46 Latent and S end-Latent Characters. 
leaves. And it is obvious that such leaves will extend 
both above and below the maximum of the period in 
direct proportion to their number. \Yhence it again fol- 
lows that the better nourished the plant is, the earlier 
will the abnormality appear. And this is true both of 
the individual L ranches and of the rosette of radical 
leaves, and therefore of the whole plant. 
From these conclusions the converse rule may be de- 
duced that the earlier a seedling produces its first tetram- 
erous leaf, the greater will probably be the number of ab- 
normal leaves on the adult plant. The most abnormal 
plants will probably be those which in the seedling stage 
had a compound primary leaf. Experience has proved 
the truth of this rule throughout my experiments. 
If we now take another glance at the table on page 38 
we see that the character recorded has gradually shifted 
in the course of generations and as a result of selection. 
The more the improvement advanced the earlier could 
selection be effected. In the third generation I kept 300 
plants in the beds to be selected from; since the fourth 
generation I have carried out the selection in the seed- 
pans and only planted out the few best (e. g., 10-20) to 
act as seed-parents. 
It is possible, therefore, within the limits of such a 
race, on the one hand to effect an increase in the number 
of multipartite leaves, and on the other to reduce it In- 
reversed selection. In both cases we go as far as pos- 
sible from the mean of the race, without, however, suc- 
ceeding in overstepping its definite boundaries. Let us 
see what selection is able to effect in the two cases, and 
let us begin with the former. It is the question of in- 
tensifying the anomaly to its extreme limit. 
/ o J 
A striking peculiarity of my race is the fact that leaves 
