e word "correlation" must be
carefully distinguished; it would be a downright paralogism to adopt one
of them in the premisses of the reasoning, and the other in the
conclusion. And this is just what is done when the principle of
correlation is invoked in explanations of _detail_ in order to account
for complementary variations, and then correlation _in general_ is
spoken of as if it were any group of variations provoked by any
variation of the germ. Thus, the notion of correlation is first used in
current science as it might be used by an advocate of finality; it is
understood that this is only a convenient way of expressing oneself,
that one will correct it and fall back on pure mechanism when explaining
the nature of the principles and turning from science to philosophy. And
one does then come back to pure mechanism, but only by giving a new
meaning to the word "correlation"--a meaning which would now make
correlation inapplicable to the detail it is called upon to explain.

To sum up, if the accidental variations that bring about evolution are
insensible variations, some good genius must be appealed to--the genius
of the future species--in order to preserve and accumulate these
variations, for selection will not look after this. If, on the other
hand, the accidental variations are sudden, then, for the previous
function to go on or for a new function to take its place, all the
changes that have happened together must be complementary. So we have to
fall back on the good genius again, this time to obtain the
_convergence_ of _simultaneous_ changes, as before to be assured of the
_continuity of direction_ of _successive_ variations. But in neither
case can parallel development of the same complex structures on
independent lines of evolution be due to a mere accumulation of
accidental variations. So we come to the second of the two great
hypotheses we have to examine. Suppose the variations are due, not to
accidental and inner causes, but to the direct influence of outer
circumstances. Let us see what line we should have to take, on this
hypothesis, to account for the resemblance of eye-structure in two
series that are independent of each other from the phylogenetic point of
view.

Though molluscs and vertebrates have evolved separately, both have
remained exposed to the influence of light. And light is a physical
cause bringing forth certain definite effects. Acting in a continuous
way, it has been able to produce a