Aksum: An African Civilisation of Late Antiquity
by Dr. Stuart Munro-Hay.
© 1991 by Dr. Stuart C. Munro-Hay.
Online edition with the author's permission, by Alan Light (alight@vnet.net).
Chapters 4-5.
Back to Table of Contents.

4.   Aksumite History

A useful division for the study of the history of Aksum, in our present state of knowledge, is to separate the historical sequence into a number of periods based as far as possible on the coinage. The latter (Ch. 9) is the best criterion we have for suggesting a chronology. Since the history of Aksum obviously overlaps the issue of the coinage at both ends, the following divisions have been employed.

Pre-Aksumite. Northern Ethiopia before the rise of Aksum.
1. Early Aksum until the reign of Gadarat. 1st and 2nd centuries AD.
2. Gadarat to the first issues of coinage under Endubis. 3rd century AD until c270.
3. The Pagan Kings; Endubis to Ezana. c270-c330AD.
4. Ezana (after his conversion) to Kaleb. c330-c520AD.
5. Kaleb to the end of the coinage. c520-early seventh century AD.
6. The Post-Aksumite period. From the early seventh century AD.
The period begins with the reign of Ashama ibn Abjar and continues until the accession of the Zagwé dynasty c1137.

1. The Pre-Aksumite Period

This period is not of major concern to us here, and in any case we have very little information about it; but some consideration should be given to the situation in Ethiopia before the rise of Aksum, since the source of at least some of the characteristics of the later Aksumite civilisation can be traced to this earlier period. Perhaps the most interesting phenomenon in this respect is that by around the middle of the first millenium BC — a date cautiously suggested, using palaeographical information (Pirenne 1956; Drewes 1962: 91), but possibly rather too late in view of new discoveries in the Yemen (Fattovich 1989: 16-17) which may even push it back to the eighth century BC — some sort of contact, apparently quite close, seems to have been maintained between Ethiopia and South Arabia. This developed to such an extent that in not a few places in Ethiopia the remains of certain mainly religious or funerary installations, some of major importance, with an unmistakeable South Arabian appearance in many details, have been excavated. Among the sites are Hawelti-Melazo, near Aksum (de Contenson 1961ii), the famous temple and other buildings and tombs at Yeha (Anfray 1973ii), the early levels at Matara (Anfray 1967), and the sites at Seglamien (Ricci and Fattovich 1984-6), Addi Galamo, Feqya, Addi Grameten and Kaskase, to name only the better-known ones. Fattovich (1989: 4-5) comments on many of these and has been able to attribute some ninety sites altogether to the pre-Aksumite period.

Inscriptions found at some of these sites include the names of persons bearing the traditional South Arabian title of mukarrib, apparently indicating a ruler with something of a priest-king status, not otherwise known in Ethiopia (Caquot and Drewes 1955). Others have the title of king, mlkn (Schneider 1961; 1973). Evidently the pre-Aksumite Sabaean-influenced cultural province did not consist merely of a few briefly-occupied staging posts, but was a wide-spread and well-established phenomenon. Until relatively recently South Arabian artefacts found in Ethiopia were interpreted as the material signs left behind by a superior colonial occupation force, with political supremacy over the indigenes — an interpretation still maintained by Michels (1988). But further study has now suggested that very likely, by the time the inscriptions were produced, the majority of the material in fact represented the civilisation of the Ethiopians themselves. Nevertheless, a certain amount of contact with South Arabia is very apparent, and had resulted in the adoption of a number of cultural traits (Schneider 1973; 1976).

