64(> I' chicles of the Hereditary Characters. 
to say, in the formation of the germ-cells the dominant 
and the recessive characters may perhaps not separate 
fully, leaving, either always or only exceptionally, a trace 
of the dominant character in the germ-cell which has the 
recessive one, and vice versa. This trace may then be 
latent during the course of a number of generations, until 
at some later moment, and for some unknown reason, 
atavistic phenomena in such hybrid races awaken the 
memory of the original cross. Experience does not as 
yet support this view; it wants a much larger number of 
generations before a final verdict may be expressed. But 
it is obvious that an atavism of this kind, if it occurred, 
would suggest that the Mendelian units were of a com- 
pound nature. 
These Mendelian factors maintain their independence 
during vegetative life and fertilization. According to 
previous conclusions, such crosses are always concerned 
with elementary characters which occur in a different 
condition in the one parent from that in which they occur 
in the other. There are mainly four distinct conditions : 
the active and the latent, the semi-active and the semi- 
latent. Their vehicles do not only separate in the forma- 
tion of the sexual cells, but occasionally also in the vege- 
tative life of the plant, as is demonstrated by the occur- 
rence of so-called bud-variations in hybrids. 1 They are 
therefore in such cases only loosely associated and not 
blended together. 
Fluctuating variability is due to variation in the num- 
ber of equivalent pangenes : this explains why it is only 
linear (Vol. I, p. 118) and why it is manifested in two 
directions only. It goes in the plus direction by a multi- 
plication and in the minus direction by a diminution of 
1 See pp. 619-620. 
