CHAPTER IV

GENESIS OF THE EARLY MOHAMMADAN PALACE

The palace of Ukhaiḍir is not an isolated phenomenon. It belongs to a group of buildings which exhibit in varying proportions the characteristic features of the fortress and of the pleasure-house of princes. These buildings are scattered over the western frontiers of the Syrian desert; Ukhaiḍir is as yet the sole example of the type which has been discovered upon the eastern side. They are a logical outcome of the period of cultural transition during which they arose, the difficult and distasteful passage from nomadic to settled life; they attest the abiding call of the open wilderness, to which the poets and chroniclers of the first century after the Hidjrah are faithful witnesses. To the Arab the desert is more than a habitation; it is the guardian of traditions older and more deeply rooted than those of Islâm; of traditions which are sacred to his race; of his purest speech, and of his finest chivalry. It is for him the natural theatre of his actions, and there is no other stage on which he can play out his part. To this day I have heard the Beduin speak of themselves as the Ahl al-Ba’îr, the People of the Camel, just as they spoke of themselves in the early centuries as Ahl al-Ḍar’, People of the Udder.[49] The authority of the Prophet was powerless to stay the current of his race. ‘Periodically the Arabs succumbed to the allurement of the camel, to the need to drink of its milk. The Prophet himself was not exempt, since he prayed God to preserve him from it. For his nation, said he, he dreaded the diet of milk. When his companions expressed their astonishment at his fears, he replied: “The passion for milk will lead you to abandon the centres of reunion and to return to nomad existence.”‘[50] His immediate successors followed the example set by him, but the national inclination was not to be restrained, and the Umayyad khalifs returned to the habits of their forefathers. Their capital was Damascus, but their residence was the Syrian desert. They escaped to the bâdiyah, the spring pasturage in the rolling steppes, where the tents of the Ṣukhûr still cover the plain when the winter rains are past; they transported their courts to the ḥîrah, the palace camp.

The word ‘ḥair’ denotes a camp, a castle, or a villa.[51] The original signification does not seem to have implied solid constructions, but rather the headquarters of a desert princeling and his retainers. Such an assemblage must necessarily have been mobile. The exigencies of pasturage and the uncertainties inherent in tribal predominance, where the limits of authority cannot be expressed in terms of geographic definition, were alike unfavourable to stable residence. Joshua the Stylite[52] talks of the ḥertâ of Nu’mân ibn Mundhir as having withdrawn into the inner desert before the attack of the Tha’labites--it must therefore have been a movable camp; on the western borders there is no certain evidence that the Ghassânid princes possessed either fenced cities or garrisoned fortresses.[53] But before the dawn of the Mohammadan era the ḥîrah had begun to change its character, and the nomad encampment to develop into the standing camp and even into the city. The Ghassânids must have had a fixed establishment in the Djaulân,[54] and some of the existing ruins on the eastern frontiers of the Ḥaurân may date from their time. At Khirbet al-Baiḍâ, for example, I could find no certain trace of Roman handiwork. The plan might date from the age of Diocletian, but the decorations betray a different origin.[55] Yet I cannot place them as late as the Umayyad period. Djebel Sais I have not seen.[56] The plan of the bath recalls the arrangement of the chambers at Qṣair ‘Amrah, and it may therefore be Mohammadan. At Qaṣr al-Azraq, Dussaud found a dedication to the emperors Diocletian and Maximian, but the fortress would seem to have been rebuilt in the thirteenth century A.D.[57]

Similarly upon the eastern side of the desert, the Lakhmid camp had grown into an important town, which absorbed the generic title and was known as al-Ḥîrah, the standing camp _par excellence_, the capital of Persian Arabia. But no sooner did the Lakhmid princes find themselves enclosed within the walls of a city than they threw out fresh ḥîrahs into the desert: palaces, the magnificence of which haunted the imagination of Beduin poets of the Days of Ignorance and gave birth to legendary tales and to moral aphorisms which were recorded with pious, if uncritical, exactitude by the historians of Islâm. We know the site of the most famous of these pleasaunces, Khawarnaq.[58] Ibn Baṭûṭah, in the fourteenth century A.D., saw the remains of its immense domes on the edge of a canal which was fed by the Hindiyyeh branch of the Euphrates. In his day it was still inhabited. The existing ruin mounds, standing upon the brink of the Sea of Nedjef, are covered with the sherds of mediaeval pottery. The canal has now silted up and the Sea of Nedjef is dry. I was told at Nedjef that thirty or forty years ago the lake was full of water, and that the climate of the town, never very much to boast of, had been considerably affected for the worse by the change. Below the town, the bed of the lake is occupied by palm-gardens and cornfields, watered by a canal recently constructed. What was its condition in Sasanian times I do not know. The lake was dry in the Middle Ages,[59] but ‘Adi ibn Zaid speaks of the Nu’mânid lord of Khawarnaq as having looked from his palace walls and rejoiced at the sight of the sea.[60] It is difficult to imagine that any one could have rejoiced in the Baḥr Nedjef if it had worn its present aspect. The extent of the mounds of Khawarnaq is not large, though my impression is that part of the steep earth cliff overhanging the Baḥr Nedjef has fallen away and carried the castle walls with it. The ancient canal from the Hindiyyeh lies about a quarter of a mile to the north of the mounds. Legend has been busy in accounting for the origin of the castle. It is said to have been built by Nu’mân ibn Imra’ al-Qais, by order of the Sasanian king Yazdegerd I, who desired that his son, Bahrâm V Gûr, should be brought up in the salubrious air of the desert above Ḥîrah. This would place its foundation in the early part of the fifth century A.D.[61] The architect was a certain Sinimmâr, a Byzantine (Rûmi) according to some authorities,[62] nor need this assertion excite surprise. A century later Justinian lent workmen to Khusrau I, when the latter was engaged in building the new Antioch near Ctesiphon. Other Lakhmid ḥîrahs are mentioned besides Khawarnaq, but they are to us nothing but a name. Al-Sadîr stood in the desert ‘that lies between al-Ḥîrah and Syria’,[63] presumably not far from Khawarnaq, since the two castles are frequently mentioned together. We hear also of al-Ṣinnîn, where ‘Adi ibn Zaid was imprisoned.[64] Of greater importance was al-Anbâr on the Euphrates, which was rebuilt by Shapûr II in the early part of the fourth century.[65] None of the Lakhmid ḥîrahs in the desert, except Khawarnaq, have been identified. In 1911 I rode out across the Baḥr Nedjef from Khan Muṣallâ to see a ruin called al-Ruḥbân, which was reputed to be ancient, but found nothing except a mud-built wall erected by the Bani Ḥasan. A few palm-trees had been planted near it. My guide, a sheikh of the tribe, was much distressed when I denied to Ruḥbân the antiquity which had been claimed for it. ‘Mistress,’ he expostulated, ‘before my beard was grown, I saw it here.’ His age I should judge to have been no greater than my own, and Ruḥbân may have had the advantage of us by a decade. After this disappointment I declined to visit other quṣûr of the Bani Ḥasan (qaṣr = fort, is the name which is applied to any walled village or palm-garden) though he mentioned a considerable number. Subsequently a mullah of the Nedjef mosque told me that there were ancient remains at Ḥiyyadhiyyeh, which lies somewhere between the Baḥr Nedjef and Ukhaiḍir, to the south of the line across the desert which I had followed. Ḥiyyadhiyyeh is mentioned by Niebuhr in his itinerary from Baṣrah to Aleppo by the desert road--Meshed ‘Ali, el Tukteqâne or el Heiadîe, el Hossian, el Chader (Ukhaiḍir) Ras el ‘Ain.[66] I doubt whether there is much to be found on the surface at Ḥiyyadhiyyeh, for the Bani Ḥasan have planted palm-groves there, and in so doing, they have probably destroyed most of what was old, but the mullah asserted that a Lakhmid castle had stood at that spot and another at Ruḥbeh, which he said was identical with Qâdisiyyeh.[67] I give his opinion for what it is worth, which is very little. There are, however, no doubt old ruins at Ruḥbeh, whether Lakhmid or of a later time, if it occupies the site of Qâdisiyyeh--a very possible hypothesis. It was a large village in A.D. 635, when the Mohammadan invaders defeated the Persians close to its walls. Muqaddasi knew it as a walled town on the pilgrimage road. Mustaufi (fourteenth century) describes it as mostly in ruins, while Ibn Baṭûṭah speaks of it as a large village.[68] The Sâl Nâmeh of the Vilâyet of Baghdâd mentions a ruined qaṣr at Ruḥbeh.[69] The sheikh of the Bani Ḥasan gave me the names of ‘Izziyyeh,[70] and ‘Atiyyah as quṣûr of his tribe, but he did not think that there were ruins at either place.

To our scanty information concerning the pre-Mohammadan ḥîrahs one other item is to be added. Mas’ûdi gives an account in the following terms of a palace built at Sâmarrâ by the khalif Mutawakkil (A.D. 847-861) in imitation of a Lakhmid ḥî ah: ‘Mutawakkil in his days raised a building such as no man knew, it is that which is called the _ḥîri_ and the two wings (literally sleeves) and the porticoes (_arûqah_). And that was because a companion of his vigils related to him upon a certain night that one of the kings of Ḥîrah, a Nu’mânid of the Bani Naṣr, erected an edifice in his capital, which was al-Ḥîrah, after the model of an army in battle. (The word I have translated by _army in battle_ is _ḥarb_ = war or campaign; Dr. Herzfeld suggests that it must be taken here to mean military camp--a somewhat hypothetical emendation)[71]. For such was his infatuation for war and his love of it; so that the memory of it might never vanish from him under any condition. In this edifice the portico was the audience chamber of the king, and this was the centre (literally the _breast_); and the two wings (_sleeves_) lay to right and left. In the two dwellings which formed the wings lodged those who stood nearest to him among his courtiers. In the right wing was the wardrobe, and in the left wing was kept such wine as was needed. The open court of the portico was common to the centre and to the two wings. The doors, three in number, led to the portico. To this day this building (i.e. Mutawakkil’s copy) is called the _ḥîri_ and the two wings in allusion to al-Ḥîrah. And the people followed Mutawakkil, imitating his creation, which is famous to the present time.’[72] The word _riwâq_, which I have translated ‘portico’, does not necessarily imply the existence of columns, though it is used for the porticoes which surround the court of a mosque. Its primary signification is a roof in front of a tent, supported by a single pole in the middle.[73] I shall have occasion to return later to this important passage (see below, p. 86).

But if we have little knowledge of the Lakhmid ḥîrahs which were the precursors of Ukhaiḍir on the eastern frontiers of the desert, we have another and a richer source of information in the Sasanian palaces. The Lakhmid princes stood in close relations with the Sasanian empire. Among the officials of the Persian court there was an Arab secretary whose special duty it was to conduct the correspondence with ‘the land of the Arabs’. Moreover, it is related that the Arab phylarch paid a yearly visit to the court of the Chosroës.[74] To a Lakhmid the education of a Persian prince was entrusted, and Lakhmid armies placed Bahrâm V upon a contested throne. The Christians of Ḥîrah belonged to the Nestorian church, the church of Assyria; we hear of one, the poet ‘Adi ibn Zaid, who was Arab secretary and enjoyed great influence with Khusrau Parwêz. Half allies, half vassals, the Lakhmid phylarchs fought side by side with the Persians against Rome;[75] they were sufficiently independent to receive an embassy from the Byzantine emperor, and sufficiently important to warrant an attempt on his part to buy them over from the Sasanians. Finally, at the beginning of the seventh century, Khusrau Parwêz set the Lakhmid dynasty aside and established in place of Nu’mân III an Arab of the Ṭayy, who lived and held his court at ‘Ain al-Tamr near Ukhaiḍir. Possibly the huge walls of Qaṣr Sham’ûn, on the outskirts of the oasis,[76] may date from the time when ‘Ain al-Tamr was the residence of the phylarch. But he was no longer an independent ruler; a Persian adviser was appointed to assist him, and a few years later the state was converted into a province of the Sasanian empire under a Persian regent. Independent or subject, the civilization of Ḥîrah must have been modelled upon that of Ctesiphon; Persian influence must have been predominant in its arts and its architecture, and the Lakhmid ḥîrahs must have reflected the glories of Sasanian palaces. It is to these palaces that we should look first for an explanation of the architectural scheme of Ukhaiḍir. One reservation must, however, be made. It is true that Ukhaiḍir cannot be regarded as primarily a fortress. The absence of any sufficient provision of water would have been a fatal weakness in time of siege. No cistern exists within the palace; no ancient well has been found, and if the conditions were the same of old as they are now (which is, however, by no means a safe assumption), any water within the palace would have been too brackish to drink, as is the case in the modern well in the palace yard. Moreover, the outer ring of walls, which encloses the northern annex, was obviously too weak for defence; it is more like the garden wall of a pleasure-ground. Nevertheless, considerable care has been lavished upon the defences of the main building. They were, and they are to this day, adequate for the spasmodic warfare of the Arab tribes. In the very act of construction the architect seems to have bethought him that such protection was necessary and to have added a strong girdle to his palace plan. On the other hand, the Sasanian palaces, so far as they are known to us, are either unfortified, or they stand within a fortified park, the walls and towers of which are not in direct structural relation with the residential buildings. At the same time Sasanian military works, where they have been examined, do not differ materially from those of Ukhaiḍir; the fortress of Qala’-i-Khusrau at Qaṣr-i-Shîrîn is an excellent case in point (Plate 73, Fig. 1). It is a rectangular enclosure, about the size of Ukhaiḍir (roughly 180 metres square), surrounded by a wall which is strengthened by rounded towers. The towers are somewhat differently disposed from those of Ukhaiḍir; they are larger and they are set twice as far apart, but the scheme is the same in both places. The interior buildings are much ruined. A row of chambers, or more probably, from the width of the ruin heaps, a row of small courts with chambers grouped round them, adjoined the inner side of the walls, leaving a central court which was partly filled by a large building, rectangular in plan. The town wall of Dastadjird was also furnished with rounded towers.[77]

Almost without exception the plan of the Sasanian palaces is a development of the lîwân type, the origin of which is to be sought in the southern Hittite sphere, northern Syria and the mountain lands north of the Mesopotamian plain. The architecture of this region is known to us best through the excavations at Zindjirli, where the evolution of the southern Hittite palace can be traced over a period of close upon a thousand years.[78] It is an evolution which is dominated

from the first to last by the monumental gateway. At Zindjirli the type appears in its earliest and simplest form in the gateways of the inner city wall, which Professor Koldewey places approximately in the thirteenth century before our era.[79] A doorway set back between a pair of solid towers leads into a narrow

court, placed latitudinally, with a second doorway opposite to the first (Fig. 5, D). Three hundred years later this structure is adapted, in the earliest khilâni palace, to residential purposes (Fig. 5, G).[80] The solid towers remain, but the space between them has been converted into a covered portico, or lîwân, and the inner latitudinal court has become a latitudinal hall with a small chamber at either end. The further development is characterized by the multiplication of chambers and the disappearance of features proper to the fortress. In the khilâni palace erected after Asarhaddon’s destruction of the city in the first half of the seventh century (it appears in Fig. 5 to the north-west of G), the arrangement of the subsidiary chambers is conceived on freer lines, the walls are thinner, the flanking towers of the lîwân have disappeared, and in their stead are set tower chambers; in short the fortress towers have given place to a purely decorative motive, the towered façade, which was destined to have a long and honourable history in Christian architecture.[81] That the Hittite khilâni was imitated by the Assyrians during the eighth and the seventh centuries we know both from inscriptions and from excavations.[82] To it the Assyrian builders owed the introduction of the column, which was foreign to their architecture. At Pasargadae the khilâni reappears in a form which bears testimony to its Hittite parentage.[83] The façade towers, the columned lîwân, the orthostatic construction, and more significant still, the latitudinal disposition of the chambers, are all to be found in the Pasargadae palaces, but the greater depth which was given to the principal room necessitated the introduction of a double line of columns to support the roof (Fig. 6). At Persepolis and at Susa the same scheme is carried out in colossal dimensions. It is found alike in the gigantic apadanas and in the palaces, in the one case adapted to the ceremonial magnificence of the Persian king of kings, in the other to the requirements of the dwelling-house. In the apadana, the lîwân was deepened and a second row of columns was added to the first; the hall of audience was magnified into a huge quadrangular chamber, the roof of which was supported by a forest of columns; solid towers of unburnt brick flanked the lîwân, and subsidiary lîwâns occupied the space behind them on either side of the audience hall (Fig. 7). In the palaces the towers were hollowed out into rooms correspondingly in depth with the lîwân, and the audience hall was flanked by side chambers. Where space permitted, as in the palace of Darius at Persepolis, additional rooms were disposed round a courtyard at the back of the edifice. So constituted, the Achaemenid palace reproduced the traits of the later khilânis at Zindjirli in a form adapted to new requirements (Fig. 8).

