esent danger of substituting
one's personal impressions for the text there is only one real
safeguard; it should be made an invariable rule never on any account to
make an extract from a document, or a partial analysis of it, without
having _first_ made a comprehensive analysis[134] of it mentally, if not
on paper.

To analyse a document is to discern and isolate all the ideas expressed
by the author. Analysis thus reduces to _interpretative criticism_.

Interpretation passes through two stages: the first is concerned with
the literal, the second with the real meaning.

III. The determination of the literal meaning of a document is a
linguistic operation; accordingly, Philology (in the narrow sense) has
been reckoned among the auxiliary sciences of history. To understand a
text it is first necessary to know the language. But a _general_
knowledge of the language is not enough. In order to interpret Gregory
of Tours, it is not enough to know Latin in a general way; it is
necessary to add a special study of the particular kind of Latin written
by Gregory of Tours.

The natural tendency is to attribute the same meaning to the same word
wherever it occurs. We instinctively treat a language as if it were a
fixed system of signs. Fixity, indeed, is a characteristic of the signs
which have been expressly invented for scientific use, such as
algebraical notation or the nomenclature of chemistry. Here every
expression has a single precise meaning, which is absolute and
invariable; it expresses an accurately analysed and defined idea, only
one such idea, and that always the same in whatever context the
expression may occur, and by whatever author it may be used. But
ordinary language, in which documents are written, fluctuates: each word
expresses a complex and ill-defined idea; its meanings are manifold,
relative, and variable; the same word may stand for several different
things, and is used in different senses by the same author according to
the context; lastly, the meaning of a word varies from author to author,
and is modified in the course of time. _Vel_, which in classical Latin
only has the meanings _or_ and _even_, means _and_ in certain epochs of
the middle ages; _suffragium_, which is classical Latin for _suffrage_,
takes in mediaeval Latin the sense of _help_. We have, then, to learn to
resist the instinct which leads us to explain all the expressions of a
text by their classical or ordinary meanings. The gramma