Version classiqueVersion mobile
OpenEdition Books

Les bâtisseurs sur les deux rives de la mer Rouge

Jean-François Breton

Chapter 8. The architecture of the late periods in South Arabia and in Tigray (1st-6th century AD)

Texte intégral

8.1. Himyarites and Axumites

1In Egypt, after the death of Alexander the Great and the break-up of his empire, the general Ptolemy took control of Egypt. From that period we know that there were contacts between Egypt and the kingdoms of South Arabia and indirect contacts with India. His successor Ptolemy II promoted commercial activity in the Red Sea by constructing ports at Arsinoë, Myos Hormos, Philoteras and Berenike in Egypt, and his successors continued to found ports all along the Red Sea coasts. This initial contact soon expanded to include trade in aromatics and other commodities from South Arabia.

8.1.1. The Red Sea, winds and ports

2The Red Sea wind regime explains the localisation of the ports. The climate makes the difference between the northern and the southern part of the Red Sea. The difference of seasonal winds in the northern and the southern halves divides the Red Sea into two parts at roughly the latitude of modern Jiddah. The southern part benefits from the monsoon winds and rains from the Indian Ocean. In the northern part above Jiddah, the prevailing winds blow from the north the entire year. In the southern half of the Red Sea, below Jiddah, from October to March/April, the wind blows from south to north; in summer it blows from north to south. To the north, on the Egyptian coast, Myos Hormos and Berenike, under the Ptolemies and the Romans, developed as new ports. Farther south, on the African coast, no early port is recorded at the latitude of medieval ‘Aydhâb (lat. 22° N) although this is the region where the southerly wind blows reliably. In the southern Red Sea, Adulis (lat. 15° N) was the major African port (from 1st cent. BC) across from Muza (lat. 13° N) on the coast of Yemen.

3From the reign of Augustus on, the increase in maritime activity (noted by Strabo) may have been due, in part, to the exploitation of the monsoons by western merchants. The policy of Tiberius and his successors seems to have been mainly to consolidate Augustan conquests and the further development of the trade in the Red Sea.

8.1.2. The “Periplus” and the South Arabian trade (fig. 8.1)

4The “Periplus of the Erythrean Sea” (the Red Sea), an anonymous handbook for sailors and merchants, written in Greek probably around 40-50 AD, lists four major cities on the Arabian coast, namely Muza (modern-day Makhâ), Okélis (modern-day Shaikh Saîd), Arabia Felix (‘Aden) and Qana’ (modern-day Bir Alî). As the Periplus tells us, the first two cities belonged to the King of Saba’ and dhû-Raydân, while Sapahar (Zafâr) was the capital of Himyar. There is no mention of the rulers of Aden in the Periplus and it is uncertain whether the Qatabanians still held it at this time. The Periplus also mentions Adulis on the western coast of the Red Sea.

5The Periplus is a trove of information about ancient geography, ports, and maritime activity in the mid-first century AD. Indeed, it is the primary source of information on commercial relations in the region during this period, since other available sources are sparse and widely scattered. On the Egyptian coast of the Red Sea at Leukos Limên inscriptional evidence has been found of a Himyarite merchant and a trader from Aden. At Alexandria, which had monopoly control of all Red Sea traffic, it is likely that South Arabians were engaged in shipping along with Syrians and Indians. Overall, it is clear that the commercial horizons of South Arabia and Ethiopia began to expand significantly in this period, and an influx of foreign cultural influences naturally followed.

6Although we have little documentary evidence concerning Roman exports to South Arabia, archaeological excavations and surveys provide more information. Different varieties of ceramics (Terra sigillata, Dressel and Rhodian amphorae), of glass, and of bronze have been found in quantity in the ports and the capitals of South Arabia, Shabwa, Tamna‘, Mâ’rib.

8.1.3. The Rise of Himyar (fig. 8.1)

7In the central Yemeni highlands, in the region of Dhâmar and Yarîm, among a loose confederation of tribes, a modest tribe of Qatabanian origin gradually emerged as the dominant tribe. Their leaders were based at the fortress of Raydân in Zafâr, near Yarîm. Gradually they became a regional power, the Himyarites, headed by the line known as “dhû-Raydân” or “of Raydân” which controlled the entire plain of Dhâmar. This new Himyarite confederation differed from its predecessor in that it was organized around allegiance to a common ruler, rather than the cult of a divinity. The territory was established as an independent principality sometime around 110 BC, and thereafter began a steady expansion westwards into the mountains beyond their watershed, and northwards, where it clashed with tribes that were still under Sabaean domination. The break with Saba’ came in the final decades of the first century AD, when the Himyarite leaders of Raydân proclaimed themselves “King of Saba’ and dhû-Raydân”.

