her rare cases in which similar hypotheses have had to be set
up, experiment has always in the long run enabled us to discover some
phenomenon which had escaped the first observers and which corresponds
exactly to the variation of energy first made evident.

One difficulty, however, arises from the fact that the principle ought
only to be applied to an isolated system. Whether we imagine actions
at a distance or believe in intermediate media, we must always
recognise that there exist no bodies in the world incapable of acting
on each other, and we can never affirm that some modification in the
energy of a given place may not have its echo in some unknown spot
afar off. This difficulty may sometimes render the value of the
principle rather illusory.

Similarly, it behoves us not to receive without a certain distrust the
extension by certain philosophers to the whole Universe, of a property
demonstrated for those restricted systems which observation can alone
reach. We know nothing of the Universe as a whole, and every
generalization of this kind outruns in a singular fashion the limit of
experiment.

Even reduced to the most modest proportions, the principle of the
conservation of energy retains, nevertheless, a paramount importance;
and it still preserves, if you will, a high philosophical value. M.J.
Perrin justly points out that it gives us a form under which we are
experimentally able to grasp causality, and that it teaches us that a
result has to be purchased at the cost of a determined effort.

We can, in fact, with M. Perrin and M. Langevin, represent this in a
way which puts this characteristic in evidence by enunciating it as
follows: "If at the cost of a change C we can obtain a change K, there
will never be acquired at the same cost, whatever the mechanism
employed, first the change K and in addition some other change, unless
this latter be one that is otherwise known to cost nothing to produce
or to destroy." If, for instance, the fall of a weight can be
accompanied, without anything else being produced, by another
transformation--the melting of a certain mass of ice, for example--it
will be impossible, no matter how you set about it or whatever the
mechanism used, to associate this same transformation with the melting
of another weight of ice.

We can thus, in the transformation in question, obtain an appropriate
number which will sum up that which may be expected from the external
effect, and can give