quarters building. The grounds for supposing them to be granaries or storehouses have been well stated by Mr Bosanquet in his account of the excavations conducted under his supervision at Housesteads.1 In some respects the Bar Hill example represents a departure from the normal type. In particular, its outside walls were less thick than is usual -being only about 2 feet-and they were not supported by buttresses. Again, compared with the great majority of similar structures elsewhere, it was remarkable for its relative breadth (32 feet). With a length of 85 feet, we should not have expected it to be more than from 20 to 25 feet wide.2

A stone partition divided the Storehouse longitudinally into two slightly unequal halves. Probably this is the explanation of the peculiarities just enumerated. If there were to be two divisions, the whole would require to be broader than is usual. On the other hand, the partition could be so utilised as to relieve the side walls of much of the pressure of the heavy roof with which we must suppose the granary to have been provided. Buttresses would thus be rendered unnecessary. Of the two halves, the eastern was the larger. It had an interior width of 13 feet, and had evidently been paved with flags, as pieces of flagstone were found lying undisturbed in the bottom. The western half was only about 11 feet wide. Its floor was doubtless also formed of flags. In this case, however, recourse had been had to a method of construction that is frequently associated with such buildings. In order to guard against damp, the flags had been supported by three dwarf walls that ran from one end of the division to the other. A good many ashes were observed in the northern portion of the free spaces so provided, but there was nothing to indicate when or how they had accumulated there. About 17 feet from the N. end were traces of what appeared to be a cross wall. It became obvious during the excavations that the two most easterly of the dwarf walls, taken along with the stone partition,

1 Arch. Ael., xxv. pp. 237 f.

2 See Bosanquet, .c. His statistics are entirely borne out by sites examined since the publication of his paper, e.g. Castlecary, Rough Castle, and Gellygaer.

represented the "three rows of ruins" shown very prominently in Gordon's plan, and spoken of by Horsley as being still visible "within the Praetorium."1 Fig. 15 gives a view, looking south, of the northern end of the "three rows," as they appeared when uncovered in 1903. The third dwarf wall and the main wall on the W. are barely distinguishable. Towards the right the stone gutter on the E. side of the street is very well seen.2

(c) The Workshops.-To the E. of the Storehouse were the remains of yet another building of stone. It had been sadly mutilated. None of its details were ascertainable. Even the limits of its foundations could not be certainly fixed, although it must have covered an area of not less than 41 feet by 33 feet. The fact that it had contained the workshops seemed tolerably clear from the nature of the objects found within what was left of its walls. These included quantities of ashes and other indications of large fireplaces, the remains of flues, many pieces of wrought iron, a number of iron nails, and-most significant of all-much iron-slag and glass-slag. Near the S.W. corner a wellpreserved flue entered the building from the E.

(d) The Baths and Latrines.-After the Praetorium itself, the most extensive stone structure discovered was a range of buildings that stretched nearly the whole way from the N. gate to the N.W. angle of the fort, at a distance of not more that 4 feet from the rampart. Measured over the foundations, it was about 15 feet broad, and rather less than 150 feet long. The general view, looking eastwards (fig. 16), conveys a good impression of its dilapidated condition. While it had evidently been a continuous suite of apartments, three clearly marked divisions had existed. Before entering on a particular description, we may mention that the N. ditch, opposite the two higher or more easterly

1 See supra, p. 407.

2 The pool of water in the centre of the foreground marks a hole dug to verify the line of the Agricolan ditch.