CHAPTER III

QAṢR-I-SHÎRÎN

The general disposition of the Sasanian ruins at Qaṣr-i-Shîrîn has been given by M. de Morgan, and the plan of the two principal buildings, the palace of Khusrau and the palace (if palace it were) of Chehâr Qapû, both of which I examined, appear in the same volume.[45] It is quite possible that the ruins may have suffered to a certain extent during the years which elapsed between M. de Morgan’s visit and my own, and this may account for the omission in my plans of some features which are shown by him. Nowhere did I observe stucco decorations in so good a state of preservation as that which is depicted in his Figure 208. I have, however, compared my photographs with those published by him and found no very noticeable differences. Moreover, it will be observed that such details as are absent from my plans are usually indicated hypothetically on those of the French mission, and it is therefore doubtful how much of them was actually seen and how much was conjectural. A very little excavation would determine whether these conjectures are correct. It is much to be regretted that I had not the French plans with me, as I might have been able to form some more definite opinion as to the value of the proposed restorations. As it is I must content myself with recording that which I saw above ground.

THE PALACE OF KHUSRAU

The larger edifice, which is known as the palace of Khusrau (i.e. Chosroës II, Parwêz, A.D. 590-628), is not built upon a single level. The central part is raised above the plain by means of a solid platform of earth some 3 metres high. The terribly ruined state of the buildings made it difficult to take elevation measurements which should approach to accuracy; I have therefore endeavoured to give a correct impression of the structures upon the two levels by reproducing the plan in two parts. In the one (Plate 53) the upper rooms and courts are given; the uncovered areas on the upper level are lightly tinted, the covered rooms are dotted, while the buildings on the lower level are shown only in outline. In the other (Plate 54) the upper level is left in outline and the covered and open areas of the lower level are fully indicated.

The palace is exactly oriented, the main rooms and entrance facing east. The building materials are undressed stones laid in a thick bed of gypsum mortar. The stones are used exactly in the shape in which they were furnished by nature, a shape which happened to be that of large rounded pebbles. With such materials accurately coursed masonry is not to be expected. The core of the walls is no more than a mass of concrete with stones bedded at haphazard in the strong gypsum mortar. On the outer surface of the wall, particularly in important chambers, the pebbles are, however, coursed with considerable care, but the face of the walls is necessarily very rough and must always have been covered with plaster. The vaults are constructed of the same unfavourable materials. They were built over a centering on which was laid an inner skin of stones and mortar; when this had hardened it was strong enough to bear the mass of concrete which was built round and above it. Construction of this kind would have been impossible but for the excellent qualities of the mortar. I observed that the vaults both in this palace and at Chehâr Qapû had almost invariably a slight outset from the wall (Plate 52, Fig. 2), as is generally the case in Sasanian vault building, whether in brick or in stone. The vaults are round or slightly ovoid, except in the lower corridor, under the margin of the platform (Plate 54, Corridor 103). Here the vaults are very markedly pointed (Plate 51, Fig. 1), but I should attribute this form not to any conscious predilection for the pointed arch--an arch which was, so far as I am aware, unknown to Sasanian architects--but to an accident inherent in the rude construction of an unimportant part of the building. Occasionally brick was used. I saw fragments of brick among the ruins of the palace of Khusrau, and in Chehâr Qapû some brick vaults are still standing. The walls which were intended to support these massive stone roofs were seldom less than 1·30 metres thick, and sometimes considerably thicker. (In Chehâr Qapû, however, they are not infrequently reduced to a thickness of little over a metre.)

