part ii. p. 126, where Dr. Herzfeld points out that a chamber somewhat

similarly placed in the palace of Mshatta may also be a mosque.)

No difficulty will be found in following on the plan the arrangement of the upper floors in the northern part of the palace. In the second storey, the space marked B^{2} is occupied by the vault of the great hall B (Fig. 81). At A^{2} three windows open into the hall from the room in the second storey. R^{2} and S^{2} correspond with the two courts R and S. In the third storey the rectangular space A^{3} is unroofed, and the space B^{3}, below which lies the vault of the great hall, is also unroofed (Fig. 82). The eastern part of this storey is completely ruined, but there would appear to have been rooms

round R^{3} similar to the rooms round R^{2}. The _chemin de ronde_, T T´, is on a level with this storey.

Between the main palace block and the eastern fortification wall there lies a group of rooms which is clearly an addition to the original scheme. It is interesting to observe that these rooms are in all essentials of their plan a repetition of the group of rooms to the south of court D. Room U corresponds with the antechamber E; room V with the square room F; W with the cloister J; X, Y, and Z to G, H, and T. But the columns in I I´ are not repeated in the small rooms, Z Z´; room V is covered with a groined vault instead of the barrel vault of F, and the court A is not closed with a wall like the court K. I make no doubt that both these groups of rooms, which are so strikingly similar in arrangement, were intended for the same purposes, and I conjecture that they were ceremonial reception rooms. Herzfeld has compared E and F with the throne room of Mshatta (_Der Islâm_, loc. cit.).

All the rooms and corridors of the palace are vaulted. Some of the finer vaults are built of brick tiles (for example, over the great hall B and over rooms E, F, I, and I´), but as a rule the vaults are constructed with stones set in mortar, the stones being cut into thin slabs so as to resemble bricks as closely as possible. (_Cf._ the Sassanian palace of Firûzâbâd, Dieulafoy, _L’Art Ancien de la Perse_, vol. iv.) All the vaults, whether of brick or stone, are built without centering, and all are set forward slightly from the face of the wall. (The same construction is found at Ctesiphon, see below, Fig. 109.)[83] The groined vault occurs seven times in the corridor C C´ (Fig. 98), and it is also found in room V. (See my article in the _Hellenic Journal_ above cited.) The fluted dome over room A is bracketed across the corners of the rectangular substructure (Fig. 88). In several cases where a barrel vault terminates not against a head wall, but against another section of barrel vault, it is adjusted to the angles of the substructure by means of squinch arches (Fig. 99). A noticeable feature of the vault construction of Ukheiḍir is the presence of masonry tubes running between the parallel barrel vaults (Fig. 100). The structural purpose of these tubes is to diminish the mass of masonry between the barrel vaults. Whenever two barrel vaults lie parallel to one another, a tube will be found between them, and similar tubes exist between the vault of the cloister O O and O´ O´ and the outer wall. (See too Fig. 95, which shows a tube between a barrel vault and a straight wall.) Over the vaults of the rooms of the annex in the eastern part of the court, and also over the vaults of the fourteen parallel chambers outside the enclosing wall to the north, a false roof is laid (Fig. 103). It serves as a protection against the heat of the sun. Under the eastern annex there are some much-ruined subterranean chambers. A staircase at the south-eastern angle of court D leads down into similar cellars (serâdîb).

The arches over the doorways are usually of an ovoid shape, sometimes slightly pointed. When the door-jambs take the form of engaged columns, the capitals of the columns, roughly blocked out in masonry, carry an arch slightly narrower in width than the opening of the doorway beneath it. But when the door-jambs are formed merely by the straight section of the wall, the span of the arch is wider than the opening of the doorway (Fig. 102 illustrates both types). This set-back of the arch was doubtless employed in order to facilitate the placing of centering beams. Three wide doorways with round arches, b b´ and c, lead from the main block of the palace building into the surrounding court. The arches are usually characterized by double rings of voussoirs (_cf._ Ctesiphon and other buildings of the Sassanian and early Mohammadan period), the inner ring laid so as to show the broad face of the stones or tiles, while the narrow end shows in the outer ring. (See the arch in Fig. 102.) The arch construction in the eastern annex is, however, much rougher in style. The outer ring of voussoirs is omitted there, nor is it invariable in other parts of the palace.

