the continuation of which it proved impossible to trace. The courtyard itself had been covered with fine gravel. In the absence of doorways it was not easy to determine the exact number of rooms that the house had contained, as some of the divisions may have been merely sleeper walls for the support of floors. It seems probable, however, that it had consisted of some nineteen apartments. That opening upon the entrance passage was perhaps inhabited by the porter.

On the south side facing the rampart were what appeared to be the kitchen and the servants' quarters. A black deposit lay on the floors of several of these rooms. In the floor of the second room from the south-east corner of the building was a small pit which yielded an iron pickaxe. On one side of the room next to it on the west was a built cellar or stoke-hole, shown in Plate XI., Fig. 2, thirteen and a half feet long by two feet one inch wide. Except on the south side the walls remained in fair condition to a height of about four feet. They were built of hammer-dressed sandstone courses. On the west side a depression in the wall suggested the position for steps. A flue or drain issued from the north end, but it had disappeared beyond the wall. It is possible that the house had rooms with hypocausts which were heated from here. Four bricks of a type commonly used for hypocaust pillars were found lying together in the corridor, but these were the only traces of any arrangement of the kind. Indeed, no hypocaust was discovered in any of the buildings within the fort itself. In the built cellar there was found a small globular pot of white ware, covered with a black engobe, like Castor ware.[1] In the room adjoining it were the remains of what had probably served as a hearth. It was placed in one corner, and was more or less circular in shape, twenty-one inches by thirty inches, and terminating in a flue-like arrangement; the whole measured five feet ten inches in length. The sides were formed by sandstone blocks. The suggestion that the offices of the house had been here is perhaps confirmed by the presence of the base of the cistern at this side of the courtyard as noted above. The entrance from the corridor would be arranged conveniently for drawing water. The rooms on the west gave little indication of their use. In one there was found a bronze key, in another a fragment of a vessel of yellow-brown castor ware; yet another showed traces of a floor of opus signinum—a pavement formed of broken brick and lime.

The apsed room projecting into the courtyard had been destroyed down to its cobble foundations, only a single stone of the scarcement course

1 Plate xlviii. fig. 40.