8The factors behind the success of the Himyarites remain obscure. One clear advantage was their location in the heart of so-called “Green Yemen”, the prosperous land of the Dhâmar plain and its environs, where abundant rainfall (averaging more than 400 mm and sometimes as much as 900 mm) and fertile soil are capable of sustaining intensive agriculture and animal breeding. The introduction of the horse sometime in the first century AD doubtless had significant consequences: the rapid development of cavalry units.

8.1.4. Axum and South Arabia

  • 1 Bowersock 2014: 8-17.

9At the beginning of the third century the Axumite king controlled an army that occupied parts of southwestern Arabia. About 270, the Ethiopians left South Arabia when Saba’ and Himyar recovered their independence, but for a long time the kings of Axum not only issued coinage but arrogated to themselves titles that asserted sovereignty over places in Arabia where they no longer ruled1.

10In the fourth century the titulature of the negus is known from the Aezanas or ‘Eizana inscription where the whole process of his conversion to Christianity is played out. At about the same time, in the late fourth century, the Himyarite kings converted to Judaism.

  • 2 Gajda 2009.

11According to G. W. Bowersock, the Aezanas text served as the inspiration for the new negus of the early sixth century, Ella Asbeha or Kâlêb. The massacre of Christians in Najrân in 523 triggered an invasion from Axum to central and south western Arabia that restored Christianity for a period of nearly fifty years2. In 522 Kâlêb was already poised to invade Yemen, and in 525 he overthrew Yûsûf As’ar Yath’ar and appointed Sumyafa Ashwa, himself quickly toppled by Abraha. Abraha succeeded in strengthening his own position in Himyar by distancing himself from Axum. By 570 the Axumite presence in South Arabia was over and the Persians took over Himyar without a struggle.

8.2. Himyarite architecture (Jean-François Breton)

12Himyarite architecture is spread mainly over the highlands of Yemen, between the region of Khawlân in the north and the region of Hujariah in the south, altogether some 300 km north to south. These highlands are over 2,000 m in altitude, with peaks reaching up to 3,760 m (Jabal Nabî Shu’yab). Further west, a hilly piedmont zone ranges in height up to 700/800 m. Rocky outcrops are separated by high plains or deep valleys. This terrain, though fine for defence, was disadvantageous for the unification of the Himyarite tribal confederation.

13Various types of stone are easily accessible here, and limestone and granite quarries were numerous; some are still in use today. Semi-precious stones, mentioned by Hasân al-Hamdâni, include different kinds of quartz-like stones such as agate, chalcedony, and soft calcite (alabaster) (see section 3.1 of chapter 3).

8.2.1. Materials

14A large variety of building stone was available to the architects. Various limestones (some coloured grey, red or green) and basalts were used in the northern regions. It may be assumed that prestigious buildings, like the Ghumdân Castle in Sanaa, used limestone layers of different colours. Granites were common in Hujariah and in the regions around al-Baydâ’, north to Wâdîs Hârib and Bayhân. Timber was also available in large quantities. The existence of half-timber civil and religious buildings is attested on the highlands. The historically recorded burning of castles and churches points to a large amount of wood used in the construction.

8.2.2. Layouts

15Most of the Himyarite buildings have now vanished due to continuous occupation of the villages or to destruction over history. The currently inhabited one or two-storey buildings show ancient platforms of ashlar masonry with large blocks (about 0.50 m to 0.70 m) tightly laid and finely dressed, contrasting with the modern rubbled masonry of the upper courses.

16Though with some difficulty, a number of main features can be mentioned. Rectangular or square high stone platforms are the most common remains. No average dimension can be given: platforms can be of small size (8 m by 8 m), medium (ca. 10/12 m by 10/12 m), large (16 m by 10 m) or truly enormous (37 m by 31 m in Sirwâh (Arhab)). Average dimensions of the platforms in al-Mî’sal are 8 m by 10 m, with a maximum of 20 m in length. In Qaniya al-Husn, near al-Mi’sâl, the castle is 16 m by 11.5 m, in Silâ’ (near Jabal Kânin) one castle (“Gebäude A”) is 22 m by 15 m (fig. 2.13). The platforms range between 2 m and 3 m high, exceptionally 4 m.

17These platforms are cellular platforms built to support the superstructures; they are composed of a grid of stone walls delimiting rubble-filled compartments, so as to create an elevated podium. The internal frame consists of two main walls, parallel to the length of the building, and orthogonal walls dividing the platform into compartments: the two main walls were used as foundations for the main corridor and the secondary walls as foundations for the rooms. The majority of the large civil and religious buildings in the regions north of Sanaa, such as Arhab, Khawlân, have such lines of stone pillars. Most of these platforms are foundations of multi-storey houses whose appearance was not very different from that of the current houses.

