Vous êtes sur la page 1sur 191

The Palace of Diocletian at Split

A Thesis ----------------Presented to the Faculty of the Graduate School University of Missouri Columbia -----------------

In Partial Fulfilment Of the Requirements for the Degree Master of Arts

----------------By

Guy Dominic Robson Sanders

Kathleen Warner Slane August 1989

THESIS SUPERVISOR

i CONTENTS

Page numbers

List of figures.................. Abbreviations .................. Chapter 1 Introduction............... Chapter 2 Diocletian's Palace at Split Chapter 3 Camps Chapter 4 Palaces Introduction............... Antioch.................... Thessalonika............... Summary.................... Chapter 5 Villas Mogorjelo.................. Gamzigrad.................. Maxentius' Villa........... Piazza Armerina............ Chapter 6 Conclusions................. Bibliography.......................... Figures ..........................

ii-iv v-vi 1-11 12-35 36-63 64-95 64 70 74 89 93-120 94 97 106 111 117-130 131-142 End plates

ii LIST OF PLATES. 1.1 1.2 1.3 1.4 1.5 2.1 2.2 2.3 2.4 Jacob Spon's view of Split. Bulic, pl. I. Hebrard and Zeiller's plan of Split. Hebrard -Zeiller, fig. 1. Gerasa. MacDonald, Architecture II, fig 35. Philippopolis. Butler, fig. 135. Split - sectors of excavation 1968-1974. McNally et al. Drawing 1. Modern town of Split showing area of Diocletian's Palace. Marasovic et al. Drawing 15. Split - State plan of Diocletian's Palace Late Roman Walls. Marasovic et al. Drawing 28. Split - Reconstructed plan of Diocletian's Palace. Marasovic et al. Drawing 28 and GDRS. Split - East-West cross sections through mausoleum and temple (above), through gates (below). Niemann fig. 5. Split - Isometric drawing of the peristyle and surrouding monuments. Niemann pl. 15. Split - Elevation of Mausoleum and section through vestibule. Niemann pl. 7. Split - Mosiac floor and walls in N.W. quadrant. Niemann fig. 114-115. Split - State plan of mausoleum. Niemann pl. 9. Split - Elevation of Vestibule facade and section through mausoleum. Niemann pl. 13.

2.5 2.6 2.7 2.8 2.9

2.10 Split - Plan of peristyle court east of the mausoleum. McNally et al. Drawing 2. 2.11 Split - Elevation of Vestibule facade. Niemann fig. 56. 2.12 Split - North-south longitudinal sections through mausoleum (above), through vestibule and hall (below). Niemann fig. 136, 137.

iii

3.1 3.2 3.3 3.4

Traditional legionary fortress according to Polybius. Fabricius, fig. 1. Housesteads, Yorkshire. 2nd cent. traditional castrum. Sear fig. 22. Lambaesis. M. Janon, "Lambaesis," Antike Welt 8.2 (1977) fig. 1. Porchester castle, Hants. Saxon Shore castrum of the late 3rd century. Cunliffe, AJ 52 (1972) pp. 70-83, fig. 1. Drobeta. Left: phase 2 = 2/2 3rd. century, right: last phase = late 4th. century. Lander, figs. 155, 274. el-Leggun, Syria. Castrum of c. 300 A.C. Brunnow and Domaszewski, pl. XLII. Tamara. Encyclopedia of Archaeological Excavations in the Holy Land, p.1148. Qasr Qarun. Carrie, MEFR 86, fig. 1. Palmyra, Camp of Diocletian. Palmyre VIII, map 1.

3.5

3.6 3.7 3.8 3.9

3.10 Palmyra, Camp of Diocletian. Palmyre VIII, map 2. 3.11 Luxor, Temple of Ammon. I. Kalavrezou Maxeiner, fig. A. 4.1 4.2 4.3 4.4 4.5 4.6 Rome, Plan of central monuments. Architecture II, fig. 206. MacDonald,

Milan. Modern city showing Late Roman monuments. Enciclopedia dell' arte antichita, Vol. V, fig.2. Trier. Plan of Late Roman city. MacDonald, Architecture II, fig. 34. Aquileia. Plan of Late Roman city. Enciclopedia dell' arte antichita, Vol. I, fig. 692. Sirmium. Modern city with location of Late Roman monuments. Sirmium I, fig. 3. Nikomedeia. Modern Izmit with location of ancient monuments. Ozture, map 1.

iv 4.7 4.8 4.9 Nikomedeia. Sketch plan of Late Roman city. GDRS. Antioch. Reconstructed plan of Late Roman Monuments. GDRS. Antioch. Late Roman city plan. Downey, History of Antioch, fig. 11.

4.10 Thessalonika. Plan of city with Late Roman monuments. Spieser, gatefold. 4.11 Thessalonika. Plan of Galerius' Palace. GDRS after Moutsopoulos, pl. 10. 4.12 Thessalonika. Arch of Galerius, phase 1 (above), phase 2 (below). Velenis, figs. 14 & 15. 4.13 Thessalonika. Octagon and central court. Moutsopoulos, pl. VIII. 5.1 5.2 5.3 Mogorjelo. Plan and elevation. Wilkes, fig. 10, after Dyggve. Gamzigrad. Plan of monuments. 22-3, fig. 5. Srejovic, Arch. Jug.

Maxentius' Palace on the Via Appia. Villa, early 4th. century phase. Pisani Sartorio and Calza, pl. 48.

5.4 Piazza Armerina. Wilson, fig. 1.

v ABBREVIATIONS. Abbreviations follow those listed in AJA 82 (1978) 3-10 with addenda and corrigenda in AJA 84 (1980) 3-4. Other frequently cited works are listed below. Brunnow and Domaszewski: Brunnow, R.E. and Domaszewski, A. V. Die provincia Arabia. vol. 2, Strassburg, 1905. Bulic: Bulic, F. Kaiser Diokletians Palast in Split. Translated by L. Karaman. Zagreb, 1929. Downey, Antioch: Downey, G. A History of Antioch in Syria. Princeton, 1961. Downey, "Libanius": Downey, G., "Libanius' Oration in Praise of Antioch (Oration XI)," ProcPhilAs 103 (1959) 652-86. Downey, "Palace": Downey, G. "The Palace of Diocletian at Antioch," Les annales archeologiques de Syrie 3 (1953) 106-16. Fellmann, Die Principia: Fellmann, R. Die Principia des Legionslagers Vindonissa und das Zentralgebaude der romischen Lager und Kastelle. Vindonissamuseum, 1958, pp. 75-92. Fellmann, Diokletianspalast: Fellmann, R. "Der Diokletianspalast von Split im rahmen der spatromischen Militararchitektur," Antike Welt 10 (1979) 47-55. Fellmann, "le Camp,": Fellmann, R., "Le `Camp de Diocletien' a Palmyre et l'architcture militaire du Bas Empire," in Melanges d'histoire anciennes et d'archeologie offerts a Paul Collart. (Cahiers d'archeologie romande. no.5). Lausanne, 1976, pp. 173-91. Lander: Lander, J. Roman Stone Fortifications. Oxford, 1984. Marasovic et al.: Marasovic, J., Marasovic, T., MacNally, S. and Wilkes, J.J. Diocletians Palace: Report on Joint Excavations in Southeast quarter. Part 1. Split, 1972. Niemann: Niemann, G. Der Palast Diokletians in Spalato. Vienna, 1910.

vi

RIA:

Ward-Perkins, J.B. Roman Imperial Architecture. Harmondsworth, 1981.

Vickers, "Hellenistic Thessalonike,": Vickers, M. "Hellenistic Thessaloniki," JHS 92 (1972) 156-70. Vickers, "Hippodrome,": Vickers, M. "The Hippodrome at Thessaloniki," JRS 62 (1972) 25-32. Vickers, "Octagon,": Vickers, M. "Observations on the Octogon at Thessaloniki," JRS 63 (1973) 111-20.

1 CHAPTER 1. INTRODUCTION.

Recent research accepts the view that the plan of Diocletian's residence at Split was derived from the design of Roman castra.1 This assumption stems from the visual resemblance between the plan of Split and that of a traditional Roman camp such as Lampasas or the city of Timgad. In such a camp the principal streets, the via praetorian and via principalis, met at right angles in the form of the letter T in front of the praetorium. The praetorium has been understood by most scholars to be the residence of the garrison commander. At Split, however, the decumanus and cardo were designed to meet at right angles in the form of a Latin cross in front of what is thought to be the residence of the former emperor. Parallels drawn between the T plan of castra and Latin cross arrangement at Split on the one hand, and the location of the praetorium of camps and Diocletian's private apartments at Split on the other, seem to indicate a close affinity between the two designs.2 The idea that Diocletian's residence was derived from the military camp is firmly established in literature.
1 2

RIA 454-56; F. Sear, Roman Architecture. Ithaca, 1982, p. 265. Various reconstructions of the street plan attempt to demonstrate the similarity of Split and the traditional castra plan. These follow the basic scheme outlined above. F. Weilbach, "Zur Rekonstruktion des Diocletians Palast," in Strena Buliciana , pp. 123-125. Edited by E. Wiegand. Zagreb-Split, 1924, proposed that a via quintana bisected the northern part of Split and that the plan thus paralleled that of the Polybian more closely than formerly thought.

2 The earliest academic reference to the plan of Split seems to be that of Serlio (b.1475) who sketched a walled Polybian castra in Dalmatia described to him by Marco Grimani, the patriarch of Aquileia. According to Dinsmoor this illustration is an early representation of the residence at Split.3 Late in the 16th century Palladio made detailed drawings of parts of the residence.4 Spon and Wheler, who visited Split in 1677, were the first to provide eyewitness accounts of the remains. Both outlined the major visible remains and drew sketches of Split (fig. 1.1). Although there are notable differences

in their plans, each describes the peristyle, the vestibule, the mausoleum, the temple and the colonnaded gallery on the south wall. There is some evidence in their accounts that they understood the residence to have been a development of a military camp plan.5

W.B. Dinsmoor, "The Literary Remains of Sebastiano Serlio," ArtB 24 (1942) 55-91. Dinsmoor refers to Munich MS. cod. iconogr. 190, fol.1, dated to ca.1558. Dinsmoor may have assumed that the Trajanic camp was Split because of its castra like plan and its location in Dalmatia rather than for objective reasons. In other words, he may have identified Serlio's drawing as Split because of the plan and concluded that Serlio thought that the residence appeared to be a castra. 4 D. Keckemet, "Louis Francois Cassas i njegove slike Istre i Dalmacije 1782," RAD. Jugoslavenskie Akademije Znanosti i Umjetnosti 379 (1979) 7-200. 5 Spon drew the site as a square with square corner towers, four gates with the main gate to the west, and the colonnaded gallery running between the west gate and the south-west tower. Inside he drew the peristyle, with the mausoleum, the vestibule and the temple, all described as temples, disposed around it; for this plan see Bulic, pl.6. Wheler drew a rectangular site oriented east to west with square angle towers and three gates with the main gate at the north. His internal arrangement is much the same as that of Spon. In the text he mentions the gallery on the south side and Spon's conjecture that there was originally a door in the

3 In 1751 Farlati devoted four chapters of his book on Illyricum to Diocletian and Split,6 which with verbal accounts of other travellers, encouraged Robert Adam to visit the site. In the company of his assistant, Clerisseau, and two draftsmen, Adam surveyed the standing remains and produced the first scale plan of Split. His reconstruction was based on both the visible remains and on the writings of Vitruvius and Pliny.7 Throughout he was clearly influenced by the Renaissance and Baroque ideas on Roman architecture developed by Palladio and others. Eighteen years later, Cassas travelled to Split to correct Adam's drawings and subsequently published a state plan of the remains.8 New finds encouraged Lanza to publish a revised state plan in 1855,9 and at the beginning of this century, Niemann and Hebrard and Zeiller independently resurveyed, illustrated and described the site (Fig.

south side, see G. Wheler, Voyage de Dalmatie, de Grece et du Levant. vol.I, Paris, 1723, pp.25-9 and fig. on p.17. 6 D. Farlati, Illyricum Sacrum. Venice, 1751. 7 R. Adam, Ruins of the Palace of the Emperor Diocletian at Spalatro in Dalmatia. London, 1764. The drawings are all the work of Clerisseau and the draftsmen, while the text was largely written by William Robertson the historian. Adam provided the money and inspiration but little of the actual work but gave no credit to his co-workers. The plates are very accurate with minor errors of detail. For an account of Adam's work at Split see J. Fleming, "The Journey to Spalatro," Architectural Review 123 (1958) 103-107. 8 L.F. Cassas, Voyage pittoresque et historique de l'Histrie et de la Dalmatie, Paris 1802. Cassas not only used some of Clerisseau's drawings but aslo recognised his his hitherto unacknowledged role in the production Adam's book. D. Keckemet, op. cit. illustrates many of Cassas' plates and discusses his work in detail. 9 F. Lanza, Dell' antico palazzo di Diocleziano in Spalato. Trieste, 1855.

4 1.2).10 The most complete description to date is that of Bulic, a cleric and amateur archaeologist who lived in Split at the turn of the 20th century. Bulic's familiarity with the town, its remains, the fragments found by excavation and by accident, and the ancient and modern literature is evident in his text.11 Each of these scholars emphasised the relationsip between Split and the plan of traditional castra and indicating that the structure was at once both camp and villa. The scholarship on the origins of the plan of Diocletian's residence at Split falls into four broad categories. The most widely held theory is that Split was planned using military archetypes, but the influence of regional palace design, city planning and fortified villas has been upheld by a few. The interpretation of Split as a glorified camp was strongly reinforced by Downey who saw a superficial resemblance between the New City at Antioch, with its streets crossing at a tetrapylon and gallery overlooking the Orontes, and similar features at Split. He argued that the palace at Antioch was built on the site of a military camp laid out by Valerian, and that the palace plan closely followed the earlier lay-out. Downey concluded

10

Niemann, passim; Diocletien. Paris, 11 F. Bulic, Palaca Kaiser Diokletians

E. Hebrard and F. Zeiller, Spalato, le palais de 1912. Cara Dioclecijana u Spilitu. Zagreb, 1927; idem, Palast, trans. L. Karaman, Zagreb, 1929.

5 that since the palace had the appearance of a camp and since it resembled Split therefore the Antioch palace was the inspiration for Diocletian's residence at Split.12 Since Downey, numerous other scholars have referred to the palace at Antioch as further proof of the influence of military planning on the plan of Split. To L'Orange the plan was "clearly influenced by Roman military architecture."13 MacMullen thought that the mix of civil and military architecture was natural and that Diocletian, sentimental for his campaigning days, sought to surround himself with a nostalgic architectural reminder of them.14 The textbooks of Roman architecture promote the association. Ward-Perkins goes further than his colleagues and denies any major influence from city plans in the Roman east such as Gerasa and Philippopolis, stating that the plan is of purely military inspiration,15 and in this is endorsed by Sear.16 Some scholars have interpreted the remains of Split as a palace and take the plan as the prototype of later constructions such as the so-called Palace of Theodoric at Ravenna.17 With this end in mind,

12

G. Downey, "The Palace of Diocletian at Antioch," Les Annales de Syrie 3 (1953) 111, n.2. 13 H.P. L'Orange, Art Forms and Civic Life in the Late Roman Empire, Princeton, 1965, p. 73. 14 R. MacMullen, Soldier and Civilian in the Late Roman Empire , Cambridge, MA, 1963, p. 42-6. 15 RIA, p. 454-59. 16 Sear, op. cit.. pp. 261-65. 17 K.M. Swoboda, Romische und Romanische Palaste, Vienna, 1919. Swoboda gave Split an important position in the development of the militarisation of palace plans of which he gives Mshatta and Kasr Ibn Wardan as examples.

6 Dyggve discussed the southern extension of the cardo beyond the central crossing in terms of an "open air basilica". This he considered to be part of the complex of architectural spaces that eventually led, through the monumental porch and vestibule, to the "throne room" in the great basilica hall. Implicit in this interpretation is the idea that Diocletian did not retire to private life as an ordinary citizen but retained both the trappings of oriental kingship and the aura of his godhead, with which he was supposedly invested early in his reign.18 This grandiose approach to the "throne room" had immense appeal to architectural iconographers such as Swoboda, who included Split prominently in his survey of the Late Roman and Medieval palaces. Swoboda reconstructed the southern extension of the cardo as a peristyle approaching the "Throne Room" in the domed vestibule, with the tablinum to the south.19 Other scholars such as Smith portrayed Split as a sacred castrum cum palatium,20 while

l'Orange identified the central elements of the residence

18

E. Dyggve, Ravennatum Palatium Sacrum, Det Kgl. Danske Videnskabem Selskab, Archaeologisk-Kunsthistorische Meddelelser, III, Copenhagen, 1941. 19 K. Swoboda, "The Problem of the Iconography of Late Antique and Early Medieval Palaces," JSAH 20 (1961) 78-89. Swoboda agrees with Dyggve that the peristyle was an area of importance in the Imperial cult and draws parallels between elements of Split and palaces throughout the Late Antique, Byzantine and Islamic world. 20 E. B. Smith, Architectural symbolism of Imperial Rome and the Middle Ages. Princeton, 1956 pp. 141-42. Smith followed Downey in thinking that the plan of Split was copied from the castra plan of Antioch and suggested that it ultimately influenced the plan of the Great Palace at Constantinople. In each of these the castra form of the palace was symbolic of the centre of divine authority.

7 and peristyle as the Palatium Sacrum around which the remaining structures in the camp-palace were arranged.21 Finally, Strzygowski saw parallels between the arcade of the peristyle and Syrian and Mesopotamian examples. He acknowledged the resemblance to castra but theorised that the plan of Split was ultimately derived from eastern city plans such as those of Philippopolis and the New City on the island at Antioch. He compared the fortification, the crossing streets with axial tetrapylon and the location of the palace at the end of the cardo to Split, and observed that the gallery existed at both Split and Antioch.22 New information on the plan of Split has been added by various reports of rescue excavations undertaken since World War II and by a joint project conducted by the University of Minnesota and a Yugoslav team in the 1970's (fig. 1.3).23 As a result of the most recent research, numerous details can be added to the basic plan produced by Adam and Clerisseau's survey. These details along with modern knowledge of Roman architecture afford a radically different reconstruction of the original form of the complex. However, according to the Yugoslav and University

21 22

H.P. l'Orange, op. cit., pp. 69-85. J. Strzygowski, "Spalato, ein Markstein der romanischen Kunst bei ihrem Ubergange vom Orient nach dem Abendlande," Studien aus Kunst und Geschichte Fr. Scneider gewidmet, Fiebourg-en-Brisgau, 1906 from which extracts were translated and published by N.Duval in the Bulletin de la Societe Nationale des Antiquaires de France (1961) 110-17. 23 Marasovic et al. and S. MacNally, J. Marasovic and T. Marasovic, Diocletians Palace. Report on Joint Excavations Part 2. Split, 1976.

8 of Minnesota excavators, Split's plan should still be viewed as a combination of villa and castra.24 The idea that Split is derived from Roman military camps has been enlarged upon by Fellmann, who draws parallels not only between the plan of Split and the traditional Polybian camp, but also with border fortlets of the Late Empire found principally in the Balkans and along the eastern limes. Unlike Ward-Perkins, Fellmann considered the peristyle at Split to be part of the overall design and compared the resulting cruciform lay-out to the camps of Portchester, Cardiff and Palmyra, among others. He also observed that use of the curtain wall for the arrangement of barracks at Split echoed the use of the walls of small forts as the back walls of strigae, stables and stores.25 The sole dissentor to the views outlined above is Duval. Acknowledging that little is known of Late Roman palaces, Duval observed that Split, castra and palaces have virtually nothing in common. He contended that Strzygowski's arguments associating Split with the city plans of Antioch and Philippopolis are in error because the walls of Split are quite unlike city walls, the residential zone is not limited to the southern part and the plan lacks a tetrapylon. He also denied that Split can be compared with camp plans and pointed out that the

24 25

Marasovic et al., p. 3. Fellmann, Diokletianspalast.

9 arguments using as proof the lay-outs of Luxor, Palmyra and castra are circular in nature. For Duval Split was merely a rich fortified villa described best as a "Chateau". He felt that no `Royal' interpretation could be given to the Peristyle, vestibule and "tablinum" if Split was not a palatium and Diocletian was no longer emperor.26 The most recent commentary, written by Wilkes, one of the principal investigators of the Minnesota and Yugoslav excavations, recognises the importance of Duval's interpretation but gives equal consideration to that of Fellmann. While praising both Duval's and Fellmann's interpretations, Wilkes stresses the importance of the Peristyle and central block for the interpretation of the site. He concludes that clarification of the idea behind the plan will come with a better understanding of Galerius' fortified retirement villa at Gamzigrad, which he feels resembles Split in intent and plan.27 The plan of Split can be redrawn and recast. What was previously held to be the residence proper is only one part of the whole; the temple, the Mausoleum and the
26

N.Duval, "La place de Split dans l'architecture aulique du basempire," Urbs 4 (1961-62) 67-95. Duval points out that palaces are urban administrative centres whereas Split is a rural residence, that the walls are unlike urban fortifications, that the Peristyle was not symbolic of imperial authority but rather a traditional facade and that the street plan is typical of Roman cities especially in the east. His primary argument is published in a minor journal put out by the Town Planning Institute of Split. This paper was presented orally to the Societe Nationale des Antiquaires de France of which the text was subsequently published: N.Duval, "Le palais de Diocletien a Spalato a la lumiere des recentes decouvertes," Bulletin de la Societe nationale des Antiquaires de France (1961) 71-109. 27 J.J. Wilkes, Diocletian's Palace, Split. Sheffield, 1986, pp.56-70.

10 northern blocks are all integral parts of the villa which occupied the entire area within the walls rather than just the southern portion. Instead of streets meeting at a central point in front of the "residence", the streets run through the residence and cross at a central point marked by a tetrapylon, with the cardo continuing to the entrance of the private apartments.28 In the following pages I have examined the design of the various agglomerations that are supposed to have influenced the lay-out of Split. It will be seen that Duval was essentially correct in rejecting the association between Split, palaces and castra. In fact, many of assumptions about Diocletian's godhead, kingship, and the iconography of architectural programmes are open to doubt. It will be seen that castra had developed new forms by the time of the Tetrarchy, that Antioch cannot be used freely as a parallel for Split, that Late Roman palaces are fundamentally different from Split in scale, function, location and plan and that Split was not, by definition, a palace. The following chapters include a brief description of the extant remains of Split and a likely reconstruction of the original appearance. This is followed by a discussion

28

The foundations of the tetrapylon, considered by Duval to be significantly absent, may have been found in the early 1960's; see B. Gabricevic, "Decussis Dioklecijanove Palace u Splitu." Vjesnik za Archeologiju i Historiju Dalmatinsku 53-54 (1961-62) 113-24.

11 of the various building types starting with castra and continuing with palaces and villas. A final chapter places Diocletian's retirement villa in its true perspective.

12 CHAPTER 2. DIOCLETIAN'S PALACE AT SPLIT.

In May of 305 Diocletian retired from public office to a residence he had built at Split near his birthplace at Salona. His prior intention to retire is implicit in his creation of the joint rule by four emperors,29 and indicates that the residence was probably under construction and perhaps even finished by 305. The site is described in the late 4th and early 5th centuries as "his villa at Split, not far from Salona".30 Split (fig. 2.1) is located on the Dalmatian coast on a peninsula opposite the large island of Bras. It was built on previously undeveloped land31 at the landward end of a peninsula separating the ria on which Salona stands from the sea. Neither Salona nor Split had much strategic or economic importance; Salona was 150 kilometres from the main road from Aquileia to Sirmium, but both sites had

29

According to a contemporary source, Lactantius 18-19.6, Diocletian retired at Nicomedeia after his vicennalia in 304. Lactantius suggests that Diocletian had intended to retire for some time, but that his decision was precipitated by pressure from his Caesar, Galerius. Lactantius says that after abdicating Diocletian returned to his native country but mentions nothing of Split or of Diocletian going to Salona. Lactantius may have exaggerated Galerius' role, and it is generally accepted that retirement was part of Diocletian's long-range plans. Indeed, Lactantius 20.4 says that Galerius himself intended to retire after his own vicennalia "and at that stage....he in his turn could lay down his power". Later, at Carnuntum, Diocletian expressed his happiness in retirement; Aurelius Victor Epitome 39.6. 30 "haut procul a Salonis in villa sua Spalato," Jerome, chronicle, p. 230. "Diocletianus haud procul a Salonis in villa sua Spalato moritur," Prosper Tiro, Epit. Chron. (Chronica Minora vol. I p. 428). "in villa quae haud procul a Salonus est," Eutropius, IX.28. 31 Recent excavations found no trace of earlier building activity on the site: Marasovic et al., p. 41.

