m the first copies, good or bad, that came to hand, combined and
corrected at random. Editions of ancient texts are nowadays mostly
"critical;" but it is not yet thirty years since the publication of the
first "critical editions" of the great works of the middle ages, and the
critical text of some ancient classics (Pausanias, for example) has
still to be constructed.

Not all historical documents have as yet been published in a form
calculated to give historians the security they need, and some
historians still act as if they had not realised that an unsettled text,
as such, requires cautious handling. Still, considerable progress has
been made. From the experience accumulated by several generations of
scholars there has been evolved a recognised method of purifying and
restoring texts. No part of historical method has a more solid
foundation, or is more generally known. It is clearly explained in
several works of popular philology.[68] For this reason we shall here be
content to give a general view of its essential principles, and to
indicate its results.

I. We will suppose a document has not been edited in conformity with
critical rules. How are we to proceed in order to construct the best
possible text? Three cases present themselves.

(_a_) The most simple case is that in which we possess the original, the
author's autograph itself. There is then nothing to do but to reproduce
the text of it with absolute fidelity.[69] Theoretically nothing can be
easier; in practice this elementary operation demands a sustained
attention of which not every one is capable. If any one doubts it, let
him try. Copyists who never make mistakes and never allow their
attention to be distracted are rare even among scholars.

(_b_) Second case. The original has been lost; only a single copy of it
is known. It is necessary to be cautious, for the probability is that
this copy contains errors.

Texts degenerate in accordance with certain laws. A great deal of pains
has been taken to discover and classify the causes and the ordinary
forms of the differences which are observed between originals and
copies; and hence rules have been deduced which may be applied to the
conjectural restoration of those passages in a unique copy of a lost
original which are certainly corrupt (because unintelligible), or are so
in all probability.

Alterations of an original occurring in a copy--"traditional variants,"
as they are called--are due either to f