Before the khilâni palace was taken up again by Persian hands, an immense revolution had swept over western Asia. Alexander’s invasion is a turning-point

in history. The Mesopotamian arts emerged from the period of Greek rule profoundly modified by direct intercourse with the West; for the Seleucid kingdom, with one capital on the Tigris and another on the Orontes, had bridged the gulf between Babylonia and the Mediterranean coast-lands. Greek culture, Greek artistic conceptions were carried across Asia by the invaders; but the further they penetrated, the less they overmastered local tradition. Babylonia, Assyria and Persia were never Hellenized in the sense in which Syria was Hellenized. The ancient East, with 3,000 years and more of a highly elaborated civilization behind her, assimilated what was brought to her, but she used it after her own fashion. She turned the Greek kings into oriental despots, and translated Greek ideas into her own forms of expression. The architectural remains of this period are as yet scanty. Seleucia and Antioch are unexplored, and except for the Greek theatre at Babylon, the excavation of Mesopotamian sites has yielded little but fragments.[84] But if the Seleucid era is comparatively unknown, the new elements which the Greek conquest had introduced into oriental architecture stand out with an amazing vividness in Parthian buildings. Loftus, whose excavations at Warka were the first to reveal a great Parthian settlement on a Babylonian mound, was not slow to appreciate the significance of his discoveries.[85] Together with capitals which bore an obvious relationship to the Ionic, and walls enriched with Ionic half-fluted engaged columns, he found plaster ornaments and fragments of wall-surface decoration covered with continuous geometric patterns in which he recognized an art that was essentially oriental. The Chaldaean monuments at Warka were covered with mosaics set in geometric designs which are the prototypes of the Parthian coloured reliefs.[86] Hellenistic houses of the Parthian period have been unearthed in the Amrân mound at Babylon. The small Parthian palace at Niffer, with its columned hall of audience, opening through an anteroom, which is in the nature of a closed lîwân, into a square peristyle, resembles a Greek dwelling-house seen through a Babylonian medium[87] (Fig. 9). At Assur, together with a temple (if temple it were) which is almost peripteral,[88] and a stoa,[89] we have a palace on a lîwân plan, with ionicizing capitals and a façade of stucco mock-architecture

which indicates the road that led from the Hellenistic façade in two orders[90] to the stucco façades of Ctesiphon and Ukhaiḍir.[91] At Hatra a building which looks like the Parthian conception of a temple in antis stands in the court of a monumental lîwân palace,[92] but so far as can be judged without excavation the Hellenistic house is conspicuous by its absence. Not only the royal palace (Fig. 10) but also such of the smaller palaces as are known to us through the admirable publication of the Deutsche Orient-Gesellschaft, show a strongly characterized lîwân plan. To the Parthian interpretation of the venerable khilâni scheme the Moslem East has remained unswervingly true. The lîwân, as it is to be seen at Hatra, dominated the fancy of the Sasanian and of the early Mohammadan architects, and it continues to be an indispensable part of the modern house of Damascus or Baghdâd--except indeed the post-modern, which are wretched imitations of the worst European styles, but these are found more often in ultra-civilized Syria than in Mesopotamia. The huge Parthian lîwân was possibly a result of the introduction of the vault. The great hall, in which, no matter what its size, the interior space was unbroken by pier or column, was a setting for princely state which could not be enhanced by any

architectural device. Portico and audience chamber were blended together, and the columns of the one served to enrich the walls which flanked the monumental archway of the other.

The vault itself was not a new feature. It was well known to Babylonian and to Assyrian builders, by whom it was used to cover spaces of narrow span.[93] Vaulted drains and tombs are of frequent occurrence, and Place found a barrel vault with a span of 4 metres in the gateways of Sargon’s palace at Khorsâbâd.[94] But though the principles of vault construction were familiar, the vault does not seem to have been developed to any notable extent before the second Babylonian empire at the earliest. Félix Thomas claims to have found the remains of monumental vaults in Sargon’s palace, but the proofs which he adduces are not convincing. There is no direct evidence for the domes which Place reconstructs over the rectangular chambers adjoining the temples, the area of the palace which was known in his days as the Harâm.[95] Layard found no trace of monumental vaults in his excavations of Assyrian palaces,[96] nor have any been discovered by the German excavators at Assur. Professor Koldewey is of opinion that the great hall at Babylon was vaulted, since, in the absence of all trace of columns, no other way of covering it is conceivable; and though direct evidence is not forthcoming, there is a strong likelihood that the proportions of the vault may have been greatly increased, and its structural value much more fully realized towards the end of the seventh or the beginning of the sixth century before Christ.[97] There are no data for its employment in Mesopotamia during the Hellenistic period, but it may safely be assumed that the absence of vaulted buildings in the eastern parts of the Seleucid kingdom is fortuitous. From the fourth century B.C. onwards western Asia shows a continuous series of cut stone vaults of small span,[98] many of which exhibit traits which point to their derivation from the sun-dried brick vaults of Assyria or from the cut stone vaults of the Saitic period in Egypt, themselves a derivation from sun-dried brick construction. In the second half of the third century, vaults with similar characteristics appear under Hellenistic influence in central Italy, where, after the middle of the second century, they underwent a development to which the Hellenistic East can offer no parallel.[99] At the end of the second century, while Latin builders threw their stone vaults securely over a span of 14·50 metres, as in the Ponte di Cecco in the Via Salaria, and even of 18·50 metres, as in the Pons Mulvius,[100] the Greeks of Asia Minor did not venture upon a span wider than 7·10 metres,[101] and confined themselves as a rule to vaults under 4 metres in span. It was now the part of the East to learn from Imperial Rome. Western Asia took back its own creation from the hands of Roman builders in the vast proportions which the proficiency of the latter had given to it, and over the whole of the Roman Empire the monumental vault sprang into being. The earliest extant examples on Mesopotamian soil are the great vaults of the palace at Hatra.[102] Throughout the city, so far as our knowledge goes, the vault is systematically used, and for the first time it is constructed of dressed stone, not of brick. For it must be borne in mind that the expansion in Asia of the Roman Imperial stone and mortar vaulted architecture encountered a similar expansion of brick vaulted architecture in which both material and structure point to an ancient oriental tradition and an independent Asiatic origin.[103] If Hatra is the oldest example of the systematic use of the vault in a monumental building, the very presence there of a method so fully developed postulates a long evolution. That this evolution was oriental is suggested by the fact that the forms which the vault assumes at Hatra can be traced back, almost without exception, to Asiatic brickwork, while the systematic employment of the vault is foreshadowed in hollow substructures which date from the Hellenistic era, and even from earlier times.[104] In Babylon such substructures, several stories high, roofed with stone slabs, would seem to have been devised before Alexander’s conquest, while Strabo’s description, which probably applies to a Hellenistic reconstruction, mentions terraces in which the vaults rested on cube-shaped piers, vaults and piers being built of burnt brick with a mortar of asphalt. Moreover, Strabo notes that in Seleucia, the capital of the Hellenistic kingdom on the Tigris, all the houses were vaulted on account of the want of timber.[105] That these vaults were of brick goes without saying; stone was even more difficult to obtain at Seleucia than wood. In this connexion the possibility that Nebuchadnezzar’s great hall at Babylon may have been covered with a vault should not be overlooked.

The vaults of Hatra fall into five groups.

1. A primitive vault, composed of oversailing horizontal courses of stone is found in the small chambers of tombs (_Hatra_, ii, Figs. 93, 111, 155). Sometimes the walls incline smoothly inwards from base to summit until the space between them is narrowed sufficiently to admit of the imposition of a covering slab (_Hatra_, ii, Figs. 99, 118, 120, 155. In Fig. 155 the slope begins in the fourth course above the base). The vault built of oversailing horizontal courses was an obvious expedient for the roofing of narrow spaces, and it is, as might have been expected, widely distributed.[106] There is one instance at Hatra of a dome constructed in the same manner. It covers a rectangular chamber, 1·50 × 1·70 metres, and it is the solitary known example of an attempt on the part of Parthian builders to solve the problem of a circular vault over a rectangular substructure (_Hatra_, ii, Fig. 93).

2. The true vault oversailing the wall occurs in numerous tomb chambers (_Hatra_, ii, Figs. 100, 105, 125, 130, 144, 145, 149, 152, 163), as well as in most of the smaller rooms of the inner palace (_Hatra_, ii, Figs. 225, 226, 237, and Plate 8) (Plate 74, Fig. 2). It is a form which originated in brick building. It is found in Assyrian brick tombs,[107] but never, so far as my knowledge goes, in any dressed stone vaults save in those of Hatra. It appears at Ctesiphon in the side vaults,[108] and in the rough stonework of Qaṣr-i-Shîrîn (Plate 52, Fig. 2, and Plate 68, Fig. 1). It is constant at Ukhaiḍir and in early Mohammadan architecture,[109] and it is used invariably in the brick vaulted constructions of Mesopotamia at the present day. It is perhaps the triumphant survival of the old brick vault of horizontal oversailing courses, represented by Mughair, and it bears, at Hatra and elsewhere, another indubitable mark of its brick origin in the horizontal or almost horizontal joints of its lower courses.[110]

3. The vault springing flush with the walls is used in tombs (_Hatra_, ii, Figs. 103, 118, 128, 139, 159), in the southern and in the northern lîwâns of the main palace and in the two lîwâns which were added at the northern end (_Hatra_, ii, Plate 8), in the western annex, the so-called temple (_Hatra_, ii, Plate 9), and in building B (_Hatra_, ii, Fig. 183). The moulded cornice, which usually divides this vault from the walls below, is absent in most of the tombs. The high stilt formed by the horizontal lower courses, which is especially remarkable in the larger of these vaults, differentiates them from western Hellenistic vaulting and connects them more closely with brick forms. In one of the smaller palaces there is a striking example of the survival of brick building methods (_Hatra_, ii, Fig. 74). The stone vault is composed, almost to its whole height, of horizontal courses, and only the very top of the arch is filled in with radiating voussoirs. Nor is the elliptical vault, which is the form naturally assumed by oriental uncentered brickwork[111] wanting at Hatra (_Hatra_, ii, Figs. 108 and 162, Fig. 162 being a primitive example, where the vault is carried down to the floor of the chamber).

4. One room on the upper floor of the palace shows a fuller comprehension of the thrust and buttressing of the vault (room No. 12, _Hatra_, ii, Plate 10 and Fig. 226). The space to be covered is diminished by placing two arched niches on either side, a system which points the way to the breaking up of the wall into buttressing piers. This principle was carried out yet further by Sasanian builders. In the palace of Sarvistân the lower portion of the piers was detached from the body of the wall and further lightened by being divided into two small columns,[112] while angle piers terminating in a single detached column bore the dome of a chamber situated at the back of the palace (Plate 74, Fig. 1). The advance in structural knowledge thus gained was carried little further in these regions; indeed it is curious to observe that Ukhaiḍir exhibits a movement in the opposite direction. Although in rooms 33 and 40 the vaults are set upon columns which stand absolutely free, the vault of the great hall rests upon arched niches whereof the piers are connected with the wall, and the principle of the detached column is recalled only by the engaged columns which form part of the pier. The arcade on free standing columns with a vaulted corridor behind it is of frequent occurrence, but the fact that in all the palace only one, and that one the shortest, of these arcades remains standing (No. 20) shows that the skill of the builders was at fault. Again, in the church of Mâr Ṭahmâsgerd at Kerkûk the engaged columns are present, as in the great hall of Ukhaiḍir, but in the same manner they are structurally one with the piers behind them[113] (Plate 75, Fig. 1); and in the churches of northern Mesopotamia, where deep niches under the vault are a constant feature, the engaged pier of Hatra returns in all its primitive simplicity.[114] Whether the data afforded by extant monuments in Mesopotamia and Persia are conclusive would be hard to determine. The setting of arch, vault, and dome on free standing supports would seem to have been a conception deeply rooted in Hellenistic art, but for actual examples we can adduce only the evidence of relief architecture or the disposition of rock-cut tombs and temples. The blind order under the vault of the men’s caldarium near the forum at Pompeii,[115] the rock-cut dome on engaged columns of the Hellenistic tomb of Akeldama at Jerusalem[116] exhibit a motive to which the architecture of a later age was to give fully developed plastic execution. Yet more explicit are the indications afforded by the rock-cut monuments of Egypt and of India. At Memphis one of the graves of the Persian period shows a vaulted nave resting on piers,[117] and the rock-cut temples of Hellenistic India, with their long vaulted naves resting on columns,[118] point to similar achievements in the Seleucid architecture of Mesopotamia from which they are derived. The existence of an underlying desire to solve statical problems which were of the highest importance to the spatial interior is attested by the sporadic survival of such buildings as the Praetorium at Musmiyyeh and a room in the Golden House of Nero,[119] where the four-sided and the round dome were placed respectively on piers and on columns; but the final mastery was reserved for early Christian builders of the Hellenistic coast-lands, or developed in the same age in Rome out of methods which were specifically Roman, such as the intersecting barrel vault and construction in concrete. In Rome also the original impulse may have come from the East.[120]

5. In three of the upper rooms in the palace (Nos. 13, 15, and 16, _Hatra_, ii, Figs. 227 and 228, and Plate 10) the roof is formed by means of transverse arches (respectively five, three, and one in number) carrying stone slabs which cover the space between them. This type of roof was universally employed in Syria from Nabataean times until the Mohammadan invasion.[121] It was a simple and a satisfactory method of roofing in stone in a country where centering beams, sufficiently massive to sustain a stone vault, were difficult to obtain. I know no other Mesopotamian example of it in stone, but it was copied in Sasanian brickwork, where the stone slab was replaced by a brick vault running at right angles to the main axis.[122] In this form it finds a place at Ukhaiḍir in room 32, and it continued to be used by Mohammadan builders in the Middle Ages, the most renowned example being that of Khân Orthma, at Baghdâd.[123]

