404 
J. ARTHUR HARRIS 
Having justified the use of as a constant describing the degree of 
interdependence between p and o and s, I turn to the constants them- 
selves. 
The correlations are conspicuous for their irregularity. Practically 
speaking, however, they are positive throughout. This is strictly true 
for the relationship between pods per plant and ovules per pod. It 
holds for 29 of the 32 measures of the relationship between pods per 
plant and seeds per pod ; of the 3 negative correlations all are small and 
neither can be safely regarded as statistically trustworthy with regard 
to its probable error. 
Computing the physical constants of these varying values of cor- 
relation I find 
The relationship between number of pods and ovules per pod is 
apparently higher than that between the number of pods and the num- 
ber of seeds matured. The first of these may be considered a more 
strictly morphogenetic relationship, the second a more truly physiolo- 
gical one. The difference, .0684 i .0153, is nearly four and a half 
times its probable error, and so perhaps significant. 
Further evidence may be obtained upon the problem of the relative 
values of the two correlations by considering the differences in the two 
constants for each experiment separately, as shown in table I. In 7 
cases Tps ^ ^poy 
but in 25 cases fpo > r^s- Taking the ratio of the 
differences to their probable errors, it appears that 3 of the cases in 
which the correlation for seeds is higher than that for ovules may pos- 
sibly be considered significant ( > 2.5) while there are 18 cases in which 
the correlation for ovules is higher that may be taken to be trustworthy 
statistically. The mean value of the ratios of the positive differences, 
fps > ^po^ to their probable errors is 2.56, while that for the negative 
differences is 4.96. 
It is, therefore, clearly demonstrated that although the values of the 
correlations are both low and irregular, those measuring the relation- 
ship between the number of pods and the number of ovules are sensibly 
higher than those measuring the relationship between the number of 
pods and the number of seeds matured per pod. 
^ Even if regression is not strictly linear, r is the best constant to be used, since 77 
would be too largely affected by the errors of sampling in the numerous small arrays. 
Mean 
Standard deviation. . . 
Coefficient of variation 
.1948 ± .0107 
.0899 ± .0075 
46.16 
.1264 ± .0109 
.0915 ± .0077 
72.38 
