he 'critical' problem. Hence they taught dogmatically as true a
theory of scientific method which Hume himself had elaborately proved
impossible. It was just because Hume had seen so clearly that no
universal scientific truths can be derived from premisses which merely
record particular facts that he professed himself a follower of the
'academic' or 'sceptical' philosophy. He recognized the impossibility of
constructing scientific knowledge out of its material constituent alone,
but did not see where the formal constituent could come from, and so
resigned himself to regarding the actual successes of science as a kind
of standing miracle.

The men of the 'seventies were, after all, in many cases more anxious to
damage theology than to build up Philosophy. They read Hume without any
delicate sense for his urbane ironies, and believed in good faith that
he and John Stuart Mill between them had shown that by a mysterious
process called 'induction' it is possible to prove rigorously universal
conclusions in science without universal premisses. A scientific law,
according to them, is only a convenient short-hand notation in which to
register the 'routine of our perceptions'. Thus we have known of a great
many men who have died, and have never known of any man who lived to
much over a hundred without dying. The universal proposition 'all men
are mortal' is a short expression for this information, and it is
nothing more. It ought to have been obvious that, if this is a true
account of science, all scientific 'generalizations' are infinitely
improbable. The number of men of whom we _know_ that they have died is
insignificant by comparison with the multitude of those who have lived,
are living, or will live, and we have no guarantee that this
insignificant number is a fair average sample. So again, unless there
are true universal propositions which are not 'short-hand' for any
plurality of observed facts whatever, we cannot with any confidence,
however faint, infer that a 'regular sequence' or 'routine' which has
been observed from the dawn of recorded time up to, say, midnight,
August 4, 1919, will continue to be observed on August 5, 1919. How,
except by relying on the truth of some principle which does not depend
itself on the validity of 'generalization', can we tell that it is even
slightly probable that the nature of things will not change suddenly at
the moment of midnight between August 4 and August 5, 1919? What is
calle