Tricotylons Races Do Not Arise by Selection. 411 
But when the values of several parents do not differ 
so greatly from one another that the differences fall out- 
side the limits of probable error, individuals should be 
planted out every year from all of the best parents. This 
would, however, necessitate an increase of the cultures on 
too large a scale. Nevertheless I extended my culture 
as far as possible in the summer of 1898 on this principle, 
and saved the seeds of 300 plants separately. 
The result of the 300 counts which were made on 
these lots of seeds, embracing 300 seedlings for each 
lot, corresponded with my expectations to this extent 
that one of the grandparents proved to be by far the best. 
Its hereditary value was 14%, that of its offspring had 
a mean of 20%, and for two individuals even 25 to 27%. 
Moreover, outside the limits of this group, this race also 
showed an increase in hereditary capacity. 
The harvest of 1898 contained the seeds of 300 tri- 
cotylous plants, which in their turn had been raised from 
15 parents with values from 10 to 15%. The results 
were as follows (expressed in percentages) : 
Parents 10 10 10 10 10 10 11 11 12 12 13 13 14 15 15 
Offs rin ^ Mean ' 8 10 10 H 14 19 10 l5 U 16 10 13 20 14 14 
' < Max. . 8 11 20 20 27 26 19 25 20 24 21 21 27 21 22 
Number of offspring 1 2 19 28 11 6 8 9 27 6 36 32 12 64 39 
The mean of the whole series of counts was 12% and 
denoted a considerable increase in the character of the 
race. 
Let us now compare this series of figures with that 
given on page 405 for Amarantus speciosus. That series 
relates to the hereditary coefficients of 140, this to those 
.of 300 plants, which in the former case belong to the 
8th, and in this one to the 9th generation. Both series, 
therefore are the result of a selection which extended 