The absence of the dome at Hatra is significant. The small square chambers of the palace were well suited to dome construction, yet nothing but the barrel vault is present. Moreover, it is the barrel vault in its simplest expression; not even an intersection is attempted. In the vaulted passage surrounding the central chamber of the western annex, the ‘temple’, one end of the vault terminates on each of the four sides against a transverse arch, whereby the insuperable difficulty of intersection was avoided[124] (Plate 75, Fig. 2). Hellenistic builders had attacked the problem as early as the second century B.C. in Asia Minor,[125] and yet more boldly in Rome.[126] I know no single example of the intersection of barrel vaults in Sasanian buildings; even at Ukhaiḍir the system is sparingly used, and never without careful abutment. Where two barrel vaults meet at right angles, they are either joined together diagonally, without intersection, as in the chemin de ronde, or they terminate against transverse arches, and not infrequently in the rectangular space thus formed, a semi-dome takes the place of the intersecting vault, as in the mosque and in the upper gallery No. 134. The rock-cut temples of India exhibit a similar termination of the barrel vault in a semi-dome.[127] The dome, though it is at Ukhaiḍir of frequent occurrence, the chambers of the chemin de ronde in all the round towers being domed as well as the two chambers north and south of the great hall, Nos. 4 and 27, is never placed over a span wider than 3·10 metres. The square rooms, Nos. 30 and 141, behind the two lîwâns 29 and 140, where, on the analogy of the Sasanian palaces (see below, pp. 74, 76 and 78) a dome might be expected, are covered in one case by a barrel vault, and in the other case by a groined vault. There was no question here of a dome on free standing columns; where the opportunity occurred, in rooms 33 and 40, it was set aside in favour of parallel barrel vaults. The domed chambers in the towers have a circular ground-plan, and when the problem presented by the rectangular substructure arose, it was met in a fashion which is applicable only to very small edifices. The dome in No. 4, and all the calottes over rectangular niches, are set over the angles upon horizontal brackets of masonry. On the octagon, or half-octagon, thus formed, a circle or segment of a circle of small diameter could be placed without any difficulty. It was an expedient which had been adopted by early dome builders both in Syria and Asia Minor,[128] but it was inadequate when the space to be covered assumed larger dimensions and, before the date of Ukhaiḍir, Byzantine and Sasanian architects had elaborated solutions of the problem. In the West the great dome of Santa Sofia had already been placed securely upon stone pendentives; in Persia the use of the arched angle niche, or squinch, had enabled Sasanian builders to throw their domes over a span of 16 metres. The three domes of Firûzâbâd, the earliest of the Sasanian palaces, have a diameter of 13·30 metres; the larger of the two domes at Sarvistân is about 12 metres across, the dome in the smaller palace at Qaṣr-i-Shîrîn covered a chamber 16·15 metres square.[129] If the audience chamber in the larger palace at Qaṣr-i-Shîrîn was domed, as I suspect, it covered an area about 16 metres square. Under this dome, at each angle, at a distance of 2·90 metres from the walls stands a corner pier 1·40 metres square, terminating on the two inner sides in an engaged column 1 metre in length. The distance between the piers is thus about 16 metres, that is to say that the dome would have been no larger in diameter than that which covered the principal chamber in the neighbouring palace. The walls there are 3·90 metres thick, whereas the side walls of the chamber in the palace of Khusrau are never more than 2 metres thick, but in the one case the wall was the only support, whereas in the other the thrust would have been taken first by the angle supports and by them transferred to the outer wall. Moreover, the walls themselves were buttressed by vaulted rooms. The piers are buried about 1 metre in the ruins with which the hall is filled (the ruin heaps lie deepest along the walls and reach almost to the height of a doorway arch which remains in place on the south side); the best preserved of the four piers projects less than 1 metre out of the present surface; that is to say that its whole height is at present under 2 metres. It is conceivable that the piers may at no time have been carried very much higher. Like the columns under the small dome at Sarvistân, they may have been bound into the wall at that level by arches carrying a barrel vault, which would in this instance have had a span of 5·20 metres, and the dome placed upon the square substructure thus formed would reproduce the Sarvistân dome in magnified proportions.[130] It is clear that Ukhaiḍir shows a retrogression in the art of dome building, both in point of span and in point of distribution of thrusts, nor is the fact surprising. The desert ḥîrah of an early Mohammadan prince need not be expected to rival in architectural achievement the summer palace of the Sasanian king of kings, situated upon one of the high roads of his empire.

Firûzâbâd affords the earliest extant example of the dome in Persia. In Babylonia and Assyria no dome is standing which can be dated earlier than Ukhaiḍir. Possibly the Lakhmid ḥîrahs would have provided us with other instances, but the tentative nature of dome building at Ukhaiḍir throws doubt upon the proficiency of Lakhmid construction in this respect.[131] In the Babylonian cultural sphere the dome does not seem to have played an important part in monumental building until a late period, and in my opinion too much significance has been attached to the celebrated relief exhibiting domed buildings which Layard found at Quyundjik.[132] We have here a representation of village architecture, and it is natural to suppose that the domes were of small dimensions. They are to be found to this day in the village architecture of northern Syria and northern Mesopotamia, indeed no other form of roof exists; and they take the shapes depicted upon the relief. They are built of sun-dried brick held together by a mortar of clay. The high ovoid domes which appear upon the relief and in modern villages are built of oversailing rings, like the solitary dome at Hatra. I imagine that the summit of the round domes is constructed over a light centering, but I have not actually seen them in process of being built. The difficulties presented by these methods are practically nil, owing to the light and malleable material and the smallness of the span. The translation of this primitive dome into larger diameters was a very different matter, and there is no evidence for the belief that this step was taken in Mesopotamia in an early age.

The Sasanian conquerors came out of lands on which Hellenism had made an impression less deep than on Mesopotamia, lands where Rome had never penetrated; and they came of a stock more tenacious of its own traditions and less eclectic than the Parthians. To a large extent they re-orientalized the territories which they occupied. No doubt there was less for them to copy, for in the interval of some 300 years during which the Parthians were predominant, Seleucid monuments must have disappeared, and the blurred Arsacid copy of Greek or Roman models had taken their place. The Sasanians created an art of their own, less dependent than that of Parthia on Western forms, and more potent to influence those who came into contact with it, not excluding the Byzantines. In the earliest of their palaces, so strongly marked is the reversion to Achaemenid types that Dieulafoy relegated it unhesitatingly to the earlier Persian period. In its general characteristics the plan of Firûzâbâd differs little from that of an Achaemenid khilâni palace (Plate 73, Fig. 2). The lîwân has deepened, and the employment of the vault has enabled the builder to dispense, as at Hatra, with the columns that sustained its roof. The greater depth of the lîwân, combined with a desire to keep the vaulting span within moderate bounds, have led to the breaking up of the tower room on either side into two narrow chambers. In order to counteract more effectually the thrust of the main vault (13·30 metres wide) the side chambers are placed at right angles to the lîwân, a principle which was not adopted at Hatra, but which rules at Ctesiphon, and at Ukhaiḍir. The towers themselves have disappeared, and though their place remains in the plan, in the elevation it is probable that the façade presented an unbroken line. The audience hall of the khilâni palace is reduced to a domed chamber, and the clumsy construction of the dome makes it evident that the builder would not have ventured to stretch its diameter further. Finally, round the posterior courtyard are grouped, besides the living-rooms, two smaller lîwâns, placed, like those in the Ukhaiḍir courts, so that they may serve respectively for winter and for summer.

The resemblances in detail between the Achaemenid palaces and Firûzâbâd are no less striking. The high fluted gorge and narrow torus of stone which cover the doorways and niches of the one are repeated in the plaster-work of the other. The plain fillets which surround the openings at Persepolis reappear at Firûzâbâd, but in the latter case all the openings are arched, and the moulded archivolt is set within the rectangle formed by the fillets. The ṭâqchah niches, which, so far as my knowledge goes, are found for the first time in the palace of Darius, are present also at Firûzâbâd,[133] and henceforth assume a permanent place in Persian architecture, from which they were borrowed by Mohammadan builders.

The building material at Firûzâbâd is undressed stone, very roughly coursed and set in a bed of mortar. In the domes the stones are cut thinner, more carefully coursed and provided at intervals with a bonding course; in the vaults the thin slabs are laid vertically, parallel with the main axis of the chamber. Exactly the same principles are observed at Ukhaiḍir. Nor do the resemblances end here. Tubes are not absent from the vaulting system,[134] and most of the archways are set back from the jambs to facilitate the placing of centering.[135] The arches are semicircular as at Qaṣr-i-Shîrîn. In the vault of the big lîwân centering would seem to have been used, for it is set back from the face of the walls, doubtless in order to leave a convenient ledge for the centering beams. The vaults and domes here and in all other Sasanian buildings have the ovoid shape common to Ukhaiḍir and to subsequent Mohammadan work in Mesopotamia. It is the old Mesopotamian vault contour. The exterior walls of Firûzâbâd are broken into a continuous series of recessed and arched blind niches divided by engaged columns carrying an entablature of modest proportions.[136] The appearance of this decoration is to my eyes so entirely un-Hellenistic that I have difficulty in connecting it with any classical influence, and in point of fact an arched niche from one of the reliefs from Quyundjik, in the British Museum (Fig. 11), is nearer akin to it than such

façades as those of Ctesiphon or Ukhaiḍir. But it must be admitted that while the recessing of Babylonian and Assyrian wall surfaces is in no sense an imitation of architectural forms, least of all an imitation of the column, which was an element unknown to the designers of these recessed buildings,[137] and that while on the Quyundjik relief the architrave is placed directly upon the piers without the intermission of impost or capital, the engaged columns of Firûzâbâd are true columns carrying an impost, and the whole scheme is no longer a pattern, but a copy in relief of a colonnade in the round. In the courtyard the rectangular niching is retained, but without the engaged columns.[138] On the façade of the palace a series of seven arched niches is set high up in the wall, on either side of the arched opening of the lîwân.[139] It is a motive which recalls the open loggias in the façade of an Assyrian palace.[140]

The palace of Sarvistân bears an obvious relationship to that of Firûzâbâd, but the strict symmetry which regulates the latter is not so closely adhered to, and the construction is handled with greater freedom and skill (Plate 76). The principal lîwân happens, it is true, to have resumed the old latitudinal disposition, but the longitudinal lîwân is present in a subsidiary position. The lateral chambers are provided with wide arched openings which, together with the arch of the lîwân, form a façade not unlike those of the Ukhaiḍir courts.[141] The breaking of the façade by doors leading into the lateral chambers of the lîwân occurs first at Hatra, and characterizes all lîwân buildings later than that of Sarvistân. Instead, however, of the piers and engaged columns of Ukhaiḍir, the three arches of Sarvistân are separated by groups of triple flutes. These flutes are far more clearly connected with ancient oriental tradition than the engaged columns of Firûzâbâd. They are derived from the reed-like flutings of Babylonia and Assyria, which are to be found as late as the Parthian counterfeit at Tellôh.[142] The motive does not disappear after the Mohammadan invasion. It occurs at Kharâneh, a ḥîrah on the western borders of the Syrian desert (see below, Plate 80, Fig. 2), and I found it upon the façade of Sultan Khân, a Seldjuk building in the heart of Asia Minor.[143] Here, as at Sarvistân, it flanks a central doorway. At Sarvistân it gives way at the angles of the palace to a single engaged column. As at Firûzâbâd, the audience hall at Sarvistân is a square domed chamber, but it opens immediately into the posterior courtyard and a single lîwân faces it on the further side. Besides the partial detachment from the wall of the supports of some of the vaults and of the columns bearing the smaller dome, there are other evidences of advance in structural knowledge. In the central lîwân, in the tower chambers, and in the central domed chamber the walls are partially hollowed out by blind niches, which add to the security of the vaults while they increase the interior space of the chambers. These blind niches lend to the supports of the dome something of the appearance of free standing angle piers, and they show a dawning apprehension of the fact that the thrust of the dome is concentrated mainly upon the corners of the substructure. In the isolated dome of Ferâshâbâd[144] the hollowing out of the walls is carried yet further.

The building material used in walls and vaults is undressed stone and mortar, but at Sarvistân the stones are more carefully coursed than at Firûzâbâd. As far as can be judged from photographs, the vaults must have been built over a centering. They oversailed the walls as at Ukhaiḍir, while the semicircular door and window arches were set back from the jambs according to Dieulafoy’s restoration, and oversailed the walls according to the restoration of Flandin.[145] The side walls of the palace are broken by frequent doorways, and in the smaller dome windows were pierced through the drum.[146] The domes are built far more skilfully than those of Firûzâbâd. The zone which contains the squinch oversails the wall, standing flush with the outer edge of a small cornice adorned with a dog-tooth. The squinches are built with a proficiency which is in marked contrast with their rude prototypes at Firûzâbâd. They are divided from the dome by a second dog-tooth cornice, and the dome itself is constructed of light brick tiles.[147] This combination of the two materials is resorted to again at Ukhaiḍir. The niches in the columned chambers are covered with semi-domes which are set clumsily over the angles on very small squinches.[148] The Achaemenidizing plaster-work of Firûzâbâd is not repeated, but the dog-tooth is copied from the cornice under the dome in the older palace. It is significant that the cornices of Sarvistân have but one fillet instead of the two fillets of Firûzâbâd. A tendency to reduce the importance of horizontal decorations is characteristic of Sasanian and of Mohammadan work in Mesopotamia (see below, p. 130).

Both for Firûzâbâd and for Sarvistân a minute re-examination is urgently needed, but the political conditions of the province of Fars are not favourable to archaeological research. Nor was the state of affairs ideal at Qaṣr-i-Shîrîn when I was there in April 1911, and I measured the palace of Khusrau to the tune of the whizzing of stray bullets. That they were not intended to hit me was due principally to the fortunate circumstance of my having been accredited by a powerful Kurdish ally on the Turkish side of the frontier to the leading Kurdish brigand, Kerîm Khân, on the Persian side. This fact rendered the situation more reassuring, but I was not tempted to prolong my stay beyond the five days which I devoted to the palaces, neither did I loiter over my work. It would have been difficult to push on further into the interior, or perhaps I should say that it would have been too expensive; for though Kerîm Khân would have provided me with an escort, he would have expected a small fortune in return for his protection, and perhaps it might fairly be urged that he would have deserved it. According to the information which has reached me from Baghdâd, matters have gone from bad to worse since the date of my visit, and the high road of the Sasanian kings has been definitely closed to traffic.

Like the Achaemenid palaces, Firûzâbâd and Sarvistân were not intended for the lodging of vast hordes of retainers. These may have been accommodated in tents or in mud-built houses of an unpretending nature. But with the close of the sixth century we come to a group of royal dwelling-places wherein provision was made for an indefinite number of women, courtiers, servants, and guards, and the type of building thus created was taken over by the khalifs of Islâm and extended to proportions vaster still. Of this type the palace of Khusrau at Qaṣr-i-Shîrîn is the best example we possess.[149] In general terms Ukhaiḍir is its fortified counterpart.

