In the sciences of direct observation, when a fact is missing from a series, it is sought for by a new observation. In history, where we have not this resource, we seek to extend our knowledge by the help of reasoning. Starting from facts known to us from the documents, we endeavour to reach new facts by inference. If the reasoning be correct, this method of acquiring knowledge is legitimate. But experience shows that of all the methods of acquiring historical knowledge, reasoning is the most difficult to employ correctly, and the one which has introduced the most serious errors. It should not be used without the safeguard of a number of precautions calculated to keep the danger continually before the mind. (1) Reasoning should never be combined with the analysis of a document. The reader who allows himself to introduce into a text what the author has not expressly put there ends by making him say what he never intended to say.[199] (2) Facts obtained by the direct examination of documents should never be confused with the results obtained by reasoning. When we state a fact known to us by reasoning only, we must not allow it to be supposed that we have found it in the documents; we must disclose the method by which we have obtained it. (3) Unconscious reasoning must never be allowed; there are too many chances of error. It will be enough to make a point of putting every argument into logical form; in the case of bad reasoning the major premiss is generally monstrous to an appalling degree. (4) If the reasoning leaves the least doubt, no attempt must be made to draw a conclusion; the point treated must be left in the conjectural stage, clearly distinguished from the definitively established results. (5) It is not permissible to return to a conjecture and endeavour to transform it into a certainty. Here the first impression is most likely to be right. By reflection upon a conjecture we familiarise ourselves with it, and end by thinking it better established; while the truth is, we are merely more accustomed to it. This is a frequent mishap with those who devote themselves to long meditation on a small number of texts. There are two ways of employing reasoning, one negative, the other positive; we shall examine them separately. II. The negative mode of reasoning, called also the "argument from silence," is based on the absence of indications with regard to a fact.[200] From the circumstance of the fact not being mentioned in any document it is inferred that there was no such fact; the argument is applied to all kinds of subjects, usages of every description, evolutions, events. It rests on a feeling which in ordinary life is expressed by saying: "If it were true, we should have heard of it;" it implies a general proposition which may be formulated thus: "If an alleged event really had occurred, there would be some document in existence in which it would be referred to."