation as a 
measure of variation. Unfortunately, it is usually said that the 
coefficient of variability is used instead of the standard deviation 
because it is an abstract measure and pounds can be compared 
with inches, etc. Standard deviations in the same concrete terms 
are usually thought comparable with each other. But is this true? 
Apply the rule to the data in-tables I and II. The range of length 
of corolla of N. forgetiana (314) is 3 classes, the standard deviation 
is 2.27 0.08 mm., and the coefficient of variability is 8.86 +0. 33 
per cent. The range of N. alata grandiflora (321) is 6 classes, its 
standard deviation is 5.38+0.20 mm., and its coefficient of vari- 
ability is 6.82+0.25 per cent. Paiecing standard deviations, 
N. alata grandiflora is twice as variable as N. forgetiana. Com- 
paring coefficients of variability, which being functions of the 
mean give weight to the size of the mean, the large-flowered type 
(321) is less variable than the small-flowered type (314). 
Let us now look at the matter from an ordinary common-sense 
biological standpoint. These pure line populations may be con- 
sidered as composed of near-homozygous individuals. The range 
of variability shown is therefore almost wholly due to environ- 
ment. In general, NV. alata grandiflora has a corolla more than 
twice as long as NV. forgetiana. Is it not reasonable to suppose 
that the unit change effected by environment and expressed as 4 