8.2.3. Late tower houses

18Most of the dedicatory inscriptions originating from the Yemeni highlands are dated from the 2nd to the 5th century AD. Late inscriptions mention “four stories” (text Gl. 1539/5 in Baynûn), “six levels with six floors |of a palace whose upper part comprises two floors” (DJE 12), and “six floors” (text Gr 27 in Zafâr). Another type of evidence is provided by bas-reliefs which depict houses with distinct floors, decorative posts framing the openings, and terraces possibly with merlons. So, even if we may overestimate the number of storeys, it seems certain that the prevalent type of ancient South Arabian dwelling was the so-called “tower-house”. In the walls of these edifices are inserted building texts, engraved by the owners, attesting to their property and their pride. In addition to the above three, a short selection of some of them is given in the accompanying table.

Text number


Name of the house


Al-Iryâni n°19


house GRT

mid-3rd cent.

Qâniya 1


2 houses

mid-2nd cent.

Qâniya 3


houses HRN and N’MM

2nd cent.

Gl 1209


house ‘TWTm

RES 4708

Nakhlat al-Hamra

house HMW

mid.-3rd cent.

CIH 37


house HY’D

mid.-2nd cent.

A 671

M. Halamlam

tower YF’M

2nd cent.

RES 4648


house Yagûr

3rd cent.

RES 403


house dhû-YKR

4/5th cent.

Istanbul 7630


house dhû-DHR DHRN

4/5th cent.

CIH 644


house Yarish

6th cent.

19Text Istanbul 7630 can be translated: “ZYN ‘SHH, independent head of a family derived from the Banû NSRm with his wife... restored and repaired their house dhû-DHR DHRN, and added to it a roof; and they dug and plastered and set in order a cistern and its retaining wall...”

  • 3 Yule 2007: 120.

20Courtyards extending before the platforms are a common architectural element. In Zafâr at Husn Raydân, its first major building, recently excavated, reveals a more than 16 m wide court with lightly coloured slabs3.

8.2.4. Stone pillars and capitals (fig. 4.21, 8.2 and 8.3)

21Hundreds of stone pillars, either square or rectangular with cut sides, or octagonal in section, were recorded by C. Rathjens (1953) and P. Griaznevich (1994) during their surveys north of Sanaa. Their monolithic capitals show alternating bands, or some smooth, some decorated with recesses, false headers or dentil patterns. Some of the capitals have 2 rows (one plain and one with false headers), some 4 rows (alternating 2 + 2), some 6 rows (either plain bands alternating with decorated bands or some with horizontal grooves). The average height of the capital is around 30 cm, the highest being ca. 50 cm.

22Round columns and their capitals appear in the Highlands of Yemen probably at the beginning of our era. Unknown from previous periods, they indicate a new style of building and decoration. These columns are conservative in style and thus difficult to date. The question arises if they were in production over the entire Himyaritic period, or if they were produced early and then reused for decades.

  • 4 Rathjens 1953, fig. 26.
  • 5 Radt 1973: pl. 9-10.
  • 6 Radt 1973: pl. 8-9, n° 23-24.
  • 7 Simpson 2002: 152.

23The buildings of Ma‘în, Mâ’rib and Tamna‘ have provided no examples of surviving columns. Only two fragments of a quarter of column were found in the royal palace of Shabwa in its last period (4/5th cent. AD). At Mâ’rib square or octagonal stone pillars are attested in large numbers but no column has been found. Most of the preserved columns come from the Himyaritic or Sabaean highlands. C. Rathjens discovered some fragments of columns in Amrân4, fluted ones in Sûq al-Shibâm and the base of a column from Shibâm-Kawkabân (fig. 8.3) and from Silâ. In the W. Radt Catalogue of Sanaa museum, fluted columns are listed from Shibâm-Kawkabân5, Baynûn, etc. Some columns are monolithic with their bases, a curious local adaptation. Limestone capitals display acanthus leaves as at Shibâm-Kawkabân6, at Tana’im, at Baynûn. These acanthus-leaf capitals had a decorative rather than a structural function and although deeply influenced by classical art they differ in their proportions, the leaves being arranged directly above each other rather than placed randomly as with the classical prototypes7.

