two different accumulations of an enormous number
of small causes is contrary to the principles of mechanistic philosophy.
We have concentrated the full force of our discussion upon an example
drawn from phylogenesis. But ontogenesis would have furnished us with
facts no less cogent. Every moment, right before our eyes, nature
arrives at identical results, in sometimes neighboring species, by
entirely different embryogenic processes. Observations of
"heteroblastia" have multiplied in late years,[36] and it has been
necessary to reject the almost classical theory of the specificity of
embryonic gills. Still keeping to our comparison between the eye of
vertebrates and that of molluscs, we may point out that the retina of
the vertebrate is produced by an expansion in the rudimentary brain of
the young embryo. It is a regular nervous centre which has moved toward
the periphery. In the mollusc, on the contrary, the retina is derived
from the ectoderm directly, and not indirectly by means of the embryonic
encephalon. Quite different, therefore, are the evolutionary processes
which lead, in man and in the Pecten, to the development of a like
retina. But, without going so far as to compare two organisms so distant
from each other, we might reach the same conclusion simply by looking at
certain very curious facts of regeneration in one and the same organism.
If the crystalline lens of a Triton be removed, it is regenerated by the
iris.[37] Now, the original lens was built out of the ectoderm, while
the iris is of mesodermic origin. What is more, in the _Salamandra
maculata_, if the lens be removed and the iris left, the regeneration of
the lens takes place at the upper part of the iris; but if this upper
part of the iris itself be taken away, the regeneration takes place in
the inner or retinal layer of the remaining region.[38] Thus, parts
differently situated, differently constituted, meant normally for
different functions, are capable of performing the same duties and even
of manufacturing, when necessary, the same pieces of the machine. Here
we have, indeed, the same effect obtained by different combinations of
causes.

Whether we will or no, we must appeal to some inner directing principle
in order to account for this convergence of effects. Such convergence
does not appear possible in the Darwinian, and especially the
neo-Darwinian, theory of insensible accidental variations, nor in the
hypothesis of sudden accidental var