Only let it be understood that in
practice we do not postpone work upon documents till we shall have
gained a serene and absolute mastery over all the "auxiliary branches of
knowledge:" we should never dare to begin.

It remains to know how to treat documents supposing one has successfully
passed through the preliminary apprenticeship.

BOOK II

ANALYTICAL OPERATIONS

CHAPTER I

GENERAL CONDITIONS OF HISTORICAL KNOWLEDGE

We have already stated that history is studied from documents, and that
documents are the traces of past events.[54] This is the place to
indicate the consequences involved in this statement and this
definition.

Events can be empirically known in two ways only: by direct observation
while they are in progress; and indirectly, by the study of the traces
which they leave behind them. Take an earthquake, for example. I have a
direct knowledge of it if I am present when the phenomenon occurs; an
indirect knowledge if, without having been thus present, I observe its
physical effects (crevices, ruins), or if, after these effects have
disappeared, I read a description written by some one who has himself
witnessed the phenomenon or its effects. Now, the peculiarity of
"historical facts"[55] is this, that they are only known indirectly by
the help of their traces. Historical knowledge is essentially indirect
knowledge. The methods of historical science ought, therefore, to be
radically different from those of the direct sciences; that is to say,
of all the other sciences, except geology, which are founded on direct
observation. Historical science, whatever may be said,[56] is not a
science of observation at all.

The facts of the past are only known to us by the traces of them which
have been preserved. These traces, it is true, are directly observed by
the historian, but, after that, he has nothing more to observe; what
remains is the work of reasoning, in which he endeavours to infer, with
the greatest possible exactness, the facts from the traces. The document
is his starting-point, the fact his goal.[57] Between this
starting-point and this goal he has to pass through a complicated series
of inferences, closely interwoven with each other, in which there are
innumerable chances of error; while the least error, whether committed
at the beginning, middle, or end of the work, may vitiate all his
conclusions. The "historical," or indirect, method is thus obviously
inferior to the metho