134 THE NATURE-STUDY OF PLANTS 
which they produced were the grandchildren of the 
original plant, and their leaves were distinctly nearer 
the type than those of their parent, the petals were 
somewhat larger and of a lighter pink, but the more 
conservative seeds were still without any threads. 
Their children, the third generation grown in 
moist soil, and the great grandchildren of the 1916 
plants from the shingle, have not yet (October, 1918) 
got beyond the autumn stage of a rosette of leaves. 
Although they are nearer the type than their 
parent and very much nearer than their grandparent 
and great-grandparent, these seedlings can be readily 
distinguished by their darker colour and by their 
more finely divided and less hairy leaves. 
The root system of all three generations exhibited 
a similar progressive reversion to the type, but in a 
much more marked degree, that of the great-grand- 
children being practically the same as in Fig. 35. 
Thus we can trace a gradual return to the typical 
moist hedgerow plant, most pronounced in the root, 
less so, although quite clear, in the leaves, but not 
nearly so marked in the petals, while the seeds 
remained unaltered for two generations. 
These observations lend some colour to the 
suggestion that the marked differences which occur 
in the extremes may be the result of growing upon the 
shingle generation after generation. 
It would be interesting, although the practical 
difficulties would be much greater, to make experi- 
ments in the opposite direction with a view to obtaining 
the shingle variety by breeding from the type, but there 
is no telling beforehand how many generations it would 
take to produce the desired result, if it could be done 
at all, though one is inclined to think that, given the 
