HARRIS: NUMBER OF OVULES FORMED 451 
number of seeds per pod, 7., while the fourth contains the correla- 
tions between the number of ovules per pod and the deviation of 
the number of seeds from their probable value. 
All of the values for 7,, are positive, and of a substantial order 
of magnitude. Both positive and negative coefficients for ros 
occur, and the values are low throughout. Of the 53 determina- 
tions, 38 are negative and 15 positive in sign. Were there no 
biological relationship between o and s an equal number of positive 
and negative values would be expected. Thus there is a deviation 
from equality of 11.5 * 2.46,* which is probably significant. 
Thus there is apparently a distinct negative relationship be- 
tween 0 and z. This conclusion is supported by restricting the 
constants upon which it is based to those which are more probably 
statistically significant with regard to their probable errors. Thus 
I find that for the 41 constants 2.5 and more times their probable 
error 28 are negative and 13 are positive. Of the 26 constants 
which are over 4 times their probable error, 21 are negative and 5 
are positive. Of the 13 which are over 7 times their probable 
error, II are negative and 2 are positive. 
Taking means of the ratio of the correlation coefficients to their 
probable errors, I find a mean ratio of 5.10 for the negative values 
and of 3.87 for the positive. Thus the negative constants are as 
a whole more trustworthy than are the positive, although some of 
the positive values must be certainly regarded as trustworthy 
statistically. 
Consider next the relative magnitude of the positive and nega- 
tive coefficients. The 38 negative constants give a mean value of 
— .0732 while the 15 positive correlations give a mean value of 
+ .0529. Thus the negative coefficients are numerically larger 
than the positive. 
I now split the materials up into the individual varieties. The 
results are conveniently summarized in TABLE Il. 
Positive coefficients occur only in Navy, Ne Plus Ultra and 
White Flageolet. In the two latter, only 3 of the 22 constants 
are positive, and neither of the three can be safely regarded as 
significant with regard to its probable error. In Navy, quite 
different conditions seem to prevail. In both strains 6 of the 9 
* Calculated from .6745 V 53 X.5 X-5- 