8.3. Stone, quarries and dressing at Axum (Jean-Claude Bessac)

8.3.1. Rocks used in Axum buildings Orthogneiss

24On the northern side of Beta Giyorgis and the Mai Qoho hills, brown orthogneiss is visible all over. It is most apparent at the Mai Shum water reservoir, at the Rock, Tomb and at Bazen Tomb. Due to the frequent cracks in the rock, it is impossible to extract large building blocks. At best, small blocks were produced, whether due to natural fragmentation, erosion or human attempts to dig through this brittle rock. These small blocks were used more or less as is without any dressing for the construction of ordinary walls. It is mainly this category of rock that we observe between the stone angles in the tombs, among them the tomb close to Stela 1 but also the Kaleb and Gabra Maskal tombs, and at Dongur Palace on the west side. Syenite

25The rock used at Axum for monoliths is a variety of syenite of a dominant bluish-grey color. Although geologists distinguish syenite from granite, stone professionals consider them as almost equivalent rocks. The variety of syenite at Axum has small or medium-size grains (diameter from 1 to 4 mm) with some clearer veins some centimeters in width. Its overall percentage of quartz seems very low, even nonexistent; it is a syenite of nepheline type. From the point of view of its physical characteristics, syenite is an anisotropic rock: its physical behavior differs in each of its three dimensions because of the presence of directional microfractures.

8.3.2. The production of syenite blocks

26According to the desired size—small or middle-sized stones for an ashlar masonry wall, large slabs for tombs or paving, very long monoliths for stelae—the appropriate place in the quarry had to be carefully surveyed and selected. For megaliths, the best quarry seems to have been Gobedra III (fig. 8.5 and 8.6). Paving slabs or covering slabs could also be extracted from Gobedra III, although erratic blocks down the slopes of Gobedra III were also used for small-sized blocks. The Adi Tsehafi quarry produced blocks for a nearby building and not for Axum.

27We can surmise that the blocks used at the Kaleb and Gebra Meskal tombs were excavated nearby and not from the Gobedra quarries, which were too far away. Here, the hillsides show many erratic blocks, some of a large size, ready to use.

8.3.3. Extraction techniques

28Generally the operations of extraction were limited to stone cutting and to on-the-spot squaring of the blocks and the megaliths after clearing them of some sediment. Most of the blocks present very rounded-off faces, especially when they are erratic and thus had undergone thousand-year-old natural erosion during their sliding on slopes. The Axumite quarrymen had thus, first of all, to square them by shaping them into an approximate parallelepiped on all sides (fig. 8.5).

29Because of the great weight of most of the blocks, this operation had to be realized with the least possible movement of the stones, or even without moving them at all, except at the last moment for their transport. Before anything else, the convexity of the superior face was leveled off, bounded by a straight line, realized by piquetage with a pick or a pointed chisel. Often this operation could be made along the natural stratification lines of the syenite because, even on slopes, the largest surface of the block usually corresponds to an initially horizontal natural fracture in the rock.

30As a separate operation, a row of wedged-shaped holes was dug at roughly to cm intervals by means of a sharp brooch or a find of punch with a very narrow sharp edge (approximately 2 to 3 mm). Into these holes, metal wedges were then driven to split the stone along the row of hides. The holes were about 15 cm deep and were cut in the shape of the metal wedges; they were shaped in such a way that the tip of the wedge would not touch the bottom of the hole, thus conveying the force laterally and with uniform pressure into the rock (fig. 8.6). The outline of the holes was carefully drawn on the rock and engraved with the head of the punch before digging them. Their digging constituted one of the most delicate operations and was certainly entrusted to very experienced quarrymen. To compensate for possible small irregularities in the digging of the holes, plaques of metal or wood may have been interposed between the metal wedges and the rock, as is done today by traditional quarrymen.

8.3.4. Megaliths dressing Tools used for dressing Three kinds of tools were used:

  1. the most effective tool is the pick, which is mainly used for the sketch and the basic works, as also for digging out the anathyrose hollows in the rising (i.e. vertical) joints. The pattern of its impacts as seen in the anathyrose hollows is similar to that seen in digging large trenches in the quarry and in rectifying the biggest irregularities before the transport of stones. Evidently the same tool was used for all these operations.

  2. the second tool used for the dressing is the punch (or pointed chisel), the head of which is very similar to the pick head.

  3. the narrow bull-nosed (or narrow round) chisel whose edge is so narrow (1 to 3 mm) that it is close to a pointed chisel. Its edge is well adapted to granite and syenite and it is also used for decoration. Rock cutting

31Three graves at Axum, all situated in the southwestern part of the massif of Mai Qoho, have been excavated in a kind of orthogneiss. Their digging was realized only by slow progressive picking away at the rock because it is almost impossible to proceed by extraction of blocks. Orthogneiss is much easier to dig than syenite, but it is very difficult to obtain a precise result. Only Bazen Tomb was finished; the other two were given up well before the end of their digging. Nevertheless, the unfinished tomb situated some meters to the northwest of Bazen Tomb reveals a real effort to obtain a precise and geometrically correct result, in particular as regards the curve of its arches. Megalith dressing (fig. 8.8)

32The process of wedge-splitting a huge stone lengthwise in the quarry to produce a crudely squared stela yielded not only the stela itself (the long central section of the stone) but also long outer sidepieces as enormous work chips. These were cut only on one side; their other sides retained the natural curve of the original stone. Like the main stelae, some of these work chips were also transported from the quarry to the Stela Field to be set up in the same way as the carefully squared stelae. Some examples are visible in the center of the Stela Field, but this category of rudimentary stelae is especially well represented on the west of the Stela Field (fig. 8.8).

