The crudest analytical procedure consists in rejecting those details in the legendary narrative which appear impossible, miraculous, contradictory, or absurd, and retaining the rational residue as historical. This is how the Protestant rationalists of the eighteenth century treated biblical narratives. One might as well amputate the marvellous part of a fairy tale, suppress Puss in Boots, and keep the Marquis of Carabas as an historical character. A more refined but no less dangerous method is to compare different legends in order to deduce their common historical basis. Grote[162] has shown, with reference to Greek tradition, that it is impossible to extract any trustworthy information from legend by any process whatever.[163] We must make up our minds to treat legend as a product of imagination; we may look in it for a people's conceptions, not for the external facts in that people's history. The rule will be to reject every statement of legendary origin; nor does this apply only to narratives in legendary form: a narrative which has an historical appearance, but is founded on the data of legend, the opening chapters of Thucydides for example, ought equally to be discarded. In the case of written transmission it remains to inquire whether the author reproduced his source without altering it. This inquiry forms part of the critical investigation of the sources,[164] so far as it can be pursued by a comparison of texts. But when the source has disappeared we are reduced to internal criticism. We ask, first of all, whether the author can have had exact information, otherwise his statement is valueless. We next put to ourselves the general question: Was the author in the habit of altering his sources, and in what manner? And in regard to each separate second-hand statement we ask whether it has the appearance of being an exact reproduction or an arrangement. We judge by the form: when we meet with a passage whose style is out of harmony with the main body of the composition, we have before us a fragment of an earlier document; the more servile the reproduction the more valuable is the passage, for it can contain no exact information beyond what was already in the source. VII. In spite of all these investigations, criticism never succeeds in determining the parentage of all the statements to the extent of finding out who it was that observed, or even recorded, each fact. In most cases the inquiry ends in leaving the statement anonymous. We are thus confronted with a fact, observed we know not by whom nor how, recorded we know not when nor how. No other science accepts facts which come in such a condition, without possibility of verification, subject to incalculable chances of error. But history can turn them to account, because it does not, like the other sciences, need a supply of facts which are difficult to ascertain.