Partial Variability of Tricolyly. 447 
quently be entirely transformed. tranches of the second 
order exhibit no more than traces of the anomaly. 
If we apply this instance to the distribution of tri- 
cotylous seeds on a plant, we should conclude that the 
lower part of the terminal inflorescence of the main stem 
would produce the highest proportion of tricotylous seed- 
lings ; but the flowers and fruits themselves are lateral 
branches, and so the question arises, how far we are 
justified in expecting this. 
In species like Ocnothcra and Antirrhinum, which 
have a primary inflorescence rich in flowers and in seeds, 
I have usually limited myself to this and, where possible, 
to its lower and middle parts. In Helichrysum to the 
flowerheads which open first, in Clarkia and Phacclia to 
as many of the earlier flowering branches as would pro- 
vide a sufficient quantity of seed, and so on. The question 
is w r hether these first fruits possess the same hereditary 
capacity as would be found from the largest number of 
fruits that could be gathered on the plant. 
With a view to determining this point I have insti- 
tuted a number of subsidiary experiments. The answer 
has been in the affirmative throughout. Certainly there 
are small differences ; these, however, seldom fall out- 
side the latitude of S c /c which is the limit of observational 
error. I shall now present the results of these experi- 
ments in a condensed form. 
A first experiment was conducted with Oenotliera 
hirtclla, which fertilizes itself in parchment bags with- 
out artificial aid. In the summer of 1898 I planted out 
seven tricotylous seedlings of one parent with a heredi- 
tary value of 66 c /r at distances of about a meter apart, 
allowed them to branch freely and to develop into vig- 
orous plants, enclosed the flowers in parchment bags and 
