to all three, but also to the keeps of the first three-quarters of the 12th century. The absence of a forebuilding is probably an early sign,[1123] and so is the extensive use of tiles.[1124] The chapel with a round apse which projects externally only occurs in the keeps of London and Colchester, and in the ruins of Pevensey keep.[1125] The absence of a plinth is believed by Enlart to be an early token.[1126] But Colchester has a plinth and so has the Tower. It is, however, very possible that in both cases the plinth is a later addition; at Colchester it is of different stone to the rest of the building, and may belong to the repairs of Henry II., who was working on this castle in 1169; while the Tower has undergone so many alterations in the course of its eight hundred years of existence that it is difficult to say whether the rudimentary plinth which it still possesses is original or not. Wide-jointed masonry is generally recognised by architectural students as a mark of the early Norman style. Even this is a test which may sometimes deceive; certain kinds of ashlar are very liable to weather at the edges, and when the wall has been pointed at a comparatively recent period, a false appearance of wide joints is produced. Moreover, there are instances of wide-jointed masonry throughout the 12th century. The use of rubble instead of ashlar is common at all dates, and depends no doubt on local conditions, the local provision of stone, or the affluence or poverty of the castle-builder. We are probably justified in laying down as a general rule that the dimensions of the ashlar stones increase as the Middle Ages advance. There is a gradual transition from the _petit appareil_ of Fulk Nerra's castle to the large blocks of well-set stone which were used in the 15th century.[1127] But this law is liable to many exceptions, and cannot be relied upon as a test of date unless other signs are present. The Tower of London is built of Kentish rag; Colchester keep of small cement stones (septaria), which whether they are re-used Roman stones or not, resemble very much in size the masonry of Langeais. It is of course unnecessary to say to anyone who is in the least acquainted with Norman architecture that all Norman walls of ashlar are of the core-and-facing kind, an internal and an external shell of ashlar, filled up with rubble; a technique which was inherited from Roman times in Gaul, but which was not followed by the Anglo-Saxons.[1128]