1910] HARRIS—FERTILITY IN CERCIS 123 
we find average deviation=o.248 mm. A discrepancy of a fourth of 
a millimeter in a character such as the pod length of Cercis may 
seem to most biologists quite negligible, but I rather suspect that 
the falling off in mean pod length in pods with seven ovules is signifi- 
cant. The point is of little practical importance for our present 
discussion. 
For number of seeds maturing per pod and length of pod we 
find the interdependence r,=0.5055-+0.0092, and the equation to 
the regression straight line 
y= 58.645 + 3.579%, 
Where the significance of y is as above, and x=seeds per pod. The 
diagram (fig. 1) shows the agreement between the line given by the 
equation and the empirical means. Throughout the central region 
of the distribution of the number of seeds developing, the fit is very 
good, but at both ends the observed means fall considerably below 
those to be expected from the equation. If these deviations are 
biologically significant, and not due merely to the small number of 
observations which fall at the extremes of the range, they indicate 
that both the pods developing only a single seed and those developing 
the maximum number of seeds are somewhat dwarfed in length as 
compared with the whole series of 3,000 pods. I am not yet ready 
to discuss the reasons for this condition. 
Comparing the constant for ovules and length, and seeds and length, 
we find 
For ovules and length, r=o.42780.o101 
For seeds and length, r=o.5055-+0 .0092 
Difference=o.0777 
Since the correlation is actually, though only slightly, higher for 
Seeds developing than it is for number of ovules formed, it would 
appear that the number of seeds developing must have some connec- 
tion with the length of the pod independent of the interdependence 
for length and ovules. | 
__ As pointed out early in this paper, one of the difficulties in assert- 
ing that there is a real physiological relationship between the number 
of seeds developing and the size of the fruit arises from the fact that 
