watercourses which were both wider and more winding than a town ditch and which were apparently open in Roman times. Perhaps, therefore, the east and south-east as well as the north-west front of Durovernum may have been skirted by river channels during at least some part of the Roman period. Burials were discovered in one of the deposits (p. 77), and this may indicate that part at least of these channels dried up in or before Roman times. Unfortunately, no one troubled to record precisely the position of the burials, and arguments from them are therefore dangerous.
   The manner in which this area was built over is no better known to us than its boundaries and defences. Of the Roman streets we are wholly ignorant. During the drainage works many old road surfaces were noted beneath the present streets. But none of them possessed decisively Roman features. Even those under Beercart Lane and Watling Street, which have been confidently styled Roman, have yet to be proved deserving of the epithet, and most of the existing streets are plainly not Roman, since Roman walling has been found beneath them. It is plain that, as a whole, the present street-plan of Canterbury bears little relation to that of Durovernum. The plans and sizes of the Roman houses are no less obscure. The house walls unearthed in the drainage works were in great part of uncertain age. Even if we accept the majority of them as Roman, they are too fragmentary and too detached to yield either any plans of houses or any clue to frontages, streets or roads. They prove that the site was built upon. Beyond that, the one fact noticeable is that their direction is not uniform. The Roman walls under High Street and the Parade run in general north-east and south-west; others, under St. Margaret’s Street, Butchery Lane, Sun Street, Stour Street, run east and west. If all these represent Roman houses, the buildings of Durovernum faced divers ways like those of Silchester.
   We can trace a few of these houses. One stood at the lower, north-western end of High Street, near the present County Hotel. Here masonry and fragments of mosaics indicate a private house on the extreme north-west of the Roman area. The chief mosaic possesses some individual interest. It is—or, rather, was—a fragment measuring 4 ft. by 8 ft., found about 1758 on the site of the Hotel, and decorated with a noteworthy design. In the centre is a doorway with semicircular head in which the voussoirs are alternately dark and light coloured, according to a not uncommon Roman fashion (p. 48). On each side is a large conventional flower, worked carefully in various shades of red and brown. These flowers are out of all proportion to the size of the door if regarded as actual blooms; but they perhaps symbolize flowering shrubs placed to flank an entrance. A little further up the street, in front of the Fleur-de-Lis Hotel, some oolite columns and cornices may indicate a small temple. At the junction of High Street and St. Margaret’s Street, Roman walls and burnt wheat and wood point to a house or shop with a wooden loft or shed for storing grain. A bit of plain tessellation found in front of the Fleece Inn, in the Parade, testifies to another house, and some remains suggestive of a hypocaust add another at the corner of Butchery Lane and the Parade. On the south-west of High Street we can cite a plain tessellated floor in front of the Fountain Hotel, some vestiges of a hypocaust at the junction of Hospital Lane and Castle Street, the floor and painted plaster of (it may be) a half-timbered house at the meeting-point of Rose