development, it is difficult to understand how an organ arrested at a very
early period of growth should acquire its full functional perfection;--how a
petal, supposed to be thus arrested, should acquire its brilliant colours, and
serve as an envelope to the flower, or a stamen produce efficient pollen; yet
this occurs with many peloric flowers. That pelorism is not due to mere chance
variability, but either to an arrest of development or to reversion, we may
infer from an observation made by Ch. Morren (13/70. In his discussion on some
curious peloric Calceolarias quoted in 'Journal of Horticulture' February 24,
1863 page 152.) namely, that families which have irregular flowers often
"return by these monstrous growths to their regular form; whilst we never see
a regular flower realise the structure of an irregular one."

Some flowers have almost certainly become more or less completely peloric
through reversion, as the following interesting case shows. Corydalis tuberosa
properly has one of its two nectaries colourless, destitute of nectar, only
half the size of the other, and therefore, to a certain extent, in a
rudimentary state; the pistil is curved towards the perfect nectary, and the
hood, formed of the inner petals, slips off the pistil and stamen in one
direction alone, so that, when a bee sucks the perfect nectary, the stigma and
stamens are exposed and rubbed against the insect's body. In several closely
allied genera, as in Dielytra, etc., there are two perfect nectaries, the
pistil is straight, and the hood slips off on either side, according as the
bee sucks either nectary. Now, I have examined several flowers of Corydalis
tuberosa, in which both nectaries were equally developed and contained nectar;
in this we see only the redevelopment of a partially aborted organ; but with
this redevelopment the pistil becomes straight, and the hood slips off in
either direction, so that these flowers have acquired the perfect structure,
so well adapted for insect agency, of Dielytra and its allies. We cannot
attribute these coadapted modifications to chance, or to correlated
variability; we must attribute them to reversion to a primordial condition of