13 direct access to the sea and thus to other cities on the coast of the Adriatic. The residence at Split stands on gently sloping ground near the sea shore. It consists of a trapezoidal enclosure covering a total area of c. 38,400 square metres. It is oriented 30 degrees west of north and measures 174.9 and 180.9 metres on the short sides and 215.5 metres on the long sides (fig. 2.2). The walls are fortified with square angle and interval towers, while polygonal towers flank the three main gates in the middle of the north, east and west sides. On the south side facing the sea there is a small postern and, at second storey elevation, an arcuated gallery. The plan of Split is a coherent whole, designed and laid out on a strictly geometrical basis. The site was divided on two axes into four parts from a point of origin at the crossing of the cardo and decumanus32 located about 3 metres north of the geometric centre of the site. This central point was monumentalised by a freestanding foursquare structure, probably a tetrapylon or tetrakionion, which also marked the point at which the site surveyor may

32

O.A.W. Dilke, The Roman Land Surveyors. New York, 1971, pp. 231-33. These terms were used by Roman agrimensores for the north-south and east-west axes respectively when laying out the limites of agricultural land holdings. Both words may have originated as augurs' terms describing the principal axes of towns. Neither was used to describe the principal streets of camps.

14 have placed his groma or dioptra.33 The walls of the complex were laid out perpendicular to the axes of the main streets from an origin at the geometric centre of the site, rather than the centre of the street intersection. A module of 55 Roman feet was used to mark off the north and south walls of the temenoi (fig. 2.3), the north line of the decumanus, the axis of the mausoleum and the temple, the wall of the residence and the north wall of the gallery. The same module can be recognised in the placement of the east and west walls of the west temenos, the northern blocks, the cardo, the temple and to a lesser extent in the complex in the south part of the site.34 Internally the plan of the residence is divided into five parts. In addition to the four quarters created by the streets, there is a block extending the width of the southern part of the site. The area between the temenoi
33

Sir H. G. Lyons, "Land Surveying in Ancient Times," Geographical Journal 69 (1927) 132-143. The groma was a surveying instrument used for laying out perpendicular axes. It consisted of a staff on which a cross-shaped instrument, with plumb lines at the end of each arm, was set. By sighting along the axes of the plumb lines a surveyor could describe lines set at 45 and 90 degrees to each other. The dioptra was an instrument similar to a modern theodolite which was used mainly in astronomy for the measurement of angles rather than for setting points. 34 I calculated the length of the foot from the dimensions of the decumanus from east to west gate (approximately 158.08 metres) and the cardo from the Porta Aurea to the north wall of the gallery (approximately 184.43 metres). The ratio of the lengths is exactly 6:7 in units of 26.35 metres. Assuming each of these units is a whole number of feet long, then each unit is 90 Roman feet of 0.293 metres. This foot length corresponds with the Roman foot of the third century and later of 0.294 metres. O.A.W. Dilke, op. cit., p. 82 notes that military surveyors used 5 Roman feet as a basic measure (2 paces = 1 passus). The Split module of 55 feet is not necessarily a military measure; W. Macdonald, The Architecture of theRoman World. New Haven, 1982, p. 140 notes that multiples of 5 Roman feet were usual in Roman buildings.

15 and the fortification walls was used for bath buildings and other structures. In the northern half of the site a perimeter street separates cubicles built against the defensive wall from large complexes of which the plan and function are unclear.35 The materials used for the construction include white limestone, quarried on the island of Bras, and small quantities of a yellow limestone, used for the Porta Aurea, from Sutilija. For architectural details, exotic stones were used. These include red and white marble, porphyry, cipollino, and Egyptian grey and red granites mainly for the columns and revetments, Proconnesian marble for the capitals, and tuff from the River Jader for the arches. Bricks for the walls of internal structures came from local factories, but those used in the dome of the mausoleum were imported from Aquileia.36 The fortification walls of the complex are just over 2 metres thick. They rise from a height of 14.5 metres on the north side to about 16.5 metres on the south side owing to the seaward slope of the terrain.37 The tops of the walls in the north part are level, but step down

35

The description of Split given here is derived from Bulic, and from recent excavation reports. Occasional recourse has been made to earlier works, notably R. Adam, op. cit. and Niemann. 36 Bulic, pp. 18-20; Marasovic et al., p. 3. 37 Niemann, pl.1; the elevations are scaled from the threshold of the Porta Aurea to the top of the wall and from ground level at the southeast corner. Bulic, p. 22 gives these dimensions as 17 and 24 metres respectively. Bulic, measured from the estimated ancient level, while Niemann measured from the present ground level.

16 slightly immediately to the south of the interval towers on the south side. The wall in the northern sector has two storeys. The upper level is pierced at regular intervals by broad arched windows, 2 metres wide by 3.8 metres high. Where the wall-top steps down, on the line of the north face of the residence block, there are three storeys. Here the window zone descends to the middle-storey elevation.38 On the south side, 9 metres above the ancient ground level, is a columnar arcade. Originally, at the centre and at each end of the arcade, were tetrastyle loggias with tall columns supporting a horizontal architrave broken in the middle by an arch. Of these the central loggia no longer stands. Between the loggias the arcade of the

gallery is interrupted by two taller bays with arcuated architraves. The gallery is divided into a total of 51 bays. The three central bays and the three at either end are united to form the loggias. The loggias are separated by 42 arched bays of which the twelfth to either side of the central loggia is slightly larger. Above the gallery the wall probably continued another 5 metres.39 In the middle of each of the sides are gates. Of these, the north, east and west are large impressive

38 39

Niemann, fig. 8. Bulic, p. 23 considers the rusticated masonry above the gallery to be ancient, whereas Niemann, pl. XVIII restores a pitched roof sloping from the back wall of the gallery to just above the architrave and with pediments above the loggias. The original wall probably continued higher and cannot be clearly traced in the present wall which has been much rebuilt.

17 portals, while that on the seaward side is little more than a postern or a door leading from the cellars of the residence to the shore. The north gate, now called the "Porta Aurea" has a single entrance measuring 4.17 metres wide by almost 5 metres tall and has a flat arch lintel. Above the lintel is a relieving arch, 1.9 metres in diameter, decorated with a double fascia, a wreath and a Lesbian cymation moulding on the archivolt.40 Flanking the arch and above the gate are a series of niches. The two niches to either side of the arch are semicircular in plan and 3.1 metres tall. The ornamented sill of each niche projects beyond the wall plane. Corinthian pilasters framing the niches are supported by ornate brackets. The frame of the niches probably supported a Syrian arch within a triangular pediment. The attic storey was embellished by an arcaded gallery on colonettes, with Corinthian capitals and a continuous cornice at the level of the springers. The colonettes, which survived as late as 1820, rested on consoles which, except the middle pair which have minotaur heads, were decorated with acanthus leaves. Three rectangular niches over the door and niches of the second elevation alternate with bays with a flat facade. These and the semicircular niches may have
40

A similar door is shown in a fresco on the wall of St. Apollinare at Ravenna depicting the palatium of Theodoric. The tympanum of the gate is in-filled with a mosaic decoration showing Constantine (?) trampling a serpent. It is possible that the Porta Aurea was an early example of a decorated relieving arch, see Bulic, p. 25.

18 contained statuary; possibly figures of the Tetrarchs. Four bases, two large and two small, that still stand atop the gate may be in situ and may also have supported sculpture.41 The west gate, the Porta Ferrea, was complete until damaged by thirteenth century construction. The single

entrance is 3.5 metres wide and originally stood about 5 metres tall. Like the Porta Aurea, it had a flat arch

lintel below a relieving arch. Semicircular niches similar to those of the north gate originally flanked the arch, but have now been in-filled.42 Instead of an arcaded gallery over the gate, the arched windows that run along the upper part of the curtain wall continue. The east gate is not well preserved, but was probably similar to the west gate. Inside each of the three main gates there is a

courtyard, 9.3 metres deep and 10.6 metres broad, with and a deep inner portal opposite the entrance. These inner courts were overlooked and defended by a parapet. The north and west gates have grooves for a portcullis in the jambs of the outer portal. It is likely that the east gate was similarly equipped.43

41

Three colonettes and bases of two others are shown still standing in Adam's elevation of the gate, while the tympanum was in-filled with brick and plastered over. For Adam's drawing see Bulic, pl. 13 and pp. 25-7 for a description. Niemann, pl. III. is an excellent state drawing of the gate. 42 Adam saw and drew the traces of one of these niches. For a reconstruction of the gate see Niemann, fig. 36. 43 Bulic, pp. 27-8; Niemann, figs. 28-40 for the west gate, figs. 4750 for the east gate. The portcullises could be drawn up in grooves

19 Flanking the main gates were fully projecting octagonal towers of which only vestiges still survive. Each of these stood three storeys high, and extended above the curtain wall. Traces of access stairs set within the wall are evidence for the third storey. The lower storeys seem to have been roofed with domical vaults of which traces of the springers are still visible. The corner and interval towers were square and also projected beyond the walls. The interval towers are fragmentary, but vestiges of these can be seen mid-way between the corner towers and octagonal towers on each side except the south. Enough remains of a tower north of the west gate to show that they measured approximately 9 by 9 metres and were three storeys high. In each of the ground storey walls were narrow embrasured windows, while in upper storeys were pairs of arched windows. The angle towers still stand in relatively good repair. These are 12 metres square and are attached to the curtain wall only at their inner corners. The north angle towers had three storeys, while the south towers had four and projected above the curtain wall. Like the interval towers there were embrasured windows in the lowest storey and larger arched windows in the higher storeys.44

extending vertically through the arches of the gates. This does not necessarily mean that the tympana were unfilled. 44 Bulic, p. 28. For drawings of the angle towers see Niemann, figs. 12-17, for the interval tower, see fig. 18.

20 On the inside, to the north of the decumanus, cubicles were attached to the fortification wall, so that the wall acted as the back wall of each unit. The cubicles have been traced, immured in the walls of medieval houses, only at the north end of the west side and at two points along the north half of the east side. Each unit was 5.5 metres deep and approximately 4.6 metres broad with a door 1.2 metres wide. Approximately 60 units of this size can be reconstructed, 13 to each side of the Porta Aurea and 17 along both the east and west walls. At the corners and adjacent to the gates, the units were larger and gave access to the polygonal and corner towers.45 These rooms seem to have been used as magazines for storage.46 The cubicles opened onto an arcuated portico, 4.3 metres wide, supported on large rectangular piers. Immediately above the keystones of the arches was a moulded cornice apparently marking the level of an upper floor.47 The pavement of the portico was perhaps brick set on a clay

45

Although physical evidence for larger corner cubicles is lacking, the dimensions along the wall would have allowed space for larger units at the corners. 46 S. MacNally et al. pp.63-65 for details of Probe `B' in Julija Nepota Street. This probe uncovered parts of two units, one gave access to the interval tower on the east side, the other contained one complete amphora and numerous fragments. The discovery of amphorae in one room suggests that some rooms were used as magazines. Others may have lodged soldiers and their equipment but there is no evidence for this. 47 See Niemann, fig.51 for a plan and elevation of the piers and arcade standing at the location of Probe `B'. The openings were 2.6 metres wide. The piers, 1.3 metres wide by 1.7 metres long, stood 1.9 metres high and were surmounted by arches.

21 and mortar base.48 Above the cubicles and portico may have been more rooms or, as reconstructed by Niemann, a broad corridor-like parapet or walkway running around the wall and lit by the arched windows in the upper part of the curtain wall. This walkway was probably covered by a longitudinally running, pitched roof (fig. 2.4).49 Running around the interior in front of the portico and cubicles is a street about 10 metres wide and paved with flagstones. This perimeter road joins the cardo and decumanus at an arch in the colonnades near the main gates and provides access to the wall, towers and magazines. In both plan and elevation the cardo and decumanus with their flanking colonnades are the most prominent features of the site. They are 12.4 and 11.7 metres wide respectively with porticoes, 5.6 metres wide, on each side. The colonnades stood approximately 4.5 metres from ground to architrave. From the cross-vault at the intersections of the perimeter street, the columns appear to have supported a horizontal architrave.50 Beyond the crossing the line of the cardo continues a further 38.5 metres to the porch of the private apartments (fig. 2.5). The street is wider here (13.25 metres) and is lined with

48

S. MacNally et al. pp.64-65. From various state plans the portico appears to have extended all the way around the north half of the site. 49 Niemann, figs.5,10. Niemann did not find any evidence that the upper portions were partitioned. See also Bulic, p.31. 50 Bulic, p.33 notes traces of the colonnades still standing immured in the bishop's palace as evidence of their appearance.

22 colonnades with arcuated rather than trabeated architraves (fig. 2.6). The red granite columns of these colonnades stand 5.25 metres high and rest on pedestals set on square plinths. The arches spring from Corinthian capitals, and have a three fascia moulding. Above the arcade the wall continues to a complex wall-top cornice.51 Cuttings in the sides of the lower parts of the columns indicate that balustrades, 2.4 metres high, ran between them.52 The cardo slopes gently down from the north gate to the porch of the residence. Farlati reconstructed two pillars supporting statues at the intersection of the main streets despite the lack of any evidence for embellishment at that time.53 In the early 1960's excavation revealed the foundations of a massive square structure at the crossing which may reasonably be interpreted as the base of a tetrapylon (figs. 2.2, 2.3).54 The foundations indicate that this

51

The cornice consisted of a string course, a torus supporting dentils, and consoles supporting a projecting cornice decorated with leaf and tongue motives. 52 Bulic, pp.33-34; Niemann, pl.V,VI, figs.53, 57-58, 61-67. The colonnade still stands and is now the Cathedral Square. It is usually styled `The peristyle' by those describing it, but is rather merely a street flanked by colonnades ending in a monumental entrance porch. 53 Bulic, pl.7. 54 B. Gabricevic "Decussis Dioklecijanove Palace u Splitu", Vjesnik za Archeologiju i Historiju Dalmatinsku 53-54 (1961-1962), 113-124 with summary in French. The excavations at the intersection revealed massive foundations 0.8 metres below the ancient paving. At the corners of the crossing were four 4 metre square foundations, set 4.2 metres apart, and linked by walls 1.4 metres wide walling. Between the top of the stones of the foundation and the ancient street level was a thick layer of cement. The excavator dismissed the possibility of a tetrapylon because of the cement layer, because the angles of the stoas did not appear to attach to the central structure and because such a monument would not be in keeping with the "open plan of the

23 structure exactly filled the intersection and narrowed the streets to as little as 4.2 metres as they passed under it. The northern half of the residence was divided equally by the cardo, separated from the southern half by the decumanus and was surrounded by the perimeter street. Traces of structures in the north half are fragmentary, and are buried in the walls of later buildings. In the northwest quadrant remains suggest that the block was surrounded by a continuous wall (figs. 2.2, 2.3), perhaps with an entrance in the middle of the west side. The interior of the block appears to have had a central court with an arcade set on massive piers, at least on the west side.55 This arcade probably supported an upper storey. Considerably more of the northeast block has been recovered. Here too a continuous wall surrounded the block. The remains suggest that the block was divided into three unequal parts and that the north and south parts were further subdivided, possibly into three large rooms

residence." He suggested that this massive substructure had some connection with the surveying of the site and was the place where the groma was set up. Professor Suic interpreted the remains as the mundus, a sacrificial pit associated with the act of foundation. One wonders what purpose the foundation could have served in the placement of the groma which was a lightweight, portable instrument. It seems more probable that a tetrapylon was intended but never built, or perhaps built and removed at a later date. 55 See S. MacNally, et al. Dr. 28. For a restored section see Niemann, figs. 117-118. The perimeter wall has been located intermittently in the north half of the area only. A file of five piers of the arcade, resembling those of the arcade in front of the cubicles, has been identified running parallel to, and circa 11 metres from the west wall.

24 (figs. 2.2, 2.3). In the middle of the south part, two edges and a corner of a mosaic pavement of interlocking circles has been found (fig. 2.7). One edge is bounded by a stone curb 1.2 metres from a wide east-west wall, the other runs up to the stub of a wall perpendicular to the first. About 64 square metres of the mosaic has been found, indicating that the space within the central room was probably completely covered. The central section of the three parts may have been open, for a small drain runs along the northern side of the east-west wall at floor level.56 The functions of these areas are unclear. Bulic, suggested that they were used for servant and military accommodations but equally they could be an integral part of the residence itself.57 The basic plan south of the decumanus is more symmetrical. To either side of the cardo temenoi face onto the street through the arcuated colonnade (fig. 2.5). The western temenos is a rectangle measuring approximately 34 by 53 metres (fig. 2.2). It contains a small podium

temple (9.2 x 20.7 metres) in the western part of the templum and two circular structures in the eastern part. The temple is set on a podium 2.87 metres high that

56 57

Niemann, pp.86-89, figs.113-115. Bulic, p.32. There are problems with this interpretation. No interior walls have been found in the northwest block, and the lay-out suggests an open atrium or peristyle court. The northeast block with its large halls and splendid carpet mosaic does not suggest quarters for domestic staff, but rather more suites of Diocletian's personal residence.

25 contains a vaulted cellar. A stairway on the east side, now destroyed, led up to the deep porch supported on six columns with Corinthian capitals, four across the facade and two aligned with the antae. The cella itself, measuring 5.86 by 7.27 metres, is roofed by a vault with decorated coffers. The architectural details, including the door jambs and lintels, the architraves, the consoles and the cornices are all highly decorated with vegetal motives, egg-and-dart, bead-and-reel, and even figural designs.58 The two round structures are adjacent to and parallel with the colonnade. They have a diameter of around 11 metres, and appear to have been circular temples.59 The eastern temenos encloses an area of approximately 32 by 42 metres and is almost completely filled by the mausoleum (figs. 2.2, 2.8, 2.9). The mausoleum is set on an octagonal podium, 5 metres tall and 25.7 metres in diameter, with roughly dressed foundations rising to a course of orthostats and a projecting wall head cymation.60 A narrow L-shaped passage in the southwest corner of the podium runs north to the cross axis of the
58

Bulic, pp.36-38 reconstructs the temple as tetrastyle prostyle with columns in antis, whereas Niemann prefers the arrangement described, as no traces of columns in antis could be found and the length of the porch suggested an intermediate pair of columns. For a reconstruction and state drawings see Niemann, pp.80-86, pl. XVI-XVII, figs.102-110. 59 J. Marasovic, et al. p.3. They could also, conceivably, have been foundations for columns like that of Trajan in the Forum of Trajan at Rome or that of Jupiter Optimus Maximus at Mainz. 60 For the identification of the structure as a mausoleum see Bulic, p.39.

26 mausoleum then turns east where it enters a low, domed crypt by descending three steps. The crypt is circular with eight deep wedge-shaped niches in the walls and a rectangular wellhead located off centre. The mausoleum is an octagonal structure with a maximum diameter of 18 metres (7.7 metres on each side) and is surrounded by a peripteral colonnade standing 3.3 metres from the wall.61 On the west side of the mausoleum the podium is extended for the porch. A flight of steps ascends between the extended arms of the parastades to a tetrastyle prostyle porch. The columns of the porch and the peristyle are of the Corinthian order and support a horizontal three-fascia architrave and denticulated cornice. The central vessel of the mausoleum is two storeys high and covered with an octagonal, pitched roof. The interior is circular, 13.3 metres in diameter, with eight niches alternately semi-circular and rectangular in plan. Each niche is surmounted by an arch and flanked by free standing Corinthian columns attached to the entablature by brackets. In the second storey the exedrae are omitted, but shorter columns duplicate the ground storey. The entablature is intricate and consists of two fasciae, an egg-and-dart moulding, a wreath band, a guilloche band,

61

See MacNally et al. pp.51. The crypt was not used for burial. The walls are unfinished and rough and there was no trace of settings for a sarcophaus. Apparently the crypt was closed after construction of the mausoleum was completed.

27 dentils, and consoles supporting a decorated cornice. The upper entablature is less ornate, but has a frieze with hunting scenes and putti holding garlands. The structure is covered with a brick dome to a height of 21.5 metres.62 Between the temenoi and the fortification wall and residence the area is filled with different structures. It was formerly assumed that cubicles flanked the curtain wall and that the perimeter street continued around and under the vestibule, but this is not the case.63 Southwest of the temple is a bath complex of which six small rooms have been excavated (fig. 2.2). These were oriented on different rectilinear axes; three had apses and were probably vaulted, two were circular with niches and were probably domed. The bath complex left no room for either a street or cubicles on the west side.64 To the east of the mausoleum is a peristyle court measuring 13 by 19 metres (fig. 2.2, 2.10). The court is entered from the south

colonnade of the decumanus at a point adjacent to the east gate and opposite the junction with the perimeter road. The court is bounded by the back wall of the portico and the mausoleum temenos and has a triple arched doorway on

62 63

Bulic, pp.39-57; Niemann, pp.62-79 pl.VII,IX-XIV figs.76-98. In earlier reconstructions the cubicles are shown running around the whole inner face of the defensive wall from the residence block north; see Niemann, fig.2 for Adam's plan, fig.3 for Clerisseau's and fig.4 for Niemann's; Hebrard and Zeillers' plan is reproduced in Bulic, pl.17. Later plans still show this arrangement despite the physical evidence, for example RIA, fig.308. The true disposition is shown in F. Sear, op. cit. fig.173. 64 MacNally, op. cit. pp.53-58.

28 piers on the east side. To the south a door led to other apartments, and finally to another bath complex. The covered portions of the court were paved with geometrically patterned mosaics. Gold tesserae indicate that the walls of the portico or the vaults on the east side were covered with mosaic decoration.65 At the southeast corner of the mausoleum temenos is a second bath building of which the apse and service area of one room have been uncovered (fig. 2.2). Again, these structures leave no space for a street or magazines.66 The private apartments occupy the southern fifth of the complex. The entrance vestibule is 55 Roman feet square (c. 16 metres) and protrudes beyond this block of buildings as far as the line of the south walls of the mausoleum and temple temenoi (fig. 2.2). The vestibule is approached on the axis of the cardo by a broad flight of steps leading to a tetrastyle prostyle porch (fig. 2.11). The columns of the porch are Corinthian and support an architrave which arches over the central span. The porch is attached to the ends of the peristyle arcade by short lengths of wall slightly offset from the line of the arcade. This effectively closes off the cardo, creating what is essentially a courtyard flanked by an arcuated colonnade, terminated at the northern end by the

65 66

J. Marasovic, et al. pp.13-16. Ibid , pp.27-30.

29 tetrapylon and at the south by a monumental entranceway. The visual unity of the court is enhanced by the continuation of the cornice above lateral colonnades as the architrave of the vestibule porch (fig. 2.5).67 The porch has a triangular pediment with a large base, perhaps for a statue group at the apex. The decoration of the entablature is identical to that of the peristyle.68 The highly decorated door of the vestibule is centrally placed and measures 3.8 by 2.35 metres. The jambs have intricate mouldings while the lintel is embellished with egg-and-dart, dentils, consoles, guilloche and a vegetal scroll. In the brickwork of the wall above the door is the chord of a segmental brick relieving arch. Evidently the surface was originally covered with marble veneer. The interior of the vestibule is circular in plan (diameter 11.2 metres) with four semicircular niches located in the corner piers and a door opposite the north entrance. This room had two storeys and measured 16 metres from the floor to the top of the domed ceiling (fig. 2.6). Within the southeast and southwest corner piers were stairways leading up through the thickness of the wall to the second storey gallery. The masonry is of roughly dressed stone coursed with zones of brick. The interior was lit by arched windows located
67 68

Niemann, pls. VI-VII. For illustrations of the porch of the vestibule see Niemann, pl.XIV,XV, and especially figs.56-58.

