How strangely different from these wild figures are the sober facts of the case! The whole of the church fiefs, as we have seen, were only liable to find 784 knights, a number which, small as it was, just exceeded the entire knight service of Normandy as returned in 1171. As to the lay fiefs it is not possible to speak with equal confidence. I have ventured to fix the approximate _quota_ of 104 (more or less), of which ninety-two are in favour of my theory: forty-eight fiefs, of five knights and upwards, remain undetermined.[200] If the average of knights to a fief were the same in the latter as in the former class, the total contingents of the lay barons would amount, apparently, to 3,534 knights; but, as the latter one includes such enormous fiefs as those of Gloucester and of Clare, with such important honours as those of Peverel and Eye, we must increase our estimate accordingly, and must also make allowance for fiefs omitted and for those owing less than five knights (which are comparatively unimportant). Making, therefore, every allowance, we shall probably be safe in saying that the whole _servitium debitum_, clerical and lay, of England can scarcely have exceeded, if indeed it reached, 5,000 knights. Indefinite though such a result may seem, it is worth obtaining for the startling contrast which it presents to the 60,000 of Ordericus, to the 32,000 of Segrave,[201] and to the 30,000 of Gneist. The only writer, so far as I know, who has approximated, by investigating for himself, the true facts of the case, is Mr Pearson;[202] but his calculations, I fear, are vitiated by the unfortunate guess that the alleged 32,000 fees were really 6,400 of five hides each. It is a hopeless undertaking to reconcile the facts with the wild figures of mediæval historians by resorting to the ingenious devices of apocalyptic interpretation. V. THE NORMAL KNIGHT'S FEE Much labour has been vainly spent on attempts to determine the true area of a knight's fee. The general impression appears to be that it contained five hides. Mr Pearson, we have seen, based on that assumption his estimate of 6,400 fees, and other writers have treated the fee as the recognized equivalent of five hides. The point is of importance, because if we found that the recognized area of a knight's fee was five hides, it would give us a link between the under-tenant (_miles_) and the Anglo-Saxon thegn. But, as Dr Stubbs has recognized, the assumption cannot be maintained; no fixed number of hides constituted a knight's fee.