J. A. Harris 
323 
develop than one which is produced in a fruit with a smaller number of ovules. 
This will " explain " nothing when we arrive at the answer to this question. If we 
find that it has a better chance of developing in a fruit of some particular type we 
shall not know why it has this advantage over an ovule formed in a different type 
of pod. But we shall have definitely isolated one of the factors — or complexes of 
factors — which have an influence upon the development of an ovule into a seed 
and shall have quantitatively determined the amount of influence which it 
wields. 
The ordinary coefficients of correlation between the number of ovules and the 
number of seeds developing per pod are shown in Table 12. 
Two points in this table are noteworthy. First, the correlations are high. 
Second, the agreement between the two series of material is very close. The 
constants for the two collections differ by less than the probable error of their 
difference. 
The equations to the regression straight lines for total ovules and total seeds 
per fruit are : — 
For 1906 series, y = 2-808 + -^Ix, 
For 1907 series, y = - -741 + ■719*-. 
The comparison of t] with the coefficients of correlation and the ^jE^ of 
Blakeman's test for the linearity of regression are given in Table 13. 
TABLE 13. 
Tests for Linearity of Regression in Co7'relation hetioeen Total 
Ovules and Total Seeds. 
Constant 
1906 Series 
1907 Series 
Coefficient of Correlation, r 
Correlation Ratio, r) 
1, =i'-r-' ... 
Blakeman's Criterion, ^jE^ 
•8141 + -0072 
•8317+ -0066 
•02896 
4-028 
•8106+ -0116 
•8420+ •OODB 
•05199 
3-438 
I am rather surprised to find ^jE^ so high as it is in these two cases. The 
indications are that the divergence of the means from a straight line is rather 
larger than one would expect to be accounted for by the errors of sampling. 
From a graph, however, there is no clear indication that any curve would give any 
better fit than the straight line. Certainly the data in hand do not justify any 
further analysis along these lines. The two constants are so nearly identical that 
for any practical purpose it is immaterial which is employed to describe the 
relationship between number of ovules and number of seeds per fruit. 
While the coefficient of correlation shows the degree of interdependence of 
number of ovules and seeds it does not yield one other item of information which 