30 above the entrance and in the lateral sides at the second storey level and by windows in the lower part of the dome. The walls were originally decorated with coloured marble, moldings, pilasters and pediments over the niches. Beneath the vestibule is a vaulted cellar, similar to that under the mausoleum but without niches, that serves as a crossing of thoroughfares. One passage originated with a steeply inclined stairway located in the middle of the porch steps, which ran under the southern block to the postern in the south wall. The second passage ran between the service areas of the bath complexes to the southeast of the mausoleum and southwest of the temple.69 Much of the plan of the southern block is known from the substructures, which are well preserved in medieval house walls (fig. 2.2). The upper storey is poorly preserved, but is sufficient to indicate that its plan closely followed that of the basement level. Access to the basement was restricted to the small postern, to the corridor originating under the main entrance and to stairs down from a hall in the upper storey. It seems probable that this part of the complex served primarily as magazines and provided access from the shore to the northern part of the site in general.70

69

For elevations and sections of the vestibule see Niemann, pp.55-61; Bulic, pp.61-62. 70 Bulic, p.24 considers the postern as an exit in case of emergency, but the inaccessibility precludes this explaination. R. Fellmann,

31 To judge from the substructures, the first floor consisted of an intricate complex of small and large rooms of rectilinear and circular plans. Access to the upper part of the residence was through the vestibule alone. On passing through the vestibule, a visitor entered a long basilical hall, 11.5 by 31 metres, divided into three aisles by two rows of four piers (fig. 2.2). Doors in the side walls open onto rooms to the east and west. The hall was probably roofed with a series of cross vaults. A door at the south end opens onto the long gallery, 8 by 137 metres, flanked on the south side by the arcade in the south wall of the enceinte. This gallery gave access to the various rooms of the complex arranged perpendicular to it. On the west side of the axial hall are three blocks of rooms separated by long narrow spaces which perhaps served as light wells. Immediately to the west and parallel to the hall is a block (W1) of six cubicles in series, each 6 by 4 metres, which open westwards onto a long corridor.71 The second and fifth units also have a door in their east sides which communicates with the central hall through an intermediate light well. This block is duplicated on the east side of the central hall
"Diokletianspalast" pp.47-55 sees the passage as a continuation of the cardo. 71 To facilitate discussion the blocks have been designated as W1-3 for the west range and E1-3 for the east range (fig. 2.3). The description of the private apartments block relies heavily on the state plan in Marasovic et al. fig.28.

32 (E1). The second block (W2) consists of a large basilical hall, oriented north to south and measuring 13 by 28 metres. This hall is subdivided into a nave and aisles by arcades supported on piers. Three doors open onto the hall from the gallery on the south side and give access directly to the nave and aisles. At the north end of the nave a semicircular apse, 7 metres in diameter, cuts into the thickness of the north wall and is separated from the nave by spur walls. Flanking the apse are square stair wells perhaps leading up to the second level galleries over the aisles and descending to the basement where they open into the cellars and onto the service area of the bath to the north. Like the central hall, this upper space was probably roofed with cross vaults. Two doors in each of the side walls communicate with the blocks to the east and west through light wells. The third block (W3) consists of fourteen interconnecting rooms. The largest room, immediately to the west of the basilica hall, is a long hall with an apse at its north end. Of the remaining rooms, six are rectangular in plan, two circular, one square, one square with a large apse, two cruciform and one quatrefoil. The rooms must have been covered with vaults appropriate to their plan, i.e. domes, semi-domes, barrel and cross vaults. Together these rooms presented a complex series of variously-shaped spaces which were perhaps highly decorated with mosaic and marble veneer,

33 designed both to impress the visitor and to please the owner. To the east of the central hall are three more blocks of rooms. The first (E1) mirrors the range of rectangular cubicles on the west side. The second block (E2) is cruciform and set within a rectangular, perhaps unroofed, space that extends as much as 36 by 30 metres and is surrounded on three sides by corridors and on the fourth by the gallery. From the gallery the main room is approached through a small rectangular vestibule flanked by and perhaps adjoining two small rooms. The main room has a cruciform interior plan with semicircular corner niches. Doors in the north, east and west walls open onto smaller cruciform rooms that in turn have doors in their outer walls giving access to the area between the corridors and the block itself. The ceilings of the side chambers must have been cross-vaulted while the central room may have had a domical vault. The wall surfaces were perhaps covered with marble veneer and decorative architectural elements. The third block (E3) is not well known and seems to have been a simple series of large rectangular rooms separated from block E2 by a light well.72

72

See MacNally,et al. pp.13-15 for excavation of the substructures of the easternmost block. A general description of the residence is provided by Bulic, pp.63-67 and Niemann, pp.95-109, figs. 129-140.

34 The function of the southern part of the site has been much discussed. It is generally thought to have been the personal residence of Diocletian whereas the area to the north is regarded as storage and housing of Diocletian's retinue.73 The domed vestibule has been granted ceremonial status, perhaps a throne room, although the location of doors in both the north and south sides makes this space unsuitable for such a use. More likely it was just a transitional room between the outside street and the apartments within, in other words, a vestibule74 like that opening onto the northwest portico of the Domus Flavia on the Palatine at Rome.75 The hall to the south of the vestibule, often interpreted as the tablinum, should rather be seen as a splendid hall giving access to the gallery and to blocks W1 and E1 on either side.76 The rows of small rooms, W1 and E1, are open onto corridors and are flanked by light wells for the larger structures on either side. Block W2 is basilical in form and closely parallels

the basilical halls acting as tablina in contemporary

73 74 75

Bulic, pp.32,61-67. J.J. Wilkes, op. cit , p.62.

RIA, fig.36; W. MacDonald, The Architecture of the Roman Empire, II , New Haven, 1982, p.54, fig. 40 #8. This room of the Domus Flavia has a square outside plan. The interior is circular with semicircular niches in the corners. Four doors, one in each of the sides, give access on one axis to the peristyle and on the other axis to flanking complexes of semicircular rooms. The room probably had a domed roof. 76 J.J. Wilkes, op. cit. p.62.

35 villas and palaces.77 The form is an element common to domestic architecture and must have served as a reception room.78 The block of 14 small rooms, W3, must have been relatively private apartments for the relaxation of Diocletian, and perhaps included his bedchamber. The oddly shaped block, E2, resembles a similar block of rooms at Gamzigrad and to a certain extent the triconch at the Piazza Armerina. It is set within a rectangular area, and presumably was well lit from outside. It was perhaps used as the triclinium or dining room.79 Finally, the gallery provided access to the various apartments and as a promenade overlooking the sea.

77

For instance at Piazza Armerina and the Villa of Maxentius on the Via Appia. Close parallels on a larger scale can also be found in the basilical halls of the palaces of Trier and Thessalonika. 78 Ibid, p.62. 79 The structure at Gamzigrad has not been fully interpreted, but according to Sear, Roman Architecture, p. 270, the triconch at the Piazza Armerina may have been a triclinium.

36 CHAPTER 3. CAMPS.

The idea that Split's plan was derived from Roman military camps has become an accepted fact in the most authoritative textbooks in English. In the introduction I have pointed out that this idea may have originated with Sebastiano Serlio in the fifteenth-century, and was certainly entertained by Spon in 1677. The idea still appears in the accounts of recent excavations at Split80 and in a recent article which compares the plan of Split to the arrangements of those found in Late Roman castra with a cruciform street plan and limes forts.81 In order to establish whether military architecture influenced the plan of Split it is necessary to examine the various types of plan used as comparanda by previous commentators. These include traditional full scale legionary camps, camps with crossing streets and limes forts. The degree to which Split was influenced by these building types will be assessed. Finally the Camp of Diocletian at Palmyra and the complex at Luxor will be discussed and their function and plan compared with Split.
80 81

J. Marasovic et al., 3. Fellmann, "Diokletianspalast", 47-55. Fellmann assigned Split to his last "catch-all" group, which displays mixed features. These include the fortifications, principia and shape of traditional camps, the arrangement of strigae found in small forts such as Alzei and Tamara, the crossing street plan found at Portchester, Palmyra and Luxor, and the overall arrangement of forts like Daganiya and Qasr Qarun. It will become apparent that Fellmann's typology is flawed and that his examples either do not have the features he describes, are anachronistic comparisons or are "odd".

37 The descriptions of a traditional camp plan provided by Hygenus (before the mid third-century) and Polybius (c.200-c.117 B.C.)82 refer specifically to the construction of marching camps (fig. 3.1). Archaeological evidence for permanent legionary camps corresponds to these descriptions and the plan abounds throughout the empire. Typical examples are the castra of Neuss, built in stone on the lines of its turf original in the late firstcentury, Housesteads (fig. 3.2), a second-century castra built for 800 men, and the Syrian camp of el-Leggun (fig. 3.5) on the eastern limes, which was constructed in the early years of the fourth-century. The camp described by Polybius (fig. 3.1) was one designed for two legions and allies (about 18,600 men). It was to be laid out in a square 2017 Roman feet to each side (596 metres), and enclosing an area of about 35.5 hectares. The camp was divided into two unequal parts by the via principalis which ran from the porta principalis sinistra to the porta principalis dextra and flanked the front edge of the principia enclosure. The via praetoria ran perpendicular to the via principalis from the porta praetoria to the entrance of the principia where it joined the via principalis. Lesser streets divided the interior of the camp into orthogonal blocks laid out for the housing of troops, for the stabling of animals and for the
82

Hygenus, De munitionibus castorum. Polybius, Historiae 6.27-42.

38 storage of weapons and supplies. The via decumana extended the line of the via praetoria beyond the principia compound, and led to the fourth of the main gates, the porta decumana. An intervallum street ringed the occupied interior and separated it from the circuit wall; this street is often called the via sagularis. The via sagularis gave free access to the ramparts and ensured free circulation from all parts of the camp to each of the four gates.83 The Hygenian camp was for three legions (approximately 40,000 men) and in most respects it closely resembled the plan of Polybius.84 Lambaesis (fig. 3.3), in North Africa, from the Hadrianic period, follows the Hygenian scheme. It was rectangular with sides 550 by 450 metres long and was divided into three parts by transverse, east-west streets running in front of and behind the principia. These correspond to the via principalis and via quintana respectively. Joining the via principia at a monumental tetrapylon in front of the principia, the via praetoria ran north to the north gate. Both these main streets were over 20 metres wide, were flanked by colonnades, and unlike the lesser throughfares which had earth surfaces, were paved. The principia was

83

E. Fabricius, "Some Notes on Polybius's Description of Roman Camps", JRS, 22 (1932) 78-87; Lander, pp. 12-15; A. Johnson, Roman Forts, London 1983, 27-29. 84 A. Johnson, op. cit., pp. 29-30.

39 particularly large, measuring 102 by 93 metres (9500 sq.m). On the south side a row of four structures with apses flanked administration buildings. In front of this range was a large court surrounded on three sides by smaller rooms which opened onto colonnades. Access to the complex was through a monumental entranceway straddling the junction of the via principalis and the via praetoria.85 The remaining area is subdivided into blocks reserved for barracks and storage. It can be seen from this discussion of traditional Roman legionary camps that the principal streets describe a "T" plan not a "+" plan as is found at Split.

A late development, starting in the "crisis" of the third-century, and put into effect largely by Diocletian, was the reorganisation of the limes to create a `frontier in depth'. As a result, the traditional legionary camp acted as a headquarters and strategic reserve behind the lines, while a large number of forts and fortlets were spread densely along the limes. The traditional camps were modified in design resulting in the reduction of the importance of certain gates. The porta decumana and porta praetoria of these camps are often small, not much more than posterns, while the porta principalis sinistra and dextra remained important. In this new scheme the via
85

F. Sear, op. cit. p 206-208; A. Johnson, op. cit. 120, fig. 92.

40 principalis became the primary thoroughfare, and the main axis of the camp thus flanked the principia rather than running up to it. This change in plan is visible in many camps, for instance the forts of the Saxon Shore which belong mostly to the late third and early fourth centuries.86 Portchester castle (fig. 3.4) is a typical example of such a fort and is of importance to Split not only because it is contemporary, but also because it was thought to have had a cruciform street plan.87 This however, is not the case. The castra is almost square, measuring 182 metres by 188 metres, with U-shaped interval and corner towers. The four gates are located in the middle of each side. The east and west gates, only 3 metres wide, are built into 11-metre wide recesses formed by the inturning of the curtain wall. They are plainly the main entrances to the camp, and the road running between them the main thoroughfare. The gates in the middle of the north and south walls were probably postern gates. Although a street led from the south gate to the centre of the camp, there is no evidence to suggest that it continued to the north postern.88 The finds and stratigraphy suggest a

86

For the forts of the Saxon Shore see R. E. Collingwood, and I. Richmond, The Archaeology of Roman Britain , London 1939, 1969, p 4756, fig. 20. 87 Fellmann, "Diokletianspalast", 49, fig. 4. 88 B. Cunliffe, "Excavations at Portchester Castle, Hants. 1961-3," AJ 43 (1963) 218-227; idem , "Excavations at Portchester Castle, Hants.

41 construction date in the late third-century under Carausius. There is no evidence at Portchester that the via principalis and via praetoria crossed, or to suppose that the principia occupied any but its expected position to the north of the intersection of the via principalis and via praetoria.89 The cruciform street plan of Drobeta has, like that thought to exist at Portchester, been compared with Split.90 Drobeta, a small traditional legionary camp measuring 127 by 144 metres, was laid out during the reign of Trajan (fig. 3.5 left).91 At some point, probably during the reign of Aurelian (270-275), the camp was modified. The gates at the ends of the via principalis and via praetoria were closed off and converted into interval towers, while fan-shaped towers were added at the corners with projecting interval towers in between. The camp was further modified under Valentinian (364-375), when the principia was either moved or eliminated (fig. 3.5 right).92 This final phase of construction radically changed the internal plan of the camp. The main northsouth street remained on the axis of the original gates,
1966-1968" AJ 49 (1969) 62-74; idem "Excavations at Portchester Castle, Hants. 1969-71," AJ 52 (1972) 70-84. 89 R. Fellmann, "Diokletianspalast" fig. 4 uses a plan reproduced by H. Von Petrikowits, "Fortifications in the North Western Roman Empire from the Third to the Fifth Centuries A.D.", JRS 61 (1971) 178-218, fig. 20. In this illustration the lines of the streets are shown running for a short distance from the four gates giving the misleading impression that they crossed at the centre of the camp. 90 Fellmann, Diokletianspalast, 49, fig. 10. 91 Lander, 171, fig. 30. 92 Lander, fig. 274.

42 while the main east-west street was moved from the line of the via principalis northwards to intersect the via praetoria at the middle of the camp. Both streets were flanked by colonnades and strigae. At Porchester therse is no evidence for crossing streets, and at Drobeta the pattern seems to have evolved in the latter part of the fourth century, long after the construction of Diocletian's Palace at Split. Comparisons between Split and cruciform plans are therefore, at best, anachronistic.93 In another part of the empire, the traditional "playing card" plan persisted, but even though the "T" shaped arrangement of streets continued to be employed, the emphasis on their relative importance shifts. ElLeggun (fig. 3.6) has only recently been explored in detail, but the Diocletianic date and apparently traditional plan have been long known.94 The camp was rectangular, measuring approximately 270 by 210 metres, with circular corner towers and projecting U-shaped interval towers. Of the gates in each of the sides the north and east have a triple entrance consisting of the main gate with flanking passageways for pedestrian traffic. Both the west and south gates have only one aperture. The principal streets lead from the north, east
93 94

Lander, 171. Brunnow and Domaszewski, fig. 42; S. Thomas Parker, "Central Limes arabicus Project", ASOR Newsletter, December, 1982 p 6-14.

43 and south gates and meet at the centre of the camp in front of the principia. Recent excavations have revealed the remains of a monumental tetrapylon marking the groma at the intersection of the principal streets.95 The principia complex consisted of a large rectangular court, 64 by 41 metres in plan, with three large buildings at the west end. The central structure of the complex had a monumental entrance 3 metres wide which still preserves fragments of a metal barred door. Around the walls a low podium was carried on barrel vaults, and opposite the door was a raised pier with a socket, perhaps to receive the legionary standard. It has been suggested that the vaults lining the wall served as strong boxes of the bank. Displaced to either side of the principia were blocks of barracks and stores. The features in the plan of Split thought to have been drawn from Legionary camps are the street plan, the form of the fortifications, and the location of the principia. The cruciform arrangement at Split was compared with that of camps such as Portchester and Drobeta by Fellmann, but these comparisons do not stand scrutiny. At Portchester there is clear evidence that the streets did not cross, whereas Drobeta did not acquire its cruciform plan until long after Split was built. In

95

For the tetrapylon see S. Thomas Parker, "Central Limes arabicus Project" Syria Vol. 63 (1986) pp. 401-5.

44 addition, the comparison of the Split plan with that of the traditional castra is anachronistic. Portchester and el-Leggun are just two examples of how, by the turn of the fourth-century, the emphasis of axis and access had changed in large camps. These examples demonstrate a reduction in the number and location of principal gates and consequently the emphasis on the major axes.96 At Split, the gates and main streets show little of this emphasis. As is proper, the cardo is larger than the decumanus and it is clear that the Porta Aurea is the principal entrance. This the form expected in the plan of both cities and the older castra but not that of contemporary castra. It must, however, be admitted that there is a visual, if illusory, resemblance in the plan and that Split's fortifications are derived ultimately from a military source. The southern block at Split has been compared with the principia of camps as both were thought to have been residences; at Split that of Diocletian, and in camps that of the commander. In camps the principia occupied a central location at the intersection of the via principalis and the via praetoria where it was accessible to all the garrison. It consisted of a open area for parades and buildings devoted to the administration of the

96

Other examples include numerous eastern forts, see Lander, figs. 185, 202, 207, 211, 212, and 220.

45 camp, including the bank, offices, and the temple of the signia. The residence of the commander was not in the principia, but in a separate smaller complex called the praetoria, also near the centre of the camp. For instance, at Dura on the Euphrates the commander resided in quarters near the principia, while the villa of the Dux ripae stood on the edge of the city some distance from the centre of the camp.97 At Split the element thought to be comparable to the principia was neither central nor accessible; the private apartments occupied the whole width of the extreme south end of the site and was accessible only from a single entrance on the north side. There was no Temple of the Signia, no bank, no offices, no mustering place, indeed it was private not public space. It is also essential to recognise, as emphasised in chapter 2, that the private apartments were only a fraction of the residence as a whole, which occupied the whole interior of the walled enclosure.

97

Fellmann, Die Principia, p 75-92, and idem. "Le Camp" 178, convincingly argues that the principia and praetorium had quite separate functions. The former had a purely military and religous function while the latter was the residence of the military commander. Fellmann places both near the centre of the camp, but puts the principium at the intersection of the principal streets. Some still prefer to refer to the central complex as the praetorium, for example, A. Johnson, op. cit.. p 29-30 and 311, n. 24.

46 The cubicles ranged around the curtain wall have been compared with the strigae of small forts in general. By the turn of the fourth-century Small forts were used increasingly frequently for outposts and garrisons on the limes. These were generally limited in scale and made economical use of all the available area within the walls. The space saving features developed by military architects included the removal of storage and living quarters from the centre of the camp and arranging them along the defensive walls. It is this innovation that was apparently used in the northern part of the enceinte at Split, and raises the question whether the plan of Split may have been influenced by fort design. Of the many forts of this type, two serve to illustrate the variety of plans employed. These have been chosen because they have been used as comparanda for Split. The simplest type is exemplified by Masad Tamar (fig. 3.7) which is typical of small of the limes in the east and in North Africa in Late Roman times. Particularly interesting is the identification of the function of the various rooms in the fort, especially the recognition of the principia, headquarters and praetorium.98 The fort is located on the limes palestinae in the northern Negev desert about 20 miles south of the Sea of Galilee, and is

98

M. Gichon, "Three Years Excavation at Masad Tamar (Tamara)", Saalburger Jahrbuch 33 (1976).

47 attributed to Aurelian (270-275) or to Probus (276-282). It was laid out in a square, 38 by 38 metres in plan, with square angle towers and a single gate in the middle of the northwest side. The buildings were arranged exclusively

along the curtain wall leaving a central courtyard occupied by a large subterranean cistern. The strigae , in two groups of six rooms, were placed along the the southwest and southeast sides only, and had colonnades in front of them. The rooms on the northeast side were larger than the strigae and have been interpreted as the sacellum, principia and, in a suite of rooms in the northeast corner, the praetorium. On the northwest side, guard chambers flanked the gate, while other rooms were apparently stores and a bakery.99 Access to the ramparts above the rooms abutting the wall was provided by steps to either side of the gate and in the southeast corner of the courtyard. Gichon suggests that each striga was occupied by as many as 10 men, and that the whole fort, only 1450 square metres, housed a total of over 100. The fort was built at the end of the third-century and was discontinuously occupied until 635.100 A variant of the basic plan represented by Masad Tamar is to be found at Qasr Qarun (fig.3.8), southeast of
99

In the room identified as a bakery there were found grinding stones for milling grain and a furnace. 100 M. Gichon, "Tamara" in Encyclopedia of Archaeological Excavations in the Holy Land, p 1148-1152. Edited by M.Avi-Yonah, and E.Stern. Jerusalem, English edition, 1978.

48 Cairo in Egypt.101 The fort is larger than Masad Tamar and incorporates interior blocks of strigae into the plan. Besides the strigae, the short street flanked by colonnades is of particular interest, for it has evoked comparison with the street leading to the vestibule at Split. Qasr Qarun had a rectangular circuit, measuring 93 by 79 metres in plan, with square angle towers and U-shaped interval towers. The solitary gate on the north side consisted of a triple portal which enclosed a narrow space in the thickness of the wall, and between the second and third portals, a small court flanked by benches. The interior of the fort was divided into two halves by an 7.5-metre wide avenue flanked by colonnades which ran from the gate to a structure against the south wall. This building, measuring 12 by 7 metres in plan, consisted of a small square cella entered from the north through a shallow porch offset from the avenue by a short flight of steps and by columns in antis. The south wall of the cella was filled by a broad semi-circular apse and doors in the west wall gave access to two small rooms. The identification of this structure as a temple was based partly on the plan and partly the discovery of a small Tyche or Nemesis statue found in the cella.

101

J. Schwartz, et al. Qasr Qarun/Dionysias II, 1950, Cairo, 1969.

49 To the west of the temple a 17.5 by 20-metre block of rooms abutted the south wall of the fort. These interconnecting rooms were arranged around a central staircase leading to an upper storey and were thought to have been the administrative block. To the east of the

temple was a long hall lined with a bench along the walls. The barracks were arranged along the walls on every side except the west end of the south wall. Two extra blocks of strigae flanked the colonnaded avenue and were entered on the side away from the street. The blocks were disposed so that the doors of the strigae opened onto porticoes surrounding two courtyards. Stairways in each block gave access to both the defensive towers and a second storey. The date of Qasr Qarun is fixed by references to the fort in papyri. A number of documents contain administrative data prepared by the commander of the camp, Flavius Abinaeus, who was appointed the prefect of a cavalry ala quingeneria (500 men) in March 340. Such texts demonstrate that the fort existed by the middle of the fourth-century, and earlier fragments may attest its existence at the beginning of the century. Its location, set well inside the southern border of the empire, suggests a role linked with administration rather than defence. The fort has attracted the attention of architectural iconographers because of its unusual colonnaded street

50 leading from the gate to the temple. The excavators concluded that this was part of a single building element which they interpreted as the archetype of a "ceremonial basilica". They compared this with the approach to the vestibule at Split and also to that part of the palace of Theodoric shown in the early sixth century mosaic in San Apollinare Nuovo at Ravenna.102 The concept of a "ceremonial basilica", however, is inappropriate because the avenue is a thoroughfare not a self-contained building. The similarity between the the approach to the temple at Qasr Qarun and the approach to the vestibule at Split is striking, particularly if the temple facade were to be reconstructed with a "Syrian arch", but not nesessarily significant. Streets with colonnades and "Syrian arches" were particularly common in the east, and it would be surprising if the two were not used together frequently. Differences between Split and Qasr Qarun are

reflected in the relative complexity of their plans; that one was a residence, the other an administrative fort, in one the colonnades lead to the vestibule of the private apartments, in the other to a temple. Similarities are the cubicles, the gate form and the street.

102

J. Schwartz, et al., op. cit.; for the mosaic in San Apollinare see E.Dyggve, Ravennatum Palatium Sacrum, fig. 49; J.-M. Carrie, "Les Castra Dionysiados et l'evolution de l'architecture militaire romaine tardive", MEFR 86 (1974) 819-850.

