412 Tricotylous Races. 
over a sufficiently long period of time to justify the expec- 
tation that a definite result would be obtained. They 
exhibit, however, one striking difference. The mean of 
the figures for Amarantus is between 2 3%, and this 
renders it not improbable that, if the coefficients of the 
atavists could have been incorporated in the curve, its 
apex would have been at 0, that is to say, that a unilateral 
or so-called half curve would have been the result. But 
the apex of the Scrophularia curve is at 12% and varies 
amongst the individual families from 8 to 20%, indi- 
cating thereby the possibility of an isolation of a separate 
apex for the tricotylous specimens. 
In other words, the tricotylous half race of Amaran- 
tus behaves, during this long period of selection, in es- 
sentially the same way as at the beginning (table p. 407), 
whereas Scrophularia behaves like other half races im- 
proved by selection. The scheme representing the in- 
fluence of selection on the half race of Ranunculus bid- 
bosus semiplcnus (Fig. 52 on page 252 of this volume) 
would apply equally to this case. 
The explanation of this difference in behavior is per- 
haps to be sought in the assumption that in Scrophularia 
tricotyly has only been indirectly improved. I am re- 
ferring to the case of Trifolium Incarnatum quadrifolium 
which I described above on page 239. In that case the 
smallest seeds gave rise to the largest number of aberrant 
seedlings. If the production of small seeds could be 
increased by selection the number of seedlings with com- 
pound primordial leaves would also be increased. It 
seems possible that in Scrophularia similar factors were at 
work, since the character involved need not, of course, 
necessarily be the size of the seeds. The repeated selec- 
tion of tricotyls would not, on this assumption lead di- 