33After the main stelae had been cut after a basic squaring in the quarry, it was necessary to give them their particular geometrical shape, namely an extremely elongated pyramid. Achieving the slope of the sides is the most delicate part of the work; any error of planning would be very visible and would break the physical and esthetic balance of the monument. As a guide, the stone-cutters would only have had a series of marks distributed along the length of the stela.

34Stela 3 was decorated only on three of its four sides; its northern face remained in the intermediate state between the crude squaring in the quarry and the realization of decorations. This aspect of the northern side is visible only in very oblique light, in particular at sunset. We can see on this stela more or less horizontal bands, about 5 to 10 cm wide and spaced approximately 60 cm apart. In the lower part of the stela, where the spaces between two horizontal bands are the widest, they are vertically divided by vertical bands similar to the horizontal ones. But these vertical bands occur only once or twice and never directly on the main axis of the stela. Such bands are well known elsewhere at various periods, in particular to guide the cutting of the edges, also to facilitate the final dressing. The dressing of the ashlar blocks

35This type of dressing concerns essentially the tombs and some remains of the platform of the primitive church of Axum. To a lesser degree, it is present in the covering slab stones above the side corridor at Nefas Mawcha. Facings are cut in the same way as with the monoliths by gradually densifying the impacts of the punch, to obtain a fine pointed dressing. But, lacking the use of the common chisel to refine edges, the stone-cutters were often forced to cut them with the edges slightly rounded.

36The vertical rising joints are very carefully dressed only at the contact zone between two adjacent anathyrose frames. This frame usually has only three well-formed sides (anathyrose bands) or frequently just two. The maximal width of the anathyrose bands is rather variable. The widest, observed on the vertical rising joints of the covering slabs of the corridors of Nefas Mawcha, are close to 10 cm, but much smaller at the Tomb of the False Door.

37Despite the unclarity which persists regarding the techniques of dressing hard stone in the neighboring areas of the Axumite kingdom, rather more similarities emerge with the North (Egypt, Syria and Jordan) during Roman and Byzantine times than with Yemen. Given the paucity of studies on the techniques of stone cutting and dressing, however, this possible northern influence at Axum can only be presented as a hypothesis, which future research will confirm or perhaps disprove.

8.4. Monumental architecture at Axum

  • 8 Munro-Hay 1989: 161.

38Axumite architecture is notable particularly for the monumental size of the monoliths used for stelae or tomb construction. The architects seem to have had a penchant for gigantism, best represented by the stelae and some of the tombs, but also expressed in a different way by the raising of important buildings on high podia reached by flights of steps for access. According to S. C. Munro-Hay8, it is not unlikely that this architectural arrangement was functional, designed to supplement in some way the royal ritual; splendid staircases ascending to what seem to have been open halls in an isolated central position, possibly for important reception occasions, are known from several sites.

39Axumite architecture could well be the result of a gradual incorporation of various regional styles as the kingdom expanded. The resultant unified style would thus have been a late fusion of regional traditions with, additionally, some imported innovations arriving at different periods. Arriving at different periods. Regional differences may also be expected according to the construction materials locally available in different areas (no granite being available in the Kaskase-Tokonda area).

40The buildings erected during the pre-Axumite Ethio-Sabaean period undoubtedly exercised some influence on Axumite architecture. These were the first examples of monumental architecture in Tigray. The architecture of Yeha, one of the most prominent sites for pre-Axumite monumental construction, exhibits a considerable variation in building styles. Starting from the 8th BC Grat Beal Guebri/Yeha high podium with an internal grid of foundation walls, supporting walls with wood reinforcements were ancestral to all buildings throughout the Ethio-Sabaean period and later into the Axumite period. In this case, there is proof of a long continuity in public architecture. On the other hand, monumental stone propylons, as in the Yeha temple, seem to have been abandoned in Axumite architecture. Recessed wooden gates were probably used in the late palaces (Ta’akha Maryam, etc.), rather similar to South Arabian ones.

8.4.1. Architectural layouts

41Stelae are known in great variety, from simple unworked stone slabs about a meter in height to huge and elaborately carved stelae like stelae 2 and 3, some 20 to 23 m high. Six of the stelae are carved in a representation of multi-storey buildings. Some stelae are said to depict buildings up to ten storeys. But except for information given in a very few inscriptions, there is no evidence that Axumite palaces had more than 4 to 5 storeys, and South Arabian palaces would have had the same number.