51 The use of the defensive wall at Split as the back wall of magazines has been compared with the similar features in small military forts. This use of the fortification enceinte for the back wall of barracks and other military structures is considered to be a relatively late phenomenon, but a number of earlier examples are known from as early as the end of the second-century.103 It now appears that this practice was not chronologically significant, but rather a question of size. The use of cubicles along the defensive wall is usually restricted to smaller defensive structures with wall lengths between 40 and 65 metres, although castra of intermediate size, such as Qasr Qarun, often display a combination of internal structures with structures built along the wall. The displacement of structures within forts was intended to save space, and a similar concern may have prompted the designers of Split to adopt this already well established practice. The plan of Split does seem to have been influenced by the design of forts in as much as the architects of

both utilised as much of the available interior space as possible. Apart from banal comparisons between the fortifications of forts and Split, similarities are hard
103

For example, Tisavar in Tripolitania, dates to the reign of Commodus (180-192); Lander, 102-4. Qser el-Uweinid, Lander, 139, fig. 132; Qasr el-Hallabat in its final phase, Lander, 139 and 260; Qasr el-Aseiklun, Lander, 139, fig. 132, all date to the early thirdcentury.

52 to find. The avenue and colonnades approaching the temple at Qasr Qarun, which are reminiscent of the Peristyle of Split, are hardly typical of military architecture, and more likely are derived from a third, unrelated source. This element seems to be tied to the function of the camp as an administrative centre. The majority of forts have a practical plan related to their military function. They are strung along the limes and are extremely small.104 Split too has a plan related to its function, but this is residential not military. Any resemblance between Split's plan and forts lies in the fortifications and the presence of cubicles along the defensive wall.

The closest parallels for the street plan of Split are to be found at Palmyra and Luxor. These three have a cruciform plan unparalleled in military architecture of the period. The argument that Luxor, Split and Palmyra have military functions is somewhat circular; when pointing to the military plan of one, the other two are invoked as proof, yet independent evidence is not forthcoming. Similarly circular is the argument that the "Camp of Diocletian" at Palmyra was a palace of Zenobia. The city of Palmyra grew rapidly during the firstcentury. The city reached its apogee of wealth in the 3rd
104

The largest fort, Da'janiya (1 ha.), Brunnow and Domaszewski, 8-13, fig.1, is only 25% of the size of Split (3.9 ha.), while the smallest, Masad Tamar (0.15 ha.) has only 4% of the area of Split.

53 century, and declined after a brief spell of independence under Septimius Odenathus and his wife Zenobia. Under the Tetrarchs the town played a part in his system of defence in depth on the limes arabicus. Diocletian constructed new walls, refurbished a number of existing buildings and perhaps added new ones.105 Numerous major monuments have survived including two Roman military camps. Of one camp relatively little is known except from late second-century inscriptions and associated finds.106 The other camp is the so-called "Camp of Diocletian" located just within the western salient of the defensive wall next to the Damascus gate off the Transverse Colonnade. The "Camp of Diocletian" lies within an irregular westward extension of the city fortification wall (fig. 3.9). The complex is separated from the remainder of the

city by a long wall running parallel with the Transverse Colonnade. This wall is pierced by a single triple gateway called the porta praetoria by the excavators. The excavated lower part of the camp is divided into four irregularly shaped blocks by two broad streets intersecting at a monumental tetrapylon. The primary street, the via praetoria, runs west from the porta praetoria 58 metres to the tetrapylon and thence 47 metres

105

M. A. R. Colledge, The Art of Palmyra, London 1976, 11-23; D. Crouch, "The Ramparts of Palmyra" Studia Palmyrenskie 6 (1975) 6-44. 106 I. A. Richmond, "Palmyra Under Rome" JRS 53 (1963) 43-54, 49. For the location see J. Starcky, Palmyre Paris, 1959, 42, pl. 3.

54 to steps leading to the gateway of the forum and the principia beyond. This street, about 12.5 metres wide, is flanked by continuous colonnades on a line with the pilasters of the tetrapylon to which they were was attached. The via principalis, which crosses the via praetoria at the tetrapylon gives the "camp" its cruciform plan; it is narrower and was not attached to the tetrapylon.107 The tetrapylon is rectangular, 14.25 by 14.45 metres in plan, and is made of white calcareous blocks. Nothing of the superstructure has survived, but sufficient evidence exists to show that it had a pediment and architrave. These were supported by facades with pilasters and columns in antis standing 7.4 metres from the ground to the architrave.108 According to the excavators the streets and the tetrapylon were not contemporary. The earliest phase consists of the colonnade in front of the temple of Allat which may predate the tetrapylon by as much as 150 years. The other colonnades were also constructed before the tetrapylon.109 Although the decoration and style of the tetrapylon suggest a date in the middle of the third-century, the presence of reused architectural, sculptural and inscription fragments built

107

K. Michalowski, Palmyre I. Fouilles Polonaises, 1959 , Paris, 1960, 41ff. 108 Ibid., 10-12. 109 Idem, Palmyre II. Fouilles Polonaises, 1960 , Warsaw 1962,20,40-1.

55 into the foundations are thought to be evidence that it was constructed after the Aurelian sack, probably during the reign of Diocletian.110 At the west end of the via praetoria a stairway leads to the entrance of the forum. This gate is almost completely preserved to the top of the pediment.111 The colonnades of the street run up to the facade and the porticoes open onto subsidiary entrances to each side of the principal gate. The central door is preceded by two columns, apparently free standing, and is flanked by pilasters supporting the pediment above the door which is broken by a "Syrian arch" with an in-filled tympanum. Two phases of the gate have been identified. The earlier was built after the middle of the second-century; the later, on the basis of its architectural decoration, was built at the same time as the tetrapylon.112 The forum is a rectangular space, 48.5 by 62.4 metres, surrounded on three sides by a colonnaded portico. On the east and west sides small rooms open onto the portico, with a small latrine in the southwest corner. On the north side of the forum, stretching its whole length, is the principia. It is entered through a triple

entranceway preceded by a tetrastyle porch and a grandiose


110 111

Fellmann, "Le Camp", 174-5. K. Michalowski, Palmyre III. Fouilles Polonaises, 1961, PWN: Warsaw 1963, 21 and fig.20. 112 Idem, Palmyre II, 38; M. Krogulska, "A Ceramic Workshop in the Western Quarter of Palmyra," Studia Palmyrenskie VIII (1985) 68.

56 flight of steps.113 The triple door opens onto a long transverse hall stretching the length of the building (62.5 metres). Off the hall a series of rooms along the

west wall are entered through broad doors divided by columns in antis. The central room of this series lies on the central axis of the complex, and has been described as the "Temple of the Signia" by the excavators. It is 14.6 metres long by 11.9 metres broad with an 9.5 metre diameter apse opposite the door. It was above this door that the inscription thought to identify the complex was found by early travellers.114 To either side of the `Temple' are flanking rooms which presumably served as scholae and archive rooms. The decoration appears to be contemporary with the tetrapylon and the entrance to the forum.115 There is controversy about the date and function of the "Camp of Diocletian". Early travellers made no connection between the inscription found in the principia and the area in which it was found, thinking that it
113

M. Krogulska, "Les principia" in M.Galikowski, ed. Palmyre VIII, Warsaw 1984, 70-90. 114 AA 21 (1906) 43. The inscription reads; [reparato]res orbis sui et propagatores generis humani DD Diocletianus [et Maximianus]/ [invicti]ssimi imp et Constantius et Maximianus nobb caess castra feliciter condiderunt/ [curam age]ite Sossiano Hieroclete V.P. Praes Provinciae D. N.M. Q. eorum/ CIL, III 133 no.6661. The name of Maximian was erased and later redrawn in purple, presumably after the death of Maxentius. The second Maximian refers to Galerius. Sossianus Hierocles was governor of Augusta Libanensis between 293 and c.300. Before 303 he became vicarius of Oriens: T. D. Barnes, The New Empire of Diocletian, 141, 153. 115 M. Galikowski, "Les Principia de Diokletien, `Temple des Enseignes'" in M. Galikowski, ed. Palmyre VIII, Warsaw 1984, 10-48.

57 referred to the refortification of the whole city by Diocletian.116 Most scholars except Schlumberger accept the identification of the area as a camp. Schlumberger concluded that the camp was actually a palace of Odenathus and Zenobia adapted for a military use at a later date. As proof he offered the unusual plan, the lack of regular ramparts, the date of the architectural decoration and the apparent similarity to Split.117 In favour of the camp interpretation are the plan of the inscription. A summary of the chronological phases demonstrates the problems of interpretation that have been encountered. Initially, part the site was used as a cemetery, and part as a pottery workshop. Sometime in the second-century, crossing streets with colonnades and a stairway up to a gate on the west side of the site were constructed, and the Temple of Allat built. This early phase of the plan was later modified by the addition of a tetrapylon at the intersection of the street, a new stair ramp and gateway, and a large building called the principia by the excavators. This phase is datable on decorational style to the middle of the third-century, but may represent reuse of decorated architectural elements during the reign of principia and the

116

R. Wood, The Ruins of Palmyra Otherwise Known as Tedmor in the Desert, London 1753, 31. 117 D. Schlumberger, Le pretendu Camp de Diocletien a Palmyre, MUSJ 38 (1962) 82.

58 the Tetrarchs. It is unlikely that reused blocks could have been used to produce such a unified and complete decorational programme as is found in these three later major structures, thus Schlumberger's suggestion, that the buildings were merely refurbished for a military use by Sossianus Hierocles, is a real possibility.118 Whatever the original and final intent for the complex, it is difficult to compare it with Roman camps, on the one hand, or with Split on the other. The Street plan, even if original and designed this way, is only exceptionally found in castra.119 The principia, however, does have close parallels with that of Lambaesis.120As the plan evidently developed over the course of a century or more, and cannot be considered typical of camp plans, the use of the "Camp of Diocletian" as proof that Split developed from military architecture is totally inappropriate. There is no evidence to support Schlumberger's contention, based on comparison with Split, that the complex at Palmyra was a palace. Both Split and the "Camp" have crossing streets, but in one it was a planned element in a residence, in the other it is a odd survivor of a pre-existing city plan later employed for a different function. The principia so closely parallels

118 119

Ibid., 82. At Luxor, and in rather a different form in the latest phase at Drobeta. 120 Fellmann, "Le Camp" 173-191.

59 military administration blocks that comparison with the private quarters at Split is difficult to accept.

LUXOR Another site compared with Split is the contemporary double castra at Luxor in Egypt (fig. 3.11). Again the point of comparison is the cruciform street plan. The camp was laid out at the beginning of the fourth century encompassing the New Kingdom Temple of Ammon at Luxor.121 Up to this time the temple had consisted of a large complex of hypostyle halls and courtyards arranged on axis and covering an area of 60 by 250 metres. The camp was laid out parallel with the east bank of the Nile and incorporated the temple as its long axis. Enough has been uncovered to show that the walls extended perhaps as far as the south wall of the temple and enclosed an area of approximately 190 by 250 metres (4.7 ha.). Excluding the pylon of Ramses II on the north side of his courtyard, three gates have been revealed and the location of a fourth is implied. Two gates are situated one to either side of the Pylon of Ramses II, and there is one in the east wall. As at Split, each of the gates is flanked by towers and opens into a rectangular court closed by a

121

See M. P. Lacau, "Inscriptions latines du temple de Luxour," Annales du Service des Antiquites de l'Egypte, 34 (1934) 17-46; V. Monneret de Villard, "The Temple of the Imperial Cult at Luxor," Archeologia 95 (1953) 85-105.

60 second gate. A difference is that connecting the court with the right side of each of the flanking towers are passages leading to sally ports. In the northwest part of the castra, streets flanked by colonnades intersect at a tetracolumnar monument similar to one at Antinoe commemorating a visit of Severus Alexander.122 On the bases inscriptions in honour of Diocletian, Maximian, Galerius and Constantius Chlorus can be reconstructed.123 These were placed on the sides facing the principal east-west street leading fron the Nile to the courtyard of Ramses II. On the east side of the Temple of Ammon the street arrangement almost duplicates that on the west side. The

bases are inscribed, on the sides facing the north-south street, in honour of Gallerius, Licinius, Constantine and Maximin Daia who ruled together from 308 to 311.124 The Temple of Ammon was reused as a principia for which the courtyards of Ramses II and Amenophis III acted as forecourts. Between the courtyard of Amenophis III and the inner sanctum of the temple is a hypostyle hall. This hall is divided into nine aisles by rows of columns, and in the early fourth-century the central aisle was separated from those flanking it by low walls. The central aisle of the hypostyle hall led into a rectangular room,
122 123 124

M. P. Lacau, op. cit. 20. Ibid. 20-35. Ibid. 20-29.

61 10.5 by 17.45 metres in plan, oriented perpendicular to the main axis of the temple. In front of a niche in the wall opposite the entrance four columns supported a ciborium.125 Frescos covering the walls were destroyed in the nineteenth century by Egyptologists wishing to uncover ancient texts beneath. These were recorded by the watercolourist J. E. Wilkinson in 1859 before they were destroyed.126 The upper register of the east wall depicted a procession of dismounted horsemen and pedestrians clad in white towards the niche in the south wall. The focal point of the programme showed Diocletian in a chariot127 and suggests that the scene commemorated the adventus of Diocletian to Luxor. In the niche four figures dressed in purple probably represent the emperors of the first college of Tetrarchs.128 The function of the painted hall is disputed. Monneret de Villard and Deckers consider it to be the sacellum of the camp129 Maxeiner identifies the room as an audience chamber which, with the
125 126

adventus scene on the

U. Monneret de Villard, op. cit. 88-9. I. Kalavrezou-Maxeiner, "The Imperial Chamber at Luxor," DOP 29 (1975) 227-51. 127 This was not recorded by Wilkinson, but page fifty-three of his sketchbook, the page reserved for this wall, bears the notation "Mr. Monier told Mr. Harris that the name of Diocletian was on one of the chariot wheels in this fresco." I. Kalavrezou-Maxiener, op. cit. 2389. 128 U. Monneret de Villard, op. cit. 94, 101-2; I. Kalavrezou-Maxeiner, op. cit. 244-6. 129 U. Monneret de Villard, op. cit. 101-102; J. D. Deckers, "Die Wandmalereien des tetrarchischen Lagerheiligtums in Ammon-Tempel von Luxor", Romische Quartalschrift , 68 (1973) 1-34; idem. "Die Wandmalerei im Kaiserkulten von Luxor", JDAI 94 (1979) 600-652.

62 walls, strongly suggests that Diocletian may have used Luxor as a base for operations in Middle Egypt in 298.130 The Temple of Ammon could well have been used as the palatium of Diocletian, and after his visit, as a temple of the imperial cult. The planned lay-out of Luxor is unparalleled by Roman camps of any period, even at Palmyra and Drobeta. Instead of via principalis and via praetoria meeting in front of the entrance to the principia as in a regular camp, here they cross at right angles with the crossing marked by commemorative columns. Moreover, the plan has been duplicated with crossing streets either side of the Temple of Ammon. This duplication of the camp street plan has been used to suggest the origin of the name Luxor in Arabic; literally, two castra.131 The temple was not used to enhance the defences of the camp, but rather as a central focal element. The date is well attested by the

monumental tetrastyles which seem to be contemporary with the construction of the walls and streets. On the basis

of the inscriptions, the camp was laid out in or soon before 300/301 and added to in 308/309.

The supposed military origin of the plan of Split has important repercussions on the interpretation of elements

130 131

I. Kalavrezou-Maxeiner, op. cit. 249. U. Monneret de Villard, op. cit. 96.

63 found elsewhere in Late Roman urban forms. The so-called

"Camp of Diocletian" on the outskirts of Palmyra, with its arrangement of streets crossing at a central tetrapylon in front of a "praetorium" has been used to reinforce the association between Split and a military camp plan. It should now be apparent that any resemblance between castra and Split is limited to the fortification wall and to the cubicles, and that the street plan is unrelated to military architecture except by accident at Palmyra, and in an oddity at Luxor. Conveniently, the "Camp of Diocletian" at Palmyra points the way to the next avenue of investigation. The "palatial" nature of Split has been used to argue the case that the Palmyrene camp was, in fact, a palace of Zenobia which was reused during the Tetrarchy as a camp and which provided the inspiration for both Split and the palace at Antioch.132 The relationship between Split and Roman palaces will be explored in the next chapter.

132

R. Fellmann "Le `Camp de Diocletien' a Palmyre et l'architcture militaire du Bas Empire," in Melanges d'histoire anciennes et d'archeologie offerts a Paul Collart, (Cahiers d'archeologie Romande. no.5). Lausanne, 1976, pp. 173-91. Fellmann sets out the parallels between the plans of military camps, the `Camp of Diocletian', and the palaces at Antioch and Split. For the theory that the Palmyrene complex was a palace see D.Schlumberger, "Le pretendu Camp du Diocletien a Palmyre," MelUSJ 38 (1962) 77-97 and Idem, "Notes sur la decor architectural des colonnades des rues et du Camp du Diokletien a Palmyre," Berytus 2 (1935) 163-67.

64 CHAPTER 4. PALACES.

Some modern scholars have sought the origin of the unusual features of Split among contemporary and later imperial palaces. A case for comparison with Tetrarchic palaces was made by Downey who analysed the ancient descriptions of the palace at Antioch and concluded that this complex served as the model for Split. As indicated in the introductory chapter, this approach was followed by Smith, Swoboda, Dyggve and l'Orange,133 each of whom drew liberally from the fragments of known and suspected palaces from a broad geographical and chronological range to illustrate their argument. In this chapter, I shall define the function and development of a Roman palace and before concluding what the nature of Split's relationship to palaces was, I shall describe the most completely understood contemporary palace complex at Thessalonika.

The usage of the term "palace" changed from the early to late empire. Initially the word referred specifically to the imperial residence on the Palatine hill in Rome, between the Forum Romanum and the Circus Maximus (fig. 4.1). This residence originated with Augustus' relatively
133

E.B. Smith, Architectural Symbolism; K.M. Swoboda, Romische und Romanische Palaste; E. Dyggve, Ravennatum Palatium Sacrum; L'Orange, Art Forms and Civic Life.

65 modest house near the spot where tradition placed the hut of Romulus. The Domus Augusti was a private dwelling, built in a traditional Republican Roman style, purchased from the family of the Hortensii.134 Tiberius built a much larger house on the Palatine to the northeast of the Domus Augusti. This house of Tiberius was subsequently enlarged by Caligula to include the Temple of Castor which then acted as a vestibule to the residence.135 Nero foreswore the Palatine hill and built a complex resembling a sprawling country residence which covered much of the area between the Palatine and Esquiline hills.136 Finally in A.D. 92, after retoring the Domus Tiberiana which had been damaged by fire in A.D. 80, Domitian started the Domus Augustiana, commonly called the Palatium, on the Palatine. This large residence, the eponymous palace, thereafter remained largely unaltered. It consisted of administrative and private residential blocks arranged around three large peristyle courts. It overlooked the Circus Maximus and was linked by a passage to the imperial box in the circus, opposite the meta secunda.137 Essentially, in the course

of some one hundred years, the palace had developed from a simple residence into a vast, opulent urban villa that stretched from the civic and religious centre of the city
134

E. Nash, Pictorial Dictionary of Imperial Rome, 2nd ed. vol. 1, London 1968, 310-15. 135 Ibid, 365-73. 136 Ibid. 339-48; RIA, 57-61. 137 E. Nash, 316-338; RIA 77-84. The passage was built under Trajan.

66 across the Palatine to the Circus Maximus. In the

process, it acquired certain administrative functions as authority devolved from the senate to the emperor. The association between the Palatine residence and the emperor was strong even though much of the emperor's time was spent either on campaign, on administrative trips, at another holding in Rome, or at one of a number of country villas. Examples of these latter include the Gardens of Sallust on the Pincius in Rome and, outside Rome, both Hadrian's extensive villa at Tivoli and Tiberius' villa on Capri. By the early third century, anywhere the emperor stayed was referred to as a palatium; according to Dio Cassius (c. 155 to c. 230) "The royal residence is called `palatium' not because it was ever decided that this should be so, but because the emperor lives on the palatium (sic) and has his headquarters there. His house also gained to some extent in the prestige from the hill itself, because Romulus had previously dwelt there. For this reason, if the emperor resides anywhere else, his stopping place receives the name of `Palatium'."138 On tour the emperor stayed in provincial governors' accommodations or wherever there was a suitable house. For instance, Marcus Aurelius stayed in the basileia at Sirmium which he used as a behind-the-front headquarters
138

Dio Cassius, Roman History, LIII, 16, 5-6.

67 during the Danubian wars of 169/170 A.C.139 The housing of the emperor and his retinue occasioned considerable expense which devolved on the local dignitaries of the cities in which he stayed. The route of the emperor on tour was publicised well in advance so that the necessary accommodations and festivities could be arranged in time. Sometimes structures had to be specially modified or even constructed for the occasion; Dio complains about the cost of Caracalla's tour in 214/215 saying "But apart from all these burdens we are compelled to build at our own expense all sorts of houses for him whenever he sets out from Rome, and costly lodgings in the middle of the very shortest journeys... Moreover we constructed amphitheatres and race courses wherever he spent the winter or expected to spend it, all without receiving any contribution from him."140 Perhaps it was to minimise this expense that Antoninus Pius limited his provincial tours.141 The turbulance of the third century required the emperor's presence near the borders of the empire, and as a result the emperor infrequently resided at Rome. The real seat of power, now that the army was the elector and guarantor of the emperor, was no longer with the senate, or even in Rome, but moved with the emperor and the army.

139 140 141

F.Millar, The Emperor in the Roman World, London 1977, 4. Dio Cassius, Roman History LXXVII, 9, 5-7. F.Millar, op. cit. 35.

68 Despite this, the Palatine residence continued to be used on occasion and was even expanded by Septimius Severus. Diocletian initially continued the former system of ruling from the provinces. With Maximian he informally divided their jurisdiction between the eastern and western parts of the empire, so that Diocletian principally resided, until 296, at Sirmium and Nikomedeia and Maximian at Trier, Milan and Aquileia. With the creation of the Tetrarchy in 293, the empire was loosely divided into four areas of imperial jurisdiction and strictly into four prefectures. In the course of his duties the emperor moved almost constantly within and between prefectures, staying at one of the principal residences cum administrative centres.142 Rather than change the system of movable administration, Diocletian institutionalised it by establishing palaces that doubled as government and judicial centres in a select few cites that were served by good communications and were close to the frontiers. There was probably a palace already in existence at Milan where the two Augusti met formally in the winter of 290/291 (fig. 4.2),143 but at Trier (fig. 4.3), Aquileia (fig. 4.4), Sirmium (fig. 4.5), Nikomedeia (fig. 4.6), Antioch (fig. 4.8) and Thessalonika (fig. 4.10) new palaces were

142

An exhaustive list of imperial journeys taken by the Tetrarchs and their successors has been compiled by T.Barnes, The New Empire of Diocletian and Constantine, Cambridge 1982, chapter V. 143 S.Williams, op. cit. 56-59.

69 built.144 These had certain characteristics in common, for instance their peripheral location, their proximity to a circus and their large size. Three definitions of the word "palace" emerge from the above discussion of the development if imperial residences and administrative centres. The eponymous palace was the residence of the emperor on the Palatine hill in Rome; it derived its name from its location and, even when it had long ceased to be the principal residence of the emperor, it continued to command a special place in the literature, tradition and culture of Rome. The Palatine hill residence, adjacent to the circus and to the Forum became more than a home, developing into the central element of a larger complex that embraced cultural, religious, judicial, administrative, economic and state functions. With time, anywhere the emperor resided, whether temporarily or permanently, was called a palace (alternatively regia or basileion). The palace was so called not because of its design but because of its temporary function as the emperor's court. The third definition of "Palace" describes a provincial
144

A good summary of the circus and associated monuments at Milan can be found in J. H. Humphrey, op. cit. 613-620; a less complete account is provided by R. Krautheimer, Three Christian Capitals, Berkeley 1983, 69-70 and by N. Duval, "Aquileia e Milano," Antichita Altoadriatiche 4 (1973) 158ff. For Trier see E.M.Wightman, Roman Trier and the Treveri, London 1970, 110; W. Reusch, "Trier" AA 77 (1962) 875-888. For Aquileia see N. Duval, op. cit. For Simium see V.Popovic, "A Survey of the Topography and Urban Organisation of Sirmium in the Late Empire," Sirmium I Belgrade 1971, 119-133. N.Duval, "Sirmium "ville imperiale" ou "capitale"?' Corsi di Cultura 26 (1979) 56-58.

