1917] 
GATES—THE GENUS TRILLIUM 85 
longed. We may therefore assume, as the other facts sug- 
gest, that this variation occurred independently of the others, 
and perhaps subsequently. On the other hand, the reduction 
in petals in Trillium (leading towards Paris) displays itself 
particularly іп T. Smallii Maxim., in which the petals may 
be more or less reduced or absent. 
The fact that certain teratological conditions in one genus 
frequently resemble the normal condition in a related genus, 
as we noted in a previous paragraph, shows that variations 
tend to follow certain paths. These variations must result 
from the structure of the germ-plasm, and may be compared 
with lines of cleavage or fracture. They apparently result 
from certain weaknesses in the structure of the germ-plasm, 
and they are apparently not environmentally produced (unless 
in the sense of large responses to small stimuli), but reappear 
generation after generation through long periods of time. 
They represent the unstable nature of certain elements of the 
germ-plasm, and are apparently, when reproduced from seed, 
themselves unstable. This is a matter on which more exten- 
sive data are urgently needed; e.g., will a 4-parted Trillium 
come true from seed, or how will its peculiarity be inherited, 
if at all? And will a partly double T. grandiflorum which 
reappears each year from the same rootstalk reproduce itself 
from seed? It is greatly to be hoped that breeding experi- 
ments with teratological plants will be undertaken to deter- 
mine this point. One is strongly inclined to believe that such 
peculiarities as polymery and doubling will be reproduced in 
some, at least, of the offspring. Experiments with double 
garden flowers of course point to this conclusion. 
On the other hand, it appears that somatic variations, such 
as fasciation, which are not at all inherited in some genera, 
have become a constant feature of the genus in other genera, 
e.g., Celosia. We have at present no means of knowing how 
the unstable and non-inherited or partially inherited terato- 
logical variations of one genus may give rise to the stable 
and completely inherited condition of a derived genus; but it 
is a legitimate interpretation of the facts to suppose that 