The palace of Khusrau is built upon an artificial platform like Persepolis and the Assyrian palaces, while additional lodgings for the king’s family and suite are placed on the level of the plain. The double ramps or stairways by which the platform is approached are exactly similar to those employed in the older prototypes. The eastern end of the platform is occupied by an immense open space lying before the entrance to the state apartments. A deep porch, possibly with columns on either side, leads into a latitudinal chamber, the details of which cannot be determined without excavation. From this antechamber a doorway communicates with the square hall of audience, which corresponds precisely with the audience halls of Firûzâbâd and Sarvistân. In the posterior wall there is a deep lîwân in which, perhaps, the throne of the Chosroës may have been placed. Behind the reception-rooms there is an open court round which the living-rooms are grouped, not singly, but in a series of subsidiary courts, some of which are placed on a lower level. The whole scheme is thus exactly parallel to the scheme of the palaces in Fars, though the reduplication and enlargement of the various parts somewhat obscures the resemblance at first sight. At Qaṣr-i-Shîrîn a porch is added to the lîwân palace and the entrance lîwân has become a closed chamber, the porch having superseded the columned entrance of the Achaemenids and the archways of the earlier Sasanians. The rectangular audience hall of the normal Sasanian khilâni palace follows. The small lîwân to the rear, with its flanking rooms, have their parallel at Firûzâbâd, but the small lîwân at Qaṣr-i-Shîrîn forms part of the hall of audience and three of the flanking rooms can be entered from that hall, as well as from the open court behind it.

I must pass from what went before to what came after and draw a comparison between the palace of Khusrau and the desert palace of Ukhaiḍir. A characteristic feature of the latter, the girdle of walls, must be left out of account. At Qaṣr-i-Shîrîn the walls were placed round the large pleasure-grounds with which the Sasanian king surrounded his dwelling. It is the wall-less Ukhaiḍir, the Ukhaiḍir as it was originally conceived by its builders, which must be taken into consideration, though even in that first design the desert ḥîrah was not left entirely defenceless, since it was compressed into the rectangle of its own enclosing walls, strengthened by towers. The space within those walls had to be utilized to the full. At Qaṣr-i-Shîrîn the guards could be lodged in the lower rooms about the stairways, at Ukhaiḍir they were gathered together within the main entrance. The great hall is, in fact, a monumental gateway. It belongs to the system of defences which is absent from the Sasanian palaces. The Mohammadan builders reverted to an older type, to the fortified palace of the ancient East. At Khorsâbâd the principal entrance to the palace lay within the walls of the acropolis, and it was not, therefore, strongly fortified, but such gates as those in the acropolis walls are the true progenitors of the Ukhaiḍir scheme (Plate 78, Fig. 1). In Sargon’s palace the long entrance passage, some 10 metres wide, represents the great hall of Ukhaiḍir; the lateral chambers on either side are divided at Ukhaiḍir into groups of smaller lateral rooms which, both at Khorsâbâd and at Ukhaiḍir, were very insufficiently lighted. In either case some additional light is obtained from a court into which the chambers open. The symmetrical arrangement of the Ukhaiḍir gate with the central court and audience rooms behind it would not have appealed to ancient authorities on fortification. Chaldaean and Assyrian gateways are seldom if ever situated opposite to one another, an asymmetrical disposition being accounted better for purposes of defence.[150] The long passage room of Khorsâbâd and Ukhaiḍir, but without the lateral chambers, exists in some of the excavated gateways at Susa,[151] and at Susa above the gateway stands a hypostyle pavilion offering a high and airy abode to the great folk who inhabited the palaces within, just as at Ukhaiḍir an open court with lîwâns on all sides occupies the high summit of the gate-house. At Ukhaiḍir there is no direct communication between the ground floor of the gate-house block and the rest of the palace, except one door out of the great hall. The gate tower and hall, with the adjoining rooms for dependants, and the mosque, which had of necessity to be accessible to all, formed the public part of the building, and the upper stories, since they too could only be reached by passing through the public rooms, cannot be regarded as containing private apartments. The better rooms may have been intended for guests; the chambers in the gate-tower, and those which were in direct connexion with the chemin de ronde, for guards.

The great open platform of Qaṣr-i-Shîrîn is represented at Ukhaiḍir by the central court. The ceremonial rooms at Ukhaiḍir recall with singular fidelity the disposition at Firûzâbâd, but the flanking chambers of the lîwân (the old tower chambers of the khilâni palace) have doors of their own, as at Hatra and Sarvistân, and the three halls are barrel vaulted instead of domed. Special care has been taken with these vaults. In the audience chamber (No. 30), as in the lîwân (No. 29), they are finely built of brick, while in rooms 33 and 40 they are set upon columns. The unequal intercolumniations in these rooms (the columns stand ·90 metre from the walls and 2·50 metres from each other) is no doubt due to a desire to secure as much space as possible in the centre of the room, but it produces a singular resemblance to Sasanian methods, where the short columns are set close to the walls that they may be the more easily bound in with them by arches. The rooms round the small court F are probably not intended for dwelling-rooms, but stand in some definite relation to the ceremonial chambers; as Dr. Reuther has suggested, the little room 37, with chimney-pipes in the vault, may have been used for the preparation of light refreshments for the prince and his guests. For what special purpose the elaborately decorated rooms 31 and 32 were intended it is of course impossible to say, but as I shall point out (p. 115) they accord with a similar arrangement at Kharâneh. The rooms of ceremony were provided with a serdâb under No. 42. Almost exactly the same grouping of chambers is found in the block which was set at a later date into the eastern part of the palace yard. The north-east angle of the yard forms the court; the façade of the annex is adorned with engaged columns and niches; even the serdâb and the stair to the roof are reproduced. It is clear that we have here a second set of reception-rooms similar to the first, but why a second set was needed it is impossible to tell. The fact that an outer stair was added to the older part of the palace, so as to place the new reception-rooms in direct connexion with the first floor of the gate-house block, the floor which I have tentatively assigned to guests, leads me to suggest that the second ceremonial lîwân, with its dependences, was intended for any visitor who was of such distinction as to need a separate audience room.

The courts B, C, H, and G can have served no other purpose than that of the ḥaram, the dwelling-places for the wives and children. Each court is a habitation complete in itself, a bait as it is called in Arabic, a house. Each is provided with a winter and a summer lîwân, with living-rooms adjoining it, and behind each lîwân lies a long narrow room partly open, with chimney-pipes in the vault--the kitchen.[152] Each bait has access to two of the chambers hollowed out of the towers, which, according to the suggestion of the authors of _Ocheïdir_, were probably closets. In two of the courts, B and H, the flanking chambers of the lîwân are provided with anterooms which open into the court through an archway resting on engaged columns. They are covered with barrel vaults running at right angles to the vaults of the chambers behind, and separated from the lîwân vault by transverse arches. The vault of the lîwân is carried straight through from the back wall to the wall of the court, but the side walls are not continued through to the court, as in C and G, but open through wide arches into the antechambers. These arches are the transverse arches against which the antechamber vaults abut. In the ground plan this group has the appearance of a short lîwân flanked by two short chambers, with an antechamber common to all three, though structurally this would not be a true description. The antechamber predicts the modern ṭarmah, which is, as a rule, either a short antechamber to the central room only, or a long antechamber common to all the three rooms (Fig. 12). In either case the modern ṭarmah is actually that which the ṭarmah of Ukhaiḍir only appears to be, an independent latitudinal antechamber cutting off part of the lîwân.

In court E the arrangement of the rooms is modified owing to the exiguous space which remained at the back of the ceremonial chambers. The elements are, however, the same, a court, a lîwân with side chambers, and a kitchen. To these are added a stair leading to the roof, which is absent from the ḥaram courts. It is reasonable to assume that court E was the private bait of the lord of Ukhaiḍir. These courts or baits are foreshadowed in the posterior courts of the Achaemenid and the early Sasanian palaces (again Firûzâbâd offers the closest parallel); in the palace of Khusrau they reach a development which was to be very little modified at Ukhaiḍir. The scheme can best be studied in the courts on the lower level O, Q, and S. Each of these courts is provided on the west side with a lîwân, flanking chambers, and a ṭarmah, while a fourth chamber to the north may be a kitchen. To the south a vaulted passage leads in each case to a posterior court P, R, and T. On the eastern side of the forecourts there is another lîwân group, much shallower than the first and without a ṭarmah or any subsidiary rooms. The flanking chambers of the eastern lîwâns have small doors into the court and into the vaulted passage behind them. As far as I could judge, the three forecourts communicated with each other, in which case the strict isolation of the baits of Ukhaiḍir is a new feature. In courts K and M the arrangement is a little different. The east end in one court only is occupied by a shallow lîwân group, the west end in both by a deep lîwân group with a ṭarmah, but the subsidiary chambers are to the rear, one small and one larger room, approached by a door through the lîwân and opening on to a posterior court. The four baits on the upper level are very similar. The subsidiary chambers are placed behind the main lîwân; in courts C and G there is a group of rooms to the side, and court G is without the shallow eastern lîwân group in its forecourt, but possesses it on the west side of its posterior court. Neither courts E nor I have the small lîwâns. All the courts communicate with one another (except perhaps courts I and H) and with the passage. These long vaulted passages are a feature of Ukhaiḍir also. The building materials at Qaṣr-i-Shîrîn are those of Ukhaiḍir and Sarvistân, undressed stones, coursed with a certain amount of care, and burnt brick tiles for the finer work.

One further step in the long history of oriental palaces can now be taken, thanks to the excavations of Professor Sarre and Dr. Herzfeld at Sâmarrâ. Part of the plan of the great complex of Balkuwârâ lies before us (Fig. 13). Just as the palace of Khusrau reproduced the khilâni palaces on a gigantic scale, so Balkuwârâ is a gigantic reproduction of Qaṣr-i-Shîrîn. The approach to the palace, through two courts, covers an area some 300 metres long (the measurements are only my approximate estimates made from the scale of Dr. Herzfeld’s outline plan) and passes under three ornamental gateways. A third courtyard, lying before the halls of audience, is over 100 metres long and is set round on two sides by a free standing colonnade (instead of the blind arcade of Ukhaiḍir), a corridor, and a long line of rooms, these last carried round the third side also. An immense lîwân, 30 metres long by 15 metres wide, with two rows of flanking chambers, occupies the centre of the fourth side. Beyond a small latitudinal room there is a group of four great chambers arranged crosswise. Meeting in a central chamber, between the arms of the cross, lies a complex of nine smaller rooms, four groups in all, and beyond this we find another latitudinal room and a great lîwân opening into a garden court.[153] On the further side of this garden pavilions stand upon the banks of the Tigris. The area to the left of the ceremonial halls is occupied by twenty-four courts, each one a bait after the manner of Qaṣr-i-Shîrîn and Ukhaiḍir. Besides the lîwân group at one end (Dr. Herzfeld speaks of the principal room as =⏊=-shaped, but judging from his outline the form is produced by the combination of the lîwân group and the ṭarmah) and the group of three shallower rooms at the opposite end, there are three rooms down either side of each court, and rooms flanking the group at either end. Some of the courts are still bigger and more complex. In the right

wing of the palace, besides a number of baits of a more or less normal character, there are a bazaar and barracks. The huge building here displayed covers only a quarter of the whole area of Balkuwârâ. It is interesting to note that the chief mosque lies to the right hand of the main entrance, just as at Ukhaiḍir it lies to the right of the gate. The smaller palace of al-’Âshiq is again composed of a central block between two wings.[154] The audience chambers appear to consist of a large lîwân with a rectangular room behind it, this room being flanked by two similar rooms (compare Firûzâbâd). The general features of the main gateway, a closed lîwân flanked by two chambers on either side, each with an antechamber, were already known, as well as the details of the wall decoration on either side of the gate.[155] M. Viollet, who did some work in 1910 on the great palace known as the Bait al-Khalîfah, has published a sketch-plan of it,[156] and Dr. Herzfeld is now engaged on further excavations there. Both he and M. Viollet have published exceedingly instructive photographs of stucco decoration from the palaces, and I gave a few in _Amurath to Amurath_. Dr. Herzfeld’s series is naturally far the most interesting, as his work has been the most thorough.

If the palace of Khusrau is unmistakably the culminating point of a long oriental tradition, and the model for future generations of oriental potentates, it serves also to illuminate the little known period during which it arose; it throws light upon the ḥîrahs of the Lakhmid phylarchs, concerning which we have practically no contemporary information. Mas’ûdi tells us that the khalif Mutawakkil copied in one of his palaces a scheme which had been adopted by a king of Ḥîrah. It consisted of a central block, wherein was situated the audience chamber, and two wings containing storerooms and lodgings for courtiers. In front lay an open court common to all three parts of the palace; the way to the audience chamber passed through three gates. Dr. Herzfeld, when he had laid bare the plan of Balkuwârâ, realized that it corresponded with Mas’ûdi’s description.[157] That Mas’ûdi believed the type of the Ḥîri with two sleeves to have been created by a Nu’mânid prince in imitation of the battle array of his army, we, who are acquainted with older monuments, know to be incorrect;[158] it is the latest descendant of a long ancestral line of oriental palaces which runs back through the Achaemenid and the Assyrian to the Hittite. The palace of Khusrau is as perfect an instance of the scheme as is the palace of Balkuwârâ; the differences between them are differences of dimension, not of kind. At Qaṣr-i-Shîrîn old oriental traits, such as the artificial platform and the double stairways, are peculiarly well marked. The three gates of Balkuwârâ are not present at Qaṣr-i-Shîrîn, or rather they are not laid out in the same relation to one another, but it is very possible that Mas’ûdi’s account of the Nu’mânid palace was coloured by a lively recollection of the glories of Balkuwârâ, which in his day was beginning to fall into ruin. Sâmarrâ was finally abandoned by the khalifs in 892, and Mas’ûdi wrote in 943. But if Qaṣr-i-Shîrîn fulfils the requirements of the tenth-century writer, so does Ukhaiḍir, and Ukhaiḍir, standing within two days’ journey of Ḥîrah, may well be taken to be the closest representation of the Lakhmid ḥîrahs until Khawarnaq itself is excavated.

The genesis of the lîwân house as it appears in the palace of Khusrau, at Ukhaiḍir and at Balkuwârâ has emerged from the analysis of a long series of more ancient buildings. The baits adhere severely, I might almost say implacably, to a type which was derived ultimately from the khilâni. It is, however, possible that in their later form another influence may have been at work. We know that, to a certain extent at any rate, the Parthians adopted the Hellenistic house. The Greek peristyle is found in Parthian houses at Babylon and at Niffer (Fig. 9); but, on the other hand, in the Parthian palace at Tellôh, ‘in spite of the penetration into the heart of Asia of the elements of Greek civilization, the constructors, contemporaries of the Seleucids, have remained in all points faithful to the traditions of ancient Asiatic civilization,’[159] and at Hatra no Hellenistic house has yet been recorded. The plan of the Hellenistic house is well known from excavation, principally at Delos and at Priene. As early as the second century B.C. it is found in combination with the Roman atrium house at Pompeii (Fig. 14). In the ordinary private house, which was too small to admit of a complete peristyle, the oecus gives into the courtyard through a prostas with an open colonnaded façade, while other less important rooms are set round the remaining sides of the court (Fig. 15). This has already something of the appearance of a lîwân group with a ṭarmah, and the resemblance is increased if oecus and prostas are reduplicated and two rooms placed in the centre (Fig. 16). The genesis of this house is totally different from that of the lîwân-ṭarmah house; the house of Priene is an abridgement of the peristyle house, the lîwân-ṭarmah house is a development of the khilâni, but it is nevertheless possible that the Hellenistic peristyle house, in its abridged form, may have given the initial impulse which led to the adding of the ṭarmah to the lîwân. We may be sure that no columned façade could have come into existence in Mesopotamia before the close of the second Babylonian empire, and indeed at Ukhaiḍir the columned façade is not applied to the ṭarmah house, though it is found in arcaded galleries--for instance in No. 20. Moreover, the rooms in courts B and H are structurally more closely related to the simple lîwân of Hatra than to the oecus-prostas house, while the modern ṭarmah house is structurally, as well as in plan, one with the latter.

