e same cells after modification, naturally go on increasing under the
same favouring conditions, until at last they become sufficiently numerous
to overpower and supplant the old gemmules.

Another difficulty may be here noticed; we have seen that {396} there is an
important difference in the frequency, though not in the nature, of the
variations in plants propagated by sexual and asexual generation. As far as
variability depends on the imperfect action of the reproductive organs
under changed conditions, we can at once see why seedlings should be far
more variable than plants propagated by buds. We know that extremely slight
causes,--for instance, whether a tree has been grafted or grows on its own
stock, the position of the seeds within the capsule, and of the flowers on
the spike,--sometimes suffice to determine the variation of a plant, when
raised from seed. Now, it is probable, as explained when discussing
alternate generation, that a bud is formed of a portion of already
differentiated tissue; consequently an organism thus formed does not pass
through the earlier phases of development, and cannot be so freely exposed,
at the age when its structure would be most readily modified, to the
various causes inducing variability; but it is very doubtful whether this
is a sufficient explanation of the difficulty.

With respect to the tendency to reversion, there is a similar difference
between plants propagated from buds and seed. Many varieties, whether
originally produced from seed or buds, can be securely propagated by buds,
but generally or invariably revert by seed. So, also, hybridised plants can
be multiplied to any extent by buds, but are continually liable to
reversion by seed,--that is, to the loss of their hybrid or intermediate
character. I can offer no satisfactory explanation of this fact. Here is a
still more perplexing case: certain plants with variegated leaves, phloxes
with striped flowers, barberries with seedless fruit, can all be securely
propagated by the buds on cuttings; but the buds developed from the roots
of these cuttings almost invariably lose their character and revert to
their former condition.

Finally, we can see on the hypothesis of pangenesis that variability
depends on at least two distinct groups of causes. Firstly, on the
deficiency, superabundance, fusion, and transposition of gemmules, and on
the redevelopment of those which have long been dormant. In these cases the
gemmules thems