92 MELANDRA CASTLE. 



what overlooked on the other, yet this fact does not seem to 

 have caused any change in the regular formation of the enclosure 

 With the single exception of a marl-pit, somewhat supplying the 

 place of a fosse on the southern side, there appears to have been 

 no special strength of defence in that direction, while the same 

 nature of wall and rampart appears to have enclosed the whole. 



The nature of this chief defence is somewhat remarkable. 

 It was doubtless faced around on its outer side with a stone wall, 

 though the traces of this now remain near the chief gateways only. 

 This was seemingly backed by a mound of rubble, earth, and 

 marl. Several sections made through each side, well into the 

 original surface, have one and all failed to reveal any sign of 

 an inner retaining wall. A form of rampart unusual in Roman 

 works is thus disclosed.* The outer shell of masonry has a 

 thickness of little more than a foot, which the backing of rubble 

 increases to four or five feet at its lowest course. With the base 

 of the mound included the width is increased to twenty feet or 

 more. The nature of the rampart-walk, if any such there were, 

 and its association with the towers which surmounted its four 

 corners, remains an unsolved problem. The top of the mound, 

 which probably might be gained from any point of the interior, 

 may have been used by the sentries, to whom the wall, rising 

 somewhat higher on the outer side, would thus serve as a 

 protection. 



The outer wall having been previously stripped from around 

 the three comers where the towers are otherwise well shown, 

 and at the fourth (the western) the tower itself being not clearly 

 defined, it is not now possible to examine the exact connection 

 between these features of the masonry. The mound seems, in 

 one or two instances (at least), to have been piled against the 

 side walls of the towers, and in no case had a tower, whether in 

 a corner or flanking a gate, a masoned floor at the ground level, 

 nor any definite appearance of an entrance. This bears out 

 the conjectures made elsewhere in the restoration of similar 



* The ramparts of the lately excavated fort at GelHgaer, near Cardiff, are 

 somewhat similar. Earthen ramparts seem to have been commonest in the 

 first century A.D. (F. Haverfield.) 