What is the principle which determined the arrangement of the rooms or groups of rooms within the bait, and of the baits within the palace? Professor Koldewey, in one of those generalizations, as profound as they are brilliant, which we owe to his learning and acumen, has laid down a law touching architectural grouping which will be of service in considering this question. Speaking of the intentional separation of the main chamber of a Babylonian temple from the encompassing wall, he says: ‘This intentional separation is perhaps connected historically with the origin of the Babylonian house, which must be dealt with in another place. In my view, a view which rests upon the study of Babylonian ground-plans in historic and in prehistoric times, the grouping of chambers in ground-plans throughout the Babylonian cultural sphere proceeds from the interior. The embracing wall, Duru, is the primary, the indispensable essential. Within the compass of the wall, the single chambers are set in such fashion, and in such fashion are they linked together, that ultimately a court remains over. In the Greek house, on the other hand, the single chambers, Megara, are so placed, and joined together in such manner, that ultimately a court results. The Italic house creates for itself a kind of court by sundering a roof which was originally continuous. It is therefore possible to distinguish between the different types of houses with courtyards by defining the Babylonian ground-plan as injunctive, the Greek as conjunctive, and the Italic as disjunctive.’[160]

With the disjunctive plan Mesopotamian archaeology is not concerned; nor do I believe that the conjunctive plan was either widely or permanently of importance, at any rate up to the period to which Ukhaiḍir belongs. The Greek scheme cannot be brought into sharper contrast with the Mesopotamian than by laying a plan such as that of the Pergamene palace (Fig. 17) beside a plan such as that of the smaller palace at Niffer (Fig. 9). I select with intention a building wherein Hellenism has influenced the details, but left the fundamental principle unchanged. At Pergamon the court results from the manner in which the isolated chambers are placed and linked together; at Niffer a court remains over from the manner in which the chambers or groups of chambers are placed within, and linked to, the encompassing wall. In the baits of Ukhaiḍir it is no less the encompassing wall which is the indispensable essential, and it may even be surmised that the latitudinal chamber which lies behind the lîwân is a survival of the intentional separation of the principal room from the wall. But it is not only the bait, the unit, which must be considered, it is the grouping of units. Now these units are so placed round the encompassing wall, and joined together in such fashion, as to leave a court over. In detailed and in general disposition Ukhaiḍir exhibits the injunctive plan.

Before considering the Umayyad ḥîrahs of the western desert three other Sasanian buildings must be passed briefly under review. I will deal first, though it is not first in date, with the second palace at Qaṣr-i-Shîrîn, Chehâr Qapû.

Is it a palace? A glance at the plan is enough to prove that it does not fall precisely within the four corners of the scheme to which Khusrau’s palace belongs. This divergence of plan, and the peculiar character imparted to the ruins by the isolated quadrangular chamber which dominates the whole complex, have led to the suggestion that Chehâr Qapû may have been a fire temple. In support of this view two buildings have been cited, the rectangular western annex at Hatra, and a ruin excavated by Dieulafoy at Susa. The last-named instance carries little weight.[161] Its resemblance to Hatra depends upon the reconstruction proposed by Dieulafoy upon data too slight to be convincing. Until a further examination has been made, the ruin at Susa offers too frail a substructure for the lightest of theories. As regards Hatra (Fig. 10), the western annex blocks a window in one of the smaller rooms of the south lîwân and is therefore certainly a later addition. But the learned author of the Deutsche Orient-Gesellschaft publication has given us two plans of smaller palaces, found among the ruins in the city, of which one certainly, and the other probably, is composed of a lîwân with its flanking chambers, and a posterior rectangular room with, however, the interposition of a narrow latitudinal room between them (Fig. 18). Dr. Andrae has pointed out that while a lîwân group combined with a rectangular chamber, but without a latitudinal chamber, exists in the main palace (south lîwân), two lîwâns with a latitudinal chamber but without the rectangular chamber are found in the northern annex, which, like the western annex, is a later addition to the palace. The fact that the dispositions observed in the main palace are not entirely isolated examples is of the highest significance, but it does not solve the problem connected with the so-called ‘temple’. In all these palaces the posterior quadrangular chamber may have been a sanctuary, or it may equally well have been a living-room. The theory that in the main palace it is indeed a sanctuary rests mainly upon the symbolic representations carved upon the lintel of one of its doorways.[162] The motives there used are familiar elements of Parthian decoration. The dragon occurs upon the façade of Hatra itself and was found by Loftus among the Parthian fragments at Warka,[163] as well as upon a lintel excavated by George Smith at Quyundjik,[164] but there is no saying whether the lintel belonged to a sanctuary or to a private dwelling. Nor is there much to be learnt, with regard to fire temples, from literary sources. Herodotus declares that it was not the practice of the Persians

to erect statues, temples, or altars;[165] Strabo that they erect neither statues nor altars, but, considering the heaven as Jupiter, sacrifice on a high place. Strabo goes on, however, to state that they have large shrines called Pyraetheia, in the middle of which the Magi, entering daily into the shrine, maintain an inextinguished fire.[166] Trustworthy architectural data for such buildings we do not possess, and as Dr. Andrae has observed, the rectangular chamber at Hatra is unlike any other temple known to us, either in the East or in the West.[167] In the outer court of the palace he found a ruin which he calls tentatively an âteshgâh (fire altar).[168] It is a block of masonry almost square which stood 10 to 12 metres high and has traces of a stair that may either have wound round three sides of the tower, or have zigzagged up the face on one side only. He compares it with the tower some 28 metres high at Djûr, near Firûzâbâd, which was published by M. Dieulafoy[169]. The Djûr tower may date from the time of Ardeshîr Bâbagân, A.D. 227-240. Here, too, there was a stair, which must have wound three times round the tower in order to attain the platform at the summit. M. Dieulafoy was struck by the resemblances that existed between the tower at Djûr, the ziggurat at Khorsâbâd, and the minarets at Sâmarrâ and at Cairo.[170] A ramp winding round the ziggurat to the summit of the pyramid is described by Herodotus, but has not yet been assured by excavation, and even the existence of pyramids with platforms at various heights among the ruins hitherto examined is doubtful.[171] The whole question of fire altar and fire temple is therefore very obscure. The towers at Djûr and at Hatra may have been sacrificial altars, and Strabo bears witness to the fact that the Persians sacrificed in a high place; but I find it difficult to believe that they can have been intended for an inextinguished fire. To keep a fire alight in so exposed a spot would have taxed the ingenuity of the Magi beyond endurance. The shrines in which the perpetual fire burnt must have afforded better shelter, but what shape they assumed we do not know. No help can be expected from this quarter, and the problem presented by Chehâr Qapû must be considered on its merits. It is slightly cleared by a recognition of the fact.

The quadrangular chamber of Chehâr Qapû, viewed impartially, does not offer any serious difficulty. If the audience hall in the palace of Khusrau were standing, its aspect would be much the same, for it too was a large square chamber with a dome rising above and dominating the rest of the palace. At Sarvistân a parallel structure exists to this day. But it is the surrounding buildings which are different, and the question is further complicated by the circumstance that the rooms in the immediate vicinity of the domed hall are so much ruined that their exact arrangement cannot be decided without some excavation--it is provoking to think how little excavation would be needed. So far as can be observed at present Chehâr Qapû is a rectangular complex with the main entrance to the east; the gateway is flanked to the south by two courts, to the north by one, each court being furnished with small rectangular rooms. I conjecture that these were guard-rooms, and they may be compared with the rooms under the ramps in the palace of Khusrau. The main entrance opened into a long quadrangular court with a monumental gate at the further end. To the north of this court, and communicating with it by a door at the eastern end, there is an almost quadrangular area, formed by rooms set round the courtyard numbered E on the plan. The rooms are latitudinal, and they bear no resemblance to the lîwâns of the palace of Khusrau. To the west lies another court, F, with latitudinal rooms on two sides and an independent communication with the entrance court; still further west are two smaller courts, G and H, with rooms on two sides; and finally, to the north of the domed hall, there seems to have been a fifth court or open space with rooms on two sides. The south wing is not symmetrical with the north wing and it is considerably wider. There are three large courts here. Court I has chambers on three sides; those on the south side resembling a lîwân group with a ṭarmah. Court J has on the south side a latitudinal chamber, with a ṭarmah on the north side, and a passage communicating with the entrance court, A. Court K has a lîwân group with a ṭarmah on the south side; the north and west sides are ruined. Beyond this lies a totally ruined area, to the west of which stand two rooms, apparently with a ṭarmah, and at the south-west end of the palace there is a series of four rooms. With the exception of the small courts on either side of the main gate, all the courts seem to have had some direct intercommunication; this was probably the case in the palace of Khusrau also. The grouping of the rooms in the court is, however, almost entirely unlike that which has been described in the larger palace at Qaṣr-i-Shîrîn, at Ukhaiḍir, or at Sâmarrâ. Courts I and K alone, with their lîwâns and ṭarmahs, offer shadowy resemblances to the others. The arrangement of the rooms, the irregularity of the areas covered by the courts, and the tendency towards an asymmetrical disposition, point to a reversion to the methods of the ancient East. Symmetry plays no part in the palace-planning of Babylonia and Assyria. From the earliest to the latest, from the Chaldaean palaces[172] to the palace of Nebuchadnezzar at Babylon,[173] through all the intervening palaces in Assyria, at Nimrûd, at Quyundjik, at Khorsâbâd and at Assur, no principle of symmetry is to be observed. Nor yet is it to be found, except quite fortuitously, in the Hittite khilâni palaces (the late khilâni, north-west of G in Fig. 5, is one of the few instances), although they originated in the symmetrical gateway; and it is markedly absent in the northern Hittite palaces and temples at Boghâz Keui, though in other respects they have little in common with the southern Hittite monuments.[174] Assyrian temples more nearly approach to a symmetrical disposition, but only under influences foreign to Assyria, influences which can be traced back to the end of the twelfth century before Christ in the Anu-Adad temple at Assur. The old Assyrian scheme, of which we have one example in the temple of Assur, at Assur, built by Shamshi-Adad, was derived from the Babylonian temple plan and, like the Babylonian, it was asymmetrical. The imported plan is characterized by the substitution of longitudinal for latitudinal chambers.[175] But these foreign, probably Western influences (for they were responsible also for the creation of Solomon’s temple, apparently a symmetrical building),[176] could not reduce Assyrian architecture to an ordered plan, and the temples in Sargon’s palace at Khorsâbâd fall far short of symmetry,[177] while in Babylonia the longitudinal chamber, i.e. the imported plan, was never adopted, and until the latest period, the temples, like the palaces, remained entirely unsymmetrical.[178] The plan of Quyundjik, which is the most complete record of any Assyrian palace which has yet been published, throws considerable light upon Chehâr Qapû (Plate 77). Courts XXVII and XXX in the temple area, courts XVIII, XIX, XX, and XXII in the domestic quarters, exhibit an unsymmetrical grouping of latitudinal and longitudinal chambers very much akin to that of the courts of Chehâr Qapû. In court XVI we have a foreshadowing of the ṭarmah scheme. (Place believes the rooms in court XVI to have been storehouses for wine, from the quantity of jars found in them.)[179] It would be ridiculous to push a minute comparison too far, seeing that a period of over 1,000 years separates the two buildings, but a certain resemblance in details and, still more, a general correspondence on the fundamental principle of asymmetry leads me to suspect that a primaeval tradition survived through all the innovations of Greece or Rome, Parthia or Persia, and that, at the end of the sixth century, it had sufficient vitality to guide the craftsmen to Khusrau Parwêz in the composition of a monumental building. Survivals of this nature are not infrequently connected with hieratic tradition, and if my conjecture is correct it might serve in some measure to support the claim to a non-secular character which had been put forward for Chehâr Qapû, although the domed hall, which we must assume to have been the sanctuary, bears no resemblance to the cella and anteroom of the Babylonian or of the Assyrian temple. It would be necessary to postulate that while the Sasanian builder retained in the courts and chambers of his temenos something of an ancient tradition which had come to be regarded as sacred, he gave to the shrine wherein the holy element burned with a perpetual flame the form which had been assumed by the ceremonial dwelling of the divine Chosroës.

The two remaining Sasanian buildings which it will be necessary to mention are Ctesiphon and Karkh. Ctesiphon is the most famous of all the later Persian palaces (Fig. 19). It was erected by Shapûr I (A.D. 242-272)[180] and is therefore about 100 years later than Hatra, and earlier than Qaṣr-i-Shîrîn by some 250 years. Not only chronologically, but also in plan, it is closely related to the Parthian palace. It reproduces in yet more striking dimensions the simple lîwân scheme, of which Hatra offers the earliest monumental example. The lîwân at Ctesiphon is covered by a vault spanning 25·80 metres, a dimension which was not exceeded in Rome itself. On either side of the lîwân five vaulted chambers were set at right angles; rising in stories their vaults abutted the main vault, as at Firûzâbâd and Ukhaiḍir. The side chambers had an independent entrance in the façade, a system which was first employed at Hatra. The masonry is of brick, chained with wooden beams as at Ukhaiḍir; but at Ctesiphon the beams are placed parallel with the coursing of the masonry, whereas at Ukhaiḍir they are inserted at right angles into the walls.

The second building is at Karkh, the town known in Syriac as Karkhâ de Lâdan. It was founded by Shapûr II (309--379)[181] when he rebuilt Susa, from which it is not far removed. Of this palace we have nothing but a fragment, possibly a monumental entrance (Fig. 20). The central chamber is covered by a dome which was set over squinches upon four wide archways.[182] The cutting away of the walls under a dome is thus very highly developed at Karkh. Four transverse arches span each of the wings, and the space between the arches is covered by a vault. In connexion with Ukhaiḍir this scheme of the wings at Karkh is of special interest because it is repeated in room 32, where even the windows under the vaults are reproduced by blind niches. The material used at Karkh is brick, and it may here be noticed that at Susa and in Babylonia, where brick was the only available local material, it is invariably used by Sasanian architects; in Fars and in the Qaṣr-i-Shîrîn district, where stone was more easy to obtain than brick, they constructed in unsquared stones, roughly coursed, using brick only for the larger vaults and domes and for those portions of the walls which were finely finished. The latter system was employed at Ukhaiḍir. Vault construction in stone was facilitated there by the fact that the stone broke naturally into thin slabs and could be made to assume more or less the proportions of brick tiles. For this reason stone vaults could be built without the use of centering. At Qaṣr-i-Shîrîn this was not the case. The stones are smooth rounded blocks like large pebbles; it would have been impossible to use them for vaults unless the cement in which they were laid had been peculiarly strong, and the vaults thus formed are of the rudest kind. Coursed and undressed stone held together by a clay mortar was used for vaulting purposes as early as Lydian times; a vault of that character covers the tomb chamber of the tumulus of Alyattes near Sardis. The same masonry is found in the terrace of the Takht-i-Mâder-i-Suleimân at Pasargadae (fifth century B.C.), and is still in common use in Asia Minor.[183] Masonry of dressed and undressed stones set in a mortar of clay or pitch has been found in Assyrian buildings,[184] but gypsum mortar was not known in Mesopotamia till the seventh century. Its earliest appearance was in the palace of Nebuchadnezzar at Babylon. In Egypt it is of much earlier occurrence, and the use of mortar in the Aegean region during the second millennium B.C. (Mycenae, Argos) was probably due to Egyptian influence.[185] Hatra is the earliest Mesopotamian monumental building in dressed stone and mortar; it was an example which was not followed by Sasanian architects. The method was foreign to local tradition; native workmen returned to their own systems and continued to construct wall, vault, and dome of brick or of undressed stone.