The eastern end of the platform is devoid of constructions. It is accessible by means of three double ramps which will be described in dealing with the lower level of the palace. Excluding the width of the ramps, the open platform is 149 metres long (reckoning it up to the east wall of chambers 21, 22, and 23) and 98 metres wide. The main gateway of the palace is much ruined. The hall or porch which is numbered 1 on the plan is indicated by two grass-grown mounds, 26·60 metres long by about 5·40 metres broad, leaving a space of about 9·80 metres between. Another mound lying north and south marks the eastern limit of No. 2. At either end of this latitudinal chamber there were traces of cross walls, which I have shown on the plan. Upon the eastern mound I saw through the grass circular patches of brick which may have been the remains of columns. Whether No. 1 was flanked on either side by columns, as M. de Morgan has represented it to have been, I have no means of determining, but I have little doubt that it was a covered porch of some kind leading to a latitudinal chamber, No. 2, which was some 45 metres long (between the cross-walls) by 17 metres wide, and that this chamber was a covered antechamber to the hall of audience, No. 3. The hall (3) is 27·20 metres square; the walls are ruined down to the level of the side door arches, and the interior is filled with ruins to the depth of about 1 metre--judging by the present ground-level in the doorways (Plate 55, Figs. 1 and 2). At each corner of the hall, 2·90 metres from the walls on either side, there are the remains of a pier, 1·40 metres square, with two engaged columns projecting about 1 metre and producing a heart-shaped ground-plan. The pier at the south-west corner is tolerably well preserved, and there can be no doubt as to its form. The eastern wall of the hall is 4·35 metres thick and is broken by a single door 3 metres wide. At the south-east corner a small doorway leads into a short passage, probably vaulted, which gives access to the open platform. On the west side of the hall lies a lîwân (4) 5·10 × 13·15 metres. A door, 1·60 metres wide, opens into court A, but there is no direct communication between the lîwân and its subsidiary chambers. Of these last there are two on either side. To the north, room 5 opens by doors into hall 3 and court A. No. 6 has only one door, opening into a narrow passage (9) which was probably covered by a vault. On the south side No. 7 corresponds exactly with No. 5, while No. 8 opens into No. 7 and not into the corridor 10. These corridors (9 and 10) lead respectively out of the north-west and the south-west corners of hall 3; they are prolonged beyond rooms 6 and 8 and open into court A. Parallel to them run a second pair of corridors (11 and 12) which are two of the main gangways of the palace. No. 11 is 1·80 metres wide. Its eastern end is, so far as I could ascertain, a cul-de-sac, but it may possibly be provided with a door into room 13 (the walls are very much ruined here). A doorway, placed immediately west of the end of corridor 9, leads into court A, and doors on the north side communicate with courts D and E. Corridor 12 is 1·70 metres wide and leads out of hall 3; the arched doorway into the hall is preserved. The only other doorway in this corridor of which I could make certain is one communicating with court I, but in both corridors (11 and 12) the walls are so much ruined that I cannot feel sure that they do not possess more doors. Beyond courts F and J both corridors drop down to the lower level and are then continued to the western limit of court B, where they turn at right angles and unite behind court B, but on the lower level. Whether the descent was accomplished by steps or by a ramp I could not determine, but in No. 12 the vault at this point was well preserved, and I noticed that, as in the stairs and ramps of Ukhaiḍir, it was built not in an inclined plane, but in sections rising one above the other like inverted steps (Plate 56, Fig. 1). East of hall 3 and of the chambers pertaining to it, the remainder of the central area of the palace is occupied by two courts, A and B, 33·90 metres wide, divided from one another by a much ruined cross wall in which there was presumably a door. Court A is 40 metres long from the west wall of the lîwân (4) to the cross-wall; court B is 71·30 metres long from the cross wall to the end of the platform.

To north and south of the central area lie a series of courts with lîwân groups, on the west side courts C and G alone offer slight variations of scheme. In court C there is a lîwân group at either end, the western group being the more important; as will be seen, this is the usual arrangement in the courts on the lower level. There are, besides, three chambers (13, 14, and 15), lying between court C and hall 3. These chambers are almost completely buried under ruin heaps overgrown with grass; I was able to see that No. 13 opened into No. 3 and into court C, but I could not determine the position of the doors in Nos. 14 and 15. Court C measures 21·60 metres from north to south and 19·20 metres from east to west. The western lîwân is 5·20 by 7·25 metres. I would here remark that in all cases the lîwâns open by their full width on to the court, whereas in the French plan the entrance arch is narrowed by short returns in the side walls. The side chambers (17 and 18) do not communicate with the lîwân (a rule which is followed throughout the palace), but have doors only into the court. A door in the west wall of the lîwân (16) leads into a latitudinally placed chamber (19) measuring 5·10 by 14·30 metres, which is separated by a wall at the south end from a small subsidiary chamber, 1·75 metres wide, with which it communicates by a narrow door. There is also a doorway between No. 19 and court D. This group of rooms (16 to 19) occurs unchanged in courts E, G, H and I, and is provided invariably with a posterior court. In one case only, court H, a shallow lîwân group is placed at the west end of the posterior court. All the latitudinal chambers (19, 28, 32, and 42) behind the lîwâns are completely ruined. I conjecture that they were vaulted, but it is possible that they were not wholly covered, like the corresponding chambers behind the lîwâns at Ukhaiḍir. On the analogy of Ukhaiḍir they must have served the purpose of kitchens. I saw no trace in court C of the columns which are placed there in the French plan. At the east end there is a shallow lîwân group (21, 22, and 23), the lîwân being 4 metres deep. To the north of this group lies a short passage leading to a door which communicates with the open platform. A corresponding passage (20), 2·30 metres wide, leads out of the north-west corner of court C, runs along the north side of courts D, E, and F, drops on to the lower level in the same manner as corridors 11 and 12, is continued as far west as they, and then turns off at right angles and joins the cross-passage which connects them. North of court C are two chambers on the upper level (106 and 24). No. 106 is a long passage room with two rectangular arched niches in the south wall, a door at the east end opening on to the platform, and a door at the west end which gives access to a ramp that descends into the exterior park, between the retaining wall to the south and the wall of a chamber on the lower level to the north. In the north wall of No. 106 there is a door leading into No. 24, a much ruined room about 7·50 metres square, and a door further west opening on to the roof of a short passage.

