240 
G. tJdny Yule » 
with the average weight of the single seed on the parental plant and 
the same character for the offspring? One would expect such a 
mode of grouping to give a greater intensity of heredity in the race 
at large and probably a sensible inheritance within the pure race. 
The main question therefore seems to be this, are the 
results ol Professor Johannsen’s experiments consistent with the 
conception of continuous variation in the character of the germ- 
plasm ? (That they are consistent with some form of discon¬ 
tinuous variation may, I think, be at once conceded). This question 
must be answered in the affirmative. 
On the conception of continuous plasmic variation one may 
picture the formation of the germ cells of a pure race as accom¬ 
panied by a process of gradual breaking up of the original germinal 
characters, controlled only by the action of selection (in the widest 
sense of that term). There must be a breaking-up of the original 
germinal characters (such as might be caused by the daughter cells 
receiving varying samples of the original germinal material), or no 
heritable variations would ever arise. There must be selection, or 
the breaking-up process would in course of time create variations 
deviating to an unlimited extent in every direction; which is absurd. 
Under the action of selection the process results in the genesis, from 
a single cell, of a series of somatic generations in which the 
variation gradually increases but asymptotes (provided all conditions 
are kept constant) towards a fixed limit. Under such circum¬ 
stances the coefficient of correlation will also gradually increase 
and asymptote towards a fixed value, but the initial value and the 
ultimate value depend (1) on the intensity of selection (2) on the 
intensity of effect of circumstance—definite or indefinite—in pro¬ 
ducing divergent somatic characters from similar germ cells. The 
greater the selection, and the greater the divergence, the lower both 
the initial and ultimate values of the coefficient of correlation 
measuring the intensity of heredity. 
To determine these twp factors two independent data of some 
sort are needed, e g. the coefficients of correlation between mother 
and offspring, and between grandmother and offspring, in the 
ultimate race. Unfortunately no suitable data appear to be given 
by Professor Johannsen for the characters with which he deals. 
From two tables he gives (p. 16 and p. 42) one may however roughly 
estimate the coefficients of regression (for offspring on parents) 
for weight and length-breadth ratio of beans as about *17 and *23 
respectively; the coefficients of correlation are probably slightly 