Evidently the arrival of Sabaean influences does not represent the beginning of Ethiopian civilisation. For a long time different peoples had been interacting through population movements, warfare, trade and intermarriage in the Ethiopian region, resulting in a predominance of peoples speaking languages of the Afro-Asiatic family. The main branches represented were the Cushitic and the Semitic. Semiticized Agaw peoples are thought to have migrated from south-eastern Eritrea possibly as early as 2000BC, bringing their `proto-Ethiopic' language, ancestor of Ge`ez and the other Ethiopian Semitic languages, with them; and these and other groups had already developed specific cultural and linguistic identities by the time any Sabaean influences arrived. Features such as dressed stone building, writing and iron-working may have been introduced by Sabaeans, but words for `plough' and other agricultural vocabulary are apparently of Agaw origin in Ethiopian Semitic languages, indicating that the techniques of food-production were not one of the Arabian imports. Clark (1988) even suggests that wheat, barley, and the plough may have been introduced from Egypt via Punt. Some of the graffiti found in eastern Eritrea include names apparently neither South Arabian nor Ethiopian, perhaps reflecting the continued existence of some older ethnic groups in the same cultural matrix. Various stone-age sites and rock-paintings attest to these early Ethiopians in Eritrea and Tigray. At Matara and Yeha, for example, archaeologists have distinguished phases represented by pottery types which seem to owe nothing to South Arabia, but do have some Sudanese affinities. The Italian archaeologist Rodolfo Fattovich, who has particularly interested himself in this study, has suggested that the pre-Aksumite culture might owe something to Nubia, specifically to C— group/Kerma influences, and later on to Meroë/Alodia (Fattovich 1977, 1978, 1989). Worsening ecological conditions in the savanna/Sahel belt might have induced certain peoples to move from plains and lowlands up to the plateau in the second half of the second millenium BC (Clark 1976), bringing with them certain cultural traditions. Evidence for early trade activity to regions across the Red Sea from eastern Sudan and Ethiopia at about this time has been noted by Zarins (1988), with reference to the obsidian trade. Extremely interesting results have lately come from work in the Gash Delta on the Ethiopo-Sudanese borderland, indicating the existence of a complex society there in the late 3rd-early 2nd millenium BC (Fattovich 1989: 21); possibly the location of the land of Punt there reinforces this suggestion (Kitchen 1971; Fattovich 1988: 2, 7). It seems that the new discoveries are of major importance to an understanding of the dynamics of state formation in the Ethiopian highlands. The latest work suggests that in the late second and early first millenium BC the eastern part of the Tigray plateau was included in a widespread cultural complex on both the African and the Arabian Tihama coasts of the Red Sea, in contact with the lowlands of the Sudan and perhaps with the Nile Valley, while the western part was in contact with peoples of the Gash Delta. These two regions of the plateau later became united culturally and politically under the D`MT monarchy (Fattovich 1989: 34-5).

It appears that there were undoubtedly some South Arabian immigrants in Ethiopia in the mid-first millenium BC, but there is (unless the interpretation of Michels is accepted) no sure indication that they were politically dominant. The sites chosen by them may be related to their relative ease of access to the Red Sea coast. Arthur Irvine (1977) and others have regarded sympathetically the suggestion that the inscriptions which testify to Sabaean presence in Ethiopia may have been set up by colonists around the time of the Sabaean ruler Karibil Watar in the late fourth century BC; but the dating is very uncertain, as noted above. They may have been military or trading colonists, living in some sort of symbiosis with the local Ethiopian population, perhaps under a species of treaty-status.

It seems that the pre-Aksumite society on the Tigray plateau, centred in the Aksum/Yeha region but extending from Tekondo in the north to Enderta in the south (Schneider 1973: 389), had achieved state level, and that the major entity came to be called D`MT (Di`amat, Damot?), as appears in the regal title `mukarrib of Da`mot and Saba'. The name may survive in the Aksumite titulature as Tiamo/Tsiyamo (Ch. 7: 5). Its rulers, kings and mukarribs, by including the name Saba in their titles, appear to have expressly claimed control over the resident Sabaeans in their country; actual Sabaean presence is assumed at Matara, Yeha and Hawelti-Melazo according to present information (Schneider 1973: 388). The inscriptions of mukarribs of D`MT and Saba are known from Addi Galamo (Caquot and Drewes 1955: 26-32), Enda Cherqos (Schneider 1961: 61ff), possibly Matara, if the name LMN attested there is the same as the .MN from the other sites, (Schneider 1965: 90; Drewes and Schneider 1967: 91), Melazo (Schneider 1978: 130-2), and Abuna Garima (Schneider 1973; Schneider 1976iii: 86ff). Of four rulers known to date, the earliest appears to be a certain W`RN HYWT, who only had the title mlkn, king, and evidence of whom has been found at Yeha, Kaskase, Addi Seglamen; he was succeeded by three mukarribs, RD'M, RBH, and LMN (Schneider 1976iii: 89-93).

Illustration 9. An inscription from Abba Pantelewon near Aksum, written in the Epigraphic South Arabian script and mentioning the kingdom of D`MT; it is dedicated to the deity Dhat-Ba`adan. It has been photographed upside down. Photo BIEA.

The Sabaeans in Ethiopia appear, from the use of certain place-names like Marib in their inscriptions, to have kept in contact with their own country, and indeed the purpose of their presence may well have been to maintain and develop links across the sea to the profit of South Arabia's trading network. Naturally, such an arrangement would have worked also to the benefit of the indigenous Ethiopian rulers, who employed the titles mukarrib and mlkn at first, and nagashi (najashi) or negus later; no pre-Aksumite najashi or negus is known. The inscriptions dating from this period in Ethiopia are apparently written in two languages, pure Sabaean and another language with certain aspects found later in Ge`ez (Schneider 1976). All the royal inscriptions are in this second, presumably Ethiopian, language. A number of different tribes and families seem to be mentioned by the inscriptions of this period, but there is no evidence to show whether any of these groups lasted into the Aksumite period. Only the word YG`DYN, man of Yeg`az, might hint that the Ge`ez or Agazyan tribe was established so early, though the particular inscription which mentions it is written in the South Arabian rather than the Ethiopian language (Schneider 1961). Some of the other apparently tribal names also occur in both groups of inscriptions. The usual way of referring to someone in the inscriptions is `N. of the family N. of the tribe N.', possibly also reflected later by the Aksumite `Bisi'-title; `king N. man of the tribe/clan (?) N.' (Ch. 7: 5).

