where omission would suppress some characteristic feature.
This operation, which is difficult in itself, is still further complicated by the state in which the facts which are to be condensed into formulæ present themselves. According to the nature of the documents from which they are derived, they come to us in all the different degrees of precision: from the detailed narrative which relates the smallest episodes (the battle of Waterloo) down to the barest mention in a couple of words (the victory of the Austrasians at Testry). On different facts of the same kind we possess an amount of details which is infinitely variable according as the documents give us a complete description or a mere mention. How are we to organise into a common whole, items of knowledge which differ so widely in point of precision? When facts are known to us from a vague word of general import, we cannot reduce them to a less degree of generality and a greater degree of precision; we do not know the details. If we add them conjecturally we shall produce an historical novel. This is what Augustin Thierry did in the case of his Récits mérovingiens. When facts are known in detail, it is always easy to reduce them to a greater degree of generality by suppressing characteristic details; this is what is done by the authors of abridgements. But the result of this procedure would be to reduce history to a mass of vague generalities, uniform for the whole of time except for the proper names and the dates. It would be a dangerous method of introducing symmetry, to bring all facts to a common degree of generality by levelling them all to the condition