Courts E and F stand in the same relation to one another as courts C and D; court E is the forecourt of a lîwân group with a kitchen (25 to 28); court F is the posterior court. The western wall of court F is the retaining wall of the mound on which the rooms and courts of the upper level are built. Court F, together with No. 28, are omitted in M. de Morgan’s plan, a fact which shows that there must be serious errors in his measurements.

Upon the southern side of the platform, court G is divided from the hall 3 by three chambers (33, 34, and 35) which, like the corresponding chambers north of the hall, are ruined and filled with débris. They appear to have had no communication with the hall. On the south side a door leads from court G into corridor 43, 2·60 metres wide, which corresponds with the northern corridor (20). The western end of court G is occupied by a lîwân group and kitchen (29-32), the latter opening into court H. Court H, 15 metres from east to west, differs, as has been said, from its counterpart court D, in that it is furnished with a shallow lîwân group at its western end. These rooms (36, 37 and 38) are much ruined, but it appeared to me that there was no communication with court I. Court I, 14·20 metres from east to west, and court J, 17·80 metres from east to west, with the lîwân group and kitchen between them, correspond exactly in their arrangement with courts E and F. I do not doubt that all the rooms above described were covered by barrel vaults, but there is no wall on the upper level that stands much more than a metre high, and therefore no vault is preserved.

In the central part of the palace the upper level is prolonged to the western end of court B, but in the wings it ends with courts F and J. Thus it is that the rooms and courts which flank the western end of court B are upon the lower level. They form two complete units, one on either side. The northern unit is composed of courts K and L and rooms 44 to 50. On the east side of court K lies a shallow lîwân group (48, 49, 50), the lîwân being 3·25 metres deep. On the west side the lîwân group differs somewhat from those which have been already described. A narrow antechamber, 2·40 metres deep, is interposed between the lîwân with its side chambers (44, 45, 46) and the court. A wide archway, corresponding with the arch of the lîwân, and two doors, corresponding with the doors of the side chambers, open into court K, but the width of the arch and doors of the antechamber is slightly greater than the width of the arch and doors of the lîwân and its side chambers. The door of 46 is 1·05 metres wide and stands 1·85 metres from the south wall; the corresponding door of the antechamber is 1·70 metres wide and stands 1·30 metres from the south wall. The arch of the lîwân has a width of 5·20 metres; the corresponding arch of the antechamber is 5·80 metres wide. Neither here nor in any other court where the antechamber occurs is it possible to determine the exact relation between the vault of the antechamber and the vault of the lîwân, but the fact that the lîwân arch seems to have been narrower than the antechamber arch (it is only in court K that the measurements can be taken with anything approaching to accuracy) leads me to suppose that the vault of the lîwân cannot have been carried through to the court, as at Ukhaiḍir. In that case the antechamber must have been roofed with a continuous vault laid at right angles to, and possibly higher than, the vault of the lîwân. The antechamber communicates with corridor 11. Courts M and N, on the south side of court B, are the counterpart of courts K and L. The southern end of the antechamber is exceptionally well preserved, and the arched doorway leading into corridor 42 is standing (Plate 56, Fig. 2). Part of the vault of corridor 42 can be seen in Plate 57, Fig. 1.

