206 
G. Udny Yule. 
mean character of the germ cells which he produces. If we con¬ 
ceive a number of individuals of the same somatic type, some will 
be abnormal developments of mediocre germ cells ( i.e . cells pro¬ 
ducing on the average mediocre types); others, but these will be 
fewer, will have sprung from germ cells producing on the average, 
or usually, more abnormal types. The odds are, therefore, that a 
given abnormal somatic type is an abnormal development of a 
mediocre germ cell rather than a mediocre (or subnormal) develop¬ 
ment of an abnormal cell. But the somatic characters of offspring 
will follow the germ cell characters and not the somatic characters 
of the parent; therefore regression occurs. Further, as the line of 
germ plasm is handed on from parent to offspring and the processes 
of nutrition and growth, if continued unchanged, need not be 
supposed to alter the average character thereof, any one individual 
in a line is an index to the character of the plasm in that line. 
Hence if ancestry as well as parents be abnormal it is more probable 
that the parents are an average development of a really abnormal 
type of germ cell, and hence more probable that the offspring will 
follow, and not regress from, the parental type, i.e. we have 
“ ancestral heredity.” 
I have put the whole matter as briefly as possible, without 
discussing details, because I wish at present simply to emphasise 
the fact that although the theory of ancestral contributions to a 
heritage implies the law of ancestral heredity, the converse is not 
true: the law of ancestral heredity need not in any way imply 
actual physical contributions of the ancestry to the offspring. The 
ancestry of an individual may serve as guides to the most probable 
character of his offspring simply because they serve as indices to 
the character of his germplasm as distinct from his somatic 
characters. 
The same line of argument applies in the case of attributes. 
The germ cells cannot in general be treated as if they rigidly 
determined whether the individual should be an A or an a, the one 
type of cell merely produces a majority of A’s, the other a majority 
of a’s —as a loaded die tends to fall the more often on the one face 
than on the others ; only in the extreme case the loading might be 
so heavy that the die would always rest on the one face. In general 
then one could not be certain that the germ cells of an A individual 
(reproducing asexually, to avoid complications) were of the ^4-type, 
i.e. the type producing a majority of A's; they might be or 
might not. If, however, the parent of the /f-individual were also 
