390 Tricotylous Races. 
the seeds of four seed-parents producing no tricotyls 
and those of one, 2 amongst 500 atavists. 
Silcne conoidea, like the foregoing species, was ob- 
tained by exchange in 1892 and gave a single tricotylous 
seedling. In the following spring I had about as many 
tricotylous seedlings as in the previous species; and in 
the succeeding generation (1894) 3% tricotylous plants 
amongst 407 seedlings from seeds of a single tricotylous 
seed-parent. Only these 12 tricotyls were then planted 
out; and their seeds were harvested separately in late 
summer. In every lot I recorded from 300 to 900 seed- 
lings and only in two cases, where the harvest had been 
too small, so few as 200. The sowings contained, as a 
rule, from 0.5 to 1.5% and only in one case, amongst 
316 seedlings, 3% tricotyls. All the vigorous tricotylous 
seedlings were planted out in such a way that the off- 
spring of the individual seed-parents stood in groups. 
Their seed was harvested from each seed-parent and 
sown separately; and the proportion of tricotylous indi- 
viduals was determined for each among 300 seedlings. 
This proportion fluctuated, for the 26 seed-parents, be- 
tween and 4.2%. The separate groups manifested no 
relation to the hereditary index of their several seed- 
parents. The offspring of parents with 0.5% and also 
of those with 2% and 3%, had, as a rule, 0.6%. On 
the other hand the offspring of seed-parents with a mean 
of 1 % exhibited this mean ratio of 1 % again, and the 
highest figure obtained in this experiment, 4.2% oc- 
curred amongst them. 
The maximum values in the three succeeding genera- 
tions were therefore 3, 3 and 4%, i. e., they were fairly 
constant. 
In Silcne noctiflora. also, I have studied tricotyly 
