410 Tricotylous Races. 
a mean of 6%. The two other parents 2 to 10% and 
3 to 14%, with means of 6 and 7%. Obviously the 
difference was not sufficiently considerable to base a se- 
lection among the grandparents upon. On the other 
hand it was evident that the race had been improved by 
the process of selection which now extended over six 
generations ; and this improvement was afterwards main- 
tained. 
For the culture of 1897 I selected the tricotylous off- 
spring of three seed-parents with 11- -11 and 15% tri- 
cotylous offspring, and planted out 100 of them, which 
were all strong plants with a ternary first whorl of leaves. 
The harvest was recorded in the spring of 1898 in the 
usual way. The three parents, however, which were now 
grandparents, again showed no essential difference in 
their offspring. 
VALUE FOR VALUE FOR THE PARENTS NUMBER 
THE GRANDPARENTS MIN. MED. MAX. OF PARENTS 
11% 27 15 20 
11% 3 7.5 15 24 
15% 4 8.5 13 28 
Here again the figures were the same as in the pre- 
vious generation. 
Although, as we have seen, the selection of grand- 
parents, in this case, afforded no reliable ground from 
which an improvement of the race could be started, yet 
this must be the case whenever the hereditary capacity 
of the grandparent is much influenced by fortuitous cir- 
cumstances which affect the selection of the seeds. There- 
fore the special object of such selection is mainly to ex- 
clude such cases as much as possible from the main trunk 
of the pedigree, by simply not planting out their seed 
lings. 
