It must be confessed that this list of Danish fortresses furnishes us with a very slender basis for generalisation as to the nature of Danish fortifications, judging from the actual remains. All we can say is that in six cases out of twenty-four (not including Tempsford or Fulham) the work appears to have been rectangular. In the case of Shoebury, about which we have the best evidence, the imitation of Roman models seems to be clear. If we turn from remaining facts to _à priori_ likelihoods, we call to mind that the Danes were a much-travelled people, had been in Gaul as well as in England, and had had opportunities of observing Roman fortifications, as well as much practice both in the assault and defence of fortified places. It may not be without significance that it is not until after the return of "the army" from France that we hear of their building camps at all, except in the case of Reading. As far as our information goes, their camps were without citadels. What evidence we have from the other side of the channel supports the same conclusion. Richer gives us an account of the storming of a fortress of the Northmen at Eu, by King Raoul, in 925, from which it is clear that as soon as the king's soldiers had got over the vallum, they were masters of the place; there was no citadel to attack.[139] Dudo speaks of the Vikings "fortifying themselves, after the manner of a _castrum_, by heaped up earth-banks drawn round themselves," and it is clear from the rest of his description that the camp had no citadel.[140] In no case do we find anything to justify the theory that mottes were an accompaniment of Danish camps. In five cases out of the twenty-four there are or were mottes at the places mentioned, but in all cases they belonged to Norman castles. The magnificent motte called the Castle Hill at Thetford was on the opposite side of the river to the borough, which we have seen reason to think was the site of the Danish winter quarters. Torkesey in Leland's time had by the river side "a Hille of Yerth cast up," which he judged to be the donjon of some old castle, probably rightly, though we have been unable as yet to find any mention of a Norman castle at Torkesey; a brick castle of much more recent date is still standing near the river, and probably the motte to which Leland alludes was destroyed when this was built. The motte at Cambridge is placed inside the original bounds of the borough, and was part of the Norman castle.[141] We have already dealt with the Boley Hill at Rochester, and shall have more to say about it hereafter. The rock motte at Nottingham was probably not cut off by a ditch from the rest of the headland until the Norman castle was built. [Illustration: FIG. 7. Willington, Beds.]