246 HARRIS: RELATIONSHIP OF OVULES TO SEEDS 
significant. No importance need to be attached to the difference 
for seeds developing, which is not twice its probable error. 
The difference for the standard deviations and coefficients of 
variation are: 
Standard Deviation Coefficient of Variation 
WVTIOS Oo 6c cia t —.1258 =.0097 —3.395 
ORG sical ve tia es —.0596 +.0117 —1.354 
The difference in S.D. for ovules is nearly 13 times its probable 
error and for seeds about 5 times its probable error. Both are 
clearly significant. 
The difference for the coefficient of correlation is 
fT, =, tI990 = .0082, 
a difference 13.5 times its probable error and undoubtedly sig- 
nificant. 
It appears, therefore, that our samples are sensibly differen- 
tiated from each other in type, variability and correlation. This 
fact is sufficient ground for considering their correlations inde- 
pendently. 
The significance of the coefficient of correlation depends upon 
linearity of regression. Using the familar equation for the re- 
gression straight line 
I find 
MOCANSEQ 000. Pcl a, 2 ee i a ee S$ = .3554+.7207 0 
For additional 4,000........... Mee aauicraly S$ = .1423-+.7970 0 
FOE Het AO O00 ei eee Sd S$ = .2712+.7504 0 
The closeness of agreement of the observed means and those 
given by the equation is evident from Table IV where the two are 
compared. If the two extreme variates, where the numbers of 
observations are so small that little weight is to be attached to 
them, be omitted, there is only a single case out of the eighteen 
where the deviation of the observed from the theoretical mean 
reaches thirteen one-hundredths of a seed. 
The average (weighted) deviations (disregarding signs) of the 
