214 BOTANICAL GAZETTE [MARCH 
per fruit and the length? The length of the three parallel placentae 
determines in large measure the length of the fruit. From the 
purely structural point of view, it would seem logical to compute 
the relationships for the fertility characters of the individual locules. 
But in testing the assumption that the developing seed exerts 
some influence upon the growth of the pod, it is important to deter- 
. mine whether there is a closer interdependence between the length 
and the total seeds developing per fruit than between the length and 
the number developing per locule. As a basis for comparison the — 
same relationship for total ovules per fruit must also be computed. 
The data are: ovules and length, 1906 (table VI) and 1907 (table 
VII); seeds and length, 1906 (table VIII) and 1907 (table IX). The 
constants may at once be compared with those calculated by the 
other methods in table X. For total ovules and length the agree- 
ment with the same relationships for individual locules and length in 
the population is very close. For total seeds and length, the results 
are somewhat irregular... The logical comparision would seem 
to be that of the populations rather than that of population for 
total seeds and the means of individuals for the numbers per locule. 
For both series this comparison shows a higher correlation when the 
total seeds developing are used. 
We may now determine whether the slope of a straight line ade- 
quately describes the change in pod length associated with different 
degrees of fertility. Only total seeds and length need be con- 
sidered. The equations are: 
For 1906, /=5.252+.505 s; for 1907, J=6.186+.256 s. 
The empirical means and the fitted lines from the above equa- 
tions are shown graphically in fig. 4. It does not require a trained 
eye to see that the agreement is not very satisfactory. For the 
lower seed classes, the length of the fruit increases much more 
rapidly, and for the higher seed classes much less rapidly than the 
average rate shown by the slope of the line. 
% As compared with the population constants for number of seeds per locule and 
length, both are considerably higher, but when compared with the means of the 
correlations calculated from the individuals, it appears that the constant for total 
seeds and length is higher than that for number of seeds per locule and length in 1906, 
but slightly lower in 1907. 
