(6) Richborough, near Sandwich, the Roman Rutupiae: p. 24.
   (7) Dover, the Roman Dubra or Dubrae: p. 42.
   (8) Lympne, the Roman Portus Lemanae: p. 55.

   (9) Pevensey, in the marshes between Eastbourne and St. Leonards, once probably approached closely by the sea. The fort is oval in shape, and more than nine acres in extent. Its walls still stand 12 ft. thick and more than 25 ft. high. They consist of a concrete and rubble core, a facing of small courses of sandstone masonry, bonding tiles which never run right through from face to face, and two kinds of mortar, white in the interior and pink on the surface. Externally they are strengthened by fifteen or more circular bastions. The coins found here belong entirely to the period after A.D. 250. Saxon names, Andredsweald and the like, to say nothing of the Roman tiles stamped HON AVG ANDRIA from the site, support the idea that this plainly fourth-century fort is the Anderidos of the Saxon Shore.¹²

   (10) Porchester, planted at the water’s edge near the head of Portsmouth harbour. It is a nearly square fort, in area nine acres; its ramparts are of concrete-and-rubble core with flint facing and bonding courses of stone and tile, and are strengthened by some thirteen round projecting bastions. Few Roman remains have been found in it; but the four or five recorded coins belong to the fourth century. No clue exists as to its ancient name.¹³

   (11) Carisbrooke Castle, in the Isle of Wight, seems to cover the remains of yet another fort of this series. In 1927 it was found that the ramparts of the rectangular medieval bailey cover the walls and bastions of a fortification which cannot differ far in time from the Roman defences of the neighbouring coastline. Little is as yet known about it in detail, but there was a rather curious in-turned entrance in the middle of the eastern side. It would seem that the Romano-British population of the island, now represented by half-a-dozen villas and numerous slighter evidences, was sufficiently wealthy to claim a closer protection in the fourth century than the fort at Porchester was able to afford.¹⁴

   (12) Plainly to the same general phase belongs the late Roman fortress at Cardiff, of which considerable remains survive in a restored condition. This fortress, eight acres in extent, has walls 10 ft. thick, armed with polygonal bastions. It was, however, presumably a countermeasure to Irish rather than to Saxon attacks.¹⁵

   On the whole, this series agrees plainly enough with the list of the ‘Notitia.’ We cannot, indeed, trace by inscriptions any of the garrisons mentioned in that list. But the actual remains suffice. Of the nine ‘Notitia’ forts, eight can be identified, and the ninth, Portus Adurni, may be placed either at Porchester, or, if Felixstowe was a fort, at Felixstowe.^15a We can
   ¹²  C. R. Smith, Report on Pevensey (London, 1858). For the stamped tiles see Epham. Epig. ix,1281.
   ¹³  V.C.H. Hants, 1, 328. Horsley suggests that it was the Portus Adurni; cf. Sussex Arch. Coll. xxxviii, 217; ii, 99; lii, 83.
   ¹⁴  Antiquity, i, 476.
   ¹⁵  J.  Ward, Arch. lvii, 335; R. E. M. Wheeler, Ant.Journ.. ii, 361.
   ^15a   The river name Adur, near Shoreham in Sussex, is of modern origin, and gives no reason to place Portus Adurni there. See Haverfield, Proc. Soc.Antiq. xiv, 112I iz; Sussex Archaeoi. Collections, xxxviii, 2 17.