their testimony has
to be interpreted according to the laws of evidence. The past needs to
be reconstructed out of reports, as in geology or archaeology it needs to
be reconstructed out of stratifications and ruins. A man's memory or the
report in a newspaper is a fact justifying certain inferences about its
probable causes according to laws which such phenomena betray in the
present when they are closely scrutinised. This reconstruction is often
very difficult, and sometimes all that can be established in the end is
merely that the tradition before us is certainly false; somewhat as a
perplexed geologist might venture on no conclusion except that the state
of the earth's crust was once very different from what it is now.

[Sidenote: Verification here indirect.]

A natural science dealing with the past labours under the disadvantage
of not being able to appeal to experiment. The facts it terminates upon
cannot be recovered, so that they may verify in sense the hypothesis
that had inferred them. The hypothesis can be tested only by current
events; it is then turned back upon the past, to give assurance of facts
which themselves are hypothetical and remain hanging, as it were, to the
loose end of the hypothesis itself. A hypothetical fact is a most
dangerous creature, since it lives on the credit of a theory which in
turn would be bankrupt if the fact should fail. Inferred past facts are
more deceptive than facts prophesied, because while the risk of error in
the inference is the same, there is no possibility of discovering that
error; and the historian, while really as speculative as the prophet,
can never be found out.

Most facts known to man, however, are reached by inference, and their
reality may be wisely assumed so long as the principle by which they
are inferred, when it is applied in the present, finds complete and
constant verification. Presumptions involved in memory and tradition
give the first hypothetical facts we count upon; the relations which
these first facts betray supply the laws by which facts are to be
concatenated; and these laws may then be used to pass from the first
hypothetical facts to hypothetical facts of a second order, forming a
background and congruous extension to those originally assumed. This
expansion of discursive science can go on for ever, unless indeed the
principles of inference employed in it involve some present existence,
such as a skeleton in a given tomb, which direct exp