A survey of Sasanian buildings leads to the conclusion that a singular want of technical skill was displayed in their vaulting system. The vault and the dome may have been born in Mesopotamia, but they lingered there in a state of immaturity. The barrel vault, the vault on transverse arches, the dome on Persian squinches, or in smaller dimensions on the horizontal bracket, these were the only forms which were employed. If an inclined plane was to be covered, the barrel vault was split up into sections and raised in steps; if the barrel vaults met at right angles, they were carefully separated from one another. At Ukhaiḍir the groined vault is added to this slender stock of forms, but it is not used in many places where it might be expected to appear, and when it is employed, it is only with the utmost precaution. As far as the invention shown in the Mesopotamian regions is concerned, we might to-day be obliged to content ourselves with the barrel vault and the dome poised carefully upon four walls (or little better); but the Greek builders of the Mediterranean coast-lands stepped into the breach, and it is primarily to them that we owe the development of the elementary principles of oriental vaulting.

I have already alluded to a series of early Mohammadan buildings which are of the utmost importance to the study of Ukhaiḍir, the Umayyad ḥîrahs which stand upon the frontiers of Syria. On the western side of the desert the authority of the khalifs had been preceded by the authority of Imperial Rome. Lands which were occupied by Roman armies were endowed with a solid heritage, more enduring than any political domination has proved to be. To this day the traveller to Petra has the paved Roman road under his feet for many a mile; he can reckon his journey by Roman milestones, and daily he will pass by shattered wall and piles of ruin which mark the site of Roman watch-tower and Roman fortified camp. After the lapse of eighteen hundred years these massive structures still offer a meagre shelter to the Beduin shepherds and their flocks, and in the seventh century, when the Umayyad khalifs fled from their cities to the beautiful solitudes of the Syrian desert, most of the castles of the Roman limes, which had been re-occupied by the Ghassânid allies of Byzantium, were standing in all their towered strength. Here indeed was an inheritance for those who loved the wilderness; where the Roman legionaries had languished in interminable exile, the children of the desert held their court.

The Arabian limes did not differ in its system of military defence from the limites of Europe, but whereas the European camps were originally laid down as stockaded earthworks and were not systematically clothed in stone till the time of Hadrian,[186] on the Syrian frontier the camps and forts were from the first built of solid stone masonry. The comparatively late date of the oriental defences was no doubt partially responsible for this peculiarity, but it must also be borne in mind that fortification by means of earthworks was foreign to the regions through which the Arabian limes ran. As early as the time of Vespasian, the camps of Flavius Silva at Masada, near the Dead Sea, were surrounded by walls of rudely piled stones,[187] while in the Flavian period the European camps were still fortified by earthworks and stockades. The Roman province of Arabia Petraea was created in A.D. 105, and the fortification of the first limes dates therefore from the time of Trajan. On this inner limes one great camp stands in ruins, the camp of Odhruḥ.

Archaeological research on the Roman frontiers in Germany, Austria, and Britain, as well as in North Africa, has made us familiar with the general disposition of the legionary camps; moreover, we have two literary sources of information. Polybius, writing in 150 B.C., has left a description of the camp in his day, and Hyginus, writing not earlier than a period shortly before the time of Hadrian, has given an accurate account of the camp as he knew it.[188] Architecturally there is no fundamental difference between the two. The camp of Hyginus was a rectangular enclosure, with a length one-third greater than its width. It had four gateways, the Porta Praetoria and the Porta Decumana in the centre of each of the short sides, the Porta Principalis Sinistra in one of the long sides, but not in the centre, and the Porta Principalis Dextra opposite to it in the other. Round the interior of the walls lay an open space, the Intervallum. The interior area was divided by thoroughfares placed in a regular order. Between the Porta Principalis Dextra and the Porta Principalis Sinistra ran a cross street, the Via Principalis. At right angles to it, the Via Praetoria ran up to the Porta Praetoria. These two were the most important of the roads; they were wider than the others, and in the later stone-built camps they were sometimes flanked by colonnades, while at their point of junction was set a tetrapylon. The colonnades and the tetrapylon are common in cities which were laid out on the Roman camp plan.[189] Opposite the point of junction of the two streets, the centre of the camp was occupied by official and public buildings. Here lay the Forum and the Praetorium, with the Sacellum wherein the eagles of legion and cohort were deposited. Behind the Praetorium, the Via Quintana crossed the camp from side to side, while numerous small roads at right angles to it gave access to the lodgings of the troops; the Via Sagularis, within and parallel to the Intervallum, was carried round the whole rectangle. To this general scheme the camps which have been excavated conform, with little divergence.[190] I give as an example the fort at Housesteads, on the Roman Wall (Fig. 21). The sanctuary, X, which is here rectangular, is not infrequently apsed.[191] As a rule not much remains of the interior buildings except the Praetorium and a few large public edifices, such as granaries and armouries. The Praetorium varies considerably in detail, but in general disposition it resembles the Greek peristyle house. A typical, well-preserved example is to be found at Wiesbaden.[192]

One of the most imposing of Praetoria is that of Lambaesis[193] in northern Africa, where a stone-built camp was constructed about the same date as Odhruḥ to replace the older earthwork. The development of the Praetorium varies with the size and importance of the station. As regards the outer fortifications the four gateways were flanked by towers which projected inwards, from the inner face of the wall, and not uncommonly had a slight salience upon the exterior also.[194] There are one or two examples in which the gate towers are rounded upon the outside and have a more considerable projection.[195] Towers are usually placed at the rounded corners of the wall, and sometimes at intervals along the wall; they have no salience upon the exterior.[196] The barracks, which were as a rule roughly built huts, were more solidly constructed in some of the great permanent camps, and the whole interior plan has been traced at Carnuntum and at Novaesium. The barracks in these camps consisted of long double rows of small chambers, more or less regularly disposed and standing back to back. A street or court, open at either end, unless it happened to terminate against one of the larger official buildings, separated each row from the row opposite. The Intervallum was left open, that free access might be given to the walls; at Carnuntum only, part of the west side was occupied by buildings.

In the Trajanic camp at Odhruḥ (Fig. 22) no trace of the interior buildings remains except a small apsed Sacellum, placed precisely in the position in which it would be found in a camp on the European frontiers. Since the four gateways compare equally well with those of the European camps, we may conclude that the interior arrangement of Odhruḥ was normal. But the fortifications are not normal. Rounded towers project some ten metres from the outer face of the wall and the angles are strengthened by circular towers of still greater salience. Thus in the earliest camp of the Arabian limes we encounter a developed system of flanking towers which is completely absent in Europe.

The second or outer limes cannot be much later in date, and in all probability it belonged to the time which saw the fortification of the road from Palmyra to Damascus. Ḍumair (Plate 78, Fig. 2), the second of the chain of forts that extended from Damascus to the desert capital,[197] is dated by an inscription in the year A.D. 162; it bears a close resemblance both to Trajan’s camp at

Odhruḥ and to Ledjdjûn, a camp on the outer Arabian limes. The salient, rounded, intermediate towers and circular angle towers of Odhruḥ are repeated at Ḍumair with unimportant variations in detail. No part of the Praetorium is standing, but there are traces of some of the rows of huts in the Praetentura, and according to Domaszewski’s plan they extended, on one side at least, over the Intervallum to the wall.[198] In the Retentura one ruined building remains, which the learned archaeologist believes to have been the Armamentarium. In the camp of Ledjdjûn the walls and towers are an exact copy of those of Odhruḥ (Fig. 23). The interior buildings belong to two periods. The greater part of the Praetorium, and a small apsed structure to the north of it, belong to the first period; and to the same date, Domaszewski assigns certain buildings placed along the walls between the towers, the largest of which he takes to have been a Horreum. The rows of barracks which fill the eastern half and a part of the western half of the camp are of later date and belong probably to the time of Diocletian.

No other legionary camps of the size of these three exist along the Arabian limes; the other fortresses which have been examined and planned are smaller, different in character, and later in date. Of these there are three which I propose to consider, Da’djaniyyeh, Bshair, and Qasṭal. Da’djaniyyeh is undated, but from its plan I should judge it to be earlier than the other two. Bshair is dated by an inscription in the time of Diocletian; for Qasṭal there is no epigraphic evidence, but the capital found among the ruins of the Sacellum can scarcely be earlier than the fifth century.[199] That the towers in the fortress of Da’djaniyyeh should be rectangular and set à cheval upon the walls, is not of any significance (Fig. 24). Round and square towers are commonly found at one and the same time, though the round tower, which is strategically an improvement upon the rectangular tower, is in fact later in origin (see below, p. 108). It is worth noting that the details of construction in the walls and towers of Da’djaniyyeh are exactly reproduced at Qasṭal, a fort which diverges much more than Da’djaniyyeh from the Roman camp scheme, but even at Qasṭal the stairs and approaches to the towers are copied from the Odhruḥ prototype. The remarkable feature at Da’djaniyyeh is that the Roman camp plan is obscured and almost lost. The greater part of the Intervallum is filled in with buildings; stables, horrea, and armamentaria are linked to the encompassing wall in a manner which recalls the ancient oriental system, a system which is perhaps foreshadowed at Ḍumair and Ledjdjûn.[200] In a wall set round with chambers there is no room for gates; the suppression of gateways is therefore a necessary corollary of the change of scheme, and at Da’djaniyyeh the Portae Praetoria and Decumana have disappeared. The postern in the south-east wall is not a survival of the Porta Praetoria; its existence is due to the fact that the main water-supply of the fort was a cistern lying outside the walls at this point. Apart from these striking innovations the interior preserves the Roman plan. The Praetorium and Sacellum stand in their accustomed place, but the Via Praetoria, besides having no independent gate, is no longer laid quite symmetrically with regard to the Praetorium. Something like the same combination of camp and oriental fortress can be seen in the Byzantine citadel at ‘Abdeh, but the features of the Roman camp are more completely obliterated and the Praetorium is probably represented by a large ruined building, placed unsymmetrically against one of the walls.[201] At Bshair the orientalizing process is carried a long step further (Fig. 25). The chambers are placed symmetrically round the enclosing wall; there is but

one gate, and the Sacellum itself (k) is set against the wall, leaving the central court clear. Bshair is no longer a Roman limes fortress, it is a military caravanserai. The same definition applies to the undated fort at Qasṭal (Fig. 26). Again, the interior buildings are set round the encompassing wall, but they are not single chambers; they are the baits of the Mesopotamian palaces, minus the lîwân. Each unit is composed of a small open court with rooms on either side (this is the normal arrangement, though three of the baits at Qasṭal have rooms upon one side only), and in the interior of the complex a court is left over. There is no room in this scheme for a Praetorium and accordingly it is given a place outside the walls,[202] but fragments of carved ornament found in the principal court make it probable that a small Sacellum occupied the centre. This principle is retained in the caravanserai fortresses of other parts of Syria. At Dair al-Kafh (A.D. 306) a small temple, which was subsequently converted

into a chapel, stood in the centre of the court;[203] in the barracks at Anderîn (A.D. 558) a chapel is similarly placed,[204] and at Qaṣr ibn Wardân (A.D. 561) a building, the uses of which have not been determined, stands in the barrack yard.[205] Beyond this small resemblance, the divergence of Qasṭal from the Roman camp type is complete. All the more noticeable is its likeness to the only Sasanian castrum of which we have any sufficient record. Qasṭal belongs to the same family as the fort at Qaṣr-i-Shîrîn (Plate 73, Fig. 1). The towered walls, the single gate, the chambers or baits placed round the interior of the walls so as to leave a central court over, all these are characteristic of the older building; but at Qaṣr-i-Shîrîn the lodging of the commandant is placed inside the court, whereas at Qasṭal it is outside.[206] In the Zohâb district there is another building of a somewhat similar type, but it looks more like the ordinary caravanserai than like a fortress.[207]

The caravanserai type, when once it had established itself on the Arabian limes, was not to be ousted, but its later application is not only to fortress and barrack, but to genuine lodgings for caravans. In the Roman or Byzantine caravanserai of Khân al-Zebîb enough remains to show that the interior buildings were placed round the encompassing wall.[208] At Umm al-Walîd this interior arrangement is clearly preserved;[209] at Umm al-Rasâs baits, not unlike those of Qasṭal, are linked to the wall,[210] and the plan of a later building at Khân al-Zebîb (it is probably Moslem) differs not at all from that of a small modern caravanserai.[211] Khirbet al-Baiḍâ (see above, p. 56) belongs to the same group, but from its geographical position it must be regarded as a military station rather than as a true caravanserai, though it may have served both purposes. To what cause is the singularly rapid change from Roman camp to Asiatic caravanserai to be attributed? The answer is obvious. On the Arabian limes the builders were brought into contact with a strong Asiatic tradition; they were probably themselves local workmen, and they orientalized the Roman scheme. They applied from the first their own system of flanking towers to the defences; they grafted an injunctive plan on to the Roman camp plan, and they ended by discarding the latter in favour of the former.