70 administrative centre such as Thessalonica, Antioch or Nicomedeia. These centres were apparently modelled on the Roman original and combined a residence with public buildings and located next to a circus for ceremonial public appearances: effectively they duplicated the Palatine original, with its associated buildings, within a provincial city. It is clear from this description of palace development that Split served a different function from Roman palaces after the early first century. Split was not the residence of an emperor, and served no administrative function. It was remote from the centres of imperial power, far removed from important lines of communication and strategic cities and lacked a circus.

The notion that Split was a palace is rooted partly in tendency of scholars to describe large, opulent buildings in grandiose terms, and partly in Glanville Downey's assertion that Split was modelled on the palace at Antioch. Downey examined the literary sources and concluded that not only was the palace at Antioch built on the foundations of a Roman camp, but that it also had the same street plan as Split and had a similar wall-top

71 gallery providing a panoramic view of the Orontes valley.145 The suggestion that the Antioch palace was built on the foundations of a castra was a conjecture. Malalas,

Downey's source, says that Diocletian built at Antioch "a great palace, finding the foundations already laid by Gallienus who was called Licinianus".146 Downey thought that Malalas not only mistook Gallienus for Valerian, as he occasioally had elsewhere, but also the nature of the building. He considered it more likely that, so soon after the Palmyrene occupation of Antioch, Valerian would be defending the city with a castra rather than embellishing it with a palace.147 The plan of the "New City" on the island is suggested by Libanius in an oration written for the Olympic games of 360;

"The form of this new city is round.

It lies in

the level part of the plain, the whole of it in an exact plan, and an unbroken wall surrounds it like a crown. From four arches which are joined to each other in the form of a rectangle, four pairs of stoas proceed as from an omphalos, stretched
145

G. Downey, A History of Antioch in Syria, Princeton 1961, 318-23. It was first suggested by Weilbach, F. Weilbach, "Zur Rekonstruktion des Diocletians-palastes," Strena Buliciana, Zagreb- Split, 1924, 125. 146 Malalas, 306.21-22. 147 G. Downey, op. cit. 259, n. 126 and 321, n. 16.

72 toward each corner of the heaven, as in a statue of the four handed Apollo. Three of these pairs, running as far as the wall, are joined to its circuit, while the fourth is shorter but is the more beautiful just in proportion as it is shorter, since it runs toward the palace which begins hard by and serves as an approach to it. This palace occupies so much of the island that it constitutes the fourth part of the whole. It reaches to the middle of the island, which we have called an omphalos, and extends to the outer branch of the river so that where the wall has columns instead of battlements, there is a view worthy of an emperor, with the river flowing below and the suburbs feasting the eyes on all sides. A person who wished to describe this part carefully would have to make it the subject of a discourse, but it cannot be part of a discourse on another subject. Nevertheless one should say at least that to the other palaces which exist in every part of the world... it is in no way inferior; but is superior to many, nowhere surpassed in beauty, and in size surpassing all others, divided into so many chambers and stoas and halls that even those

73 who are well accustomed to it become lost as they go from door to door."148

To this description may be added that of Theodoret (c.387c.458);

"The palace of the city of Antioch is washed on the north by the river Orontes: on the south there is a large portico with two storeys which touch the walls of the the city, and which have two high towers. Between the palace and the river is a public road leading from the city to the suburbs. One day as Aphraates was passing along the road on his way to the military gymnasium... he attracted the attention of the emperor, who was then on the top of the portico..."149

These suggest that the island was divided by broad emboloi characteristic of eastern cities, that the city covered a large portion of the island and that, where the palace abutted the city wall, there was some kind of gallery instead of battlements. Downey drew a neat parallel between the wall-top gallery described in the literature

148

Libanius, Orations, XI, 203-207. The oration is translated in full by Downey in G.Downey "Libanius' Oration in Praise of Antioch (Oration XI)", ProcPhilAs 103 (1959) 652-686. 149 Theodoretus, Eccl. Hist., 4,26,1-3.

74 and that in the south wall of Split and between the plan of the palace, which he supposed to have taken the form of Valerian's castra, and that of the plan of Diocletian's residence (fig. 4.9). As no trace of the palace was found in the course of excavation in and around the circus or elsewhere on the island at Antioch, his reconstruction remains theoretical150 and somewhat implausible. If, as seems probable from later historical descriptions,151 the palace complex at Antioch had a general resemblance to other Tetrarchic palaces, then an impression of its plan will be gained from a close examination of an example. Only one, at Thessalonika, is sufficiently well understood to merit description.

THESSALONIKA The most completely excavated palace of the period is that of Galerius at Thessalonika. It was built by Galerius after his success against the Persians in 298, and he used it as his principal residence except between 302 and 308/9 when he moved to Serdica for campaigns against the Carpi

150

W. A. Campbell, "Excavations at Antioch on the Orontes", AJA 38 (1934) 201-206; W. A. Campbell, "The Circus" in G. W. Elderkin, ed. Antioch on the Orontes I. Excavations of 1932, Princeton 1934, pp.3441; idem, "The Third Season of Excavations at Antioch on the Orontes", AJA 40 (1936) 1-9; C. R. Morey, "The Excavation of Antioch on the Orontes", ProcPhilSoc, 76 (1936) 637-651. 151 Evagrius, Eccl. Hist., 2.22. Malalas, 328,2-4 and 306,22-307,2.

75 and Sarmatians.152 Although after Licinius the city was never again a principal residence of the emperor, the imperial palace remained in use as an administrative centre, as the residence of the prefect and as an occasional stop for the emperor.153 The town is situated between the sea and a low hill to the north and was surrounded by a wall, much of which still stands.154 The Tetrarchic palace complex was located at the east end of the city adjacent to the walls which had apparently been rebuilt to accommodate the new buildings (fig. 4.10).155 The extant complex consists of the rotunda church of Saint George, an avenue flanked by colonnades leading south to the four-way Arch of Galerius over the via Egnatia, the circus and a series of structures lying to the south of the arch and west of the

152

T.D.Barnes, op. cit. 61-82. After Galerius, the city remained in use as an Imperial residence; Constantine remained in or near Thessalonika for two years while warring on Licinius in 323/4 and made brief visits in 327 and 336. On his abdication in September of 324 Licinius was sent to Thessalonika, where he was executed early the next year. 153 M. Vickers "Observations on the Octagon at Thessaloniki", JRS 63 (1973) 111-120 attributes the destruction of the palace to an earthquake in the first quarter of the fifth century. J. M. Spieser, Thessalonique et ses monuments du IVe au VIe siecle, Boccard: Paris 1984, 97-99 n. 117, 118 is not convinced by Vickers argument and places the destruction around 625 after years of use as the capital of the prefecture of Illyricum. 154 Little survives of the Hellenistic city, but the remains of the Roman city are extensive and much is known about the forum, the circus and the palace itself. The lines of the orthogonal street plan, oriented northwest to southeast, can be traced from what is known of the Roman plan and the medieval plan of the city as it appeared before the great fire of 1917. M. Vickers, "Hellenistic Thessaloniki", JHS, 92 (1972) 156-170; J. M. Spieser, op. cit. 82-90. 155 The excavations of different parts of the palace area have been reported by Moutsopoulos, 187-263.

76 circus. As it is now known, the Arch of Galerius forms the central focus of the area (fig. 4.11). The Arch of Galerius (fig. 4.12) was a tetrapylon marking the intersection of the two major axes of the palace, the via Egnatia and a north-south street which ran from the rotunda parallel with the circus. Only the western half of the arch survives. The complete structure was a monument with three openings over the east-west road and pierced by a single passage on the north-south axis. The central arch, measuring 12.28 metres above ground level, is larger than the lateral arches. This bay was covered with a dome or a domical vault. The side bays had barrel vaults aligned with the long axis of the monument. The piers were built of reused blocks and covered with marble panels with relief decoration depicting Galerius' campaign and triumph over the Persians in 297. The upper portion is of brick faced concrete.156 The arch is unlike a triumphal arch in that is a two axis passage, and not only is the decoration richer than is usual for a triumphal arch but it is arranged in friezes at a low level. Furthermore, the most important scenes are on the inner faces of the piers rather than the exterior faces.157 From the tetrapylon, colonnades flanking the via Egnatia led west towards the forum and east to the city

156 157

J. M. Spieser, op. cit. 99; Moutsopoulos, 212-235, figs. 23-29. J. M. Spieser, op. cit. 100.

77 wall. In plan the road appears to have been diverted from

its original straight course northwards to accomodate the tetrapylon. Excavation showed that the porticoes lining the street were contemporary with the tetrapylon, and therefore it seems that the realignment was part of the same programme as the palace construction.158 Leading north from the Arch of Galerius, an avenue, flanked by porticoes and probably contemporary with the tetrapylon, runs towards the rotunda. The rotunda, built on virgin ground, is located in an octagonal temenos with semicircular exedrae on the east and west sides. The rotunda in the early 4th century was a freestanding circular drum, about 38 metres in diameter, and entered from the south. It had a three storey elevation with a dome or domical vault roof and was built completely of brick. The interior is octagonal with deep rectangular recesses at the level of the ground storey, each measuring approximately 5 by 7 metres. Between the recesses are eight aediculae , while above them, at the level of the first storey, the walls are pierced by deep arched windows. At the top of the wall was a cat-walk running around the base of the dome. Access to the cat-walk was by spiral staircases built into the thickness of the wall on

158

J. M. Spieser, op. cit. 101.

78 either side of the door. The interior was originally decorated with veneer and marble architectural elements.159 The rotunda has been identified either as a mausoleum built for Galerius or as a temple.160 If it was a mausoleum, then Galerius was never buried here, for he died near Naissus and was buried at his birthplace Romuliana.161 The prevailing opinion is that the rotunda was a temple dedicated either to Zeus, the Kabeirioi or to all the gods.162 The structure seems to be too large for relatively minor, though locally important, deities such as the Kabeirioi. The circus lies just within the eastern wall of the city to the south of the via Egnatia.163 It seems to have been comparatively large with an interior length of at least 420 metres and width of approximately 72 metres. If the carceres were located near the via Egnatia to
159

J. M. Spieser, op. cit. 113-117; Moutsopoulos, 193-212, figs. 1-22 pl. I-III; RIA, 451-53, figs. 305-07. 160 Moutsopoulos, 194-204 and J. M. Spieser, op. cit., 117 catalogue the various opinions. Early travellers such as Leake and Texier identified the structure as a temple, probably dedicated to the Kaberioi who were widely worshipped in Macedonia. Dyggve and others described it as the mausoleum of Galerius as the plan and elevation are reminiscent of mausolea such as the contemporary Tor dei Schiavi in Rome, the Mausoleum of Romulus on the via Appia and, to a certain extent, Diocletian's Mausoleum at Split. Dyggve further considered that the rotunda after its conversion became the palatine church. 161 Aurelius Victor, Epitome XL.16. 162 Moutsopoulos, 194-204; Spieser, op. cit. 117 merely expresses doubts about the identification as a mausoleum. 163 The location was obvious in the 19th century when substructures of the west cavea were still standing, while a British Army map of Salonika, made during World War I, showed the Plateia Hippodromou which clearly followed the line of the spina, see M. Vickers, "The Hippodrome at Thessaloniki," JRS 62 (1972) 25-28. In 1971 the discovery of part of the curved end on Odos Mitropoleos to the south disproved previous notions of the circus' orientation. Moutsopoulos, 224, n. 110.

79 facilitate access to the public, then the circus may have had an interior length of about 400 metres.164 The oblique slant of the carceres may explain the change in direction of the via Egnatia at the Arch of Galerius. The location of the south end appears to have been determined by the line of the Porta Roma and a street running close to and parallel with the sea wall. The date of the circus has not been firmly established, but it probably belongs to the Tetrarchy and certainly before 350. The circus continued in use for some centuries; according to Procopius, the father-in-law of Belisarius was associated with the circus at Thessalonika in the 6th century, and it is mentioned by Eustathius in the 7th century.165 The core of the palace lies immediately to the west of the circus and covers an area extending from the via Egnatia to at least three quarters the length of the circus. No southern or western limits have been defined,

164 165

J.H.Humphrey Roman Circuses, London 1986, 627. J. H. Humphrey, op. cit. 631 is the only one to mention the relationship between the the street and the carceres, but he does not comment the implications for the date of the circus. If the carceres were responsible for the jog in the via Egnatia, then the street and circus must predate or be contemporary with the arch of Galerius. Most commentators assume the contemporaneity of palace and circus. He thinks that the fortification wall was extended to accommodate the circus, rather than being constructed later. The location of the imperial box and the finishing line have not been identified. The box at the Circus Maximus and the circus at Maxentius' Villa is at the turn of the meta secunda. If this were the case at Thessalonika, it would be close to the apsidal building on Gounaris street. The finishing line and official's box tended to be at a point two thirds of the way along the spina , and so probably just to the south of the basilica hall (see below).

80 but to judge from the preserved lines of the Roman and Hellenistic street plan preserved in the pre 1917 map of the city, the full extent was to about 200 metres west of the cavea and perhaps as far as the Porta Roma street.166 The main entrance of the palace is thought to have been through the southern passage of the tetrapylon on the via Egnatia. Immediately to the south of the tetrapylon a rectangular building, 43 by 18 metres and oriented eastwest, may have been the vestibule of the palace. This hall had two entrances on its central transverse axis. The room was richly decorated with mosaics.167

166

In other words a 10 hectare (24.7 acre) area measuring 500 by 200 metres. On the pre 1917 map the orthogonal plan could be traced from near the west wall to about 300 metres west of the east wall where the pattern degenerated into a maze of smaller streets overlying the circus, the area immediately around the rotunda and over the known area of the palace. The known Byzantine buildings are built respecting the street plan, and it would seem that the 6th century law forbidding the changing of existing street lines without permission was in effect in Thessalonika. The maze of streets overlying the palace would then seem to be a relatively late innovation after the palace complex and Hippodrome went out of use as late as the 7th century. If the maze corresponds with the extent of the 4th century palace, then the palace complex stretched as far as Odos Konstantinou Palaiologou midway between the circus cavea and the late 8th century church of Agia Sophia built by Irene and Constantine VI. Substructures of walls

found on the east side of this road may belong to the western side of the palace. The southern extent of the palace can only be guessed at, and was either at the east-west Porta Roma street, or if an overpass was built, then as far as the sea wall a short distance to the south. For a discussion of the grid preserved in the pre 1917 plan see M. Vickers, "Hellenistic Thessaloniki," especially fig. 3.
167

Excavated in the 1930's, this structure was never published. Velenis thinks that the room was a transition between a ramp up to the tetrapylon and the via Egnatia proper, but there is no evidence to support this conjecture, see G. Velenis, op. cit. fig. 16. Dyggve sees it as the palace vestibule. For an axiometric drawing see E. Dyggve, "La region palatiale de Thessalonique," Acta Congressus Madrigiani Hafniae 1954, I (Copenhagen 1958) 353-365, fig. 16. For a brief discussion see J. M. Spieser, op. cit. 101; Moutsopoulos, 216.

81 To the south of the vestibule and 100 metres from it, an odd rectangular structure with an apse on its north side was found on Gounaris street.168 The building measures approximately 24 by 13 metres. It was entered from the south through a tetrastyle prostyle porch opening into a small, rectangular vestibule paved with slabs around the perimeter and with a central diaper pattern of small square plaques. The inner chamber is oval, divided into two distinct halves by a step and rebates in the wall where the apse abuts the rectangular portion of the building. The south portion is half an octagon with deep semi-circular niches in the southwest and southeast angles. The apse is buttressed on outside by what may be

a blind arcade. The raised floor of the apse is paved with an opus sectile grid-iron mosaic and the walls were probably covered with marble revetments. The function of the structure is uncertain, but the suggestion that it was a temple of the Kabeirioi has been justified on the grounds that it has a temple-like appearance and that the Kabeirioi were associated with horses and racing. Moutsopoulos reconstructs a long colonnaded courtyard stretching 100 metres from the entrance of the "temple" south to the "basilica". This "basilica" is a large apsidal hall also flanks the circus cavea. It measures
168

For a discussion of the building see Moutsopoulos, 224-235, figs. 42-48, 50-53. Note that the scale of fig. 44 is in error by a factor of 2.

82 approximately 70 by 28 metres on a north-south axis and appears to have been entered from the north. Small doors at the northwest and southwest corners gave access from the west. The hall was divided into a nave with two side aisles separated by colonnades. At the ends of the aisles are apsidal recesses, while the nave, paved with slabs, terminates in a large apse floored with opus sectile. The west wall of the hall was originally flanked by a corridor, and was decorated with a blind arcade. At some unknown later date a four cell, two storey complex was inserted into the open corridor.169 The function of the hall is not known. In scale the structure exceeds that of even the basilica at Trier. From its form it must have been some kind of reception hall for official rather than private use. To the west of the apsed hall is a rectangular structure centered on a peristyle court (fig. 4.13). The perimeter of the complex is a covered corridor paved with mosaics. Defining the court on the west, south and east

sides are eleven small square rooms in units of one or two rooms connecting directly with the court itself. The court was open to the air and had a covered ambulatory running around all four sides. Access to the court was through small doors on the north and south sides, while the corridor could be entered from a large portal in the south
169

Moutsopoulos, pl. IV and fig. 49.

83 side on axis with the door to the court, from the basilica in the north east corner, and two small doors in the west side. South of the complex are a series of small rooms including a small nymphaium and three cubicles. These cubicles are in a block oriented east-west; the rectangular central room is flanked by octagonal rooms with semi-circular niches alternating with doors in the north, south, east and west sides.170 The building which includes the octagon lies at the southwest corner of the courtyard complex, but was entered only from the south side (fig. 4.14). The main roon has an octagonal plan, measuring about 37 metres in diameter (c. 30 metres from side to side), while a spacious entrance vestibule, 35 by 15 metres in plan, was extended beyond the ends of the facade by apses at either end. The whole is enclosed within a rectangular temenos which abuts the apses of the vestibule, but stands free of the octagon itself. The octagonal exterior of the main room reflects the interior which is embellished with semicircular niches in each side. The niches are of uniform size, measuring 5.2 metres, except for the niche opposite the entrance which has a diameter of 7.05 metres. The walls are of green schist coursed with brick, and were originally covered with coloured marble revetment of red and green

170

Moutsopoulos, 235-240 figs. 49, 55-61, 63. J. M. Spieser, op. cit. 111.

84 porphyry and white marble. Two phases of flooring can be detected; the first was a mosaic which was quickly replaced by an opus sectile pavement laid in panels. The later pavement resembled the floors of the `temple' on Gounaris street, but had in addition four small emblemalike panels arranged in a square just inside the door. The octagon had two storeys and was probably roofed with a dome or a domical vault. The second storey was probably a catwalk running around the circumference at the base of the dome, and was approached by spiral staircases set into the walls to either side of the door. A second phase is also evident from a modification to the door, which was narrowed from almost 5 metres to 2.25 metres, and by two small cruciform chambers roofed with cross vaults that were added to the exterior of the northeast and northwest facets. These could be entered from the apses inside by a door knocked through the wall and from the outside by doors with small bent-axis porches with a pair of columns in antis. Access to these chambers from the north was provided by a door punched through the wall of the corridor to the north.171 The function of the octagon is undetermined. The discovery of a Christian motif set in the brick wall in the north apse in connection with the discovery of the

171

M. Vickers, "Octagon," 111-120; J. M. Spieser, op. cit. 113-123; Moutsopoulos, 240-250.

85 east apse of the vestibule led to the assertion that it was a church with a baptistery in front. This suggestion has since been rejected. The idea that it may have been a mausoleum was put aside on the grounds that the semicircular niches were unsuitable for rectangular sarcophagi. A third possibility, that it was a throne room, has also been raised but not generally accepted.172 Clues to the function are provided by the shape and form of the building, which resembles, but is much larger than, Diocletian's mausoleum at Split, and the discovery of a rectangular Early Christian tomb in the large niche.173 Despite its later use as a mausoleum and its shape there is no conclusive evidence for this identification. Decorative details may also give some indication of use of the rotunda. Four pilaster capitals, probably originally flanking the niches of the octagon, and a marble arch were found in the immediate area. The capitals are compound

Corinthian and Ionic order and have small relief figures cut on the face between the volutes. Each of the figures is different, and represent a Dioskouros, a Kabeirios, Jupiter and an unidentified female figure. Both the Dioskouroi and the Kabeirioi were twins and sons of Zeus, and may in some way be

172

M. Vickers, "Octagon" 119. Actually the apses, despite their shape, are quite large enough for sarcophagi. 173 Moutsopoulos, fig. 75.

86 representative of the philadelphia of the Tetrarchs and their association with Jupiter. Jupiter is significant in that Galerius was a member of the gens Iovi. The arch was found a short distance to the south of the octagon. The extrados has two lunettes with portraits, one perhaps the Tyche of Thessalonika, the other perhaps of Galerius. The lunettes are held aloft by barbarians and are flaked on the inside by putti holding either end of a garland stretched over the intrados. It is possible that

the arch was set over the door to the octagon's vestibule, and that the representations shown have some relevance to the function of the structure as may the four opus sectile motif panels found inside the door of the octagon. The likeliest explanations for the function seem to be that it was either the unused mausoleum of Galerius or a temple of the imperial cult.174 The palace at Thessalonika was a huge complex which seems to have been built within a relatively short period, mainly in the reign of Galerius, though some rare later additions and changes can be noted. It consisted of numerous diverse elements built in a typical Late Roman fashion out of stone coursed at intervals with brick, with exterior facades decorated with blind arcades and

174

M. Vickers "Octagon", 119-120 lists the various arguments for the function but remains undecided on the issue. Similarly Spieser, op. cit. 118 has no definite opinion. Mousopoulos, 250 tentatively supports the mausoleum argument.

87 interiors faced with marble and mosaic or opus sectile floors. These elements covered an area stretching from the rotunda in the north to at least the turn of the circus in the south and include the basilica and the octagon, as well as other intricately planned buildings linked by corridors and colonnaded avenues. Although the plan

appears regular and orthogonal, there is no logical coherent intra-palatial network of passages. This lack of emphatic axes gives the plan a modular, or additive appearance of self-contained spaces that are joined but not interlinked. The large scale of the complex suggests that it was much more than a residence, but the lack of specific facts prohibit more than speculation on the various functions it served. It should suffice to say that the area was one

devoted to religious and administrative uses as well as the residence of the emperor Galerius. Parallels can be found for various elements of the plan but as no other palace complexes have been excavated as completely as Thessalonika, as a whole it stands alone. The basilica resembles a similar structure at Trier, while the octagon and rotunda are reminiscent of mausolea and temple forms common in Roman imperial architecture. The tetrapylon at the apex of the axes of the via Egnatia and the rotunda street is a monument type found at similar locations throughout the eastern empire, and as such seems

88 to have had a specific and important significance, far greater than its function as a triumphal arch. The location of the palace at the edge of the city, built on previously undeveloped land, reflects the unity of the complex; it was developed as a single entity for use in conjunction with the circus. The plan of the area is incomplete, but sufficient to draw broad conclusions. Two main axes divide the site into four parts along the lines of the via Egnatia and the street running south from the rotunda. The east-west axis serves as a throughfare connecting the centre of the city with the outside, while the north-south axis is purely a weak internal line of communication. One entrance of the palace was at the crossing of these two axes, and consisted of a rectangular hall set perpendicular to the north-south passage. The area to the south of the hall is unexplored, but may reasonably be reconstructed as a short continuation of the rotunda street. A second entrance may be inferred from the orientation of the octagon and the court. Within the southern part of the complex the various buildings are not grouped around one central focus, such as a courtyard or corridor, but form discrete spaces almost independant of one another. This disposition of elements is suggestive of independent functions for each component part.

