ON A ROMAN FORT AT GELLYGAER. 451 



and breadth 385 feet, so that this fort is one of the smallest of the British 

 excavated series, and it is a shorter oblong than usual. Another 

 peculiarity is the backward position of the Via Principalis. 



The various buildings, including the rampart revetements, were found 

 to rarely remain to a greater height than 3 feet above the Roman level. 

 They were constructed of the local Pennant gritstone, a hard, thinly 

 bedded rock, which is still the chief building material of the district. The 

 masonry may be described as regularly coursed rubble-work, but it varied 

 considerably in quality. The gates, for instance, were neatly constructed 

 of well-selected stones, many of which were more or less dressed. Some 

 of the walls of the buildings of the interior, on the other hand — probably 

 sleeper-walls to support superstructures of timber — were built of rou"h 

 stones as quarried, with an admixture of weathered field stones. The 

 voussoirs of the gate arches were of calcareous tufa. 



Mortar had been used in all these walls, but it was found, as a rule 

 to have decomposed into sandy loam of the same colour as the surrounding 

 soil. This proneness to decomposition was apparently due to the use of 

 'white' lime, i.e., lime derived from Carboniferous limestone, and the- 

 inability of getting a suitable sand in the district. This limestone,, 

 together with the calcareous tufa just referred to, is found in the vicinity 

 of Castle Morlais, ten miles to the north-west. 



The foundations of all the outside and of most of the divisional walls; 

 were deep and well laid, consisting of rough quarried and field stones, 

 roughly coursed in the case of the more important walls, and packed onj 

 end in the case of the less important. 



The roofs of several of the more important buildings had been 

 covered with red tiles of the usual Roman type ; but from the absence of 

 roofing tiles of any sort on the sites of others, particularly the ' long ' 

 buildings, it may be inferred that these buildings had been thatched or 

 covered with wooden shingles or planks. 



The ditch was found to be of a shallow V-shape in section, approxi- 

 mately 19 feet in width from lip to lip and 7 feet in depth ; and owing 

 to the tenacious character of the natural soil, the sides were singularly 

 well preserved. The excavation in front of the south-west gate showed 

 that each side of the ditch had been cut back so as to leave a step about 

 2 feet from the bottom, obviously to serve as a platform for the 

 supports of a bridge ; and from the absence of masonry it may be 

 concluded that the bridge was of wood. 



As already stated, the rampart was of earth (evidently derived from 

 the ditch) and confined between two walls. The outer of these walla 

 varied from 3 feet to 4 feet 3 inches in thickness, and the inner was 

 considerably thinner. Both walls had been built against the earth- 

 bank. The total width of the rampart varied from 19 feet 4 inches to a 

 trifle over 20 feet. 



All the gates, so far as could be judged from their remains, were of 

 similar de«gn and size. They were double— that is, each contained two- 

 passages, 11 feet long and wide, separated by an intervening spina. 

 Front and back these passages were narrowed to portals 9 feet 6 inches- 

 wide by projecting jambs or pilasters, which doubtless had carried 

 arches, as the fragments of voussoirs of calcareous tufa were plentifully 

 found on the sites. On either side of the pair of passages was an oblonc' 

 guard-chamber about 1 1 feet by 9 feet 6 inches (internal measurementsr. 

 entered by a narrow doorway in the back. 



gg2 



