Tricutyls, Ilcini-lricotyls, and 'J'ctraculyls. 359 
same culture. In doing so we find that they do not tend 
to reproduce their own type, but behave, as a rule, in the 
same way as tricotyls. Minor quantitative differences 
may occur, but of qualitative there are none. At any 
rate it seems impossible to isolate and fix these two sub- 
sidiary types and obtain pure hemi-tricotylous or tricotyl- 
ous races. 
I propose to deal first with the hemitricotyls. I have 
repeatedly isolated them and tested their hereditary capa- 
city, especially in Amarantus spcciosus and Cannabis 
satii'a. In Amarantus, if we plant out some hemi-tricotyls 
and some tricotyls, the highest values are sometimes ob- 
tained for the former, and sometimes for the latter; but 
only with slight differences. For instance, in the harvest 
of 1892 the value of 20,000 seedlings w^as a mean of 2% 
for the former, and 3.5% for the latter. Moreover 
there were slightly more hemi-tricotyls than tricotyls 
amongst the former, the difference, however, was only 
0.1% in 10,000 seedlings of each group, and in this 
figure the various degrees of cleavage were, moreover, 
lumped together. In 1892 to 1895 I then grew the 
hemi-tricotyls and the tricotyls separately for three fur- 
ther generations, always selecting the individuals with the 
highest hereditary capacities. In these four years the 
highest values obtained varied for the hemi-tricotyls from 
4.2 to 8.5% with a mean of 5.5%, and for the tricotyls 
from 3.6% to 7.4% with a mean of 5.7%. From these 
figures it seems to make practically no difference whether 
hemi-tricotyls or tricotyls are selected as seed-parents. 
Of Cannabis sativa I had in the summer of 1894 a bed 
of each of the two types. The hemi-tricotyls gave values 
varying from 1 to 26%, and the tricotyls from 4 to 14% : 
with means therefore of 11 and 9%. In Penstemon 