89 SUMMARY In the palaces discussed above there are common elements that if enough were known, would be seen to be common to all. The Tetrarchic capitals were stretched along the edge of the empire from the Middle Rhine to the Middle Danube to the Bosphoros to the Syrian frontier. Each was chosen for its strategic location and communications: for its proximity to potential danger to the empire, thus allowing appropriate response, and for ease of access to other similarly situated cities and to a like minded co-ruler to help or to be helped for the common good of the empire. Each city was the leading city of the region in which it was located, and was the seat of political and military power. In each of the regional capitals, whether the

emperor was in residence or not, the administration of the immediate area was carried out as usual; the presence of the emperor temporarily widened the political `field' of the city, extending the area administered for as long as the emperor was in residence. The Tetrarchic form of government required that the provincial and diocesan administration be centralised within the provincial capital for the sake of efficiency. When the emperor arrived in a provincial capital he added the diocesan government to the provincial government. This consisted of not only his sizeable bodyguard, but also advisors, officials, scribes, slaves all needing board and

90 lodging for themselves and their baggage train. The result was an extremely large area reserved for housing the emperor's entourage, for stabling, for storage, for appropriate working space for judicial courts, for ceremonial areas, for libraries of documents and for leisure. A need for close economic control to ensure a ready availability of coin meant that the mint had to be close enough to be directly controlled by the central administration. Finally there was a need for a places that the emperor could appear to the public. The result was the provincial Roman palace; a large sprawling complex including all the necessities for imperial administration - a Versailles, a Nicomedeia, a city within a city. In the individual plans of the palaces described some of these elements are apparent but the whole is not well understood. Taken together, they give a fairly clear idea of a `typical' palace. In many of the cities described the emperor added a mint, a large public baths, an arms factory and various temples. These are frequently known to have been close enough to the residence to be considered part of the city within a city, or "New City" in the case of Antioch, which together constituted the palace as a whole. So at Antioch the palace consisted of a maze of chambers, stoas and halls for private and administrative use, an imperial baths, and the circus. Later the Octogon church was added to the palace by Constantine. In some

91 places space was not available, so emperor had to create room for the administrative and residential block either by destroying part of the existing city or by adding a new quarter. At Nicomedeia, Thessalonika, Trier and Sirmium the emperor added new areas by extending the city wall, but also seems to have demolished marginal city structures. At Milan there seems to have been a general requisitioning of down town space to add to the extant palace, while at Antioch the area was open and free for development. The area required for the palace was considerable, somewhere between 10 and 25 hectares (25-59 acres).175 The palaces probably grew in stages, with each resident adding to the complex as necessity or whim dictated. The result would be a curiously episodic apparently disjointed array of different spaces, halls, corridors, atria, porticoes, temples and private rooms as well as monumental buildings for recreation and administration. The plan of a palace is also inconsistent with that of Split, despite Downey's assertions to the contrary. What is known of Thessalonika and the traces from other Tetrarchic palaces features such as the gallery, the
175

Split Antioch Thessalonika Milan Sirmium Trier

Fig. Fig. Fig. Fig. Fig. Fig.

1.2 4.8 4.9 4.2 4.5 4.3

175 x c.250 c.200 c.400 c.200 c.400

216 metres x 500 metres x 500 metres x 550 2 metres x 730 metres x 600 metres

= = = = = =

3.76 ha. 12.5 ha. 10 ha. 11 ha. 14.6 ha. 24 ha.

92 basilical hall and perhaps the mausoleum are common to both. However, the gallery is also a feature of fortified villas,176 the basilical hall can be found in large villas177 and the mausoleum type standing beside the main arteries leading out of Rome. The plans of the palaces at Thessalonika and of Antioch, as inferred by Libanius, are disjointed, modular and episodic. The plan of Split, on the other hand, is regular with spaces opening onto the broad connectiveaxes of the cardo, decumanus, via sagularis and gallery. It is impossible to view Split as a palace except in terms of being an opulent residence. Its plan, size and function are all at variance with what is known and understood of Roman palaces of the time. If Split is not an imperial residence, perhaps it should be examined in the light of contemporary private villas, both fortified and undefended such as Mogorjelo, the Piazza Armerina and Gamzigrad.

176 177

See Mogorjelo, chapter 5 infra. See the Piazza Armerina, chapter 5 infra.

93 CHAPTER 5. VILLAS.

Fourth century references to Split explicitly refer to the complex as a villa,178 but among modern commentators, only Duval rejected the palatial and ceremonial interpretations, and looked at the Split in terms of villa architecture. His concluded that Split can be best explained as a rich fortified villa and described most precisely by the French word Chateau.179 This appellation invites comparison between Split and fortified villas such as Mogorjelo in Yugoslavia on the one hand, and richly appointed villas such as the Piazza Armerina on the other. Excavation at Gamzigrad in Yugoslavia has revealed a fortified enceinte now known to have been the birthplace and final resting place of Galerius. In this regard it has much in common with Split, in that Diocletian was born near and buried at Split. Another contemporary imperial villa, considered by some to be important to the question, is the Villa of Maxentius on the Via Appia. In this chapter each of these will be discussed and the plan and function of the sites compared with Split.

178 179

Eutropius N. Duval, Yougoslavie: France 1971,

Brevicum, IX.28; Jerome Chronicle, p. "Le `Palais'" p. 90; idem, "Palais et recherches nouvelles" Bull. Soc. Nat. pp.99-104; idem, "La place de Split",

230. forteresses en Antiquaires de p.74.

94 The villa of Mogorjelo is located in southern Yugoslavia near Narona, overlooking the Neretva valley. The villa (fig. 5.1) is fortified and consists of a long block of residential buildings enclosed within an enceinte measuring 92 by 75 metres in plan. The walls are pierced by gates in the north, east and west sides, each of which is flanked by projecting square towers. The angle towers are also square, except for the circular northeast tower. Arranged around the north, west and north part of the east walls are 4-metre square magazines opening onto a portico which supported an upper storey. In the south half of the site, the residence occupies the area south of a line between the east and west gates. The residence is separated from the south wall and the southern parts of the east and west walls by a corridor. It consists of a series of rooms arranged in a U-shaped plan and open onto a portico on the north side. A stair in the north side indicates the presence of an upper storey. Although there is no archaeological evidence for this, the excavators suggested that the floor extended over the corridor as well, and therefore up to the fortification wall itself. The villa was originally built in the 1st century and was rebuilt as a fortified villa at the end of the 3rd century.180

180

E. Dyggve and H. Vetters, Mogorjelo: Ein spatantike Herrensitz in Dalmatien, Vienna, 1966.

95 In his reconstruction of the villa, Dyggve proposed that a gallery ran the length of the south wall at the second storey level. He used the south facade of Split as a model, and in comparing the resulting reconstructed plan and elevation to that of Split, described Mogorjelo as a palace. Duval objected to this reconstruction citing the lack of evidence that the second storey extended over the gallery. He also objected to the circular argument resulting in the description of the villa as a palace.181 Despite possible objections to the Mogorjelo reconstruction, fortified villas of the general type are known from North African mosaics. At Tarbaka, west of Carthage, a villa of this type is represented in the central apse of the main room of a fourth century house. The villa is shown in a rustic setting surrounded by ducks, pheasants, flowers and trees. The building has an

arched gate, towers and large square lower windows with an upper arcuated gallery.182 The villa of Dominus Julius shown in a late fourth century mosaic at Carthage has a second storey arcuated gallery running the full length of the facade between two square corner towers. Below is an arched central gateway and doors into the towers. Above

181

N. Duval, "Palais et forteresses", p.108 descibes Dyggve's plan as a result of "pure graphical imagination". 182 K. M. D. Dunbabin, The Mosiacs of Roman North Africa Oxford, 1978, p.120.

96 the pitched roof of the gallery are shown the upper parts of internal buildings.183 The North African mosaics demonstrate that this type of villa was widespread, and to a certain extent, their gallery facade justifies Dyggve's reconstruction of the south wall at Mogorjelo. The similarities between Split and the villas in the mosaics are obvious and general. They are rustic residences, more or less contemporary with Split, which are fortified by a wall and towers and have a gallery in the upper level of one facade. Mogorjelo can be compared in greater depth as its plan is known. Although lacking Split's street plan, Mogorjelo has deeply recessed gates between towers in the middle of the east, west and north walls which give an identical pattern of access and, to a certain extent, similar pattern of internal communications to that of Split. In both Mogorjelo and Split the fortification wall acts as the back wall of cubicles designed for storage, and perhaps for accommodation. These at Mogorjelo also were fronted by an arcade supported on piers. In both, the enclosure is divided into public and private space; at Mogorjelo the public space is a more or less open courtyard in the north part of the site, while at Split the public area space is considerably more complex. The private space in both is located in the south part of the
183

Ibid. pp.120ff. no.32 and pl.XLIII.

97 enclosure. At Mogorjelo as at Split the individual rooms open directly onto a long corridor, but here it is an open colonnade rather than a gallery. Whether or not Mogorjelo should be reconstructed with a wall gallery, its similarities to Split are significant. Not least of these is that public and private spaces are two indivisible parts of the whole complex. A major difference between them is the scale; Mogorjelo covers less than one fifth the area of Split (0.69 ha.) and is scarcely palatial by any definition of the word.

The discovery of a luxurious fortified villa with rich mosaics at Gamzigrad, about 40 miles north of Naissus prompted a variety of reconstructions of the plan and explanations of its function and ownership. The site has been described variously as a fortified mining center, a fortified villa, a castra with the residence of the commanding dux and an imperial palace.184 Soon after its discovery, the plan was reconstructed along the lines of Diocletian's palace at Split,185 but further excavations

184

D. Mano-Zisi, "Le castrum de Gamzigrad et ses mosaiques", Archaeologia Iugoslavica, 2 (1956) 67-84; A. Mocsy, "Aurelianus-AquaeGamzigrad", Studia Balcanica 1 (1970), 49-54; M. Canak-Medic, "Le Palais de la Basse Antiquite pres de Gamzigrad," Actes du XIVe congres international des etudes Byzantines III (1971) Bucarest 1976 pp.357362; N. Duval, "Palais et forteresses" and D. Srejovic, "An Imperial Roman Palace in Serbia", ILN 263, (Oct. 1975), 97-99. 185 N. Duval, "Palais et forteresses" p. 119 and fig. 8 based on a plan by Canak-Medic. This showed the walls as rhomboidal with gates in the north, west and east sides, and an arcaded cryptoporticus running around the interior face of the wall. The interior was shown divided

98 have shown this to be totally inaccurate. In 1984 the discovery of an inscription finally identified the site as Romuliana, and, because of the associations with the emperor Galerius, comparisons with Split were renewed.186 Very little is known from the ancient sources about Romuliana. Lactantius records that Galerius, like Diocletian, intended to retire after his vicennalia in March of 312,187 but became fatally ill in Dardania, south of Naissus, and died in April 311. His body was carried to Romuliana where he was buried.188 Although Romuliana is supposed to have been named after Romula, the mother of Galerius, and was his birth place,189 there is no historical evidence to show that he intended to retire there or be buried there.190

into three parts by streets with colonnades and by a large open courtyard. 186 D. Srejovic, "Two Memorial Monuments," pp.41-49; and idem , "Felix Romuliana, Galerijeba palata u Gamzigradu", Starinar 36, (1985), 5165. 187 D. Srejovic, "Two Memorial Monuments," pp.41-49; and idem , "Felix Romuliana, Galerijeba palata u Gamzigradu", Starinar 36, (1985), 5165. 188 Ibid. XXXV, 3-4; T. D. Barnes, The New Empire of Diocletian and Constantine , Cambridge Mass. 1982, p. 64. 189 Aurelius Victor, Epitome de Caesaribus, XL, 16; "(Galerius) ortus Dacia Ripensi ibique sepultus est; quem locum Romulianum ex vocabulo Romulae matris appellarat". 190 D. Srejovic, "Two Memorial Monuments" p.49 read Lactantius to say that Galerius intended to retire to a place where "he would be surrounded by an impregnable wall behind which he could enjoy a carefree and calm old age". He takes this to mean that Galerius built a fortified residence and identifies this residence as Gamzigrad. J. J. Wilkes op. cit. follows Srejovic and states that Gamzigrad is therefore comparable to Diocletian's villa at Split. In fact, Lactantius, De Mort. Pers. XX.4, says that after replacing himself by making his 9 year old son Caesar, Galerius was going to retire "thus, with Licinius and Severus in supreme control of the empire, and with Maximin and Candidianus in the second rank as Caesars, he would be surrounded by an impregnable wall...". In other words his successors were the impregnable wall.

99 The fortifications of Gamzigrad (fig. 5.2) consist of two concentric walls of different dates, built on ground sloping gently down in three terraces from west to east. Excavation of the earlier wall revealed a gate flanked by polygonal towers, a square interval tower to the north and a cryptoporticus along the inside of the wall which presumably supported ramparts above. Surface survey shows that this wall enclosed a roughly trapezoidal area of about 4.2 ha. measuring about 225 by 195 metres. It had angle towers at each corner, two interval towers on each side and a second gate at the re-entrant angle in the middle of the east wall. Between the two gates a colonnaded street ran parallel to the south wall.191 The later walls are still substantially preserved. They were erected outside and parallel to the early circuit at a distance of about 11 metres from them. The plan is identical to the early wall except for the use of massive dodecahedral towers throughout. The walls enclosed an area of 250 by 225 metres (5.6 ha.). The west gate is largely intact and consists of a single entrance in the centre of a curved recess in the wall. The gate is covered by a plain marble arch with a five fascia moulding. Debris

191

D. Srejovic, A. Lalovic and D. Jankovic, "Gamzigrad," Starinar 31 (1980) 65-80, fig.1. This figure is prepared from a plan by M. CanakMedic, an architect and the principal investigator at Gamzigrad. D. Srejovic, "Two Late Roman Temples at Gamzigrad", Archaeologia Iugoslavica, 19 (1978), 54-63 fig. 3, prefers a reconstruction of his own, at odds with the architect's drawing,. There is nothing in the state plans of the site to support his reconstruction.

100 found outside the gate, including Ionic and Corinthian capitals, spirally fluted columns, pilasters, architrave blocks and cornices of green sandstone and limestone suggest that there were two highly decorated galleries set above the gate.192 This gate necessitated cutting a passage through the south gate tower of the earlier circuit. Because of the dimensions of the new circuit, it is likely that the east gate required the destruction of the old east gate tower in a similar fashion. The lower 2.25 metres of the walls consist of limestone ashlars, while the upper portions, up to a total reconstructed height of as much as 20 metres, were of single courses of small blocks of green sandstone alternating with three courses of brick. The internal structures are dispersed over a wide area in three distinct clusters. In the western half are

found a large villa, a granary and a multi-roomed structure within a separate temenos. All are aligned perpendicular to the south wall and parallel to the central axis of the site. The eastern part contains two small temples and a second, perhaps later, villa. The western villa occupies the northwestern quarter of the interior, covering an area approximately 55 by 50 metres, and consists of a U-shaped series of long halls, two

192

M. Canak-Medic, op. cit., figs. 11-14; D. Srejovic, "An Imperial Roman Palace in Serbia", ILN, 263 (1975) 97-99.

101 colonnaded atria, a triclinium, and a small bath complex. The halls make up the southern half of the villa: two long rooms oriented east-west are linked at their west end by a transverse hall. The principal entrance, flanked by columns, was at the east end of the south hall (41.5 by 7.5 metres), and a visitor would progress westwards to a transverse hall, 23 by 12 metres in plan. This room in turn opened onto a room 35 by 11 metres with a raised platform at its east end, a small elevated apse in the east wall and a small octagonal chamber, with a hypocaust heating system, opening off the southeast corner.193 All the halls were richly decorated. The exterior walls were articulated by blind arcades supported on marble pilasters and were covered with serpentine, porphyry veneer. Above a marble dado the interior walls had frescoes divided from each other by a stucco moulding. The floors were covered with mosaics.194 North of the hall complex is a courtyard, 19 by 12 metres in plan, with an ornamental fountain in the middle and geometric mosaics in the surrounding peristyles. On the north side, a triclinium 18 metres long by 10 metres wide opens onto the court; it has a broad 7-metre diameter apse in its north wall. The floor of the triclinium has a

193

J. J. Wilkes, op. cit. p.69. For parallels to the octagonal chamber see Desenzano, North Italy in RIA pp. 464-5, fig. 316. 194 D. Srejovic, "An Imperial Roman Palace" pp. 97-99; J. J. Wilkes, op. cit. p. 69.

102 colourful mosaic of a youthful Dionysos enthroned, holding a kantharos and a garlanded staff, and accompanied by a leopard. East of the triclinium is a small court with geometric mosaics in the peristyle. The peristyle leads to a small, three-roomed bathing complex. This consists of two circular spaces, with circular exedrae, linked by a circular entrance hall.195 To the south of the villa, on the same orientation and probably closely associated with it, are two large structures. The larger is opposite the villa on the south side of the street. This is a long, broad hall measuring 51.2 metres by 19.4 metres and oriented north-south. The walls of the hall were buttressed on the outside in a fashion similar to those of the villa. Inside were four regularly spaced rows of six floor supports giving the overall impression that the building served as a granary196 The other structure is west and slightly to the south of the granary. It is a multi-roomed building surrounded by a temenos 32 metres square. The central structure consists of three small rectangular rooms on a north-south axis with smaller rooms flanking each of the two northern rooms. The floors are covered with geometrically patterned mosaics similar to those in the west villa. No

195 196

J. J. Wilkes. op. cit. p.69. This is similar to the horrea at the Constantinian imperial villa at Fenekpusza, see J. Lander, op. cit. p. 235, fig. 251. and A. Mocsy, Pannonia and Upper Moesia, London, 1974 p. 302ff. and fig. 50.

103 identification has been suggested for this structure, but over the door of the southernmost of the central rooms was a pediment, now fallen, with the FELIX ROMULIANA inscription.197 To the east of the horrea is a large, 33 by 24 metre, tetrastyle prostyle temple set on a 4-metre high podium with steps up on the east side. An altar is located about 7 metres to the east of the temple. Below cella, in the podium, is a two roomed crypt. Immediately to the south of the temple is a long rectangular structure oriented eastwest. The north side flanked by a colonnade and access is provided from the colonnade and though a narrow entrance hall at the east end. In the area of the temple fragments of sculpture, including a colossal cuirass figure, two statues of Hercules and a statue carrying a torch were found.198 The foundations of a second temple and an associated altar were found within a compound limited on both sides by the two villas. This structure was smaller than the other temple, measuring only 16.5 by 10.5 metres, but was also a podium temple, tetrastyle prostyle and with a crypt
197

The inscription is contained within a wreath supported by two crudely drawn peacocks and decorated with modest sprigs of ivy leaves. For the most recent discussion of the complex see D. Srejovic "Felix Romuliana". 198 D. Srejovic, "Two Memorial Monuments", pp.22-3; D. Srejovic, A. Lalovic and D. Jankovic, "Two Late Roman Temples at Gamzigrad" Archaeologia Iugoslavica 19 (1978) 54-63; D. Srejovic, A. Lalovic and D. Jankovic, "Gamzigrad" Starinar 31 (1980); D. Srejovic, "An Imperial Roman Palace" pp.97-99. Srejovic suggests that the temple served both Olympian and cthonic deities, to Hercules and to Cybele, whom he linked with Galerius and his mother Romula respectively.

104 located in the podium. The crypt was approached by a pair of stairways along the west wall of the cella.199 The temple and altar associated with it are not aligned with any of the principal axes of the fortifications or internal buildings. The Villa in the north-east quarter is built on an irregular plan on a different alignment to the other structures and consists of a peristyle courtyard with rooms arranged around it. On the west side is a triclinium, 12.5 by 11 metres in plan with an apse in the west wall and massive side walls that might have supported a second story. To the south of the second villa is a long, narrow structure, 51.2 by 11.5 metres in extent, with seven rooms and a long corridor along the south side giving access to the various rooms. Its

function is not known. Finally, buried in the south-east quarter of the compound is a large, unexcavated structure.200 The phases of the construction of Gamzigrad are confused. The excavator, Canak-Medic, dates the older wall, the villa in the north-east quadrant , the horrea and the structure to the west of it to the reign of Galerius. She considers that the small temple and the later, outer circuit may belong either to the reign of
199

This, Srejovic, ops. cit. associates with a taurobolia ceremony, again for Cybele. 200 J. J. Wilkes, op. cit. pp.69-70.

105 Constantine or Theodosius.201 Srejovic disagrees on the grounds that the larger temple appears to be a more developed form of the smaller temple, both of which he considers to have the same function. He dates the

small temple and the early wall to the period when Galerius was Caesar (293-306) and the larger temple, north west villa, and outer wall to his Augustate (305-311); other structures are attributed to Licinius after the death of Galerius, and before Licinius' loss of the region to Constantine in 314.202 Phasing on the basis of alignment is difficult, while the stratigraphy is confused and not well published. In balance, Canak-Medic's chronology is preferable to that of Srejovic. Contained in the plan of Gamzigrad are numerous elements comparable with Split. Their size is comparable, the early walls with polygonal towers flanking the gate are similar to those at Split, as are the galleries over the later gate, the colonnaded street the curious building west of the horrea and the temples. As at Split the buildings are extravagantly decorated with coloured marbles, decorative facades and mosaics. Furthermore there is a strong association between the site and the

201 202

M. Canak-Medic, op. cit. pp.361-2. In suport of his chronology uses the alignments of the various elements: D. Srejovic, "Two Late Roman Temples" Drawing 3 and D. Srejovic, A. Lalovic and D. Jankovic, "Gamzigrad" pp.77-80. This argument is considerably weakened by his misrepresentation of the alignments. Nevertheless J. J. Wilkes, op. cit. p.70 follows Srejovic's dating.

106 birthplace and final resting place of the emperor Galerius although it is not evident from the plan or the literary sources that he built the complex as a retirement villa

or for his burial. The lay out of Gamzigrad differs radically from that at Split. The internal buildings are scattered in a disjointed fashion, each standing discretely apart from each other whereas at Split the various parts are integrated in a predetermined plan with complexes designed for different functions connected to one another by streets, corridors and linking rooms. The private quarters at Split are large and extend along the south wall of the site giving a view over the sea. At Gamzigrad the older of the two villas is modest in scale and design in comparison to even modest contemporary villas, and has no view or elevated gallery. A most intriguing parallel exists between the multi-roomed structure in which the inscription was found and a similar unit in the eastern block of the private apartments at Split. Despite the obvious similarities in patronage, eventual use and in certain elements of the plan, it is difficult to consider Gamzigrad as a close relation of the complex at Split.

Although the Villa of Maxentius on the via Appia outside Rome is not a palace in the sense of those described in Chapter 4, there is reason to believe that

107 Maxentius, who was illegitimately created emperor by the people of Rome, wished to build a complex that recalled the residences of legitimate rulers. In view of its date, patron and possible function, it seems reasonable to seek parallels to Split in its plan.203 The villa (fig. 5.3) lies between the second and third milestones from the city of Rome on the Via Appia. There are three main units making up the villa, namely the rotunda, the residence and the circus. These were built over an earlier complex developed during the Republic and in the first two centuries after Christ. The excavators suggest that the early villa belonged to Herodes Atticus, and that the circus was founded on the hippodrome garden attached to the villa.204 The rotunda has a plan consisting of a circular cella with an an external diameter of 34 metres, with a rectangular porch, 18 metres deep, attached to the southeast side. Six internal buttresses are attached to the inside wall.205 The whole was set centrally within a rectangular temenos with an exterior measurement of 120 by 112 metres. An arcaded colonnade formed a portico around
203

This is convincingly argued by A. Frazer, "The Iconography of the Emperor Maxentius' Villa on the Via Appia", Art Bulletin 48 (1966), 385-392. 204 For the plans of the early villa, see G. Pisani Sartorio and R. Calza, La villa di Massenzio sulla Via Appia. Il Palazzo. Le Opere d'arte (I Monumenti Romani 6) Instituto di Studi Romani: Rome 1976, pp.113-121, pl. 53-55. 205 These presumably supported a wooden peaked or domed roof, as the walls, only 2 metres thick, could not have supported a dome of 34 metres diameter as Ward Perkins, op. cit. p.424, suggests.

108 the rotunda. Buttresses on the inside walls of the temenos supported roof beams of a pitched roof. This portico appears to have been entered only by small doors set in the northwest corner and east side of the temenos.206 The rotunda is usually assumed to be a mausoleum. Inscriptions in the area of the carceres and triumphal arch of the circus which commemorate Maxentius and his son Romulus have led to the assumption that funeral games were held in the circus for Romulus, who was buried in the mausoleum.207 Apart from parallels to the shape of other contemporary circular structures of equally uncertain identification there is little to suggest a mausoleum.208 The circus, clearly functional, started close by the temenos of the rotunda. It is well preserved with parts standing up to 16 metres in height. The carceres still stand at the south and are flanked at each end by entrance towers three storeys high. From the carceres the arena extends for a total internal length of 503 metres and was up to 79 metres wide. Three-fifths of the distance down the east cavea are the substantial remains of the judges' box at the finishing post. Also preserved are the remains
206 207

Pisani Sartorio and Calza, op. cit. pl.58. J. H. Humphrey, op. cit. p.601. The texts are given by Frazer, op. cit. p.385, n.4. 208 There are no niches, no inscriptions, sarcophagi or bones to support the suggestion. In fact only parallels with the Tor dei Schiavi at Rome (with vault and niches and pronaos), Galerius' socalled mausoleum at Thessalonika which is also set within a temenos, but is now thought to be a temple, and Diocletian's mausoleum at Split (vaulted, with niches and separate porch) support the identification. For the mausoleum argument see Frazer, op. cit. p. 387.

