IV. How, then, is it possible to imagine facts without their being wholly imaginary? The facts, as they exist in the historian's mind, are necessarily subjective; that is one of the reasons given for refusing to recognise history as a science. But subjective is not a synonym of unreal. A recollection is only an image; but it is not therefore a chimera, it is the representation of a vanished reality. It is true that the historian who works with documents has no personal recollections of which he can make direct use; but he forms mental images on the model of his own recollections. He assumes that realities (objects, actions, motives), which have now disappeared, but were formerly observed by the authors of the documents, resembled the realities of his own day which he has himself seen and which he retains in his memory. This is the postulate of all the documentary sciences. If former humanity did not resemble the humanity of to-day, documents would be unintelligible. Starting from this assumed resemblance, the historian forms a mental representation of the bygone facts of history similar to his own recollection of the facts he has witnessed. This operation, which is performed unconsciously, is one of the principal sources of error in history. The things of the past which are to be pictured in imagination were not wholly similar to the things of the present which we have seen; we have never seen a man like Cæsar or Clovis, and we have never experienced the same mental states as they. In the established sciences it is equally true that one man will work on facts which another has observed, and which he must therefore represent to himself by analogy; but these facts are defined by precise terms which indicate what invariable elements ought to appear in the image. Even in physiology the notions which occur are sufficiently clear and fixed for the same word to evoke in the minds of all naturalists similar images of an organ or a movement. The reason is that each notion which has a name has been formed by a method of observation and abstraction in the course of which all the characteristics which belong to the notion have been precisely determined and described.