The cross-passage connecting corridors 20, 11 and 12 affords communication with the western courts, which form three units, all exactly alike, except for slight variations in width. Each unit consists of a pair of courts and two groups of rooms. A shallow lîwân group lies at the east end of each of the forecourts, O, Q, and S (Plate 57, Fig. 2). Doors from the passage are placed in the side chambers of the lîwâns, and corresponding doors open into the courts. As far as I could ascertain the courts communicated with one another, but the division walls are ruined, often down to ground-level, and it is hard to decide between a doorway and a breach. At the west end of the courts stands a more important lîwân group with an antechamber (Plate 58, Figs. 1 and 2, and Plate 59, Fig. 1). In no case is there a door in the back of the lîwân, but communication with the posterior court is provided by means of a narrow vaulted passage (59, 67 and 75) placed to the south of the lîwân group.[46] There is no latitudinal chamber in the posterior courts, but a small additional chamber (58, 66 and 74), possibly for domestic purposes, lies on the northern side of each lîwân group. A corridor (79) leading out of court N bounds these courts to the south, and at right angles to it another corridor (80) bounds them to the west. The outer wall of No. 80 is ruined to the foundations, and I could not see whether there were doorways opening into the park. There were clear traces of doors leading into this corridor from courts P and T. Parallel to No. 79, but wholly separated from it, runs the continuation of corridor 43, which, after passing round the south side of court N, turns at right angles and opens at its western end into the park (Plate 59, Fig. 2). To the south of these corridors lies a large court, U, with remains of an arcade along its northern side. The space between the arcade and the wall of corridor 43 was probably vaulted; at its southern end it opens into the corridor. Court U is almost square (51 × 51·70 metres). To the west and south its walls are ruined, but on the west side great heaps of stones furnish indications of a gate. On the opposite side of the court there is another gateway of which a considerable part is standing. It is situated at the west end of a rectangular area, court V, arcaded on either side, which must have been intended for a private pleasure-ground or a place for games (Plate 60, Fig. 1). The latter is the more probable conjecture, since there is no direct communication between court V and the palace. The gateway was an important structure. From the western court (U) a porch 2·70 metres deep opened through an archway 3·70 metres wide into a rectangular vaulted chamber (83) 4·50 metres from east to west (Plate 60, Fig. 2). To the east of 83 lay a chamber (82) almost square (5·90 × 5·80 metres) having a rectangular vaulted niche, 1·50 metres deep, to north and south and an archway to the east opening into court V. No. 82 must have been covered by a dome, which was in all probability set over the angles on squinch arches (see below, Plate 69), but no part of the dome is standing (Plate 61). On either side of the gateway there are four chambers accessible only from court V. No. 85 opens into the passage, probably vaulted, which was formed by the northern arcade; No. 89 opens on to the area outside the southern arcade. It would be natural to expect that an outer wall ran parallel to this arcade, dividing court V from the park, and I looked for traces of such a wall, but did not find them. Court V (18·50 × 102·50 metres) terminates in a group of much-ruined buildings of which I could only make out the general plan. The arcaded passage (92) ends in a small vaulted and unlighted room (93) (6·55 × 3·55 metres). To the south of 93 are two large chambers (94 and 95), No. 94 terminating at the southern end in a deep niche. Nos. 93 and 94 are separated by a narrow passage from a small rectangular court (W) having two chambers at either end. Of these chambers Nos. 99 and 100 are completely ruined, but the vaults of Nos. 97 and 98, which are built partly under the upper platform, are standing (Plate 62). To the south lies another small court (X) out of which the passage 101 leads into a small rectangular chamber (102) which in turn communicates with the arcaded corridor 103. This corridor runs round the eastern end of the platform which is carried over it on a vault. The vault, which was very roughly constructed, is noticeably pointed, especially on the east side (Plate 51, Fig. 1). Three double ramps provided access to the platform, the eastern pair being the largest and most important. The eastern ramps begin opposite the fourth detached pier at either end of the arcade of the corridor, where a mass of masonry 6·60 metres long by 4·90 wide blocks the adjoining arch. Vaults carrying the ramp are placed before the seventh and eighth arches from either end of the arcade, and in front of the central arch lies a vaulted chamber 3·75 metres wide. The length of this double ramp is 48 metres (Plate 63, Fig. 1). On the west side of the corridor there are nine vaulted chambers, 5·80 metres deep, which are tunnelled out under the platform. Their doorways correspond with the arches of the corridor. A detached chamber lies at either end of the corridor. The north and south ramps are constructed in the same fashion, but they are only 30·80 metres long. Opposite the central vault there is a chamber under the platform; on either side the platform is solid, after which there are two vaulted chambers.

