[127] Even the specialists in external criticism themselves, when they do not take the line of despising all synthesis _a priori_, are almost as easily dazzled as anybody else by incorrect syntheses, by a show of "general ideas," or by literary artifices, in spite of their clear-sightedness where works of critical scholarship are concerned. [128] The sciences of observation do, however, need a species of criticism. We do not accept without verification results obtained by anybody, but only results obtained by those who know how to work. But this criticism is made once for all, and applies to the author, not to his works; historical criticism, on the contrary, is obliged to deal separately with every part of a document. [129] Cf. _supra_, book ii. chap. i. p. 67. [130] Cf. _supra_, p. 122. [131] Taine appears to have proceeded thus in vol. ii, _La Révolution_, of his _Origines de la France contemporaine_. He had made extracts from unpublished documents and inserted a great number of them in his work, but it would seem that he did not first methodically analyse them in order to determine their meaning. [132] Fustel de Coulanges explains very clearly the danger of this _method_: "Some students begin by forming an opinion ... and it is not till afterwards that they begin to read the texts. They run a great risk of not understanding them at all, or of understanding them wrongly. What happens is that a kind of tacit contest goes on between the text and the preconceived opinions of the reader; the mind refuses to grasp what is contrary to its idea, and the issue of the contest commonly is, not that the mind surrenders to the evidence of the text, but that the text yields, bends, and accommodates itself to the preconceived opinion.... To bring one's personal ideas into the study of texts is the subjective method. A man thinks he is contemplating an object, and it is his own idea that he is contemplating. He thinks he is observing a fact, and the fact at once assumes the colour, and the significance his mind wishes it to have. He thinks he is reading a text, and the words of the text take a particular meaning to suit a ready-made opinion. It is this subjective method which has done most harm to the history of the Merovingian epoch.... To read the texts was not enough; what was required was to read them before forming any convictions...." (_Monarchie franque_, p. 31). For the same reason Fustel de Coulanges deprecated the reading of one document in the light of another; he protested against the custom of explaining the _Germania_ of Tacitus by the barbaric laws. In the _Revue des questions historiques_, 1897, vol. i, a lesson on method, _De l'analyse des textes historiques_, is given apropos of a commentary by M. Monod on Gregory of Tours: "The historian ought to begin his work with an exact analysis of each document.... The analysis of a text ... consists in determining the sense of each word and eliciting the true meaning of the writer....