386 Tricotylous Races. 
beginning of the flowering period all these were covered 
with fine gauze, guarded from the visits of insects and 
artificially fertilized every day or every other day. About 
300 seedlings of each plant were recorded. The ratios 
were from to 1.2% with an average of 0.8%. Two 
parents had 1.4% and one 2%, that is to say, that here 
again there was no progress. The offspring of the six 
parents fell into groups between which the differences 
were but slight (3 with 0.7%, and 3 with 0.8% on the 
average). 
In the list which follows I have collected the highest 
values that were obtained in the course of these genera- 
tions. 
-SPRING OF- 
1893 1894 1895 1896 
Highest values 2.8 % i.l % 3.7 % 2.0 % 
Selected seed-parents 2.8% 1.0% 2.1% 
These figures show a fluctuation within fairly nar- 
row limits, but no essential advance in the course of four 
generations. It seemed therefore to be useless to carry 
the experiment further. It is certainly probable that, 
in the course of time, further selection might have 
brought about some slight improvement; but obviously 
this would have been of little significance, and at any 
rate there was no prospect of ever obtaining a race with 
50% tricotyls. 
Chenopodiinn album. A tricotylous plant flowered in 
1889, in isolation in my garden, and produced 1% tri- 
cotyls in the spring of 1890 amongst about 1000 seed- 
lings. Four of these were cultivated further, but their 
seeds gave rise again to no more than 1%. The third 
generation was therefore not better than the second. 
Dracocephahnn nwldaiicuin (Fig. 82). In the spring 