On the north side of the palace there is another group of much-ruined buildings on the lower level. The arcaded corridor (103) ends at this point in a narrow vaulted chamber (104) which lies under No. 106. Like 106, No. 104 has two arched niches in the south wall. It abuts at its western end against the ramp which descends from No. 106. A narrow passage leads out into a large enclosure, court Y, in which all the walls are ruined. Plate 63, Fig. 2, shows the eastern end of No. 106 with its vault partially preserved, and the walls and substructures of No. 24. In the south-west angle of court Y there was a large chamber (105), and the north-west corner was occupied by two groups of three rooms lying to north and south of the small court Z. Possibly there was a somewhat similar arrangement of rooms on either side of court Z[1].

CHEHÂR QAPÛ

Like the palace of Khusrau, Chehâr Qapû faces east. It covers a rectangular area 134 metres from east to west, and 82·60 metres from north to south (Plate 64). The building materials are the same as those used in the larger palace. The principal entrance is in the east end; I saw nothing of the great portico which M. de Morgan places on the south side, and as the outer wall at that point is entirely ruined, it is impossible to say whether there were a door there or no. The eastern gateway is much ruined (Plate 65, Fig. 1),[47] but the transverse arch between chambers 1 and 2 is standing. To north and south lie a series of courts and small chambers, occupying a width from east to west similar to that of the gateway buildings and apparently appertaining in some way to the entrance, since they do not communicate with the interior of the palace. The eastern wall both of the gateway and of the outer courts has fallen, so that the architectural scheme of the façade cannot be determined. It is certain, however, that it was not symmetrical, for the courts are not symmetrically disposed, nor is the north wing equal in length to the south wing. To the south of the central gate lie two courts, A and B, 10·10 metres from north to south, and 9·35 metres from east to west. Court A is provided with a pair of small rectangular chambers on either side; in court B there are two rooms upon the south side only. There are slight variations in size between these chambers, but they average about 4·10 metres square. They communicated with the court, but not with one another. They have all been covered by conical domes set over the angles on squinch arches. I give an example from No. 6 which will serve to illustrate the construction in every case (Plate 65, Fig. 3). Many of the rooms had a small niche in one wall (Plate 65, Fig. 2), the ṭâqchah, which is to be seen in all Persian houses; it appears again in numerous rooms in the body of the building. In No. 6 the niche is unusually large and, though it has broken through, the plaster decorations on the archivolt are preserved (Plate 66, Fig. 1). They consist of three fillets, and above the archivolt the small oversailing band of plaster which marks the springing of the dome is lifted so as to form a rectangular label. As can be seen from the photographs, most of the plaster has fallen from the walls; where it remains it is usually decorated with an insignificant striated motive consisting of narrow vertical and horizontal bands of five lines each, which look like the impress of some coarse matting on the wet plaster. To the north of the central gate there are two rooms, 9 and 10, communicating with one another. Further north lies a large court, _C_, 14·10 metres long, with two rooms at either end. Nos. 11 and 12 differ from the usual arrangement. No. 11 measures 6·20 by 4·05 metres and has a niche in the east wall. The north wall, which contained the door into the court, has fallen. No. 12, 1·65 × 4·20 metres, opens into the court by a narrow door in the north-west corner, part of the wall having been cut away to allow space for it. Nos. 13 and 14 are domed rooms of the customary type. In No. 14 the north-west squinch is particularly well preserved, part of the plaster fillets over the archivolt being still in place (Plate 66, Fig. 2).

The central gateway opens into court D, 31·50 × 13·30 metres. At the western end of the south wall of this court there are faint traces of plaster decoration, shallow arched niches separated by engaged colonnettes. The court terminates in a second vaulted gateway (15), which is so much ruined that the details of its structure cannot be made out (Plate 67). On either side of this gate a low archway leads into the vaulted passages 16 and 17. At the eastern end of court D a door gives access to a chamber (18) 27 × 4·20 metres, which forms the east side of court E and opens into that court by two wide doorways. To north and south of court E lie chambers 19 and 20, 12·40 × 4·20 metres and 12·40 × 4·20 metres, which open into the court by three arches carried on masonry piers varying from 2·50 to 2·80 metres in length. On the west side of the court, No. 21 corresponds with No. 18, but the greater part of its walls have fallen. Court F is flanked to the south by No. 23, 11·50 × 4·20 metres, a closed chamber with a single door, and to the north by No. 22, which is only 9·10 metres long in order to allow space for a door leading into No. 24 (11·40 × 4·40 metres). The west side of court F is partly occupied by the vaulted passage (16) and partly by No. 25, a room which no doubt communicated with the court by a door. A door leads from it into No. 26, whence a pair of doorways give access to court G. No. 27 lies to the north of court G and communicates with No. 28, to the north of court H. No. 28 in turn communicates with No. 29, lying parallel with Nos. 30 and 31, two rooms that open out of the west side of court H. Back to back with Nos. 29, 30, and 31 lie Nos. 32, 33, and 34, with doorways opening west. The vaults of these six chambers are well preserved. Plate 68, Fig. 1, shows the interior of No. 31 with an arched ṭâqchah in the wall. The vault is ovoid and oversails the wall.

