Neither at Ste-Suzanne nor at Bamburgh ([91]) is there existing stonework earlier than the twelfth century. Of the castle of Saint-Céneri-le-Gérei (Orne), which we know to have been fortified with stone walls before the end of the eleventh century,[111] only indistinguishable masses of masonry remain to-day. On the other hand, there are a certain number of castles on rocky and isolated sites, the walls of which may be fairly attributed, in whole or part, to the later half of the eleventh century. The most important example is Richmond castle in Yorkshire ([91]), on a high promontory of rock above the Swale. The shape of the enclosure is triangular. The most conspicuous feature of the castle is the splendid square tower or donjon, which was completed between 1170 and 1180, and stands on the north side of the enceinte, at the head of the approach from the town. The curtain, however, west of the donjon, contains “herring-bone” masonry,[112] and is of a rough construction which affords the greatest contrast to the regularly dressed and closely jointed masonry of the great tower. The tower, on three sides, forms an outward projection from the curtain, of great size and strength, and is a structure of one period from the ground upwards. But, on entering the castle, it is at once obvious that the lower part of the south wall of the tower is formed by part of the earlier curtain. In the middle of this section of the work is a wide doorway, with a round-headed arch of two unmoulded orders, which now forms an entrance into the basement of the tower. The capitals of the jamb-shafts of this archway are of an unmistakably eleventh century character, with volutes at the upper angles, and a row of acanthus leaves round the bell. This type of capital is seen in such buildings as the two abbey churches at Caen, the nave of Christchurch priory, the west front of Lincoln minster, and other fabrics completed before 1100, and is a sure guide to the date of the work in which it occurs. It would appear, then, that the masonry of this archway and much of the curtain is the work of Alan of Brittany, earl of Richmond, who certainly founded the castle, and died in 1088.[113] The castle contains more work of his date, of which something will be said in the sequel.

Ludlow; Entrance to inner ward

When the great tower of Richmond was built, an entrance was made on the first floor, from the rampart-walk of the curtain. It is quite clear that, up to this time, the archway just described had been the main entrance of the castle, and had probably been covered on the side next the town by a rectangular building, which formed the lower stage of a gateway-tower or gatehouse, lower than the present donjon. This is borne out by a comparison with the keep at Ludlow, where it is quite clear that an eleventh century gatehouse was converted at a later period into a keep, by walling up the outer entrance ([94]). A new entrance to the castle, as at Richmond, was made in the adjacent curtain, where it could be easily commanded by the tower. The date of the lowest stage of the donjon is revealed, as at Richmond, by the details of capitals and shafts, which in this case belong to an arcade in the east wall of the inner portion ([95]).[114]

Ludlow; Wall arcade in basement of great tower

Ludlow; Plan

The site of Ludlow ([96]), like that of Richmond, is a rocky peninsula, where a stone curtain, for which material existed on the spot, formed the obvious means of defence. There was no mount and no keep. Exeter, again, is an early example of a stone-walled castle upon a rocky site, where a gateway with a tower above formed the principal entrance. Such sites were protected naturally by the fall of the ground on the steeper sides; the side on which approach was possible was covered by a ditch cut in the rock. The ditch would be crossed by a drawbridge, let down from the inner edge, next the gateway. The gatehouse itself would be a building of two or more stages; at Ludlow the upper stage, as completed, was considerably loftier than the lower.[115] At Exeter there were probably three stages. A single upper stage remains at Tickhill. At Lewes and Porchester there is clear evidence of an upper chamber. The gatehouse at Porchester, as at Ludlow, was the entrance to an inner ward, divided by a ditch from the large outer ward,[116] which, at Porchester, represented the greater part of the enceinte of the Roman station, and contained the priory church and buildings. In these early gatehouses, the lower stage was closed at either end by a heavy wooden door, and was covered by a flat ceiling of timber. There was no arrangement for a portcullis. At Ludlow the lower stage appears to have been divided into an outer porch and inner hall by a cross wall, in which there must have been a door; but communication between these parts was also obtained by a narrow barrel-vaulted passage in the thickness of the east wall, which, opening from the outer division, turned at right angles to itself in the direction of the length of the wall, and, with another right-angled turn, opened into the inner hall ([95]). This passage was guarded by doors, which opened inwards at either end.[117] When the outer doorway of the gatehouse was blocked, the lower stage was covered in with a pointed barrel-vault.[118]