52 Latent and Semi-Latent Characters. 
much as the plant is most sensitive in its early stages. 
In this form of the experiment we can, so to speak, only 
investigate the last vestiges of its former susceptibility. 
Far more striking results are to be expected from experi- 
ments with seedlings; but here a great uniformity in the 
samples of the seeds is necessary for the results to be 
reliable. It is not sufficient to mix the seed, but it is 
advisable to harvest seed from two or three or still better 
from a single seed-parent of known and pure ancestry. 
It is even better to allow the influences that are to be 
investigated to operate during the development of the 
seed on the parent plant. 
In accordance with these considerations, therefore, 
I cut one of my plants into two parts, one of which I 
transplanted into poor sandy soil but the other into 
good garden soil, and allowed them to set seed. I was 
thus able to study both the direct effect on the plant and 
also the indirect effects on the succeeding generation. 
(See Vol. I, Part III, pp. 521-522.) 
The experiment, which was carried out during the 
years 1892-1894, was made with a single individual which 
arose from the stock plant for 1891, mentioned on page 
38 and marked S. This plant had, when it germinated 
in 1892, a bimerous primordial leaf, and in the same 
year bore seeds which, when sown in the spring of 1893, 
gave rise to about 40% seedlings with a tetramerous 
leaf. As soon as this was visible in the seeclpan the 
choice was made and the parent plant, which I had kept 
through the winter in a bed, was cut in two and trans- 
planted into the above mentioned kinds of soil. Both 
halves grew well, although not with equal luxuriance; 
they flowered in July, were pollinated from the various 
plants around them composing the main culture of that 
