the English conquest; nor is there known to be in Normandy any specimen of the polygonal or circular form of keep as early as that event. De Caumont, indeed, attributes the rectangular keep of Langeais, in which brick is largely used, to the year 992; but there is great reason to doubt this conclusion, and Du Pin and St. Laurent are probably among the oldest of this form, and do not seem to be earlier than the reign of Duke William; and this is true also of Arques and Nogent-le-Rotrou. In Normandy, as in England, the polygonal or shell keep, though on the older site, seems usually to be in masonry, which is the later construction; that of Gisors was built by Robert de Belesme in 1097, and that of Carentan at about the same time. Many even of the most considerable mounds, as Briquessart and Vieux-Conches, show no trace of masonry. The shell keep of Plessis-Grimoult was held by De Caumont to have been constructed before 1047; but if this be so, it is certainly a singular exception. Castle-building in Normandy seems to have preceded the English conquest, if at all, by but a very few years. The Romans left behind them in Britain many walled towns; but it is not known to what extent these defences were preserved by the Northmen, or in what condition they found them. At the conquest, Chester, Lincoln, Exeter, Hereford, Leicester, Oxford, Stafford, and Colchester, seem to have been already walled, and the walls of Exeter had been repaired or rebuilt by Æthelstan. Canterbury, Nottingham, and York were defended by a ditch. There were also probably some others, and possibly a few military towers in masonry of English workmanship; but there is no evidence of there having been anything like a rectangular keep, notwithstanding the special mention in 1052 of Richard’s Castle, the work of Richard, the son of Scrob. There is no reason to suppose that it possessed a tower of that character, which would have been quite out of keeping with the moated mound which even now marks the spot, and upon which the remains of the shell keep are still to be seen. Still less had the English any shell keeps constructed in masonry. What there really was in the way of military masonry and what was its character are not so clear. It was said of Dover, by William of Poictiers, that it was by Harold “studio atque sumptu suo communitum,” and that there were “item per diversa loca illius terræ alia castra ubi voluntas Ducis ea firmari jubet”; also in the account of the advance of William from Canterbury it is added, “Veniens ... ad fractam turrim castra metatus est,” pointing to a work in masonry, though no doubt it might, as at York, be Roman. Arundel, named in “Domesday” as having been a castle in the reign of the Confessor, was probably, from the size of its mound and the depth of its ditches, as strong as any castle of its type in Normandy; but no masonry has been observed there, either upon or about the mound, of a date earlier than the Conquest, if as early.