426 Tricotylous Races. 
parent with the value of 56% flowered in a cage of metal 
gauze. The values, calculated in the usual way from the 
harvest, gave a good curve, whose mean was 72 c /o. The 
lowest percentage of tricotyls was now 38%, the three 
highest 83 83 and 89%, that is to say, a very consider- 
able advance. 
Atavistic seedlings of the same parent were also 
planted out, but they naturally exhibited a somewhat 
smaller advance. 
In the following year, 1898, I did not sow the seeds of 
the plant with 89%, but from various considerations 
those of one with 66%. The reason for this choice was 
that the plant had flowered early and that the harvest 
had been increased thereby. The parents with the higher 
values had flowered too late or set too little seed, and it 
would have been very dangerous to have continued the 
experiment along this line. Moreover there was no 
longer any particular interest in improving the race fur- 
ther. The mean value has been lessened thereby to 
about 40%, and the maximum to 74%. 
Canuabis satii'a. I propose to deal now with two 
cultures of dioecious plants, Cannabis and Mercurialis. 
In these cases self-fertilization is impossible; neverthe- 
less the isolation of the intermediate races was effected 
as easily and almost as quickly as in Ocnothera. Without 
doubt self-fertilization has in such experiments, when- 
ever feasible, the high value which is usually assigned to 
it ; but the experiments now to be described show that it 
can often be dispensed with, as well. This result is very 
important, because it makes isolation and selection pos- 
sible in species in which an artificial fertilization of every 
single seed-parent would increase the labor bevond meas- 
