work which can be paralleled from Roman Kent. They occur together in the Dover Pharos (p. 48), and alternately coloured voussoirs are depicted on a Canterbury mosaic (p. 68). These arguments look better on paper than in fact. The walls of St. Martin’s nave are rude work, much more suggestive of Saxon than of the simplest Roman building. The red plaster is a Saxon as well as a Roman material. The alternation of voussoirs occurs in early medieval work abroad, if not in England, and at St. Martin’s it is carried out with so little intelligence that it differs markedly from the Roman fashion. But the decisive evidence against a Roman date for the St. Martin’s nave seems to have been provided by excavation; the nave is shown thereby to be later than the Saxon brickwork of the chancel (Arch. Cant. xxii, 21; Arch Journ. liii, 279; lviii, 415).
   No other structural remains seem to have been found near St. Martin or, indeed, anywhere in this part of the suburb. Brent, indeed, mentions a mosaic as found in ‘St. Martin’s parish’ and gives Somner as his authority (Arch. Cant. iv; 38; Cant. Olden Time, 28). But I can find no such statement in Somner, and perhaps Brent has misread his reference to the mosaic in St. Margaret’s parish. Apart from that, only one Roman find has come to light in this quarter—a fibula and some potsherds found in December 1864 ‘in the grounds abutting on St. Martin’s churchyard’ (Brent in Proc. Soc. Antiq. Ser. II, iii, 55; vi, 377; Cant. Olden Time, 30, 46, plate; C. R. Smith, Coll. Antiq. vii, 203, plate; fibula in Cant. Mus.). The fibula is an interesting piece—a circular disc of pierced work, enamelled in blue (or purple) with a red pattern in the middle. The general character of the find, whether sepulchral or other, is not recorded.
   It would seem, according to our present knowledge, that the general neighbourhood of St. Martin’s Church was not occupied by any Roman building, secular or sacred. It was, indeed, as we know (p. 79), the site of one of the principal cemeteries of Roman Canterbury. If, therefore, there is anything in Bede’s statement, we may perhaps take it to imply that there stood here in Roman times a small Christian chapel built for funerary uses or in association with the grave of some Christian martyr. For the normal uses of the Roman Christian congregation the site is an unlikely one. The analogies of Silchester and, more especially, of other towns in other provinces would lead us to look within Durovernum itself for the conventicle of local Christianity. It may be that Bede was entirely wrong and that the building of St. Martin’s was due to quite different circumstances. We may wonder whether the positions of St. Pancras and St. Martin’s were determined rather by the existence of an early Saxon settlement without the walls. The English do not appear to have occupied the town itself till a comparatively late date. Their great cemeteries lie some three or four miles off to the south-east, and they may have lived outside the Roman walls until intercourse with France and the introduction of Christianity from that country brought higher civilization in the latter half of the 6th century. It will be observed that the distance of St. Martin’s from the town—under half a mile—would permit the carriage of such Roman bricks as were needed for the little church.
   On the south side of Canterbury, between the Dover Road and Wincheap, Roman remains are common, but they are almost wholly sepulchral (p. 77). Some silver and iron rings, bronze fibulae, a bronze ‘ligula,’ and potsherds were discovered in 1876’ in the water of the Silver Spring’ in excavating for the Whitehall Swimming Bath (Cant. Olden Time, 31, 49, plates ii, 9; xviii, 5). They may conceivably represent offerings thrown into the spring in Romano-British times, but our accounts of the find are too vague to do more than suggest the possibility. The structural remains noted in this quarter are few and unsatisfactory. Pillbrow found, 9 ft. below the Old Dover Road, the tops of four parallel walls running directly across the line of the Street and separated by intervals of 40 ft. to 6o ft. (Arch. xliii, 158), but their age and use is not plain; they are too far apart for piers of a causeway. The ‘amphitheatre’ in Martyrs’ Field, the mill near the Southern railway bridge over the Stour, and the ‘ Roman camp’ on the higher ground of Whitehall, though it yielded two gold Coins of Gallienus and Valerian in 1876 (Cant. Olden Time, 24; Arch. Cant. xi, 417), seem all unproven, and the first two are highly improbable.
   We turn now to the cemeteries of Roman Canterbury.
   1. North-western quarter: St. Dunstan’s.—A cemetery covering some 20 acres and containing only cremation burials existed on the north-west of Canterbury, near the Roman road to London. 
combined with opus reticulatum, on an arch of the Aqua Claudia in the upper Anio valley; both these instances belong to the first century A.D. For another (undatable) example, also on an aqueduct, see Papers of the British & School at Rome, iii, 147, fig. 14. In Gaul the device occurs in the Imperial Palace at Trier (about A.D. 300) and in many undatable buildings; see illustrations in de Caumont, Abecedaire (ed. 1870), pp. 54, 170, 366, etc.; Bonnin, Vieil-Evreaux atlas, plate x, D and E; Blanchet, Enceintes, plate 12, etc. Its use in the later Empire perhaps represents the same feeling which prompted the ornamental masonry and brickwork of the Cologne Römerturm and the Richborough walls (p. 30). In post-Roman architecture it developed into definite Polychromy on the Continent, as in the early churches of Lorch (near Mainz) and Andernach. But it is very rare in England. The Saxon chapel of Stone, near Faversham, shows a hint of it.