The niche plays a large part in the decoration of Ukheiḍir. A row of narrow niches runs along the top of the outer face of the northern enclosing wall, but very little of it is now left (Fig. 87). The southern face of the three-storeyed block bears an elaborate niche decoration (Fig. 91). Here the lowest row of niches forms part of the series already mentioned which runs round court D. Above these, on the second storey, are remains of another row of arched niches, each of which contains three small niches. So far as I know, this feature of a large niche enclosing groups of smaller niches has not yet been observed in Sassanian architecture. It is found, however, in a certain well-known type of early Christian church (see, for instance, Ala Klisse, published by me in the _Thousand and One Churches_, p. 403). On the third storey of the palace the face of the wall has been left blank, but above the windows there are still traces of a third order of small niches. Pairs of niches flanked by engaged columns are to be seen in room G´. They are set high up in the wall between the transverse arches. On these transverse arches there is a plaster decoration, the same in character as that which occurs in the semi-domes at the ends of the vault in Court S (Fig. 101). The motives there used are the flute (in the squinch arch and in the conical segment of the semi-dome above it), and a pattern which resembles a tiny battlemented motive. Upon the transverse arches the battlemented motive is doubled so as to form diamond-shaped patterns. In the centre of each of these diamonds, and in the centre of the tiny arched niches at the bottom of the vault, and also between those niches, there are small funnel-shaped motives formed of concentric rings. Between the transverse arches there is a boldly worked ribbing. The arch round the eastern of the two doors that leads into corridor Q´ is surrounded by cusps (Fig. 94). (_Cf._ Ctesiphon, Dieulafoy, _op. cit._, vol. v. plate 6.) A blind arcade, borne by pilasters, is to be seen in courts M M´ and N N´. In the antechamber U there are shallow niches on either side of the doors.

With regard to the date of Ukheiḍir there are three possible hypotheses. It may belong--

1. To the Sassanian or Lakhmid period prior to the Mohammadan conquest.

2. To the 150 years after the Mohammadan conquest.

3. To the Abbâsid period, _i. e._ after A.D. 750.

1. In defence of the first theory can be urged the close relationship between Ukheiḍir and other places of the Sassanian age, not only in plan (_cf._ Ḳaṣr-i-Shîrîn, de Morgan, _Mission Scientifique en Perse_, vol. iv., part 2), but also in the technique of brick and stone masonry and in the principles of vault construction (_cf._ Ctesiphon, Firûzâbâd, and Sarvistân, Dieulafoy, _op. cit._). But since it is certain that the arts of the early Moslem era were dominated in Mesopotamia by Sassanian influence, these affinities do not offer a convincing proof of a pre-Mohammadan date. Even if Ukheiḍir belonged to the early Moslem age, it might, and probably would, have been built by Persian workmen. At the same time certain architectural features, such as the groined vault and the fluted dome, have not hitherto been observed in any Sassanian building. The earliest Mesopotamian example of the groined vault known to me, besides the groins of Ukheiḍir, is that of which fragments can be seen in the Baghdâd Gate at Raḳḳah.

