54 Latent and Semi-Latent Characters, 
be practically no difference between the several individuals 
in each group, I calculated the mean for the two sets in 
percentages. 
Number of leaflets per leaf 34567 
From seeds from garden soil 14 13 25 16 32 
" sandv soil 39 13 23 10 15 
Difference 25 +2 +6 +17 
The effect of the treatment in the previous year is now 
perfectly plain. The curves for both groups have become 
unilateral but in the case of the better nourished ones 
the apex is at seven, and for the others at three leaflets 
per leaf. 
Conversely we may conclude that, in the experiment 
described on page 47 and graphically exhibited in Fig. 6, 
the atavists were produced by poorly nourished and the 
maximal variants by highly nourished seeds. And the 
following generalization about anomalies seems to be 
justified: that the nutrition of the seed on the parent 
plant is the most important factor influencing the devel- 
opment of the anomaly (Vol. I, pp. 521-522). 
Let us now briefly summarize the results of this 
experiment. I began by finding in the field two plants 
belonging to a five-leaved race, which however as the 
result of indifferent nutrition for several generations 
only developed tetra-pentamerous and no 6-7-foliate 
leaves. By better cultivation and by the continued selec- 
tion of the most abnormal individuals, no doubt those 
which happened to have been best fed, a race was evolved 
in the course of a few generations with a number of leaf- 
lets per leaf varying between 4 and 7 round a mean of 
five. After this selection had been repeated four or five 
times maximal variants were produced the majority of 