109 of the imperial box opposite the meta secunda on the west side. This was approached by a covered corridor stretching from the villa along an elevated terrace. The corridor opened onto the box through a circular room, while a square stair-tower gave access to the box from ground level.209 The residential complex is composed of a series of interconnecting rooms joined by corridors and vestibules. According to the excavators the main entrance was from the north through a circular chamber with semi-circular niches let into the wall.210 The focal point of the complex is a large apsidal audience chamber oriented facing the southwest. The hall, measuring 32 by 20 metres, has a broad 12 metre diameter apse in the north wall flanked by semi-circular niches. Access to the hall is from the south via a transverse, rectangular vestibule measuring 30 by 10 metres. This in turn connects with the ambulatory leading to the imperial box.211 West of the apsidal hall is a long narrow hall or broad corridor leading to a large rectangular room to the north. This has an apse at the south end and measures 53 by 11 metres. The room off which

209 210

J.H.Humphrey, op. cit. pp. 582-602. Pisani Sartorio and Calza, op. cit. p.125-126. Actually too little of the structure has been uncovered to tell whether it is the apsidal end of a rectangular hall with a door flanked by niches rather than a vestibule comparable to that at Split as is argued by the authors. 211 Pisani Sartorio and Calza, op. cit. p.124.

110 it opens has a tranverse vestibule with an apse at one end.212 In many respects Maxentius' villa resembles provincial palaces. It possesses three features apparently typical of them, namely a circus, a rotunda and a residential complex made up of large rooms with apses, ambulatories, courts, and vestibules. It differs in a number of respects.Its location is some distance outside the city rather than within, and it lacks features associated with the administrative function of provincial palaces. Despite this, Frazer suggests that the complex was built in imitation of the regional palaces, perhaps as an attempt to legitimise his claim to the purple with an appropriate monument.213 Such a gesture would seem unnecessary since he possessed the original palace on the Palatine with the Circus Maximus, the Curia and the ancestral centre of Roman worship nearby. It is possible that the complex was an elaborate rustic villa using an architectural iconography that Maxentius thought appropriate to his standing. Comparison with Split is difficult because of the incomplete plan of the private apartments. Both have large basilical halls which with other rooms open onto a long corridor. Although the possibility is not mentioned by the

212 213

Ibid.. pp.122-124. A.Frazer, op. cit. p. 392.

111 excavators, this corridor may well have had a colonnade running the length of its outward facing side, in which case comparison with the gallery of Split would be almost irrefutable. Maxentius' villa is planned on three unrelated major axes and is closely associated with a huge circus. As at Split, public and private space are strictly separated, with the mausoleum set within a separate temenos, and circus set to one side. The complex, unfettered by defensive walls is permitted to sprawl over a large area, perhaps in order to take advantage of different views or to create a series complicated architectural spaces.

A more profitable source of comparanda may be provided by large provincial villas of the late empire, such as the Piazza Armerina.214 The Piazza Armerina, located in southern Sicily, was thought to have been built by Maximian as a eetirement villa in the same way that Diocletian built Split for this purpose. Even though the date of the complex, circa 310 to 325, precludes this suggestion, the complex is of interest to the understanding of Split. The villa (fig. 5.4) is built on a gentle slope and covers an area approximately 130 by 100 metres (1.3 ha.).

214

R. J. A. Wilson, Piazza Armerina, Austin, 1983; W. L. MacDonald, The Architecture of the Roman Empire, II, New Haven, 1986 pp. 274-280.

112 It is planned in three distinct clusters of rooms which are aligned on different axes centred on the mid-point of the atrium. The entrance, through a monumental archway at the southwest corner of the site, gives access into a trapezoidal entrance court surrounded on three sides by a colonnade. Off axis on the east side of the court, a rectangular vestibule connects with the large rectangular peristyle atrium with a central pool and a door on the north side connects with the bath complex. The atrium is the focal point of the villa. Around it are arranged blocks of rooms including a small shrine on the south side, a bath complex to the northwest and a corridor linking three suites of rooms to the east.The shrine has a fountain set in the middle of the floor decorated with a mosaic showing Orpheus charming the animals. On the south side an apse contained a statue of Apollo.215 Opposite, on the north side of the atrium, a series of one, two and three roomed units open onto the peristyle. The east block is unified by a 58 metre long corridor with apses terminating each end. The floor of the corridor is decorated with a well preserved hunt scene, and the apses with unidentified personifications.216 Of the three suites opening onto the corridor, the southern two are most interesting. The central suite consists of a

215 216

Ibid. p. 32. Ibid. p. 24.

113 single room, measuring 30 by 14 metres, with a broad apse occupying the greater portion of the east end. The hall was decorated with opus sectile on the walls, floor and apse. This space probably served as a reception room or audience hall.217 The southern suite consists of eight rooms aligned on three parallel axes. From the corridor two doors open onto a semi-circular fountain court which in turn gives access to units of three rooms to north and south and to a larger rectangular room with an apse to the east. This block of rooms is thought to be for day to day

family use, and was decorated throughout with mosaics of childish pursuits, including a hunt and a chariot race, and mythological scenes.218 The bath complex is oriented on an axis radial to the centre of the atrium.It is entered from the southwest corner of the peristyle and consists of three distinct units including an entrance hall, used perhaps for changing, an octagon with cold plunges and another, heated, hall with warm rooms and pools. The entrance hall has two apses, one at either end, and a scene from the Circus Maximus in mosaic covering the floor. In the southern apse a door communicates through another room with the entrance court. The octagonal frigidarium must have been domed. In six of the Sides aediculae open onto

217 218

Ibid. p. 25; RIA p. 461. Ibid. pp. 27-8.

114 semicircular spaces. On the north side is an elongated pool fed by an aqueduct and on the south side a trefoil cold plunge. The tepidarium is similar in shape to the entrance hall and opens onto three caldaria on its west side.219 On the south side of the site is the oval court complex. This consists of a series of structures arranged around a truncated oval peristyle. At the west end is a nymphaeum and at the other a triconch structure thought to be a triclinium.220 This building has a maximum extent of about 25 by 20 metres. It is entered from the peristyle to the west up a short flight of steps and through a triple entrance way on the west side. The main room is about 12 metres square. The east, north and south walls press outward into semi-circular exedrae, and are entered through broad doors divided into three parts by columns. The floors are covered with mosaics showing some of the exploits of Hercules. The Piazza Armerina is an extremely rich villa with a complicated multi-axial plan. It is not, however, extraordinary and compares with both Split and other contemporary villas in size and in its various elements. The entrance vestibule is square at the Piazza Armerina, but at Split, and at the early fourth century villa of

219 220

Ibid. pp. 20-23. Ibid. p. 29.

115 Valentine in Gaul it is circular in plan with semicircular exedrae.221 The basilical hall is to be found both at Split but is otherwise widely spread appearing at Gamzigrad, the Villa of Maxentius222 and Loffelbach223 and is a common element found generally in Late Roman architecture.224 The transverse corridor is also widespread in villa plans, and is found not only at Split, the Villa of Maxentius and Gamzigrad but also in villas not sponsored by imperial patrons such as Woodchester and Valentine.225 One unit at the Piazza Armerina that may help interpret a suite of rooms at Split is the triconch triclinium. In the same form it appears at Desenzano, at Rioseco de Soria, and Patti Marina,226 but may possibly be related to the block of rooms with rectilinear rooms in the east part of the private appartments at Split. Another possibility is that the southern suite opening off the corridor at Piazza Armerina, and and southwest part of

221

At Valentine the entrance to the villa opens onto a long peristyle court with a transverse corridor at one end. From the corridor access to the main block, which is arranged around an interior courtyard, is through a D-spaped space with a central pool and through the circular vestibule, Ibid. pp. 80-1, fig. 51b. 222 cf. supra. 223 Loffelbach in Austia was laid out in the early second century, but modified in the late third century. Here an entrance hall leads to a peristyle court with an apsed audience hall (18 by 9 metres) on its north side. To the east of the triclinium is a second, smaller court with a peristyle on three sides leading to a small complex of baths. The plan, orientation, date and scale of the villa at Loffelbach are all very similar to that at Gamzigrad. R. J. A. Wilson, op. cit. p.107 n.36, fig.50c. A fuller account can be found in W.Modrijan, Der romische Landsitz von Loffelbach, Graz. 1971. 224 For instance in the palace at Thessalonica (fig. 4.11) and at Trier, RIA, fig.299. 225 R. J. A. Wilson, op. cit. fig. 51c, d. 226 Ibid. figs. 48, 50; RIA fig. 316.

116 Gamzigrad is a better parallel, and that this element is a day room rather than a triclinium.

Both the Piazza Armerina and the Villa of Maxentius are unfortified, and so are not constrained by rigid boundaries. As a result they sprawl across their respective sites with groups of structures aligned on different axes. Yet in these, and in contemporary villas there are numerous elements in common with Split. Mogorjelo and Gamzigrad are villas of a different type, with axes strictly defined by walls and streets. At Gamgigrad the enclosed area is so large that the internal structures lack cohesion giving the impression that the site developed slowly over a number of years. At Mogorjelo the plan is regular and cohesive and appears to have been the product of a single building campaign. The similarities between fortified and unfortified villas of the late third and early fourth century to Split argues the case that Split is a rich villa of the same general type as the Piazza Armerina that has been regularised and constrained by its fortification walls. Split has the transverse galleries, the day rooms, the triclinia, the vestibules and colonnades of unfortified villas and the gallery, walls and strict separation of public and private space of Mogorjelo.

117 CHAPTER 6. CONCLUSION. Conclusions about the plan of Split have to take Diocletian's political and private circumstances into account. Split was Diocletian's retirement villa, so much is given by fourth century authors. After resigning his office he retreated to a place of outstanding natural beauty near to the place of his birth, but remote from the centres of imperial government. He rarely made any effort to intervene in politics thereafter except to discuss the problem of imperial succession with Maximian and Galerius in 308, at which time Diocletian declined Galerius' suggestion that he return to office, and to request of Maximin Daia a guarantee for his daughter's safety after the death of Galerius.227 It seems that Diocletian had no active interest in politics after his resignation and willingly returned to the station of a private citizen. He therefore had no need or desire for the pomp and paraphernalia of imperial office, let alone the buildings associated with it. If Diocletian really returned to private life, it is unacceptable to attempt an interpretation of Split in terms of imperial iconography without more solid evidence than that hitherto presented. Descriptions of Split as a palatium sacrum or castrum palatium, the `peristyle' as a basilica discoperta and the vestibule as a throne room should be avoided unless
227

Aurelius Victor,

Epitome de Caesaribus 39,5-6.

118 substantiated by convincing and irrefutable arguments as they suggest that Diocletian maintained his rank and position after abdication.228 The villa seems to have been started well before his retirement. Diocletian's personal interest in urban design is well documented by his intervention in the construction of the palace and associated structures at Nicomedeia.229 The villa, therefore, was presumably laid out on a plan designed in close consultation with Diocletian and contained elements that he specifically desired. The villa was surrounded by a defensive wall and had a simple orthogonal street plan dividing the interior into four parts. Different parts of the villa were separated into discrete areas linked by transitional elements. The
228

Such suggestions fall into Boethius' category of `Rash Conclusions' in vogue in the 1940's and 50's. A. Boethius, "The Reception Halls of the Roman Emperors," Annual of the British School at Athens 46 (1951) 25-31. He particularly targets the iconographers' interpretation of Split, saying that they confuse architectural tradition, use and meaning. The description of Split by these architectural iconographers as a palatium sacrum and castrum palatium presumes that Diocletian had a godlike status during life which he maintained after abdication. This ignores the statements of ancient authors that he retired to become a private citizen once more. While emperor, Diocletian's status was not dissimilar to that of his immediate predecessors. In associating himself with a divine personage, Jove, Diocletian followed Gallienus and Aurelian who both assumed an identity with Sol; see N. Hannestad, Roman Art and Imperial Policy, Jutland Archaeological Society: Aarhus 1986, pp. 295-297, 300-301. In adopting a Caesar he did no more than ensure that the succession was settled, a practice (including the title) begun under Hadrian. In wearing fabulous clothing such as jewelled slippers, he followed the fashion of third century imperial dress. Depictions of the emperor in art and on coins do not justify the assumption that he was an incarnation of god any more than earlier emperors were; rather such representations show godlike aspects of his nature, his ancestry and his right to rule. In two respects Diocletian was unusual: he intentionally distanced himself from the public by raising imperial appearances to ceremonial occasions and took the extraordinary step of putting aside his purple robe in favour of more youthful and energetic rulers. 229 Lactantius, De Mort. Pers. 7.8-10.

119 private apartments were located in the southern part of the site and elevated so as to give a panoramic view of the coast from a second storey gallery. A religous and mortuary complex stood in front of the monumental entrance to these apartments, and were screened from the street by an arcuated colonnade. Bathing establishments were tucked between the periboloi of the enclosures and the apartments.The northern part of the site were public spaces perhaps reserved for external pleasures; in the northwest quadrant was a large peristyle court, and in the northeast a series of large rooms of which at least one had a sumptuous mosaic pavement. The villa seems to have been professionally planned. A module of measurement approximately 55 Roman feet long can be recognised, and the measurement of the site seems to have originated at the crossing point of the streets. This crossing has traces of a square monument below the present street level which appear to be the foundations of a tetrapylon. Whether this structure was ever completed or not is not known for none of its superstructure is standing. The apparent similarity between Split and castra is compelling. The fortification wall, the gates, the towers and the street plan are all closely paralleled in contemporary military camps. On the other hand, these same

120 features are found on a larger scale in city plans230 and on a smaller scale in fortified villas. Given the military conditions of the third century fortified villas and towns became the norm within the provinces bordering the limes, and it is not surprising that Diocletian resolved to have an enclosed residence rather than a sprawling villa of a type preferred in more stable eras, such as Hadrian's villa at Tivoli. The decision to fortify the villa meant that military architectural vocabulary was resorted to for the walls and that the exterior would have a military appearance from outside, and to a certain extent, in plan. The magazines along the fortification wall in the north part of Split were probably adopted for the same reasons that they were in forts, namely because the space available within the enclosure was limited. It is not reasonable to assume that they were a military innovation and the certainly did not serve a military function everywhere they were used. The only direct evidence for their use is at Split where finds of amphoras suggest that at least some of the units were for storage. The existance of similar units at Mogorjelo reinforces a non-military explanation of their function. On closer examination, crucial differences between the plan of Split and the traditional castrum emerge. The
230

For instance, the Aurelian walls of Rome, RIA, fig. 22; the Porta Appia gate in these walls, RIA, fig. 279; and the street plans of Philippopolis, Brunnow and Domaszewski fig. 1039.

121 comparison with Split was based on the misconception that the principia in a camp was the commander's residence. However, the commander of a garrison usually lived in another part of the camp often close to the principia. At Dura, for example, the commander lived in apartments to the northwest of the principia while the dux had a "palace" on the edge of the encampment overlooking the Euphrates.231 What at Split is usually thought of as the residence in the south part is actually the private apartments consisting of interior domestic spaces.The remainder of the internal area of Split was devoted to other different aspects of the residence, such as the temenoi for religious use, the baths, and the structures in the north part reserved for outside activities. Furthermore, the principia is always located adjacent to the centre of the camp and surrounded by streets, stores and barracks on all sides, while at Split what was thought to be the residence abuts the fortification wall at the extreme south end of the site. The street plan at Split is not as simple as first appears. What is called The Peristyle is, in plan, the

231

M.T.Rostovtzeff, Excavations at Dura Europos. Preliminary Report of the Ninth Season of Work, 1935-1936, New Haven, 1952, pp. 206-7, figs. 5a,b. The principia at Dura is wrongly described as the praetorium in the literature, and the commander's house, properly the praetorium, is located one block west and one block north of the principia. The house of the dux closely resembles Split in that the living apartments are arranged along a gallery overlooking the Euphrates with their entrances facing away from the principal entrance of the complex.

122 extension of the main north-south avenue beyond the crossing at the centre of the complex. The street plan is thus cruciform not T-shaped. Only in two contemporary camps, Luxor and Palmyra, are the principal streets cruciform, and in both these instances there are vital abnormalities of function. Luxor was built during the Tetrarchy around the pre-existing temple of Ammon which seems to have been converted to a monument commemorating the imperial cult. The camp of Diocletian at Palmyra was constructed within the city using pre-existing streets, and possibly reusing buildings laid out as much as a generation before for a civil rather than military function. In both cases there is reason to believe that Diocletian himself visited and stayed at the camps, indeed that they were built in expectation of his visit. Consequently it is possible that the nature of the plan is more closely related to Diocletian's requirements than to a new, short lived, design of a military camp. A further important argument against conscious imitation of the traditional castrum design at Split is that by the time of the Tetrarchy the form of camps had radically changed. During the third century the defence of the empire was modified and the traditional Polybian camp was no longer the standard fortification form. The limes were defended by small forts intended to resist invasion,

123 while the full scale legionary camps, set back from the border, were simplified. In the castra of the period, the number of and emphasis on the gates was reduced so that the principal lines of communication no longer necessarily remained the same as in the Polybian camp scheme. Either the via principalis or via praetoria became the principal axis and often, for instance in the second phase of Drobeta and at Portchester, the other disappeared from the plan. Where they co-existed, one was noticeably broader than the other. Because of the change in castra design, direct comparison between Split and the Polybian plan is anachronistic. The perceived similarities between Split and palaces have been confused by the misunderstanding of Diocletian's status after retirement and Downey's reconstruction of the palace at Antioch. If it is accepted that the ancient sources are unanimous in referring to Split as a villa, that Diocletian actually retired from office and that Libanius described the New City at Antioch not the palace, then the relationship between the plans of Split and palaces becomes clear. This can be best summarised by a rapid reiteration of what a Tetrarchic palace was. Palaces were administrative and residential complexes located within a major city generally close to the limes. They covered a huge area adjacent to a circus, and were occupied, on occasion, by the emperor, by officials and

124 their retinue. The palace at Thessalonika was composed of numerous discrete blocks of rooms which, although they were placed together in a cohesive unit, appeared modular in plan because the plan lacked connective passages and strong horizontal axes. From the palace plan it is difficult to separate public from private space. From Libanius's description, Antioch seems to have had a similarly modular plan. By contrast, Split was purely a rustic residence remote from the seat of government, with no administrative function, and occupied by a private citizen, formerly an emperor. The plan has strong horizontal axes which connect the various parts of the site, but public and private space is clearly defined. Finally, Split, although relatively large, is tiny compared to a palace. The considerable similarities between villas and Split confirm the statements of the ancient sources that it was originally designed as a luxurious rustic villa. Its plan was, to a large extent predetermined by the form of the walls and the positioning of the gates. Into this frame the various elements desired by the patron were fitted. These include the standard vocabulary of the private apartments of villas: the tricinium, the reception hall, the sleeping quarters arranged along a transverse corridor and bath structures. At Split these are confined to the southern portion of the site separate from public

125 spaces such as the streets, courtyards and storage spaces. In their arrangement they recall the disposition at Mogorjelo and the fortified villas portrayed in the African mosaics; the transverse corridor is at second storey elevation with an arcade offering a panoramic view, the cubicles are arranged around the fortification wall and public and private space are kept separate and the overall plan is controlled by the axes of walls and gates. In addition to regular villa forms, Split has elements not usually found in villa architecture for which external parallels may have to be sought. The mausoleum and the temple, both within temenoi, and the broad streets flanked by colonnades seem to draw from a different medium. The mausoleum is a structure that was to be found outside the city at Rome, for instance the Tor dei Schiavi, perhaps as part of the palace complex at Thessalonika, at the Villa of Maxentius on the Via Appia and incorporated into the city plan of the of Philippopolis (fig. 1.4).232 These latter three projects were products of direct imperial patronage, and the conclusion that at Split this element is related to Diocletian's former status as Augustus is inescapable.233

232

Butler, H.C. Publicartions of the American Archaeological Expedition to Syria in 1899-1900. London 1902, fig.135. 233 A. Fraser, "The Iconography of the Emperor Maxentius' Buildings on the Via Appia," Art Bulletin 48 (1966), p. 387-8 and A. Grabar, Martyrium: Recherches sur le culte des reliques et l'art chretien antique, Paris 1946, Vol. I, pp. 220-232, see these mausolea as a dynastic tomb cum founder heroa cum temple of the imperial cult.

126 In view of the diverse settings in which the this particular mausoleum form has been found: outside Rome, in a palace complex, as part of a rustic villa and as part of a city plan, the significance of its presence at Split cannot be ascribed to the influence of any one particular architectural prototype. Rather its presence reflects the special needs of a rich important individual who wished to lie in a structure of a particular type, suitable to his status, within the confines of his villa built near his birthplace. The temple set within its own temenos is a feature more likely to be found in a city or a sanctuary than in a private villa. Modest religious edifices were found at the Piazza Armerina, including a shrine with a statue of Apollo set in a niche and a shrine of the household gods234 in the atrium opposite the entrance. These too were provided for the use of the owner and his family. At Gamzigrad two podium temples were found. These were both richly decorated and one was even furnished with sculptures of Cybele and Hercules. The interpretation of the site suggests that it was an imperial villa rather than a sanctuary or settlement, so presumably the temples were built for the private devotions of the owner, his

234

R. J. A. Wilson, op. cit. p. 17. The Penates worshipped in lararia of private houses. Little but the Lares seem to be revered ancestors, the by Aeneas from Troy. M. Grant, Roman Myths, New

and Lares were is known about either, Penates were brought York, 1971 pp.79-83.

127 immediate family and his retinue. One authority has suggested that the temples were dedicated to Hercules and Cybele who he links with Galerius and his mother respectively, as divinities especially favoured by them.235 If shrines to divinities and ancestral spirits were standard in private houses and villas as at the Piazza Armerina, and temples to what may have been ancestral divinities were found at the imperial villa at Gamzigrad, then the temple complex at Split may, with reservations, be interpreted as a glorified lararium in honour of his ancestor Jove.236 In any case, temples in private villas are not unprecedented. The street plan of is more difficult to place in its true perspective. Villas by nature have a system of interior communication that is designed for small amounts of traffic. The main arteries of villas are through corridors and passages that skirt open areas such as atria and are channeled at certain places through doorways. At Split the interior of the villa is crowded with structures each of which require at least one point of access. The starting points of the main arteries were determined as
235

D. Srejovic, "Two Memorial Monuments", pp.22-3; D. Srejovic, A. Lalovic and D. Jankovic, "Two Late Roman Temples at Gamzigrad" Archaeologia Iugoslavica 19 (1978) 54-63; D. Srejovic, A. Lalovic and D. Jankovic, "Gamzigrad" Starinar 31 (1980); D. Srejovic, "An Imperial Roman Palace" pp.97-99. Srejovic is mistaken, however, in linking Galerius to the gens Herculii; his adoption by Diocletian made him a member of the gens Jovii. 236 For Diocletian and Maximian as "sons of gods and the creator of gods" see S. S. McCormack, Art and Ceremony in Late Antiquity, Berkeley, 1971, p. 170.