42Some characteristics of the multi-storeyed stelae can be emphasized (fig. 8.9 and 8.11). Their cross-sections show central recesses on two, three or four sides resembling those of the ground-plans of Axumite buildings. At the first level, a false door provides a faithful representation of a real wooden doorway, with the projecting heads of crosspieces and with a lock or ring-handle. The upper floors show recessed horizontal bands representing the long wooden beams, above each of which are the projecting ends of crosspieces. Rows of windows are clearly marked, their outer frames similar in construction details to those of the doorways, with internal divisions. The uppermost windows are shorter.

  • 9 Munro-Hay 1989: 104-113.

43Some of the tombs in the Stela Field are of monumental dimensions: the Mausoleum is ca. 18.3 m x 16 m, the Tomb of the Brick Arches is ca. 14 x 12 m (excluding the adit), the Tomb of the False Door is ca 12 m x 3 m. The Mausoleum (fig. 8.10) is entered through a monumental portal carved in a single slab of granite (2.55 m by 1.93 m) resembling the configuration of false doors seen on the stelae. The aperture, which is open, is 1.799 m high and 0.94 m wide. The west portal is a second monolith nearly identical to that preserved at the eastern end. The covering slabs average over 1.0 m in width, but some (inside chamber K) are up to 2.80 m wide. The Tomb of the False Door9, ca. 12 m square and 2.8 m high, is a massive construction of a near-isodomic masonry; the courses are of slightly different heights. A slab carved with a false door is more than 7 m x 5 m. In front of this door is a paved courtyard. The large blocks were carefully dressed and fitted, those comprising the substructure being exceptionally massive. North of Axum, the two tombs of King Kaleb and Gabra Maskal, basically similar, are approached by a stepped adit, fully roofed, constructed of huge, carefully dressed granite blocks of irregular (trapezoidal) shape.

44In conclusion, the colossal expenditure of effort in producing tombs and stelae demonstrates the resources, power and authority of the Axumite ruling elites. The safeguarding of the contents of the tombs depended on massiveness of construction.

8.4.2. Arches: shared innovations in South Arabia and Axum

  • 10 Yule 2007: fig. 104.
  • 11 Yule 2007: fig. 92.

45The architectural element, the arch, was unknown in archaic South Arabia. When was it introduced? Probably not earlier than the second or first century BC. Architectonic arches are few in number. In Zafâr, Masnaat Mariya (fig. 8.12), an arch was discovered which is a monolith, over 2 m in width, decorated in fine relief with birds of prey and below them monograms conveying personal and tribal names10. Some rare fragments of voussoirs come from the Himyaritic areas: a famous one, 60 cm long, depicting a royal lion hunt and a monogram, is dated 4th century AD11.

  • 12 Yule 2007: fig. 98.
  • 13 Simpson 2002: n° 142.

46Arches are mainly known from their depictions on stelae or slabs. In Himyar, in Bayt al-Ashwal, a relief displays four arches supported by columns, each niche being decorated by the bust of a young woman12 (fig. 8.13). Late funerary stelae are often decorated with images of arches resting on columns with capitals, either plain or decorated with geometrical patterns, acanthus leaves, or flowers. Arches decorated with flat strips, twisted ropes, low summit arches, and ogived arches form a vast iconographic repertory (fig. 8.14). Late stelae from Jawf and from Tamna‘, dated ist-3rd centuries AD, belong to this category13.

47Rock paintings in Jarf al-Yahûdî, recently discovered 50 km east of Sanaa, depict an urban landscape with tower-houses and arched gates to the city. The two gates depicted consist of a high arch supporting one floor and an upper terrace with crenels. The height of the arch can be inferred from the size of the warriors depicted below it (6-7 m high?). It is noteworthy that these paintings, dated ca. 1st-3rd cent. AI), are the first depictions of an arched city gate.

48Arches are a very characteristic feature of Axumite architecture. Arches might have been in use at Axum from the first quarter of the fourth century. Three tombs, the Tomb of the Brick Arches, the Brick Vaulted Structure and the Mausoleum Brick Arch contain brick arches.

49The first one (Tomb of the Brick Arches) contains three horseshoe-shaped brick arches (fig. 8.15), the outer one with a relief slab above and resting on slate shelves with mud-mortared walls below. Arch n° 1 gives access to the tomb, the aperture being between 1.00 m and 1.30 m wide and 2.57 m in height. The arch, 1.00 m deep, is constructed of fired, mould-made, red-brown clay bricks (20 x 20 x 7 cm) set in lime mortar. The other arches were simply built into the cuttings of the rock, rough stone infilling around them where necessary; stone blocking of similar type was used to block the arches. The bricks employed in the arches were sometimes laid lengthwise and sometimes sideways.

