188 Observation of the Origin of Varieties. 
victuals which had 21, 21, and 22 rays respectively, in their 
terminal inflorescences. One of those with 21 served 
for the continuation of the experiment but all three had 
exhibited correspondingly high numbers in their lateral 
flowers. The fertilization in this year was therefore 
sufficiently pure. 
This was not the case in 1897. The 34-rayed seed- 
parent of that year was pollinated at first amongst all the 
other plants, and later amongst the rest of the selected 
ones which were however as many as 25 in number. It 
set so little seed that it was impossible to rely solely on 
the seed due to the later pollinations (i. e., the purest 
seed) for next year's crop. The two maxima of the 
curve of 1898 are therefore, at least to some extent, due 
to mixed pollination (Fig. 32 under 1898). 
In 1898 I selected the seeds for the continuance of 
the race in two periods on the chosen seed-parent after 
having marked the flowers separately for them. The 
first harvest was from flowers which had bloomed be- 
fore the removal of the rest, the second from those which 
had bloomed later. The latter must therefore have been 
fertilized by the seven remaining seed-parents whose ter- 
minal inflorescences, however, all had had more than 34 
rays (the numbers were 35-36-37-38-39-40 and 46). 
The two specimens were sown separately and their curves 
determined ; but no essential difference between them 
could be detected, either in their limits, or in their means, 
or in their general shape. The seed-parent of 1899 with 
66 rays and with the first 1-3 ligulate florets on its discs 
(Fig. 33), belonged to the first series, the 67-rayed plant 
shown in the table on page 180 for 1899, however, to 
the second. 
In the summer of 1899, towards the end of July, I 
