Exeter Castle stands on a natural knoll, occupying the north-west corner of the city, which has been converted into a sort of square motte by digging a great ditch round the two sides of its base towards the town.[442] That this ditch is no pre-Roman work is shown by the fact that it stops short at the Roman wall, and begins again on the outside of it, where, however, the greater part has been levelled to form the promenade of the _Northernhay_ or north rampart of the city. On top of this hill, banks 30 feet high were thrown up, which still remain, and give to the courtyard which they enclose the appearance of a pit.[443] On top of these banks there are now stone walls; but these were certainly no part of the work of Baldwin de Molis, who must have placed a wooden stockade on the banks which he constructed. One piece of stonework he probably did set up, the gatehouse, which by its triangle-headed windows and its long-and-short work is almost certainly of the 11th century. It has frequently been called Saxon, but more careful critics now regard it as "work that must have been done, if not by Norman hands, at Norman bidding and on Norman design."[444] It was no uncommon thing at this early period to have gatehouses of stone to walls of earth and wood. Of these gatehouses Exeter is the most perfect and the most clearly stamped with antiquity. One thing we look for in vain at Exeter, and that is a citadel. There is no keep, and there is no record that there ever was one, though a chapel, hall, and other houses are mentioned in ancient accounts. Mr Clark says that probably the Normans regarded the whole court as a shell keep. It certainly was, in effect, a motte; but it was altogether exceptional among Norman castles of importance if it had no bailey. And in fact a bailey is mentioned in the _Pipe Roll_ of 1 Richard I., where there is an entry for the cost of making a gaol in the bailey of the castle.[445] Now Norden, who published a plan of Exeter in 1619, says that the prison which formerly existed at the bottom of Castle Lane (on the south or city front of the present castle) was "built upon Castle grounde," and he states that the buildings and gardens which have been made on this ground are intrusions on the king's rights.[446] The remarkably full account of the siege of Exeter in the _Gesta Stephani_ speaks of an outer _promurale_ which was taken by Stephen, as well as the inner bridge leading from the town to the castle, before the attack on the castle itself. Unfortunately the word _promurale_ has the same uncertainty about it that attaches to so many mediæval terms, and the description given of it would apply either to the banks of a bailey, or to the _heriçon_ on the counterscarp of the ditch of the motte. We must, therefore, leave it to the reader's judgment whether the evidence given above is sufficient to establish the former existence of a bailey at Exeter, and to place Exeter among the castles of the motte-and-bailey type.