128 soon as the location of the gates was decided. One passage, the decumanus ran between two gates, another, the cardo ran from one gate to the vestibule of the private apartments, a third connected all three gates and gave access to the cubicles, towers, ramparts and baths. Conceivably the architects could have chosen a meandering path for the streets and a random or multi-axial arrangement of internal structures to produce a maze of passages with constantly changing architectural spaces. Constrained by limited space, by the alignment of the walls and the location of gates, the architects decided on a strictly orthogonal plan. The result was a pattern of long, straight vistas on the main axes of the site which provided direct access between points of departure and destinations. This plan is most readily compared with the street plan of cities, especially those of the eastern empire, where long streets, flanked by colonnades, dissected the built up areas. At Split, as in many Roman cities, the plan was laid out from a central point.237 The resulting street plan in

237

See Chapter 2, note 5; In laying out of a Roman city a prescribed ceremony was employed. This ceremony, borrowed from Etruscan practice, was first used by Romulus in the founding of Rome and is described in Plutarch, Vita Romuli, 11. The centre of the city was marked and streets laid out at right angles from this point of origin. The city limits were defined by a ploughed trench called the sulcus primigenius or pomerium and the gates marked by raising the plough and lifting it over the designated place. At the centre a pit was excavated and the first fruits of various crops thrown in along with earth brought from the country of origin of the various citizens and closed. This pit was the mundus and at Rome the pit was unsealed thrice yearly to allow the ghosts contained to roam the city. Roman towns and camps were also

129 cities was an othogonal lay-out parallel to the two major axes of the plan, the cardo and decumanus respectively. In eastern cities these two main streets, at least, were broad avenues flanked by colonnades. A notable feature of the major crossings of such foundations is that they are frequently marked by monuments. These monuments take two forms; the tetrapylon or quadrifons found from Spain to Syria and the tetrakionion limited in distribution to the eastern part of the empire only.238 At Split the crossing streets and colonnades are both suggestive of eastern city street types. It is also known that the site was laid out using a fixed unit of measurement239 originating at the intersection of the streets and that at this point four massive foundations possibly for a tetrakionion, which may never have been used, marked the four corners of the intersection. It seems then that Diocletian set about the
laid out in a similar fashion, the surveyor placed his groma and marked out the lines of the principal axes of the settlement. Subsidiary streets were ultimately derived from these base lines. A mundus was found at Cosa under the Capitolium see F.E.Brown, E.H.Richardson and L.Richardson, Cosa II. The Temples of the Arx, MAAR 26 (1962) pp. 9-15. 238 W. MacDonald, The Architecture of the Roman Empire Vol.II, New Haven 1986 pp. 87-91. MacDonald suggests that these monuments were perhaps influenced by Greek tetrastyle altars and the Ianus Geminus quadrifons passage shrine at Rome. To MacDonald it seems natural that the intersection of streets should be embellished and monumentalised, but he also sees a greater significance in the monumentalisation; "A four square structure stands where a surveyor drove his stake or set up his sighting instrument, a place highly charged with meaning, a place holding a whole town or city quarter in fealty. From it a governing order was laid upon the earth roundabout through the agency of two intersecting, controlling lines. Four square structures celebrate the location and significance of these spots, giving them spatial definition and visual character." The spot marked is essentially the mundus of the new foundation and the monument set up a portal. 239 See Chapter 2, supra note 34.

130 construction of his villa at Split with the correct formulae for the act of foundation of a city. This is particularly enlightening in the light of MacDonald's ideas that much of villa architecture contains much of the building typology of urban planning. For instance, at the Piazza Armerina he ennumerates aqueducts, arches, a basilica, baths, a circus, exedrae, fountains, latrines, peristyle courts, shops and shrines among the urban features either existing, implied in the plan or in mosaic and sees tha same kind of features embodied in Hadrian's Villa at Tivoli.240 At Split the a similar array of urban forms can also be found including evidence of the act of foundation. In the light of the evidence, it seems that the plan of Split, except for its walls, was not modelled on military prototypes and that palace planning had no role in its design. The complex was a fortified villa, or rather, a chateau, as Duval envisioned it. Besides the obvious influence of villa architecture, strong influence of Roman city planning can be detected.

240

W. MacDonald, The Architecture of the Roman Empire Vol.II, pp. 274283.

131

BIBLIOGRAPHY Ancient Sources. Aurelius Victor. Epitome de Caesaribus. Dio Cassius. Historiarum Romanarum. Eusebius. Historia ecclesiastica. Eutropius. Breviarum historiae Romanae. Evagrius (Scholasticus). Historia ecclesiastica. Hygenus Gromaticus. De munitionibus castorum. Jerome. Chronicon. Lactantius, De mortibus persecutorum. Libanius, Orationes. Malalas, John. Chronographia. Nicetas Choniates. Historia. P. Beatty Panop. T.C.Skeat, trans. Papyrii from Panopolis , Hodges, Figgis:Dublin 1964. Pausanias, Periegeta. Plutarch, Vita Romuli. Polybius, Historiae. Procopius, De aedificiis. Prosper Tiro, Chronica minora. Scriptores historiae Augustae. Socrates Scholasticus Historia ecclesiastica. Strabo, Geographus. Theodoretus, Historia ecclesiastica. Zonaras, Epitome historiarum.

132

Modern Sources. Adam,R. Ruins of the Palace of the Emperor Diocletian at Spalatro in Dalmatia. London, 1764. Arix, R.O. AA 54 (1939) 156-171. Beckwith, J. Early Christian and Byzantine Art. Harmondsworth, 1970. Barnes, T. The New Empire of Diocletian and Constantine. Cambridge, MA, 1982. Bissing, F.W. von "Altchristliche Wandmalereien aus Agypten," Festschrift zum Sechsigsten Geburtstag von Paul Clemen. Dusseldorf, 1926, pp. 181-88. Boethius, A. "The Reception Halls of the Roman Emperors." BSA 46 (1951) 25-31. Bouchier, E. A Short History of Antioch. Oxford, 1921. Boskovic, O., Duval, N., Popovic, V. and Vallet, G. Sirmium: recherches archeologiques franco-yogoslaves a Sirmium. Ecole Francaise de Rome, Institute Archeologique de Belgrade, 1977. Brown, F.E., Richardson, E.H., and Richardson, L. Cosa II. The Temples of the Arx. Memoirs of the American Academy in Rome 26 (1962) 9-15. Browning, I. Palmyra. New Jersey, 1976. Brunnow, R.E. and Domaszewski, A. v. Die provincial Arabia. vol. 2, Strassburg, 1905. Bulic, F. Kaiser Diokletians Palast in Split. Translated by L. Karaman. Zagreb, 1929. Bury, J.B. "The Great Palace". BZ 21 (1912) 210-25. Butler, H.C. Publications of the American Archaeological Expedition to Syria in 1899-1900. Part II: Architecture and Other Arts. London, 1904. Calderini, A. Storia di Milano. Milan, 1965. Campbell, W.A. "The Circus," in Antioch on the Orontes, I. The excavations of 1932, pp.31-34. Edited by G.W. Elderkin. Princeton, 1934.

133

________. "A Byzantine Stadium," in Antioch on the Orontes, I. The excavations of 1932. p.32. Edited by G.W. Elderkin. Princeton, 1934. ________. "Excavations at Antioch on the Orontes." AJA 38 (1934) 201-6. ________. "The Third Season of Excavation at Antioch on the Orontes," AJA 40 (1936) 1-9. ________. "The Fourth and Fifth Seasons of Excavations at Antioch on the Orontes: 1936-1936," AJA 42 (1938) 205-8. ________. "The Sixth Season of Excavation at Antioch on the Orontes: 1937," AJA 44 (1940) 417-27. Canak Medic, M. "Le palais de l'epoque de la Basse Antiquite pres de Gamzigrad," in Actes de XIVe congres internationale des etudes Byzantines vol. III, Bucarest, 1976, pp. 357-62. Carrie, J-M. "Les Castra Dionysiados et l'evolution de l'architecture militaire romaine tardive," MelRome 86 (1974) 819-50. Cassas, L.F. Voyage pittoresque et historique de l'Istrie et Dalmatie. Paris, 1802. Castagnoli, F. Orthogonal Town Planning in Antiquity. Cambridge, 1971. ________. "Roma Quadrata." in Studies Presented to David Moore Robinson. vol. I, pp. 389-99. Edited by G.E. Mylonas. St. Louis, 1951. Colledge, M.A.R. The Art of Palmyra. London, 1976. Collingwood, R.E. and Richmond, I. The Archaeology of Roman Britain. London, 1969. Crouch, D. "The Ramparts of Palmyra," Studia Palmyrenskie 6 (1975) 6-44. Cunliffe, B. "Excavations at Portchester Castle, Hants. 1961-3," ArchJ 49 (1969) 62-74. ________. "Excavations at Portchester Castle, Hants. 1966-68," ArchJ 49 (1969) 62-74.

134 ________. "Excavations at Portchester Castle, Hants. 1969-71," ArchJ 52 (1972) 70-84. Dagron, C. Naissence d'une capitale: Constantinople et ses institutions de 330 a 451. Paris, 1974. Daressy, C. "Notes sur Luxor de la periode romaine," Annales du Service des Antiquites de l'Egypte 19 (1920) 166ff. Dawkins, R.M. and Wace, A.J.B. "Greek Island Embroideries," Burlington Magazine 26 (1915) 99-107. Deckers, J.D. "Die Wandmalereien des tetrarchischen Lagerheiligtums in Ammon-Tempel von Luxor". RomQ 68 (1973) 1-34. _________. "Die Wandmalerei im Kaiserkulten von Luxor". JdI 94 (1979) 600-652. Delvoye, C. "Sur quelques aspectes des relations entre l'architecture palaeobyantine de l'Anatolie, de Constantinople et de la Grece," RA (1976) 143-152. Dilke, O.A.W. The Roman Land Surveyors. New York, 1971. Dinsmoor, W.B. "The Literary Remains of Sebastiano Serlio," ArtB 24 (1942) 55-91. Dorner, F.K. "Inschriften und Denkmaler aus Bithynien," Istanbuler Forschungen 14 (1941) 44-51. Downey, G. "The Palace of Diocletian at Antioch," Les annales archeologiques de Syrie 3 (1953) 106-16. ________. A History of Antioch in Syria. Princeton, 1961. ________. "Libanius' Oration in Praise of Antioch (Oration XI)," ProcPhilAs 103 (1959) 652-86. Dunbabin, K.M.D. The Mosiacs of Roman North Africa. Oxford, 1978. Duval, N. "Sirmium `ville imperiale' ou `capitale'?" Corsi di cultura sull'arte ravennate e bizantina 26 (1979) 56-8. ________. "Aquileia e Milano," Antichita Altoadriatiche 4 (1973) 158ff.

135 ________. "La place de Split dans l'architecture aulique du bas-empire," Urbs 4 (1961-62) 67-95. ________. "Le `palais' de Diocletien a Spalato a la lumiere des recentes decouvertes," Bulletin de la Societe nationale des antiquaires de France (1961) 71117. ________. "Palais et forteresses en Yougoslavie: recherches nouvelles," Bulletin de la Societe nationale des antiquaires de France (1971) 99-129. Dyggve, E. Ravennatum Palatium Sacrum. Det Kgl. Danske Videnskabem Selskab, Archaeologisk-Kunsthistorische Meddelelser III, 2 (1941). _______. "La region palatiale de Thessalonique," Acta Congressus Madrigiani Hafniae 1954. Copenhagen, 1958, pp. 353-65. _______ and Vetters, H. Mogorjelo: Ein spatantike Herrensitz in Dalmatien. Vienna, 1966. Ebersolt, J. Le grande palais de Constantinople et le livre de ceremonies. Paris, 1910. Elderkin, G.W. ed. Antioch on the Orontes I. Excavations of 1932. Princeton, 1934. Fabricius, E. "Some Notes on Polybius's Description of Roman Camps," JRS 22 (1932) 78-87. Farlati, D. Illyricum Sacrum. Venice, 1751. Fellmann, R. Die Principia des Legionslagers Vindonissa und das Zentralgebaude der romischen Lager und Kastelle. Vindonissamuseum, 1958, 75-92. ________. "Le `Camp de Diocletien' a Palmyre et l'architcture militaire du Bas Empire," in Melanges d'histoire anciennes et d'archeologie offerts a Paul Collart. (Cahiers d'archeologie romande. no.5). Lausanne, 1976 173-91. ________. "Der Diokletianspalast von Split im rahmen der spatromischen Militararchitektur," Antike Welt 10 (1979) 47-55. Firlati, N. Izmit Sehri ue Eski Eserleri Rehberi. Istanbul, 1971.

136 Fisher, C.S. "Bath B, House A and the Roman Villa," in Antioch on the Orontes I. Excavations of 1932. Edited by G.W. Elderkin. Princeton, 1934, 7-13, ________. "Bath A," in Antioch on the Orontes I. Excavations of 1932. Edited by G.W.Elderkin. Princeton, 1934, 5-6. ________. "Bath C," in Antioch on the Orontes I. Excavations of 1932. Edited by G.W. Elderkin. Princeton, 1934, 19-31. Fleming, J. "The Journey to Spalatro," Architectural Review 123 (1958) 103-107. Frazer, A. "The Iconography of Maxentius' Villa on via Appia," ArtB 48 (1966) 385-92. French, D. Roman Roads and Milestones of Asia Minor. vol. 1: The Pilgrims' Road. Oxford, 1981 Frothingham, A.L. "Circular Templum and Mundus. Was the Templum only Rectangular?" AJA 18 (1914) 302-20. Gabricevic, B. "Decussis Dioklecijanove Palace u Splitu," Vjesnik za Archeologiju i Historiju Dalmatinsku 53-54 (1961-62) 113-24. Galikowski, M. "Les Principia de Diokletien, `Temple des Enseignes'," in M.Galikowski, ed. Palmyre VIII. Warsaw, 1984, 10-48. Gichon, M. "Three years Excavation at Masad Tamar," Saalburger Jahrbuch 33 (1976) ________. "Tamara," in Avi-Yonah, M. and Stern, E. eds. Encyclopedia of Archaeological Excavations in the Holy Land. Jerusalem, English edition, 1978, 1148-52. Guilland, R. Etudes de topographie de Constantinople Byzantine. Amsterdam, 1969. Hanfmann, G. From Croesus to Constantine. Ann Arbor, 1976. Hannestad, N. Roman Art and Imperial Policy. Aarhus 1986. Hebrard, E. and Zeiller, F. Spalato, le palais de Diocletien. Paris, 1912. Humphrey, J.H. Roman Circuses. London, 1986.

137 Janin, R. Constantinople Byzantin. Paris, 1964. Johnstone, A. Roman Forts. London, 1983. Jullien, M. "Le Culte chretien dans les Temples de l'antique Egypte," BSA Copt 6 (1940) 163-4. Kahler, H. Das Funfsaulendenkmal fur die Tetrarchenauf dem Forum Romanum. Cologne, 1964. Kalavrezou-Maxeiner, I. "The Imperial Chamber at Luxor," DOP 29 (1975) 227-51. Kennedy, D.L. Archaeological Explorations on the Roman Frontier in North East Jordan. Oxford, 1982. Kinch, K.F. L'arc de triomphe de Salonique. Paris, 1890. Kraeling, C.H. Gerasa. New Haven, 1938. Krautheimer, R. Three Christian Capitals. Berkeley, 1983. ________. Early Christian and Byzantine Architecture. Harmondsworth (4th. rev. ed.), 1981. Krekemet, D. "Louis Francois Cassas i njegove slike Istra i Damacije 1782," RAD. Jugoslavenskie Akademije Znanosti i Umjetnosti. 379 (1979) 7-200. Krogulska, M. "Les principia," in M. Gaalikowski, ed. Palmyre VIII. Warsaw, 1984, 70-90. Lacau, M.P. "Inscriptions latines du temple de Luxour," Annales du Service des Antiquites de l'Egypte, 34 (1934) 17-46. Lander, J. Roman Stone Fortifications. Oxford, 1984. Lang, S. "A Few Suggestions towards a New Solution of the Origin of the Early Christian Basilica," RACrist 30 (1954) 200-204. Lanza, F. Dell' antico palazzo di Diocleziano in Spalato. Triest, 1855. Lassus, J. "Antioch" in PECS, 62. Legrain, G. "Rapport sur les travaux executes a Luxor a l'ouest de Temple d'Amman," Annales du Service des Antiqites de l'Egypte 17 (1917) 49-75.

138 Lehmann, K. "The Dome of Heaven," ArtB 27 (1945) 1-27. Levi, D. Antioch Mosaic Pavements. New Haven, 1947. L'Orange, H.P.L. Art Forms and Civic Life in the Late Roman Empire. Princeton, 1965. ________. Studies on the Iconography of Cosmic Kingship in the Ancient World. Cambridge, MA, 1953. Lugli, G. "Studi topografici intorno alle antiche ville suburbane." BullComm 52 (1924) 120ff. MacCormack, S. Art and Ceremony in Late Antiquity. Berkeley, 1981. MacDonald, W.A. The Architecture of the Roman World, I. New Haven, 1982. ________. The Architecture of the Roman World, II. New Haven, 1986. MacMullen, R. Soldier and Civilian in the Late Roman Empire. Cambridge, MA, 1963. McNally, S. "Der Diokletianspalast in Split," Antike Welt 10 (1979) 35-46. Marasovic, J., Marasovic, T. and McNally, S. Diocletians Palace. Report on Joint Excavations. Part 2. Split, 1976. Mango, C. The Brazen House. Copenhagen, 1959. ________. Byzantine Architecture. London, 1986. Mano-Zisi, D. "Le castrum de Gamzigrad et ses mosaiques," Archaeologia Iugoslavica 2 (1956) 67-84. Marasovic, J., Marasovic, T., MacNally, S. and Wilkes, J.J. Diocletians Palace: Report on Joint Excavations in Southeast quarter. Part 1. Split, 1972. Michalowski, K. Palmyre I. Fouilles Polonaises, 1959. La Haye-Paris, 1960. ________. Palmyre II. Fouilles Polonaises, 1960. Warsaw, 1962. ________. Palmyre III. Fouilles Polonaises, 1961 , Warsaw, 1963.

139

Millar, F. The Emperor in the Roman World. London, 1977. Miller, K. Itineraria Romana. Stuttgart, 1916. Mocsy, A. "Aurelianus-Aquae-Gamzigrad," Studia Balcanica 1 (1970) 49-54. ________. Pannonia and Upper Moesia. London, 1974. Modrijan, W. Der romische Landsitz von Loffelbach. Graz, 1971. Monneret de Villard, V. "The Imperial Cult at Luxor," Archaeologia 95 (1953) 85-105. Morey, C.R. "The Excavations of Antioch on the Orontes," ProcPhilSoc 76 (1936) 637-51. ________. The Mosaics of Antioch. London, 1938. Mouterde, R. and Poidebard, A. Les Limes de Chalkis. Paris, 1945. Moutsopoulos, N.C. "Contribution a l'etude du plan de ville de Thessalonique a l'epoque romaine," Atti del XVI Congresso di Storia dell Architettura, Athens 1969. Rome, 1977, 187-263. Muller-Wiener, W. Bildlexikon zur Topographie Istanbuls. Tubingen, 1977. Nash, E. A Pictorial Dictionary of Ancient Rome. New York, 1961. Naval Intelligence Division Geographical Handbook, Turkey. vol. 1 Niemann, G. Der Palast Diokletians in Spalato. Vienna, 1910. ________. "Zur Porta Aurea in Spalato." OJh 12 (1909) 340-42. Ozture, A. Izmit Tarihi. Istanbul, 1981. Parker, S. Thomas. "Central Limes Arabicus Project," ASOR Newsletter December issue (1982) 6-14. Palmer, R.E.A. Roman Religion and Roman Empire. Philadelphia, 1974.

140

Petrikowits, H. von. "Fortifications in the North Western Roman Empire from the Third to the Fifth Centuries A.D.," JRS 61 (1971) 178-218. Pisani Sartorio, G. and Calza, R. La villa di Massenzio sulla Via Appia. Il Palazzo. Le Opere d'arte. Rome, 1976. Reusch, W. "Die Spatantike Kaiserresidenz Trier im Lichte neuer Ausgrabungen," AA 77 (1962) 875-88. Popovic, V. "A Survey of the Topography and Urban Organisation of Sirmium in the Late Empire," Sirmium I. Belgrade, 1971, 119-33. Rey Cocqais, J.P. "Palmyra". PECS, 669. Richmond, I.A. "Palmyra Under Rome". JRS 53 (1963) 43-54. Rostovtzeff, M.T. Excavations at Dura Europos. Preliminary Report of the Ninth Season of Work, 1935-1936. New Haven, 1952. Schlumberger, D. "Notes sur la decoration architectural des collandes des rues et du Camp du Diocletien a Palmyre," Berytus 2 (1935) 163-67. ________. "Le pretendu Camp du Diocletien a Palmyre," MelUSJ 38 (1962) 77-97. Schwartz, J. et al. Qasr Qarun/Dionysias II, 1950. L'Institut francais d'Archeologie orientale du Caire. Cairo, 1969. Sear, F. Roman Architecture. Ithaca, 1982. Seston, W. Diocletien et la Tetrarchie. Paris, 1946. Smith, E.B. Architectural Symbolism of Imperial Rome and the Middle Ages. Princeton, 1956. Spieser, J.M. Thessalonique et ses monuments du IVe au VIe siecle. Paris, 1984. Srejovic, D. "Two Memorial Monuments of Roman Palatial Architecture," Archaeologia Iugoslavica 22-23 (1982-83). ________. "Two Late Roman Temples at Gamzigrad," Archaeologia Iugoslavica 19 (1978) 54-63.

141

________. "An Imperial Roman Palace in Serbia". ILN 263 (Oct. 1975) 97-9. Srejovic, D., Lalovic, A. and Jankovic, D. "Gamzigrad," Starinar , 31 (1980) 65-80. ________. "Two Late Roman Temples at Gamzigrad," Archaeologia Iugoslavica 19 (1978) 54-63. Starcky, J. Palmyre. Paris, 1959. Stillwell, R. Antioch on the Orontes, II. The Excavations of 1933-1936. Princeton, 1938. Strzygowski, J. "Spalato, ein Markstein der romanischen Kunst bei ihrem Ubergange vom Orient nach dem Abendlande," Studien aus Kunst und Geschichte Fr. Scneider gewidmet. Fiebourg-en-Brisgau, 1906. Swift, E.H. Roman Sources of Christian Art. New York, 1951. Swoboda, K.M. Romische und Romanische Palaste. Vienna, 1919. ________. "The Problem of the Iconography of Late Antique and Early Medieval Palaces," JSAH 20 (1961) 78-9. Texier, C. Description de l'Asie Mineure. Paris, 1839. Velenis, G. "Architektonische Probleme des Galeriusbogen in Thessaloniki," AA (1979) 249-63. Vickers, M. "Hellenistic Thessaloniki," JHS 92 (1972) 156-70. ________. "The Hippodrome at Thessaloniki," JRS 62 (1972) 25-32. ________. "Observations on the Octogon at Thessaloniki," JRS 63 (1973) 111-20. Ward-Perkins, J.B. Roman Imperial Architecture. Harmondsworth, 1981. ________. "Constantine and the Christian Basilica," PBSR 9 (1954) 69-90. Warde Fowler, W. "Mundus patet," JRS 2 (1912) 25-33.

142 ________. The Roman Festivals of the Period of the Republic. London, 1933. Wheeler, M. "Roman Cardiff: Supplementary Notes," ArchJ 2 (1922) 392. ________. Roman Art and Architecture. Oxford, 1964. Weilbach, F. "Zur Rekonstruktion des Diocletians Palast," in Strena Buliciana pp. 123-25. Edited by E. Wiegand. Zagreb-Split, 1924. Wheler, G. Voyage de Dalmatie, de Grece et du Levant. vol. I. La Haye, 1723. Wiegand, E. "Die Stellung Dalmatiens in der romischen Reichskunst," in Strena Buliciana. Edited by E. Wiegand. Zagreb-Split, 1924, 77-104. Wightman, E.M. Roman Trier and the Treveri. London, 1970. Wilkes, J.J. Diocletian's Palace, Split. Sheffield, 1986. Williams, S. Diocletian and the Roman Recovery. London, 1985. Wilson, R.J.A. Piazza Amerina. Austin, 1983. Wood, R. The Ruins of Palmyra Otherwise Known as Tedmor in the Desert. London, 1753. Zeiller, J. "Sur la place du palais de Diocletien a Spalato dans l'histoire de l'art," Byzantion 6 (1931) 565-69.

Plates 1.1 and 1.2

Plate 1.3

Plate 2.1

Plate 2.2

Plate 2.3

Plate 2.4

Plate 2.5

Plate 2.6

Plate 2.7

Plate 2.8

Plate 2.9

Plate 2.10

Plate 2.11

Plate 2.12

Plate 3.1

Plate 3.2

Plate 3.3

Plate 3.4

Plate 3.5

Plate 3.6

Plate 3.7

Plate 3.8

Plate 3.9

Plate 3.10

Plate 3.11

Plate 4.1

AA

Plate 4.2

BB

Plate 4.3

CC

Plate 4.4

DD

Plate 4.5

EE

Plate 4.6

FF

Plate 4.7

GG

Plate 4.8

HH

Plate 4.9

II

Plate 4.10

JJ

Plate 4.11

KK

Plate 4.12

LL

Plate 4.13

MM

Plate 5.1

NN

Plate 5.2

OO

Plate 5.3

PP

Plate 5.4

Vous aimerez peut-être aussi