therefore proceed with confidence to describe from existing remains the characteristics of the forts of the Litus Saxonicum. In shape and size they are not uniform; they vary from rectangular to oval, and from five to nine or ten acres. Their fortifications conform more nearly to one type. The walls usually have rubble-and-concrete cores, facings of flint or of stone in small regular courses,¹⁶ bonding courses of tile or stone, and slight and shallow foundations; they are thick and high, pierced by few and variously constructed gateways, and defended externally by projecting towers or bastions, usually of circular form. They betoken a period when the defensive was all-important, and in this point, just as in their lack of uniformity and in various details, they exemplify the fashions of late Imperial fortification.¹⁷
   The forts were meant for small garrisons of auxiliaries; of those mentioned in the ‘Notitia’, Richborough alone had legionaries, and probably not more than 1,000 of these. But the numbers of the troops and the internal arrangements are otherwise unknown. Except at Richborough and Lympne, hardly any Roman building has been found within their walls, and the remains in those two forts tell us little. Of barracks, such as appear in the fourth-century frontier forts of Mocsia and Arabia, the Saxon Shore shows no single vestige.
   The positions of the individual forts seem to have been chosen for various reasons.¹⁸  In Kent the desire is plain on the one hand to maintain communications with the Continent and to hold the Straits of Dover, which must be a strategic point in any British naval war, and on the other hand to block the navigable arm of the sea which then made Thanet an island, and to protect the dwellers in east Kent. On the East Anglian and southern coasts we trace rather an attempt to watch those entrances by which barbarian raiders have so often reached the interior of Britain. Thus the Wash is guarded by Brancaster, the waterways behind Yarmouth by Burgh Castle, the estuaries below Colchester by Bradwell, which incidentally protected Camulodunum itself, and the creeks of south-eastern Hampshire by Porchester. But the actual sites may suggest a further reason for their choice. Every fort—except Felixstowe, if that was a fort—stands on the shore of some inlet or harbour. Where, as at Richborough or Burgh Castle, the main part of the fort crowns a cliff high above the water, the walls on the water-side were in all probability carried down to the shore, or as near to it as the lie of the land would permit. The significance of this is plain. Whatever became of the classis Britannica in the fourth century, these forts were evidently connected with ships. One fort, indeed, has no other obvious reason for existing. Pevensey had nothing to defend. It was a good landing-place for invaders. But it had no attraction for pirates. North and north-east of it stretched the dense thickets of the Weald, west the lonely downs of Lewes, bare save for a scattered peasantry. There was no inlet here leading far into the interior, and save for one villa on
   ¹⁶ A curious feature in the facing of some of the masonry is a use of light or dark coloured stones to produce the effect of ornament, as at Richborough (p. 30). This may be copied from Roman brickwork: compare the bastion at Cologne. (See also below, pp. 74, 75, etc.)
   ¹⁷ Fox, Arch. Journ. liii, 373, argues that the absence of uniformity in the shapes of the forts points to different dates of building. Richborough (he says), being rectangular, was built before Pevensey and Lympne, which are irregular-shaped. It is doubtful if this view can be maintained. It would seem that in the fourth and fifth centuries, rectangular and other shapes were used simultaneously and almost arbitrarily. Note the want of uniformity in shape and detail of the fourth-century frontier forts on the upper Rhine attributed to about A.D. 285—300: Wesideutsche Zeitschrift, xxv, 169, and plate 3.
   ¹⁸ Lewin’s account of these, Arch. xl, 362, is somewhat one-sided and theoretical.