The mode of construction cannot be regulated by the ideal plan of the science we desire to construct; it depends on the materials we have at our disposal. It would be chimerical to formulate a scheme which the materials would not allow us to carry out; it would be like proposing to construct an Eiffel tower with building-stones. The fundamental defect of philosophies of history is that they forget this practical necessity. I. Let us begin by considering the materials of history. What is their form and their nature? How do they differ from the materials of other sciences? Historical facts are derived from the critical analysis of the documents. They issue from this process in the form to which analysis has reduced them, chopped small into individual statements; for a single sentence contains several statements: we have often accepted some and rejected others; each of these statements represents a fact. Historical facts have the common characteristic of having been taken from documents; but they differ greatly among themselves. (1) They represent phenomena of very different nature. From the same document we derive facts bearing on handwriting, language, style, doctrines, customs, events. The Mesha inscription furnishes facts bearing on Moabite handwriting and language, the belief in the god Chemosh, the practices belonging to his cult, the war between the Moabites and Israel. Thus the facts reach us pell-mell, without distinction of nature. This mixture of heterogeneous facts is one of the characteristics which differentiate history from the other sciences. The sciences of direct observation choose the facts to be studied, and systematically limit themselves to the observation of facts of a single species. The documentary sciences receive the facts, already observed, at the hands of authors of documents, who supply them in disorder. For the purpose of remedying this disorder it is necessary to sort the facts and group them by species. But, for the purpose of sorting them, it is necessary to know precisely what it is that constitutes a _species_ of historical facts; in order to group them we need a principle of classification applicable to them. But on these two questions of capital importance historians have not as yet succeeded in formulating precise rules.