(Уот. 11 
442 ANNALS OF THE MISSOURI BOTANICAL GARDEN 
the number of seeds per locule may be deduced from the con- 
densed correlation table appearing as table ту.! 
The correlations between the number of fruits and the number 
of seeds per locule as given in table хуш are negative in all 4 of the 
cases but are of a very low order, ranging from — 0.043 to 
— 0.127. "Two of the 4 are over twice as large as their probable 
errors. 
TABLE XVIII 
CORRELATION BETWEEN FLOWERS on ae NOIRE oe AND NUMBER 
SEEDS PER LOCULE AND BETWEEN NUMBER OF FRUITS PER 
INFLORESCENCE AND момака ОЕ ара PER LOCULE 
"um rd e T Fruits т Tar 
ries an E. and E- ifference |--- 
seeds ~ seeds = Еги. 
Meramec High- 
lands 1906 . .| +.0321 +.0389 | 0.825| —.0429 +-.0389 | 1.102} —.0750--.0387 | 1.938 
MeramecHigh- 
lands 1907 . .| +-.0382+.0388 | 0.985| —.0569 2-.0388 | 1.466| —.0951 2-.0387 | 2.457 
Differen --.0061 +.0548 |0.111| --.0140--.0548 | 0.255 — 
— Ba 
+.0952 +.0423 | 2.250} —.0867 4-.0423 | 2.050| —.1819 4-.0424 |4.290 
--.0202--.0426 | 0.474| --.1265--.0420 | 3.012! —.1063-+.0600 |1.772 
J Қ. жері Ваг- 
racks 1907 
Difference. . .| —.1154 2-.0600 | 1.923} —.0398-+.0600 0.663} MEE, yen 
These results show that unless there are statistical considera- 
tions which invalidate the coefficients of correlation as measures 
of interdependence in these cases, the relationship between the 
characteristics of the inflorescence and the number of seeds ma- 
tured is very slight indeed. Apparently inflorescences which are 
initially large tend to have capsules with slightly larger numbers 
of seeds. Inflorescences which mature a large number of capsules 
tend to have a slightly smaller number of seeds in these capsules. 
The difference between the correlations for number of fruits and 
seeds and the number of flowers and seeds shows that in all 4 
1Tha f slightly 
different from those shown here because of the fact that 10 locules i in series IV and 
14 locules in series V1 had a questionable number of seeds. The difference in results 
obtained from calculations from this table and those actually used in obtaining our fh 
constants can hardly be significant, since ig depend merely on differences due to 
10 out of 5211 and 14 out of 4176 locul 
