have only one, for the mere mechanical
comparison of the different readings is often enough to remove
obscurities which the uncertain light of conjectural criticism would
never have illuminated. However, an abundance of manuscripts is an
embarrassment rather than a help when the work of grouping them has been
left undone or done badly; nothing can be more unsatisfactory than the
arbitrary and hybrid restorations which are founded on copies whose
relations to each other and to the archetype have not been ascertained
beforehand. On the other hand, the application of rational methods
requires, in some cases, a formidable expenditure of time and labour.
Some works are preserved in hundreds of copies all differing from each
other; sometimes (as in the case of the Gospels) the variants of a text
of quite moderate extent are to be counted by thousands; several years
of assiduous labour are necessary for the preparation of a critical
edition of some mediaeval romances. And after all this labour, all these
collations and comparisons, can we be sure that the text of the romance
is sensibly better than it would have been if there had been only two or
three manuscripts to work upon? No. Some critical editions, owing to the
apparent wealth of material applicable to the work, demand a mechanical
effort which is altogether out of proportion to the positive results
which are its reward.

"Critical editions" founded on several copies of a lost original ought
to supply the public with the means of verifying the "_stemma codicum_"
which the editor has drawn up, and should give the rejected variants in
the notes. By this means competent readers are, at the worst, put in
possession, if not of the best possible text, at least of the materials
for constructing it.[74]

II. The results of textual criticism--a kind of cleaning and
mending--are purely negative. By the aid of conjecture, or by the aid of
conjecture and comparison combined, we are enabled to construct, not
necessarily a good text, but the best text possible, of documents whose
original is lost. What we thus effect is the elimination of corrupt and
adventitious readings likely to cause error, and the recognition of
suspected passages as such. But it is obvious that no new information is
supplied by this process. The text of a document which has been restored
at the cost of infinite pains is not worth more than that of a document
whose original has been preserved; on the contrary,