SOMATIC CHARACTERS AND FERTILITY 
Table II 
Equations for Regression of Ovules per Pod and Seeds per Pod on 
Number of Pods per Plant 
Series 
Regression Equation 
End Poin 
s of Line 
Range of Pods 
HH 
0 
= 4.940 + .027 p 
= 4-97» 
38 = 5.96 
1-67 
HH 
s 
= 3.388 + .050 p 
= 3-44. 
38 = 5-28 
1-67 
HD 
0 
= 4.375 + .070 p 
= 4-44, 
14 = 5-35 
I-I4 
HD 
s 
= 3.072 + .011 /> 
= 3-08, 
14 = 3-23 
I-I4 
GGH 
0 
= 5.300 + .031 p 
= 5-33. 
22 = 5.99 
1-33 
GGH 
s 
= 3-743 + -046 P 
= 3-79. 
22 = 4.76 
1-33 
uss 
0 
= 5-173 + -022 p 
6 
= 5-31. 
27 = 5-77 
1-50 
uss 
s 
= 3-457 + .026 p 
6 
= 3-62, 
27 = 4-17 
1-50 
FSS 
0 
= 5.202 + .030 p 
3 
= 5-23, 
30 = 6.11 
1-54 
FSS 
s 
= 2.536 + .047 p 
3 
= 2.58, 
30 = 3-95 
1-54 
GGD 
0 
= 4.561 + .089 p 
I 
= 4-65, 
8 = 5-27 
1-8 
GGD 
s 
= 2.709 + .022 p 
= 2.73, 
8 = 2.88 
1-8 
LL 
0 
= 4.950 + .013 p 
I 
= 4-96, 
27 = 5.30 
1-27 
LL 
s 
= 2.515 + .014/) 
I 
= 2.53. 
27 = 2.88 
1-27 
graphical methods of testing for linearity are equally valid, or even 
stronger, for formal tests, for example those based on a comparison of 
r and rj. For o correlation v may have a substantial value when the 
arrays are numerous; it is in just such cases as this that very mis- 
leading results might be obtained through a careless application of 
Blakeman's test. 
In the graphs the theoretical lines are drawn according to the 
equations calculated from the whole material, but are not extended 
beyond classes of pods per plant represented by only five individual 
plants or fewer. 
The slope of all the lines is slight and the scatter of the empirical 
means decidedly large. This is precisely the condition to be expected 
in widely varying and but slightly correlated characters. For the 
benefit of those who may not be familiar with either the biological or 
the statistical phases of the study it may be pointed out that the ir- 
regularities seen in these diagrams are neither due to carelessness in the 
biological experiments nor to inadequacies of the statical methods, but 
solely to the difficulties inherent in the material. 
With regard to the critical question, that of the sensible linearity or 
non-linearity of the empirical means, the diagrams do not — and largely 
because of this great irregularity — furnish a final answer. But while 
there is some indication of significant non-linearity, for practical pur- 
poses a straight line describes the change in 0 and s associated with an 
increase in p with sufficient accuracy. 
