Tricotylous Races Do Not Arise by Selection. 413 
rectly to the increase of this character, but to a supposed 
change in the structure of the seeds which would favor 
the anomaly. Whether or no this supposition is correct 
I do not know, but it is a fact that the fruits and seeds 
of my race have gradually become smaller; and that it 
is the plants producing the largest number of tricotyls 
which bear the smallest fruits and seeds, I find repeatedly 
noted in my records. 
At any rate we are here in touch with a principle 
which may be applicable to other cases also. A selection 
may produce its effect on an unknown character which in 
its turn will affect the character actually dealt with. 
With regard to the extent of my culture of 1898 it 
is further worth remarking that it shows that, in general, 
the extent of the cultures is by no means so important a 
factor as is usually supposed. If I had confined myself to 
experimenting with three or five seed-parents, as for in- 
stance in Amarantus in 1897 (p. 406), I should have 
chosen the best ones according to their hereditary co- 
efficients ; and it was exactly amongst these that the best 
of all occurred as the series on page 41 1 shows. Increased 
extent of the experiment deepens one's insight into the 
processes involved, but does not hasten the improvement 
of the race ; although it is never advisable to confine one- 
self to experimenting with one single seed-parent, if this 
can be avoided. 
The next generation, the 10th and at present the last, 
has repeated the progress observed in the ninth. Here 
again the race of Scrophularia behaved differently from 
that of Amarantus. I confined myself to the progeny 
of the plants of 1897 which exhibited the value of 14%, 
and amongst the offspring of which the mean attained 
20%. From these I selected five specimens with 22 23 
