ference to the effect, so long as the later part of the process which
is the cause remains unchanged. Suppose, for example, that a man dies
of arsenic poisoning, we say that his taking arsenic was the cause
of death. But clearly the process by which he acquired the arsenic
is irrelevant: everything that happened before he swallowed it may be
ignored, since it cannot alter the effect except in so far as it alters
his condition at the moment of taking the dose. But we may go further:
swallowing arsenic is not really the proximate cause of death, since a
man might be shot through the head immediately after taking the dose,
and then it would not be of arsenic that he would die. The arsenic
produces certain physiological changes, which take a finite time before
they end in death. The earlier parts of these changes can be ruled out
in the same way as we can rule out the process by which the arsenic was
acquired. Proceeding in this way, we can shorten the process which we
are calling the cause more and more. Similarly we shall have to shorten
the effect. It may happen that immediately after the man's death his
body is blown to pieces by a bomb. We cannot say what will happen after
the man's death, through merely knowing that he has died as the result
of arsenic poisoning. Thus, if we are to take the cause as one event and
the effect as another, both must be shortened indefinitely. The result
is that we merely have, as the embodiment of our causal law, a certain
direction of change at each moment. Hence we are brought to differential
equations as embodying causal laws. A physical law does not say "A will
be followed by B," but tells us what acceleration a particle will have
under given circumstances, i.e. it tells us how the particle's motion is
changing at each moment, not where the particle will be at some future
moment.

* The theory of quanta suggests that the continuity is only
apparent. If so, we shall be able theoretically to reach
events which are not processes. But in what is directly
observable there is still apparent continuity, which
justifies the above remarks for the prevent.

Laws embodied in differential equations may possibly be exact,
but cannot be known to be so. All that we can know empirically is
approximate and liable to exceptions; the exact laws that are assumed in
physics are known to be somewhere near the truth, but are not known to
be true just as they stand. The laws that we