163
ovipositors of the tiny Pteromali ; while in attempting to breed
caterpillars taken in the field, the chance is so greatly against the
evolution of a butterfly that hymenopterists actually choose this
method of supplying their cabinets. ‘Of two hundred larvae of
Pieris Brassicae] Mr. Drewson, of Denmark, writes to me, £ I ob
tained only twenty pupae ; all the rest were attacked by Microgaster
glovieratus, and my own attempts with the larvae of Pyrameis
Atalanta, both in America and Europe, have been even more
unavailing. These caterpillars seem to be peripatetic banquetting
halls of Microgasters and Tachinae.”—(Nature. vol. Hi., ft. 147.)
Weismann says :—
“It is clear that in such animals as insects we can only speak
figuratively of normal death, if we mean by this an end which is
not due to accident. In these animals an accidental death is
the rule; and is therefore, strictly speaking, normal.”—{Essays
upon Heredity, p. 22.)
A sphere in which accidental death is normal can
scarcely be regarded as the most favourable for the illus
tration of the action of Natural Selection, for two reasons :
in the first place, because there will be little certainty as
to the survival of favourable variations ; and secondly,
because in some cases there might be a danger of exter
minating a group by selection applied after so great an
amount of accidental death. On this latter point Mr. J.
Jenner Weir says :—
“ It appears to me rather that as so few specimens become
imagines in proportion to the eggs produced, the more need is
there that these few should survive.”—(Nature. vol. Hi., ft. /66.)
We may regard the difficulty of “accidental” death from
another point of view. The analogy of transmutation by
Artificial Selection, or, in other words, by the isolation of
similar variants from among the variations necessarily con
nected with reproduction, teaches us that there must be
one object to be attained, and one process of selection
strictly dominated by that object. Therefore, if Natural
Selection is to work in nature, there must be one risk, one