50The arches of the Tomb of the Brick Arches are extremely similar to the brick-arched entrance to chamber R of the Brick Vaulted Structure, and almost certainly, like the latter, are to be dated to around the first half of the fourth century.

8.5. Architecture and statues in South Arabia and Tigray

51In the lowlands of Yemen, recent archaeological discoveries emphasize the importation of architecture and decoration, from the beginning of our era. Some such elements have been discovered at Shabwa and Tamna‘.

8.5.1. The so-called Temple of Shabwa (fig. 8.16)

52In Shabwa, the building erected against the al-‘Aqab outcrop which dominates the southernmost end of the main street is often presented as “the Main Temple”. It is commonly considered to have been dedicated to the main god of Hadhramawt, Siyân dhû-Ilim. This attribution to Siyân is too facile not to raise questions. On the one hand, new reconstructions call into question the nature of this building; on the other hand, its architectural design does not fit in the series of South Arabian temples.

53In its last phase, the building displays several terraces set up in front of a rear wall, almost 19 m long and east-west oriented, and connected by staircases. The role of these staircases seems essential in this building, as the terraces are nearly 6 m high. In its western part, a 7.50 m wide staircase is topped by four stone bases, east-west aligned; each is preceded by a smaller base for a statue. This staircase was flanked by one monumental stone base which doubtless supported an enormous statue. A certain number of elements allow preliminary reconstructions. Most notably, the main stone base, 3 m long by 2.40/2.60 m wide, displays two divergent grooves 2.20 m long, which have the shape and configuration of two large human feet.

54The Periplus mentions the trade in statues for the court of the king of Shabwa. Unable to identify with precision the origin of the statues which were found in Shabwa, we are reduced to informed speculation. The monumental statue in the main temple could have been made in separate parts imported and then assembled together on a wooden frame. This building technique is well attested in the Greek and Roman world. In the Himyaritic kingdom, two important bronze statues, some 2.40 m high, came to light in 1931 in Nakhlat al-Hamra. King Dhamar Ali Yuha’birr and his son, co-regent Tharân, kings of Saba and Dhû-Raydân, are represented heroically nude. These two statues illustrate in a striking way the relation with Syria, to which Yemen was bound by tight economic links.

55South Arabia provides some examples of monumental statues. Unfortunately, only some faces are preserved. One is a head, originally entirely gilded, 20.8 cm high, part of a bronze statue from Ghaymân and dated ca. 2nd century AD (fig. 8.18). It was found among statues of horses and riders but the original building where they were erected is unknown. There is a large rearing horse—originally with a rider—in the Dumbarton Oaks collection which appears to have been found in 1929 (fig. 8.19). Part of a second bronze horse was found in 1931, along with a second bronze head. The heads could not belong to the missing riders, given their size. These heads, very idealised, are not portraits, and may represent gods rather than people.

  • 14 Krencker 191 3: 81-84.

56In the city of Axum, probably in one of the courtyards of Enda Sem’on, a monolith was discovered (1.65 m by 1.36 m, and 0.30 m thick) topping a 1.37 m high stone base14. This monolith likewise has grooves in the shape of two long feet, pierced with two sockets for the heels and two others at the end of each foot (fig. 8.20).

  • 15 Breton & Darles 1998: fig. 22.

57At Shabwa as at Axum, the grooves for the feet seem disproportionate with the (supposed) height of the statue but ensure the stability of its base. At Shabwa the layout of the mortises and the metal clamps makes it possible to restore a monumental bronze statue with an interior wooden framework15. This could be a human figure but not an animal one.

  • 16 Breton & Weber 2009: 224-228.

58At Shabwa, the lower eastern terrace comprises four thick slabs provided with clamps and dowels for hooves. Taking into account Hellenistic archaeological and artistic elements of the first centuries AD, we have proposed to reconstruct riders rather than bulls16. The recent discovery of two bronze horse statues on both sides of a monumental staircase, in a late annex of the Awâm temple in Mâ’rib, provides key elements of comparison.

59This Shabwa building does not fit anyway in the series of the sanctuaries of Hadhramawt. On the other hand, it can be compared with monuments recently discovered in the steppes of eastern Syria, belonging to the very beginning of the Roman period. These monuments consist of a high podium topped with monumental statues in basalt, a king associated with horses, chariots drawn either by deer or by lionesses, and riders.

60One of these Syrian monuments is the Arab sanctuary at Sahr al-Leja (fig. 8.21). On the eastern face of the base stands the statue of a life-size rider dressed in a tunic and a heavy coat. In front of him, riders of a much smaller size occur in a line. The monument of Sahr is far from unique in southern Syria. At Trachonitide, at the sanctuaries of Menara/Henu, at Sha’rah, at Dar’a, at Masmiyia in Shaqqa, there also stand statues of riders, horses and carts pulled by lions. The large number of riders and soldiers seems one of the characteristics of these temples at the edge of the Syrian desert.

