52 THE EXCAVATIONS 



be followed for some distance, it might afford some evi- 

 dence, but tbe occurrence of a few boulders at intervals 

 under so much rubble would bardly be conclusive. 



Will tbe argument from analogy help us here? The 

 ramparts of the Scottish forts are, almost without excep- 

 tion, made of earth. The later forts were of stone, 

 and apparently the rampart of earth and stone marks 

 a transition. The neighbouring forts of Mancunium 

 and Brough had a stone rampart 6 to 7 feet thick. 

 The exact history of the transition, however, has 

 not yet been made out. In his valuable note on this 

 subject,^* which I am glad to be able to use, Mr. Haver- 

 field mentions the case of a fort in the Carpathians built 

 not earlier than a.d. 110, which had at first earthen walls, 

 and was given stone ramparts in 201. A similar case is 

 reported by Arrian as occurring oil the Armenian fron- 

 tier. Mr. Haverfield concludes : " It is exactly the same 

 development as that by which the early earthen tumuli of 

 Rome grew into stone structures like the tomb of Caecilia 

 Metella, ... in these cases, as in the ramparts, there was 

 a period of transition when earth and stone were both in 

 use." As far as Melandra is concerned, I know of no 

 evidence to show whether the wall was added to the clay 

 bank, or whether the two were raised simultaneously, but 

 Professor Conway sends me the following note on this sub- 

 ject:— 



My knowledge of walls and earths is far too slight for me 

 to venture to set any opinion of my own on a practical 

 matter against a definite judgment of either Mr. Bruton's 

 or Dr. Haverfield's. But as every general description of the 

 rampart is inductive and to some extent constructive, it 

 seems one's duty to state what one believes one's self to have 

 seen. Mr. Bruton's descriptions of what is now visible 



64. The Roman Fort of Gcllygaer, p. 38. 