The covering of dead ground by means of flanking towers and crémaillères goes back in western Asia to the earliest times. The plan of the acropolis of Gudea, drawn upon a tablet which is placed in the lap of a statue of the patesi of Lagash, exhibits, in the middle of the third millennium B.C., a system of fortification so fully developed that scarcely a dead angle exists in the whole circuit of the walls (Fig. 27). In the science of military engineering even Egypt would seem to have lagged behind Chaldaea, for the advantage of flanking towers was not understood there until the Asiatic expeditions of the Eighteenth Dynasty had taught the Pharaohs how to correct the defects in the unbroken lines of their massive defences.[212] In the Assyrian reliefs, double and triple rings of walls set thick with towers surround the towns; towered walls are represented in the ground-plans,[213] and excavation has proved the existence of rectangular towers in the walls of Khorsâbâd and of Assur.[214] A chemin de ronde, loopholes, and machicolations have been found _in situ_ in the walls of Assur, together with traces of crenellation,[215] and all these features, as well as hourds projecting from the battlements, and the ladders and battering-rams which they were intended to counteract, are familiar upon Assyrian reliefs. Rounded towers have not been revealed by Babylonian or Assyrian excavations. They belonged to a later age or perhaps to a different sphere of culture, the Hittite or Syrian. But Dieulafoy observed them on the Achaemenid fortifications of Susa;[216] and at Hatra, while the inner walls of the town were flanked by rectangular towers, solid or casemated, and casemated bastions, on the outer wall a rounded tower has been recorded, and Dr. Andrae conjectures that it was one of many.[217] In this particular, as in the approximately circular outline assumed by its walls, Hatra may exhibit traits borrowed from the civilization of the southern Hittites. There are rounded and rectangular towers in the larger Parthian palace at Niffer.[218] In Sasanian fortifications the rounded tower seems practically to have displaced the rectangular.[219]

Flanking towers strengthened the walls of Hittite cities. At Zindjirli the gradual development of more scientific methods can be traced in the successive walls which encompassed the town and the acropolis. The inner city wall, which was the first in date (it was probably built in the thirteenth century), is provided with rectangular towers which have a salience of 2 metres. The outer acropolis wall (Fig. 5), built about 900 B.C., has semicircular towers with a salience of 3½ metres; the strategic disadvantages of rectangular towers had been realized and corrected. A further improvement was effected in the inner cross wall, behind the main gate of the acropolis. The wall is built in retreating angles, and set with towers alternately rounded and rectangular; the rectangular towers project 1·80 metres from the face of the wall, while the rounded towers cover them with a projection of 4·50 metres. The outer city wall was built after the destruction of the city by Asarhaddon in 681 B.C. and is no more than a copy of the earliest wall, but at the same period casemates were added to the walls of the acropolis.[220] The Hittite capital of Qadesh on the Orontes, as depicted in the frescoes at Abû Simbel, a temple built by Rameses II (1388-1322), was protected by a wall with towers, the height of which must be due partly to the imagination of the Egyptian craftsman.[221] These towers have the appearance of being round, but the absence of architectural records of round towers at so early a date throws doubt upon the matter. In Asia Minor rectangular towers have been found upon the outer and the inner walls of Boghâz Keui;[222] they do not as a rule exceed a projection of 2½ metres. At Troy the earliest walls had towers 3 metres wide, and 2 metres salient; the curtain wall was in some places not longer than 10 metres, and the city gates were flanked by deep bastions. In the walls of the third period at Troy three towers were uncovered on the south-east side; they are 3·20 metres wide, 2·35 salient, and are separated from one another by a distance of only 6·40 metres.[223] But on the Greek mainland, at Tiryns, and at Mycenae, the fortifications are characterized by crémaillères and by deep bastions rather than by towers.[224] Much more lavish is the use of towers in the pre-Hellenic cities of Asia Minor, other than Troy. The very ancient acropolis on the Yamanlar Dâgh above Smyrna possessed rectangular towers.[225] In Caria the fortification known as the Wall of the Leleges opposite Iassos had rounded towers and crémaillères,[226] and the walls of Alinda rectangular towers à cheval.[227] The Lycian towns depicted upon the bas-reliefs in the tombs at Pinara, discovered by Benndorf and Niemann, exhibit salient rectangular towers,[228] while fortified towers of the same character are depicted on the monument of the Nereids at Xanthos,[229] and we have a plan of the ancient walled town of Pydnai in which the features portrayed on the reliefs are clearly to be recognized.[230] Nor must the towns of the Phoenicians be forgotten, the towered walls of Mount Eryx in Sicily, of the acropolis of Lixos in Mauritania Tingitana, of Thapsus, of Carthage, and of Tyre.[231]

With such a wide development of fortifications by means of flanking towers, extending from the cultural spheres of the Babylonians and the Hittites over all the western parts of Asia, and carried by the Phoenicians into the furthest limits of the Mediterranean, it is not surprising that the fortifications of Greek towns in the fifth century should exhibit the same features. Assos, the finest example of this period, carries on the tradition in the crémaillères and rectangular towers of its walls;[232] and Messene, with its rounded and rectangular towers, shows in the succeeding century a yet more complete understanding of military architecture.[233] The acropolis of Selinus, with semicircular towers, bears witness at a like age to the carrying over of the Greek system of defences into Sicily.[234] The walls of Ephesus, built by Lysimachus towards the close of the third century, ‘one of the greatest monuments of fortification which have been left to us by antiquity,’[235] show the towered wall of the Hellenistic age, while Mantineia, with its circular outer wall, is like an isolated reversion to the round cities of Hittite lands.[236] Philon of Byzantium formulated the laws which governed Greek fortification in the Alexandrian age. Towers, crémaillères, and casemated walls combined to make a system of defence all the elements of which had been familiar to the Hittites and to the Assyrians, and the methods of attack which he sought to counter were the same as those which can be seen on the Assyrian reliefs.[237] Vitruvius advocates the flanking of walls by round or polygonal rather than by rectangular towers, but his words should be taken as a counsel of perfection, not as representing the practice of his day, for the systematic use of rounded towers by Roman engineers is later than Augustan times and polygonal towers are unusual before the age of Diocletian. At Aosta, which was fortified soon after 25 B.C., the towers are rectangular,[238] but at Fréjus and at Autun, both of which were fortified in the Augustan age, we have two of the rare instances of circular or semicircular towers.[239] As Schultze has pointed out, the planning of towers varies with time and place, but not infrequently rounded and rectangular towers can be seen on buildings of the same date.[240] As at Zindjirli the rounded tower denotes a technical advance, though the rectangular tower is not necessarily displaced by it. The typically Roman conception of frontier defences, the fortified limes, was definitely abandoned in Europe about the year A.D. 360, but a century earlier the invasion of Gaul and Spain by the Franks had proved that the long line of strongholds was powerless to check the inrush of barbarian hordes, and in the last half of the third century the fortified town was virtually substituted for the fortified frontier. Towered walls sprang up about the cities of Roman Gaul, and the work of fortification begun by Probus was carried on by Diocletian.[241] The same process can be observed throughout the empire during the course of the third century, and almost without exception these later fortifications were strengthened by circular or semicircular towers.

But if the walls of Roman cities can claim to have inherited, through Greece and the civilizations of the Aegean, the formulae of the ancient East, the fortified camp was essentially the creation of Rome herself. The stockaded earthwork, with rounded corners and lines devoid of flanking defences, determined the plan of the stone wall which replaced it in Europe and in Africa,[242] and it was not until the Romans applied their system to lands which had seen the birth and development of a science of warfare different from their own that they modified their design. The difference was fundamental. The Roman camp was intended primarily for purposes of attack. It was the camp of an army on the march, indispensable, in the eyes of commanders as wary as they were daring, to a halt that lasted no longer than a single night, but in its essence impermanent. The oriental fortress displays a contrary intention. It was defensive and abiding, a stronghold provided with few exits (since the gateway is the weakest point of a fortified position), but with high walls, heavily flanked by towers which would give the garrison every advantage against the besiegers.

By the time of Diocletian the transition upon the Arabian limes from camp to fortress had been completed. The Umayyad khalifs, when they in turn strewed the fringes of the Syrian desert with the creations of their architects, copied, not the Roman plan which had been imported under Trajan and had survived, in broad outline at any rate, at least, as late as the year A.D. 162 (the date of Ḍumair), they copied its oriental counterpart, adapting it to the use of princes by methods borrowed from Byzantium and from Persia. We know that the Umayyads, like the Ghassânids before them, repaired and re-occupied the Roman fortresses. Hamza al-Iṣfahâni believed that Qasṭal and Odhruh had been built by Djabala ibn al-Ḥârith;[243] Yâqût mentions that Yazîd ibn ‘Abd al-malik (Yazîd II) lived at Muwaqqar, and judging from the existing remains it is probable that he either built or rebuilt it.[244] His son Walîd occupied Qastṭal and Azraq.[245] But princes whose passion for magnificent construction was so great that the subjects of Yazîd III could see cause for exacting from him, when he came to the throne, a promise that he would not lay stone to stone or brick to brick,[246] were not likely to content themselves with the forts of the Roman limes. The poets, who were welcome guests at their palaces in the wilderness, have left descriptions of the luxury of their surroundings,[247] and the picture has been completed by the discovery of some of the buildings themselves. None of the ruins which have been examined are mentioned by contemporary writers under the name by which they are known to the Beduin, but a palace or palaces are recorded in the Wâdi Ghadaf, and it is in that district that Ṭûbah, Kharâneh, and Qṣair ‘Amrah stand.[248] Mshattâ, which was the first to be visited by archaeologists, bears a name which is probably modern.

Qṣair ‘Amrah lies somewhat outside the architectural type to which the other three buildings belong. It is a small unfortified pleasure-palace with a reception hall and throne-room on a basilical plan, and a bath. Very closely related to it is the early Mohammadan ruin of Ḥammâm al-Ṣarakh, discovered by the Princeton Expedition.[249] The bath at Djebel Sais is not dissimilar, but in the light of our present knowledge it requires re-examination.[250] Both at Qṣair ‘Amrah and at Ḥammâm al-Ṣarakh there is a small dome over a square chamber. At Ḥammâm al-Ṣarakh this chamber is 2·15 metres square; the dome is set on pendentives and lighted by windows. It is laid up in gores with projecting ribs constructed of long, thin, wedge-shaped bits of shale, entirely undressed and completely covered by plaster. When intact it must have presented an appearance not unlike that of the ribbed dome at Ukhaiḍir, except that the ribs were set wider apart and the pendentive substituted for the primitive bracket. Concerning the structural features of the dome at Qṣair ‘Amrah, the publication of the Viennese Academy, which leaves much to be desired, is not explicit. Dr. Musil, who is always the best guide in matters architectural and archaeological, describes it as being set on pendentives and lighted by windows in the dome.[251] Here and at Ḥammâm al-Ṣarakh two semi-domed niches are placed opposite to one another, one at either end of the domed chamber, and a room (3·30 metres square at Ḥammâm al-Ṣarakh) next to the domed chamber is roofed with a groined vault. We have a similar use of the groined vault in the east annex at Ukhaiḍir. At Ḥammâm al-Ṣarakh some of the doors are covered by straight lintels, others (together with all the windows) by semicircular arches. Some of the wider arches are slightly pointed, but the vaults and transverse arches in the reception-room are semicircular. At Qṣair ‘Amrah straight lintels are the rule for doors and windows, but over the architrave of the wide door leading into the audience chamber there is a shallow relieving arch. The three parallel barrel vaults of the audience chamber are visible upon the exterior, and the absence of the flat roof obviates the need of tubes between the vaults. In both of these bâdiyahs the walls were decorated with frescoes. Qṣair ‘Amrah was built between the years 711 and 750, when the house of Umayyah came to an end, the earlier date being determined by the presence among the frescoes of a representation of Roderick, the last king of the West Goths, who came first into contact with the Arabs at the battle of Xeres in 711.[252]

To the same group belong a small ruined bath at ‘Abdeh[253] and the bath at Rḥaibeh,[254] the first being possibly Byzantine. At ‘Abdeh the dome placed between two semi-domed niches is set on horizontal brackets. In the palace of Qaṣr ibn Wardân the dome between two semi-domed niches is the basis of the plan, but it is further elaborated by the placing of a semi-domed chamber on the alternate sides. These two chambers are not, however, an integral part of the domed chamber, for they are separated from it by solid walls broken only by doorways. Fortunately we are not reduced here to conjecture concerning the date. On the lintel of the south gate an inscription gives the year A.D. 564.[255] It is clear, therefore, that the dome between semi-domed niches is an architectural scheme which was taken over by the builders of the Mohammadan age from their Byzantine predecessors, and all the evidence points to the conclusion that in both periods the artificers were Syrians.

Al-Ṭûbah is the southernmost of the Wâdi Ghadaf palaces[256] (Fig. 28). Its plan is that of Qasṭal repeated three times, with the addition of projecting rectangular chambers on either side of the gates. When the three main courts adjoin one another the side chambers against the dividing walls are omitted. The individual baits are very similar to those of Qasṭal, but only one row of chambers is interposed between each of the small courts. Thus at first sight it looks as if the Ṭûbah bait consisted of a court with rooms on one side only, except in the north-east and north-west angles, where the courts have chambers on both sides, that the corner spaces may be filled in. Actually, however, the

bait centres round each alternate court, which communicates with the two chambers on either side, and the intermediate court is merely a yard common to two baits. The bait of Ṭûbah is therefore the same as the typical bait of Qasṭal. The enclosing walls and the foundation of all other walls are of stone, the rest of the building is constructed of brick tiles. The western end of the palace, and most of the northern side were completed; the eastern and south-eastern parts were never carried above the foundations. The doorways are covered by brick and stone arches, but a stone or wood lintel was placed under the arch. Where the lintel is of stone its outer side is adorned with an interesting early Mohammadan pattern, which has affinities with the carving on the eastern end of the façade at Mshattâ. The stone lintels are not carried through to the inner side of the arch. The arches, which are round, are built of stone, as is the wall below them. The wooden lintels have rotted away or have been removed by the Arabs. They were laid in brick walls and covered by brick relieving arches composed of two rings of brick tiles. In the inner ring the bricks are set vertically, parallel to the main axis of the arch, with the broad side outwards; in the outer ring they are laid horizontally, at right angles to the main axis, with the narrow end outwards. It is the principle on which many of the smaller arches at Ukhaiḍir are constructed. The brick arches at Ṭûbah are a stilted, slightly pointed oval; that is to say that the transition from the ovoid to the pointed arch is illustrated here in much the same manner as at Ukhaiḍir.

