Partial Variability of Tricotyly. 449 
unavoidable errors in the samples of seed. In Mcrcit- 
rialis annua I have for several years recorded the seed- 
lings which came up from the first seeds that fell out 
on the bed. In this species, as is well known, the fruits 
open suddenly and scatter their seeds about. These counts 
gave essentially the same results as those obtained later 
from seeds harvested and sown by hand. I often made 
use of them to obtain a preliminary idea concerning the 
extent of the improvement to be expected. 
My custom is to limit the separate harvests to the 
desired quantity of seed, by cutting back during the 
period of growth, and by stopping the saving of seed as 
soon as a sufficient quantity has been secured. For this 
purpose I have estimated this amount from the result 
obtained in the first generation, in the case of each species. 
Moreover I have frequently compared such gatherings 
with much more extensive ones, obtained either by not 
cutting the plants back or by not ceasing to save the seed 
before all of it was ripe. In this way Amarantus spc- 
ciosus often gave higher values from a smaller harvest, 
but the differences were slight and the exceptions many. 
Caunabis sativa gave the same values from large plants 
bearing more than 100 cubic centimeters of seed each, 
as from average or weaker individuals. In such large 
crops, saved without limitation and amounting to 80- 
110 cc. of seed the average was 11%, but in crops of 
20 35 cc. it was 14%. The cultures were made in 1894 
and the same happened in other years. In Ooiothcra 
rubrinerzis the value was seen to decrease as the size 
of the harvest increased, but only in degrees of one-tenth 
per cent, on the average, amongst numerous individual 
trials. The same occurred in Scropliularia nodosa. 
This latter species, as well as Silcnc inflata, is a per- 