  • 17 Breton & Weber 2009: 228-231.
  • 18 Ghirshman 1976.

61At Shabwa, the statues were set up in a line on the lower terrace as on the eastern side of the podium at Sahr. As inscriptions are missing, it is difficult to identify the statues. But according to Th. Weber17, it is possible that the equestrian statues at Shabwa were set up under the influence of Syrian models attested at Sahr. Such monumental statues erected on open-air terraces are, however, unknown in the religious context of South Arabia; they give to the visitor the impression of a public representation. In the Seleucid, Parthian and Sassanid Empires some sanctuaries similarly consisted of high terraces supporting monumental statues18.

8.6. A unique decor from South Arabia without any equivalent in Tigray (3rd century AD)

62The “royal palace” at Shabwa displays a unique architectural and artistic program.

  • 19 Seigne 1992: 113-127.

63After a first destruction, the palace was rebuilt on its old foundations. In the west wing, certain rooms were refitted (fig. 8.23). Furthermore, the entire building was raised by one additional storey, which was now reached by two symmetrical placed stairways located in the corners of the courtyard. On the ground floor of the western main building, the installation of a staircase implies a reorganization of the patterns of movement in this wing of the portico. The portico was now covered by an upstairs gallery, supported by wooden octagonal pillars resting on the stone base of the portico19 (fig. 8.22). The upstairs gallery opens onto the central courtyard through windows, each provided with a central stone mullion topped with a two-faced capital. All these decorations, both stone capitals and painted panels (see below), can be assumed to date to the mid-3rd century AD.

  • 20 Dentzer-Feydy 2009: 132-137. Simpson 2002: 151, n° 194.

64This architecture and this decoration show a curious mixture of Eastern and Greek traditions. Let us consider initially the mullion. Porticoes surrounding a central courtyard become frequent in the Hellenistic period all around Mediterranean. But at Shabwa, the layout seems an exclusively local tradition. The octagonal pillars and the capitals in the shape of a reversed truncated pyramid are neither Hellenic nor Graeco-Roman. The capital is partly decorated with vine scrolls, vine leaves and a central winged lion (fig. 8.24). The motif of a winged lion (not a griffin) with one raised forepaw standing in front of an urn derives from Hellenistic art. It is related to depictions on third-century AD Roman architecture at Palmyra in Syria, Parthian cornices from Hatra and Niniveh (in Mesopotamia)20.

65In the “royal palace” at Shabwa, on the first floor of building “B”, the walls of the porticoes were decorated with paintings on panels. Each panel, inserted between the wooden beams of the framework, measured 36-41 cm wide, which is the normal interval between two posts. Less than a half-dozen such panels could be reassembled from thirteen fragments of painted panels.

66A small panel depicts the bust of a young woman with a high headdress (fig. 8.25). The veil which envelops her deploys to the right over a brown pyramidal ornament. It is front view, with the eyes turned to the right. The garment sweeps back in a symmetric way, rising both to the left and to the right. The fibula visible at the right should belong to a second bust; the drapery at the middle of its breast is very recognizable. Indeed, according to multiple examples, the fibula is always attached on the right shoulder; that of Shabwa thus belongs necessarily to a second character placed to the right of the first one, whose own fibula does not appear. The image thus depicts a couple.

67Thus far, no paintings have been discovered at Axum; if there were any climatic conditions (rainfall and humidity) could be responsible for their disappearance. Or perhaps no Syrian or Middle Eastern artist was ever invited to Axum?


1 Bowersock 2014: 8-17.

2 Gajda 2009.

3 Yule 2007: 120.

4 Rathjens 1953, fig. 26.

5 Radt 1973: pl. 9-10.

6 Radt 1973: pl. 8-9, n° 23-24.

7 Simpson 2002: 152.

8 Munro-Hay 1989: 161.

9 Munro-Hay 1989: 104-113.

10 Yule 2007: fig. 104.

11 Yule 2007: fig. 92.

12 Yule 2007: fig. 98.

13 Simpson 2002: n° 142.

14 Krencker 191 3: 81-84.

15 Breton & Darles 1998: fig. 22.

16 Breton & Weber 2009: 224-228.

17 Breton & Weber 2009: 228-231.

18 Ghirshman 1976.

19 Seigne 1992: 113-127.

20 Dentzer-Feydy 2009: 132-137. Simpson 2002: 151, n° 194.

© Centre français des études éthiopiennes, 2015

Conditions d’utilisation :


Volume papier