are shorter than the width of the tile (7½ in.), and fill up the gap by recurring. The recurring letters, so far as can be judged by the surviving fragments, are identically the same as the others, as if the same stamp had been used over again to impress them, but there is no visible sign of any edge to the stamp. Further, the words do not begin or end at the beginnings or ends of the lines. On four fairly complete specimens the second line appears as abriabantu ca (twice), riabantucabi, and bantu. cabriab, and on less perfect pieces it begins abanu-, and ends -abii and -u.cab. Nor do the words begin and end together. Lines (1) and (2) are in fixed relation to each other, but the final m of (1) is opposite the hr of (2). We have no obvious choice but to suppose that the stamp was of very irregular shape, and that the tiles were laid out side by side for stamping and marked consecutively without any care for coincidence for the divisions of words and tiles. The whole may have been meant primarily as rude ornament. But it appears to contain a trader’s mark or advertisement in Latin.  Parietalem probably denotes a tile made for use in a house-wall (paries), a common use of box-tiles; Cabriabanu (or whatever be the ending) seems to be a Celtic name,⁵²  presumably that of the brickmaker, and the whole sentence would probably run ‘Cabriabanus made this box-tile.’ The use of Latin is significant.⁵³  We may compare these tiles with the better-known Latin graffiti on Silchester tiles, satis ‘enough,’ the groan of a weary workman, or fecit tubum Clementinus, the sprawl of a fresher colleague. In each case we have evidence that the lower classes in Roman Britain spoke and wrote Latin. But whereas the Silchester examples illustrate only the customs of the town, our present instance is the more interesting, since it occurs in a villa and illustrates the condition of the rural population. We do not, of course, know that it is a local product. But it is not likely to have been brought from far. Its rude character forbids that idea, and, indeed, brick-earth is common in Kent. We may regret more sharply than usual that we possess no materials for dating accurately the tiles or the villa.
   41. ST. MARY CRAY.—Tesseae, bricks, a coin of Faustina II, and other indications of a villa were found in the old bank and bed of the River Cray, so yds. from the present stream. The building probably was destroyed in making a new road.⁵⁴
   42. SALTWOOD, NEAR HYTHE.—Roman foundations, bricks and tiles were discovered in 1864 at the south-east corner of Harp Wood, close to the Brockhill stream, about ½ mile west of Saltwood, on the hill above Hythe. No detailed account of them exists. But they are apparently the same as some ‘Roman foundations in Carp Wood,’ recorded by Mr. Payne. A cinerary urn with a saucer serving as a lid, both of coarse dark ware, and a red earthen jug, which (if one may judge by its neck) is of late Roman date, were found in a stone cist at Saltwood some time before 1874, and are now in Folkestone Museum. ⁵⁵
   43.—SNODLAND.—Traces of a building were noticed here in 1844 and very cursorily uncovered, in two fields called Church Field and Stone Grave Field, close to Snodland church, on the west bank of the Medway, and immediately overlooking the river, a site now occupied by Gas Works. The remains examined consisted of a floor of large tiles, another of concrete of lime, sand, pounded tile and stones, and some walling, one bit of which is described as ‘a well-built wall of stone with alternate layers of red and yellow tiles.’ There were also some walls in the bank of the river, taken to be a passage leading down to the water, and tradition adds that a bath was found here early in the 19th century. Much debris of a house, tesserae, roof and flue tiles, and potsherds lie scattered about the site, and the walls of the church contain may Roman tiles.
   In building a new retort house at the Gas Works, in 1927, pieces of shapeless foundation and a length of wall showing a flint course, resting on a footing of two layers of chalk boulders, were found. Near it a terra-cotta mask, and a thin bronze plate, the counter-plate of a buckle, ‘chip-carved,’ and inlaid with niello (P1. XXVI). Mr. Reginald Smith points out the rarity of portrait-medallions in this form of decoration which belongs to the close of the Roman period. The coins found (5) include a brass of Domitian A.D. 87, a worn Pius, a Constantine I, an’ Urbs Roma,’ and a small brass of Gratian; the pottery dates from the late 1st to the late 3rd century, and includes a stamp of the La Graufesenque potter, Frontinus.⁵⁶
   44. SWANSCOMBE.—’ Foundations’ of a Roman building have been noticed by Mr. Spurrell
continued from page 123   lowest line of the same stamp above it. One can see how the line of the pre-existing I has caused an elevation, but no break in the A stamped over it. For the tile, see Proc. Soc. Ant. Loud. xxiii, 108 ff. and Eph. Ep. ix, 1289.
   ⁵²  It is rash to speculate where the lettering is so rude as that of these tiles. But, if the word was meant to end in u, we may compare the similar terminations of potters’ marks on Samian ware—Agedillu, Cintusmu, for Agedillus, Cintusmus, or Cotu, Criciru, for Coto, Criciro—due very probably to Celtic contamination. Compare Bohn’s remarks in Corpus Inscr. Lat. xiii (3), pp. 119—120.
   ⁵³  On the general question of the use of Latin in Roman Britain, see Haverfield Romanization of Roman Britain (Oxford, 1923), pp. 29—3         ⁵⁴  Journ. of R. Studies, xviii, 208.
   ⁵⁵  Six-inch O.S. Maps, Sheet No. lxxiv S.E. and Name Books; Payne, Coll. Cant. 199; objects in Folkestone Museum. See also Topographical Index.
   ⁵⁶  C. R. Smith, Arch. Journ. i, 164.; excavations, ibid. 262, and Wright, Wanderings, p. 189. Antiq. Journ. vii, 521-2; Arch. Cant. xl, p. xlviii, 79—8 2. The buckle is in Rochester Museum. The O.S. six-inch, xxxi, N.W., records tiles found here in 1893 also. For barrow, see Topographical Index.