But in proportion as a body of knowledge is more nearly concerned with the invisible facts of the mind, its notions become more confused and its language less precise. Even the most ordinary facts of human life, social conditions, actions, motives, feelings, can only be expressed by vague terms (_king_, _warrior_, _to fight_, _to elect_). In the case of more complex phenomena, language is so indefinite that there is no agreement even as to the essential elements of the phenomena. What are we to understand by a tribe, an army, an industry, a market, a revolution? Here history shares the vagueness common to all the sciences of humanity, psychological or social. But its indirect method of representation by mental images renders this vagueness still more dangerous. The historical images in our minds ought, then, to reproduce at least the essential features of the images which were in the minds of the direct observers of past facts; but the terms in which they expressed their mental images never tell us exactly what these essential elements were. Facts which we did not see, described in language which does not permit us to represent them in our minds with exactness, form the data of history. The historian, however, is obliged to picture the facts in his imagination, and he should make it his constant endeavour to construct his mental images out of none but correct elements, so that he may imagine the facts as he would have seen them if he had been able to observe them personally.[183] But the formation of a mental image requires more elements than the documents supply. Let any one endeavour to form a mental representation of a battle or a ceremony out of the data of a narrative, however detailed; he will see how many features he is compelled to add. This necessity becomes physically perceptible in attempts to restore monuments in accordance with descriptions (for example, the Temple at Jerusalem), in pictures which claim to be representations of historical scenes, in the drawings of illustrated newspapers. Every historical image contains a large part of fancy. The historian cannot get rid of it, but he can take stock of the real elements which enter into his images and confine his constructions to these; they are the elements which he has derived from the documents. If, in order to understand the battle between Cæsar and Ariovistus, he finds it necessary to make a mental picture of the two opposing armies, he will be careful to draw no conclusions from the general aspect under which he imagines them; he will base his reasonings exclusively on the real details furnished by the documents.