each proved imperfectly, yields a kind of certainty; the facts do not, in the strict sense of the word, prove, but they confirm each other. The doubt which attached to each one of them disappears; we obtain that species of certainty which is produced by the interconnection of facts. Thus the comparison of conclusions which are separately doubtful yields a whole which is morally certain. In an itinerary of a sovereign, the days and the places confirm each other when they harmonize so as to form a coherent whole. An institution or a popular usage is established by the harmony of accounts, each of which is no more than probable, relating to different times and places.
This method is a difficult one to apply. The notion of harmony is a much vaguer one than that of agreement. We cannot assign any precise general rules for distinguishing facts which are sufficiently connected to form a whole, the harmony of whose parts would be conclusive; nor can we determine beforehand the duration and extent of that which may be taken to form a whole. Facts separated by half a century of time and a hundred leagues of space may confirm each other in such a way as to establish a popular usage (for example, among the ancient Germans); but they would prove nothing if they were taken from a heterogeneous society subject to rapid evolution (take, for example, French society in 1750, and again in 1800, in Alsace and in Provence). Here we have to study the relation between the facts. This brings us to the beginnings of historical construction; here is the transition from analytical to synthetic operations.
VII. But it remains to consider cases of discordance