There is, further, a passage in Yâḳût’s Dictionary which might help to support the theory of a pre-Mohammadan origin (vol. ii., p. 626, under Dûmat ej Jandal). In the accounts given by the Arab historians of the invasion of Mesopotamia in 12 A.H. (A.D. 633-4), by Khâlid ibn u’l Walîd, frequent mention is made of ’Ain et Tamr, which Yâḳût expressly states to be the same as Shefâthâ (Shetâteh is the modern colloquial form of the name). When Khâlid ibn u’l Walîd had taken the oasis, which was inhabited by Christian Arabs, and appears to have been the one place that offered him serious resistance (Teano: _Annali dell’Islam_, vol. ii., p. 940), he is said to have marched on Dûmat ej Jandal, which he captured, putting to death its defender, Ukeidir ’Abdu’l Malik el Kindî.[84] It is generally admitted that the name Dûmat ej Jandal in this account is an error, and that the fortress which was taken by the Mohammadans in the year 12 A.H. was Dûmat el Ḥîrah. (For the reasons for substituting Dûmat el Ḥîrah for Dûmat ej Jandal in Ṭabarî’s text, see Teano, _op. cit._, vol. ii., p. 991.) Now Yâḳût gives two conflicting traditions concerning the foundation of Dûmat el Ḥîrah, but he expresses no uncertainty as to its position. It was near to ’Ain et Tamr, and its ruins were known in Yâḳût’s day (thirteenth century). According to the first tradition given by Yâḳût, the Prophet sent Khâlid ibn u’l Walîd in the year 9 A.H. against Ukeidir, who was lord of Dûmat ej Jandal. Khâlid captured Dûmat ej Jandal and made a treaty with Ukeidir, but after the death of Mohammad, Ukeidir broke the treaty, whereupon the Khalif ’Umar expelled him from Dûmat ej Jandal. He retired to Ḥîrah and built himself a palace near to ’Ain et Tamr, which he called Dûmah. This Dûmah, near ’Ain et Tamr, is no doubt Dûmat el Ḥîrah which Khâlid besieged and took in the year 12 A.H. The second tradition is substantially the same as the first as far as the Mohammadan invasion is concerned, but Yâḳût here implies that Ukeidir dwelt in the first instance at Dûmat el Ḥîrah, and was accustomed to resort to Dûmat ej Jandal for the purposes of the chase, and he adds that Ukeidir named Dûmat ej Jandal after Dûmat el Ḥîrah. Prince Teano (_op. cit._, vol. ii. p. 262) has exposed the improbabilities which attend this explanation, and he concludes that both traditions are equally untrustworthy, and doubts the authenticity of any part of the story of Ukeidir. It does, however, appear to me to be possible that the ruins of Dûmat el Ḥîrah which were standing in Yâḳût’s day were no other than the abandoned palace of Ukheiḍir, though it is not necessary to accept either of Yâḳût’s versions of the story of its foundation.

2. If the palace is to be ascribed to the period immediately succeeding the conquest, it would be a Mesopotamian representative of the group of pleasure palaces which were built upon the Syrian side of the desert by the Umayyad princes (Lammens: _La Badia et la Ḥîra, Mélanges de la faculté orientale_, Beyrout, vol. iv., p. 91). But whereas it was natural that the Umayyad khalifs should have constructed hunting palaces in that part of the desert which lay on the direct road between their capital of Damascus and the spiritual capitals of their empire, Mecca and Medina, it is difficult to see why they should have selected a site so far from any of their habitual residences as Ukheiḍir. It is true that the Khalif ’Alî made Kûfah his capital for five years. He was assassinated there in A.D. 661. But during those years he was ceaselessly occupied in quelling rebellions, and I dismiss the possibility that he should have found leisure to build or to use the palace of Ukheiḍir.

3. I am not disposed to place Ukheiḍir as late as the Abbâsid period. The Abbâsid princes had lost the habit of the desert which was so strong a characteristic of their Umayyad predecessors. When they moved away from their capital of Baghdâd they built themselves cities like Raḳḳah and Sâmarrâ. Moreover, the architectural features of Ukheiḍir, both structural and decorative, present marked differences from those of the ruins at Raḳḳah and at Sâmarrâ, and on architectural as well as on historical grounds I am inclined to ascribe Ukheiḍir to an earlier age.

Whether that age be immediately before the Mohammadan conquest, or whether it fall shortly after the conquest, during the Umayyad period, I do not think we are as yet in a position to determine. It is to be borne in mind that the ruins of the palace bear witness to two different dates of building. The eastern annex and probably the edifice outside the enclosing wall to the north are an addition to the original plan and must be of a slightly later date.