Sometimes it is required to reproduce a work the author of which is dead, and the autograph manuscript of which cannot be sent to the printer. This was the case with the _Mémoires d'outre-tombe_ of Chateaubriand, for example; it is of daily occurrence in regard to the familiar correspondence of well-known persons which is printed in haste to satisfy the curiosity of the public, and of which the original manuscript is very fragile. First the text is copied; it is then set up by the compositor from the copy, which comes to the same thing as copying it again; this second copy is lastly, or ought to be, collated (in the proofs) with the first copy, or, better still, with the original, by some one who takes the place of the deceased author. The guarantees of accuracy are fewer in this case than in the first; for between the original and the ultimate reproduction there is one intermediary the more (the manuscript copy), and it may be that the original is hard for anybody but the author to decipher. And, in fact, the text of memoirs and posthumous correspondence is often disfigured by errors of transcription and punctuation occurring in editions which at first sight give the impression of having been carefully executed.[65] Turning now to ancient documents, let us ask in what state they have been preserved. In nearly every case the originals have been lost, and we have nothing but copies. Have these copies been made directly from the originals? No; they are copies of copies. The scribes who executed them were not by any means all of them capable and conscientious men; they often transcribed texts which they did not understand at all, or which they understood incorrectly, and it was not always the fashion, as it was in the time of the Carlovingian Renaissance, to compare the copies with the originals.[66] If our printed books, after the successive revisions of author and printer's reader, are still but imperfect reproductions, it is only to be expected that ancient documents, copied and recopied as they have been for centuries with very little care, and exposed at every fresh transcription to new risk of alteration, should have reached us full of inaccuracies. There is thus an obvious precaution to be taken. Before using a document we must find out whether its text is "sound"--that is, in as close agreement as possible with the original manuscript of the author; and when the text is "corrupt" we must emend it. In using a text which has been corrupted in transmission, we run the risk of attributing to the author what really comes from the copyists. There are actual cases of theories which were based on passages falsified in transmission, and which collapsed as soon as the true readings were discovered or restored. Printers' errors and mistakes in copying are not always innocuous or merely diverting; they are sometimes insidious and capable of misleading the reader.[67]