190 Observation of the Origin of Varieties. 
given rise immediately to flowerheads with central ligu- 
late florets if it had been on a sufficiently large scale. 
But it would not have produced them in a proportion 
which could be predicted by QUETELET'S law, but accord- 
ing to the principles of discontinuous variation which 
are still unknown to us. 
The course of the improvement is different in the 
two cases. The results obtained with the maize conform 
to the law of regression, the increase in the number of 
rows in the ears becoming slower and more difficult to 
secure, the further we get from the starting-point. Ex- 
actly the reverse is the case in the Chrysanthemum. The 
progress was continuous and did not materially change 
until 1899, when the first central ligulate florets appeared. 
Then it took a leap, all the offspring of this plant having 
more or less double flowers. More strictly speaking, 
the leap had already taken place, the plant with the first 
central ligulate florets (Fig. 33) having already crossed 
the threshold. Its offspring behaved like the offspring 
of a pure race, such as for instance C '. inodonun plenis- 
simum. 
A break therefore occurred, and obviously before 
1899; either in the origin of the seeds of 1898 from 
which the plant in question arose, or even earlier. 
And since C. Inodornm plenissimum has maintained 
itself for many years without selection, it is probable 
that the new C. segetnm plenum will do the same. But 
the reverse was the case with the maize which reverted 
to the old form within a few years after the cessation 
of selection (Vol. I, p. 125). 
Hitherto I have taken the number of ray-florets in the 
terminal inflorescence -almost exclusively as a character 
of the whole plant, and the curves have been plotted 
