The practice of the established sciences teaches us the conditions of an exact knowledge of facts. There is only one scientific procedure for gaining knowledge of a fact, namely, _observation_; every statement, therefore, must rest, directly or indirectly, upon an observation, and this observation must have been made correctly. The set of questions by the aid of which we investigate the probabilities of error may be drawn up in the light of experience, which brings before us the most common cases of error. (1) The author was in a situation to observe the fact, and supposed he really had observed it; he was, however, prevented from doing so by some interior force of which he was unconscious, an hallucination, an illusion, or a mere prejudice. It would be useless, as well as impossible, to determine which of these agencies was at work; it is enough to ascertain whether the author had a tendency to observe badly. It is scarcely possible in the case of a particular statement to recognise that it was the result of an hallucination or an illusion. At the most we may learn, either from information derived from other sources or by comparison, that an author had a _general_ propensity to this kind of error. There is a better chance of recognising whether a statement was due to prejudice. In the life or the works of an author we may find the traces of his dominant prejudices. With reference to each of his particular statements, we ought to ask whether it is not the result of a preconceived idea of the author on a class of men or a kind of facts. This inquiry partly coincides with the search for motives of falsehood: interest, vanity, sympathy, and antipathy give rise to prejudices which alter the truth in the same manner as wilful falsehood. We therefore employ the questions already formulated for the purpose of testing good faith. But there is one to be added. In putting forward a statement has the author been led to distort it unconsciously by the circumstance that he was answering a question? This is the case of all statements obtained by interrogating witnesses. Even apart from the cases where the person interrogated seeks to please the proposer of the question by giving an answer which he thinks will be agreeable to him, every question suggests its own answer, or at least its form, and this form is dictated beforehand by some one unacquainted with the facts. It is therefore necessary to apply a special criticism to every statement obtained by interrogation; we must ask what was the question put, and what were the preconceptions to which it may have given rise in the mind of the person interrogated.