ubt that numerous texts which
have been preserved, in corrupt form, in unique copies, have resisted,
and will continue to resist, the efforts of criticism. Very often
criticism ascertains the fact of the text having been altered, states
what the sense requires, and then prudently stops, every trace of the
original reading having been obscured by a confused tangle of successive
corrections and errors which it is hopeless to attempt to unravel. The
scholars who devote themselves to the fascinating pursuit of conjectural
criticism are liable, in their ardour, to suspect perfectly innocent
readings, and, in desperate passages, to propose adventurous hypotheses.
They are well aware of this, and therefore make it a rule to draw a very
clear distinction, in their editions, between readings found in
manuscripts and their own restorations of the text.

(_c_) Third case. We possess several copies, which differ from each
other, of a document whose original is lost. Here modern scholars have a
marked advantage over their predecessors: besides being better informed,
they set about the comparison of copies more methodically. The object
is, as in the preceding case, to reconstruct the archetype as exactly as
possible.

The scholars of earlier days had to struggle, as novices have to
struggle now, in a case of this kind, against a very natural and a very
reprehensible impulse--to use the first copy that comes to hand,
whatever its character may happen to be. The second impulse is not much
better--to use the oldest copy out of several of different date. In
theory, and very often in practice, the relative age of the copies is of
no importance; a sixteenth-century manuscript which reproduces a good
lost copy of the eleventh century is much more valuable than a faulty
and retouched copy made in the twelfth or thirteenth century. The third
impulse is still far from being good; it is to count the attested
readings and decide by the majority. Suppose there are twenty copies of
a text; the reading A is attested eighteen times, the reading B twice.
To make this a reason for choosing A is to make the gratuitous
assumption that all the manuscripts have the same authority. This is an
error of judgment; for if seventeen of the eighteen manuscripts which
give the reading A have been copied from the eighteenth, the reading A
is in reality attested only once; and the only question is whether it is
intrinsically better or worse than the reading B.