The courts in the south wing of the palace correspond neither in size nor in disposition with those of the north wing. Opposite to the door of No. 18 a door leads into No. 35, which is an isolated chamber with a deep niche at the south end. Court I can be approached from court D only by a circuitous route through passages 17 and 45. Upon the east side of court I lie the two rooms 36 and 37, 4·40 metres wide and respectively 7·85 and 8 metres long. On the south side there is a group of rooms preceded by an antechamber, of which nothing is standing but a return at the east end of the wall or arcade. Three doors lead out of the antechamber into rooms 39, 40, and 41. In the central chamber (39) there is an arched niche at either end leaving a space 4·15 metres square which was covered by a dome set on squinches (Plate 68, Fig. 2). To east and west, the dome rested upon the arches of the doors leading into Nos. 40 and 41. Beyond 41 there is another room, 42, which was accessible from 41 only. On the north side of court I are two small rooms, 43 and 44, about 4·15 metres square and much ruined. Further west is the entrance to corridor 45. Court I is separated from court J by a wall which is ruined to its foundations. On the south side there is a single long chamber (47) with an antechamber; the north side is occupied by corridor 45, which is accessible from court J by a door in the north-west corner of the court. Corridor 45 communicates with corridor 17, a transverse arch separating the two. I call attention to the fact that the vault builders were always careful to avoid intersection; when two barrel vaults meet at right angles, the one is always divided from the other by a transverse arch. This is very noticeable in corridor 17, where the vault is standing. In the eastern arm of the corridor, opening out of court D, the east and west vault terminates against a transverse arch so as to allow the north and south vault of the western arm to run straight through to the head wall at the northern end.

The western arm of corridor 17 opens into court K. The north and west sides of this court are completely ruined and represented only by grass-grown heaps of stones. On the south side there is a true lîwân group (49, 50, 51) with an antechamber, the lîwân (49) opening into the antechamber through a wide archway, the side chambers (50 and 51) by means of doors. To the west of these chambers there is an open space with no buildings standing upon it; even the outer wall is completely ruined. It is here that the south gate is placed in the French plan. Some 19 to 20 metres west of No. 50, two chambers (52 and 53) with an antechamber are partially preserved. A mound of stones and grass runs northward, continuing the west wall of Nos. 51 and 53. East of this mound, at any rate at its northern end, there were ruin heaps indicating chambers, but I was not able to discern their exact form or extent, nor yet their relation to the hall 54. This hall is a chamber 16·15 metres square, with walls 3·90 metres thick which carried a dome set upon squinch arches (Plate 69, Fig. 1). No part of this dome is standing, but it is safe to conjecture that it was built of brick.[48] The method of constructing the squinches can be seen best at the south-west angle (Plate 69, Fig. 2). An archway, 5·70 metres wide, breaks the centre of each wall. The round arches were built of brick, but on the south side only is any considerable portion of the brickwork preserved (Plate 70, Figs. 1 and 2). The bricks are laid horizontally, not vertically, i.e. with the narrow face outward. Above each archway there is a small round-headed window. On the exterior the face of the walls has perished to a considerable extent. Between the top of the archways and the bottom of the windows the wall would seem to have been recessed back slightly (Plate 71), and at this level the corners of the building appear to have been sliced off, thus reducing the mass of masonry behind the squinches. This effect may, however, be produced merely by the decay of the masonry, for the lower part of the walls also has invariably broken away at the angles. At the north-east and north-west corners I noticed some brickwork embedded in the stone masonry. No. 54 stands 9 metres from the western outer wall, of which at this point nothing but foundations remain. At the north-west angle there are ruins of four chambers (55, 56, 57, 58) placed two deep, and to the south four chambers (59, 60, 61, 62) lie parallel to one another along the wall. No. 62 breaks off abruptly with a high peak of masonry (Plate 72), possibly part of an upper story. I saw no trace of any building further to the west.