The argument rests on two propositions: one is general, and is derived from experience of human affairs; the other is particular, and is derived from the documents. In practice, we begin with the particular proposition, the historical fact: Salamis bears a Phoenician name. We then look for a general proposition: the language of the name of a city is the language of the people which founded it. And we conclude: Salamis, bearing a Phoenician name, was founded by the Phoenicians. In order that the conclusion may be certain, two conditions are necessary. (1) The general proposition must be accurately true; the two facts which it declares to be connected must be connected in such a way that the one is never found without the other. If this condition were completely satisfied we should have a _law_, in the scientific sense of the word; but in dealing with the facts of humanity--apart from those physical conditions whose laws are established by the regular sciences--we can only work with empirical laws obtained by rough determinations of general facts which are not analysed in such a manner as to educe their true causes. These empirical laws are approximately true only when they relate to a numerous body of facts, for we can never quite know how far each is necessary to produce the result. The proposition relating to the language of the name of a city does not go enough into detail to be always true. Petersburg is a German name, Syracuse in America bears a Greek name. Other conditions must be fulfilled before we can be sure that the name is connected with the nationality of the founders. We should, therefore, only employ such propositions as go into detail. (2) In order to employ a general proposition which goes into detail, we must have a detailed knowledge of the particular fact; for it is not till after this fact has been established that we look for an empirical general law on which to found an argument. We shall begin, then, by studying the particular conditions of the case (the situation of Salamis, the habits of the Greeks and Phoenicians); we shall not work on a single detail, but on an assemblage of details. Thus, in historical reasoning it is necessary to have (1) an accurate general proposition; (2) a detailed knowledge of a past fact. It is bad workmanship to assume a false general proposition--to suppose, for example, as Augustin Thierry did, that every aristocracy had its origin in a conquest. It is bad workmanship, again, to found an argument on an isolated detail (the name of a city). The nature of these errors indicates the precautions to be taken.