210 BOTANICAL GAZETTE [MARCH 
B. DISCUSSION OF DATA FROM GENERAL SAMPLES 
In the foregoing section, samples from each tree were treated 
individually, because heterogeneity of material may have a con- 
siderable influence upon correlation. The constant for any single 
individual cannot be taken to typify Staphylea trifolia as a species. 
Not only is each plant in some measure individual, but the con- 
stants based upon a small series of fruits may be too large or too 
—, are mR 
=—_ = 
= — 
sr na — | 
eal = Fae 
Ee — | 
——= = mee 
i a — 
=) —) — | 
i—3 2 TRE — 
D RRR — 
— a) ) 
Set —= — | 
= amma, 
—= a =a 
— OER —— 
— ee —s 
= eee — 
vem 
nae eee — 
Tio Tis Difference, ris—Tio 
Fic. 1.—Diagram to show intensity of correlation between ovules and length 
(r,,.), seeds and length gt and their difference lag in 20 individuals of 
Staphylea; the individuals are numbered 11 to 30 from the top of the diagram; the 
length of the bars shows i intensity of i pra A and the shaded area the 
statistically significant amount, that is, the excess over 2.5 times the probable error. 
small by an amount known as the probable error of random 
sampling. 
To free our constants as far as possible from the influence of 
probable errors, and to gain more definite conceptions of the actual 
intensities of our relationships, we may (a) examine the means of 
the constants for the individuals, and (b) combine all our subsamples 
to form one large “population” of 2059 pods (6177 locules) for 
1906, and 1218 pods for 1907. Table V contrasts the values 