Kharâneh lies a few hours to the west of Qṣair ‘Amrah[257] (Fig. 29). It is two stories high and about 35 metres square, and it consists of baits grouped round a central court (Plate 79, Fig. 1). A rounded tower is set at each of the four corners, a semicircular tower in the middle of each of three sides, and in the fourth side stands a gate between semicircular towers, which are cut away on the interior face, like the towers on the south, east, and west gateways of Ukhaiḍir (Plate 79, Fig. 2). The rooms on the ground floor are ill lighted, and were probably intended for stables, storehouses, and guard-rooms. The court was surrounded by a cloister, the roof of which rested on arches springing from angle piers. On the upper floor this roof, which was constructed of stone slabs, provided a passage or gallery into which the baits of the first floor opened (Plate 80, Fig. 1). The rooms on the upper floor correspond with those below, but in some of the larger chambers (three, according to Musil’s plan) the vault is divided into sections by means of transverse arches borne on slender engaged columns in groups of three (Plate 80, Fig. 2). The column groups recall with singular fidelity the triple reed-columns on the façade of Sarvistân. Beyond the evidence afforded by Dr. Moritz’s photograph, we have no information regarding structural details, though they must be well worth a careful study. The vaults and transverse arches seem to belong to the same family as those of room 32 at Ukhaiḍir. The end of the chamber at Kharâneh is closed by a semi-dome reaching from the back wall to the first transverse arch--the same arrangement as has been described in the mosque and in gallery No. 134 at Ukhaiḍir. It is also extremely significant that the semi-dome at Kharâneh should be carried over the angles of the walls on squinch arches. The arches spring over the angle instead of being filled in with a small semi-dome. The fillets round the arches and round the rectangular windows must be compared with the fillets round the arched niches in room 32 and round the archivolts of squinch and niche at Chehâr Qapû. Another very important point is mentioned by Dr. Moritz. To the right of the audience chamber, which he photographed, and connected with it by a door, is a small rectangular room, beyond which lies another rectangular room of about the same size. Round this last room runs a moulding, above which stand circular plaques of stucco decorated with formal plant-motives in Sasanian style, and with late Syrian leaf-motives. One of the plaques Dr. Moritz detached from the wall, and it can be seen standing upon the floor in his photograph (Plate 80, Fig. 2), and is now in the Kaiser Friedrich Museum in Berlin. It is more than a coincidence that to the right of an audience hall there should be found both at Ukhaiḍir and at Kharâneh a chamber, the elaborate ornamentation of which points to its having some special ceremonial significance. At Ukhaiḍir this side chamber is carried through to the audience hall, at Kharâneh it is divided from it by an interposed room, but the principle is the same in both cases, and in both cases it must be connected with laws of etiquette of the Umayyad courts with which we are unacquainted. Over the above-named doorway, leading from the audience hall into the first right-hand chamber, Dr. Moritz found a graffito inscription in which a date corresponding with November, A.D. 710, is mentioned. Kharâneh, therefore, must have been standing at that time. The archway he describes as an ordinary round arch; in the photograph the door appears to be set within a niche, whereof the arch oversails the wall, like the larger archways at Ukhaiḍir. The door itself is covered by a lintel, and a lintel of solid stone covers the door of the main entrance (Plate 79, Fig. 2). In his section Dr. Musil represents some of the doors as round-arched and some with a lintel and a relieving arch above it; the latter follow a scheme which is common to most of the buildings in the west side of the Syrian desert and exists at Ctesiphon, but is unknown at Ukhaiḍir and unusual in the later Mohammadan buildings of Mesopotamia. Of the loophole windows in the outer wall at Kharâneh, those on the ground floor are finished in precisely the same manner as the loopholes in the towers at Ukhaiḍir, the opening is filled in with an upright stone against which two bricks are placed diagonally. On the upper floor the loopholes show the same method somewhat simplified. There is but one main door, as in the original scheme of Ukhaiḍir. The masonry is of undressed stones set in mortar, with an occasional bonding course as at Ukhaiḍir. All round the castle, between the two upper rows of loopholes, runs a decoration consisting of two horizontal courses of brick with a brick zigzag between. On the towers this band of ornamental brickwork is repeated lower down. The presence of brick used decoratively leads one to suspect that it may be used also in the finer vaults, but like all the technical questions at Kharâneh, this cannot be answered without further observation. Over this main gate there appears to have been some kind of hourd, corresponding in level with the upper story; above it the wall between the towers is decorated with five perpendicular bands of late Syrian leaf-motives. Dr. Musil’s reconstruction of the gate[258] cannot be correct; it does not take into account the horizontal floor-line below the opening which gave access to the hourd, and it covers the bands of ornament. The Kharâneh gateway must be reconstructed in much the same fashion as the three gates in the outer wall at Ukhaiḍir. A vaulted chemin de ronde seems to have crowned the walls.

The rooms of the upper story are grouped into five baits. Over the entrance an additional chamber is interposed between two baits (compare the courts at Ṭûbah which are common to two baits) and on the opposite side there are two extra rooms to fill up the angles. These two additional rooms communicate with the baits on either side, and the gate-house chamber communicates with either bait; otherwise the baits are kept distinct from one another. The scheme is in fact that of Ṭûbah or Da’djaniyyeh, but with the small courts vaulted over and turned into audience halls or big living-rooms, and here we may seek the explanation of the difference between the baits of the palaces on the eastern side and of those on the western side of the Syrian desert. The normal bait on the Mesopotamian side consists of two lîwân groups with a court between, and the lîwân is derived, as has been shown, from the khilâni. The domestic arrangements of the East, where the women are lodged apart from the men, and if possible the several wives apart from each other, make the bait system in some form indispensable to every dwelling-house, but in Syria the khilâni plan was adopted only for monumental façades, such as that of Solomon’s temple, and from it, through temples of the pagan era, to Christian churches. The normal bait on the Syrian side has therefore no connexion with the khilâni; the lîwân is absent. The group of chambers consists of two pairs of rooms with an intervening court, or in complexes more closely knit together, an intervening hall. The group thus formed is the half of a new unit, and may either share a central court with other half-baits, as at Kharâneh, or be provided with a small court of its own and another half-bait, as at Mshattâ. This distinction apart--it is a distinction which is due to local custom and local architectural tradition--the close relationship which exists between Kharâneh and Ukhaiḍir cannot be insisted upon too strongly, for it helps to determine the date of Ukhaiḍir.

Mshattâ lies a few hours to the west of Kharâneh (Plate 81).[259] It is the best known of the Syrian ḥîrahs, and its magnificent carved façade is now in the Kaiser Friedrich Museum. All that concerns me here, however, is its place in the architectural group of which Ukhaiḍir is the eastern representative. It was perhaps built by Yazîd II,[260] and it was left unfinished at his death. It may therefore be a little later than Kharâneh, for Yazîd died in A.D. 724. As at Ṭûbah and Ukhaiḍir, the materials used in it are brick and stone. It is surrounded by a wall set with towers, of which, as at Ukhaiḍir, more than the half-circle projects. The towers on either side of the main gateway are octagonal. Of the buildings immediately within the gate we have nothing but the ground-plan. Roughly speaking they correspond to the three-storied block at Ukhaiḍir, and as Dr. Herzfeld has pointed out,[261] a further correspondence lies in the fact that the oblong court to the right of the gate-house group, with a niche in the qiblah wall, was probably a mosque. The mosque at Ukhaiḍir occupies much the same position with regard to the gate, but since the orientation of the two buildings is different, the qiblah niche at Mshattâ is hollowed out of the main outer wall, while the niche at Ukhaiḍir is hollowed out of an opposite wall. (It must be noted that the big mosque in the palace of Balkuwârâ occupies the same position relatively to the gate.) The conclusion which Dr. Herzfeld reaches, namely that neither palace was a copy of the other, but that both were reproductions by different hands of the same general scheme, is borne out in all other particulars. Beyond the gate-house block lies the central court; beyond the court the hall of audience. At Mshattâ, where the lîwân was unknown, its place was taken by an aisled hall on a basilical plan. Instead of the simple apse there is a trifoliate chamber covered by a dome. The most renowned example of the trifoliate apse is in the church at Bethlehem. The learned disagree as to whether that apse was built by Constantine or by Justinian, but in either case it was earlier than Mshattâ. For the rest, the trifoliate or quadrifoliate chamber covered by a dome is a familiar Hellenistic motive which occurs frequently in palaces and in the baptisteries of early Christian churches. At Ukhaiḍir we have, in the same position as the trifoliate chamber, the quadrangular room No. 30. The throne-room, if I may so term it, at Mshattâ bears comparison with the throne-room at Qṣair ‘Amrah, where two small apsed rooms correspond to the apsed side niches. On either side of the ceremonial chambers of Mshattâ lies a bait, the unit, now complete, which was foreshadowed at Kharâneh and at Qasṭal. Such is the arrangement of the central part of the palace. The two wings (to return to Mas’ûdi’s definition) were never built. Schultz’s ingenious reconstruction gives in each wing a row of baits, all adhering more or less closely to the norm, with subsidiary courts, and chambers at either end to fill up the space. When we come to structural details, the materials are sadly lacking. Either the buildings are too much ruined to afford the necessary information, or the photographs which have been taken are insufficient.[262] Those given by Brünnow and Domaszewski are the best. From them, and from the reconstruction of Schultz, it is possible to see that the vaults oversail the walls[263] and that they are built of a double slice of tiles laid vertically, parallel to the main axis, so as to dispense with centering. The only photograph of a doorway which has been published[264] shows a relieving arch constructed of the same double slice of tiles, with place for a lintel below it. Schultz was able to determine that the lintel was composed of a wooden beam carrying a straight arch of stones. The straight arch occurs at Ukhaiḍir, but without the support of a lintel. The relieving arch has the form of the brick arches at Ṭûbah, a stilted and slightly pointed oval, and from the photograph it would seem that it was set back from the face of the jambs below the lintel, but Schultz in his reconstructions gives it the same width as the door opening.[265] Brünnow and Domaszewski reconstruct the doorways in the domed chamber without lintels, and the doorways in the small chambers of one of the baits without arches--that is to say, they are arch-shaped, but the arch is merely cut out of the solid wall. Schultz places lintels and relieving arches over all the doors. _Kim belir?_ The windows are small and round-arched. The closets were in the towers as at Ukhaiḍir, and Schultz in one of his drawings[266] places over the niche a fluted semi-dome. We know no more.

It now remains to sum up the conclusions reached with regard to the origin of ḥîrah and bâdiyah on either side of the desert. And first it is clear that Ukhaiḍir stands in the closest relations to the Syrian group, not only in general conception, but in details of construction. But Ukhaiḍir reflects the older Lakhmid ḥîrahs, those palaces that were supposed to represent an army in battle with two wings, and through them it re-echoes the Sasanian palaces which were contemporary with them. These too, as we know from the palace of Khusrau at Qaṣr-i-Shîrîn, were composed of a centre and two wings. Again, allowance must be made for Byzantine influence in the Sasanian palaces and the Lakhmid ḥîrahs. Justinian lent artificers to Khusrau I; Khawarnaq was built by a Greek. The intercourse, friendly and unfriendly, between the Sasanian and the Byzantine empires was unbroken. When it was friendly it took the form of commerce, and architects were among the exchangeable commodities; when it was unfriendly it took the form of prisoners of war. Khusrau I must have captured a large and varied selection of artificers when he removed the whole population of Antioch to Seleucia. It is improbable that they should have sat idle in their new abode. They exercised their crafts, and they exercised them in their own manner. It may well have become the fashion among the citizens of Ctesiphon to shop in the Greek Bazaar, just as the citizens of Damascus shop in the Greek Bazaar of their own town. Greek influence, as we know, did not begin with Justinian. It began with a mightier figure than that of the imperial lawgiver--with the mightiest of all, with Alexander. I have already shown that the Mohammadan lîwân took to itself a part of the Greek peristyle and uses it still under the name of ṭarmah. Who can tell when this process began? The Greek peristyle exists in a Parthian palace at Niffer and in Parthian houses at Babylon. Hatra fronts the desert with a Hellenistic façade; so does Ctesiphon; it adorns the central court of Ukhaiḍir. But that Byzantine or earlier Western influence affected in any fundamental manner the plan of palace or ḥîrah is not borne out by this evidence. No fundamental change can be observed at any time, but on the contrary a steady, continuous growth of oriental methods, on oriental lines, and a steady development based on developing needs, ceremonial and social. From the days of the Hittites the palace was composed of a centre and two wings. The khilâni palaces of Zindjirli were laid out on a small scale; the khilâni palaces of Pasargadae and Persepolis covered a wide area, but provided little better accommodation; for the courtiers and guards were lodged elsewhere, in buildings of a less permanent character. Persepolis was the capital of an empire; all the needs of the time were fulfilled there. But this is not the case at Firûzâbâd and Sarvistân. Of the capital seats of the Sasanian kings we know but two, in any real sense, Ctesiphon and Qaṣr-i-Shîrîn, and at Ctesiphon we know only the great hall of audience--together with a fairly accurate guess at its flanking chambers. Before we can say that the extensive wings, which at Qaṣr-i-Shîrîn were added to the khilâni palace, were not a natural development (and they are planned on principles which are entirely oriental) we must have a clear conception of that which lay about the great hall at Ctesiphon, of the palace at Dastadjird which Heraclius committed to the flames, and of the palaces in the Zohâb district. The oriental palace, in the form which it had received from Chosroës and Nu’mânid, laid a strong hold upon the imagination of the East. In the Days of Ignorance the Arab of the desert entered its courts with praise; in the days of conquest he divided its spoils with his fellow soldiers, and sent a part to Mekkah, glorying in the God-given strength which had dispossessed the kings of the earth. Not by literary evidence alone can the deep impression which it created be measured. It gave birth to the Syrian ḥîrahs and to the stupendous residences of the Abbâsids.

On the Syrian side of the desert there is another element to be considered, the Roman legionary camp, and this too had a centre and two wings. The truth is that any complex of buildings laid out on an ordered plan falls almost inevitably into this disposition. The palace of the Flavians on the Palatine had a centre and two wings, yet it was not for that reason derived from the khilâni or related to the oriental palaces. Its ancestor was the Greek peristyle house which goes back in turn to the megaron palaces of Mycenae and Tiryns and Troy. Neither were Qaṣr-i-Shîrîn and its offspring in the Syrian desert derived from the limes camp. Gradually but surely, while Rome still held the Syrian frontier, or rather while Rûm--the Hellenistic, the Byzantine Rome, itself half-orientalized--held it, the ancient Asiatic scheme invaded the limes fortress, pushed out the Praetorium, or pushed it back against the encompassing wall, which had become an indispensable requisite, and having grouped its baits after its own fashion, left a court over. The union of both sides of the desert under the hand of a single ruler quickened the process. Neither the Roman Qasṭal nor the Umayyad Ṭûbah are palaces on the Persian ḥîrah plan; then suddenly Kharâneh and Mshattâ spring into being, uniting the oriental traditions of the Mesopotamian side of the desert with oriental traditions which had developed independently from the same root on the Syrian side. The Syrian architects were masters of a more scientific technique, for they had been trained in a Graeco-Roman school. They taught their Mesopotamian brothers, and even the builders of remote Ukhaiḍir had learnt how to lay a cross vault.

But if the legionary camp is powerless to affect the ancient palace plan, it did not wither away, unnoticed, like a plant upon uncongenial soil. It bloomed again in the cities of the eastern Roman empire, in Boṣrâ, in Damascus, in Apamea. Towns such as Diyârbekr, where not one Roman stone remains upon another, still betray a Roman origin in their crossed thoroughfares and quadruple gateways.[267] And therewith it returned, remodelled, to the West. The palace at Antioch was built on the plan of the Roman limes camp. Diocletian copied it at Spalato, and Constantine’s palace in his new capital was in some respects an echo of that of Diocletian, though the true oriental palace was not without effect upon Constantinople.[268] The imperial residence, stereotyped by him, went on into other phases, too complex, and often too obscure, to be followed here, but it is curious to note that five hundred years later, Theophilus, himself an Asiatic, since his father, Michael II, was a Phrygian by birth, built for himself a palace on the Bithynian coast which was modelled avowedly on the palace of the khalifs at Baghdâd.[269] A few years later Mutawakkil laid out Balkuwârâ--what sister _ḥîri_ with two sleeves stood at Bryas, on the shores of Marmora?

One other point remains. The palace of Ukhaiḍir is contained within a towered wall which is wholly distinct from it. This is not the encompassing wall of the ancient East, the primary condition of the structure. It has the four gateways of the Roman camp, though the unneeded cross-roads have dropped away. Here at last Imperial Rome has come to her own. For all its oriental system of fortification, its towers and its hourds, its machicolations and its loopholes, its casemates and its crenellations, this wall is perhaps no other than the wall which surrounded the legionary camp. But I doubt whether the camp itself, which made so fleeting an apparition on the Asiatic frontiers, was the deciding factor. The camp lived on in the city and made a far deeper impression through the city than through the limes fortresses. The scheme is repeated at Sâmarrâ. Balkuwârâ forms part of a great enclosure similarly disposed, with three gates, like the gates of Ukhaiḍir, the palace taking the place of the fourth.[270] The area covered by the enclosure is so extensive that it resembles a town rather than a royal dwelling, and through this town run the crossed thoroughfares which were once the Via Principalis and the Via Praetoria.