Whether or no the defences of the fortress followed in detail the lines which we have tried to visualize for them, the evidence is at least adequate to show that Dover of the Saxon Shore lay in or about the Market Square, on the western bank of the Dour. So far we have depended solely Upon archaeological evidence—mostly evidence recovered by a single observer within the last dozen years. But Mr. A. W. Clapham has pointed out^83a that there is one particle of documentary evidence which is relevant to our problem and happily confirms our general conclusions. It is said that Eadbald, King of Kent, before 640, founded a monastery ‘in the castle of Dover,’ and it has been assumed that this monastery is represented by the surviving Saxon church of St. Mary amongst the castle buildings on the Eastern Heights. But the structure of St. Mary’s-in-Castro cannot be earlier than the end of the tenth century; nor is there a particle of evidence that as early as the seventh century a ‘castle’ of any kind existed on that site. Indeed, a reference to a 'castle’ at Dover in the seventh century can only be interpreted as a reference to the remains of the Roman fortress; and that fortress, built, as abundant evidence seems to show, in connexion with the Channel Fleet, must have been placed within effective distance of the harbour—i.e., somewhere on the site of the medieval town.⁸⁴  Here, in fact, was an early Saxon foundation, that of St. Martins-le-Grand, of which the rebuilt Norman church still yields stubbornly to the modern builder on the western side of the Market Square.⁸⁵  The church lay within the area of intensive Roman occupation, and well within our conjectural limits of the Saxon-Shore fortress. It is consistent with all the evidence, therefore, to suppose, with Mr. Clapham, that King Eadbald’s church was the first church of St. Martin’s; and, as slight confirmatory evidence, the discovery of a tombstone with cross and Runic inscription of early Saxon date near the east end of St. Martin’s (No. 4) points with probability to the early Christian use of the site. Confusion between the Roman ‘ castle’ and the Norman castle on the Eastern Heights seems, however, subsequently to have arisen. According to Tanner, Wictred, King of Kent moved the monastery from the castle to the town in 696, and hence rose the church of St. Martin. This may be dismissed as a late attempt to bring the medieval castle ‘into the picture‘ long after the Roman fortress had been forgotten.
   The total disappearance of the Roman defences is not difficult to explain. In Roman times they seem to have looked eastwards upon a small lagoon or harbour formed by an expansion of the Dour, and perhaps aided artificially by a timber-framed quay on its southern side. After the Roman period the wind-blown sand and river silt gradually clogged this harbour and the seaward approach to the fortress. When the medieval builders arrived, the old site was obsolete. It was now beyond the reach of shipping, and it was too small. Roman Dover had been a useful signpost for Channel shipping, and possibly a permanent station of the Channel Fleet. But Roman traffic, orientated, perhaps, rather on the Rhine than on the Boulogne-Calais coast,
    ⁸³  Arch. Journ. lxxxvi, 56.
   ⁸⁴  One writer has questioned whether the place had any walls in Roman times, thinking a supposed fort on the Castle Hill enough defence. But, as already remarked, the existence of that fort is wholly doubtful, and, if it existed, it was too far off to help in days before artillery; so exposed a site as Dover Harbour must have had its walls.
    ⁸⁵ See Arch. Cant. iv, I. A fragment of the church still stands behind the Canton Club.