448 Tricotylous Races. 
harvested and examined the seeds for each individual 
branch separately. Only slight differences were pre- 
sented by the values obtained, and I shall therefore only 
give the means. I determined the values of the inflores- 
cences of the main stem and found the average to be 
38% ; for the strong lateral stems produced from the 
axils of the rosette (see Fig. 55, Vol. I, p. 302) I found 
it to be 45% ; for the upper branches of the stem (see 
Fig. 49, Vol. I, p. 282), however, 47% (calculated from 
24 determinations), and for the lower branches of the 
stem, which in iihis species tend to be very much weaker, 
52% (in eight counts). The distribution of the differ- 
ences, therefore, was different from what would have 
been expected. They show that in this case, the harvest 
from the primary inflorescence gives a somewhat lower 
value than the whole harvest of the plant in question 
would have given. In Dracocephalum moldavicum, where 
the values are always small, I collected, in the summer of 
1895, the seeds of all specimens from the main stem and 
lateral branches separately, but found no difference (0.4% 
for both). In Amaranhis spcciosus the seeds from the 
terminal panicle regularly gave somewhat higher values 
than those from the lower branches, but with very slight 
differences only (1892). The average calculated from 
20 plants was 2.8% for the former and 1.7% for the 
latter. 
In many of my experiments I have saved the seeds 
which ripened first, separately from those which ripened 
later: e. g., Amarantus speciosus, Scrophularia nodosa, 
Mercurialis annua, Antirrhinum majns, Silene inflata, 
and others. No differences of any importance were 
found in this way. Deviations are sometimes found in 
larger series, but only such as can be attributed to the 