It seems that these `inscriptional' Sabaeans did not remain more than a century or so — or perhaps even only a few decades — as a separate and identifiable people. Possibly their presence was connected to a contemporary efflorescence of Saba on the other side of the Red Sea. Their influence was only in a limited geographical area, affecting the autochthonous population in that area to a greater or lesser degree. Such influences as did remain after their departure or assimilation fused with the local cultural background, and contributed to the ensemble of traits which constituted Ethiopian civilisation in the rest of the pre-Aksumite period. Indeed, it may be that the Sabaeans were able to establish themselves in Ethiopia in the first place because both their civilisation and that of mid-1st millenium Ethiopia already had something in common; it has been suggested that earlier migrations or contacts might have taken place, leaving a kind of cultural sympathy between the two areas which allowed the later contact to flourish easily. The precise nature of the contacts between the two areas, their range in commercial, linguistic or cultural terms, and their chronology, is still a major question, and discussion of this fascinating problem continues (Marrassini 1985; Avanzini 1987; Pirenne 1987; Isaac and Felder 1988).

Jacqueline Pirenne's most recent (1987) proposal results in a radically different view of the Ethiopian/South Arabian contacts. Weighing up the evidence from all sides, particularly aspects of material culture and linguistic/palaeographic information, she suggests that "il est donc vraisemblable que l'expansion ne s'est pas faite du Yémen vers l'Ethiopie, mais bien en sens inverse: de l'Ethiopie vers le Yémen". According to this theory, one group of Sabaeans would have left north Arabia (where they were then established) for Ethiopia in about the eighth or seventh century BC under pressure from the Assyrians; they then continued on into south Arabia. A second wave of emigrants, in the sixth and fifth century, would reign over the kingdom of Da'amat (D`MT), and would have been accompanied by Hebrews fleeing after Nebuchadnezzar's capture of Jerusalem; an explanation for the later Ethiopian traditions with their Jewish and Biblical flavour, and for the Falashas or black Jews of Ethiopia. These Sabaeans too, in their turn would have departed for the Yemen, taking there the writing and architecture which they had first perfected in Tigray. In the fourth and third century BC the remaining Sabaean emigrés would have left Ethiopia for the Yemen, leaving elements of their civilisation and traditions firmly embedded in the Ethiopian's way of life. This ingenious mise en scène, so far only briefly noted in a conference paper, must await complete publication before it can be fully discussed; but it is expressive of the highly theoretical nature of our conclusions about pre-Aksumite Ethiopia that so complete a reversal of previous ideas can even be proposed. Isaac and Felder (1988) also speculate about the possibility of a common cultural sphere in Ethiopia and Arabia, without giving either side the precedence.

It has also been suggested that the progress of the youthful Ethiopian state brought it into conflict with Meroë in the reigns of such kings as Harsiotef and Nastasen from the fourth century BC. Whilst there must have been some contact later, there is no real evidence from this early date (Taddesse Tamrat 1972: 12).

The altars, inscriptions, stelae, temples, secular structures, tombs and other material left by the Sabaean-influenced Ethiopian population occur in considerable numbers even from the few excavated sites; those attributed to the Sabaeans themselves occur more rarely. The monuments are dated from the 5th century BC by study of the letter-forms used on them (palaeography), and seem to appear in Ethiopia at about the same time as they do in South Arabia (nb. the reservations about the dating expressed by Fattovich 1989). The disc and crescent symbol used on some of the monuments (and very much later by the pre-Christian Aksumites) was also familiar on some South Arabian coins, and South Arabian altars; many of the same deities were being worshipped in the two regions. It was also during this period that iron was introduced into the country. In the present state of our knowledge, it is unclear how much of Aksumite civilisation was a direct continuation of a cultural heritage from pre-Aksumite times, or how much any South Arabian aspects might be better attributed to a renewal of overseas contacts in the period after the consolidation of Aksum as an independent polity in the first and second centuries AD. No clear evidence of connexions between the pre-Aksumite, Sabaean-influenced, period, and the earliest Aksumite period is at the moment available, though it seems intrinsically more likely that Aksum in some way was able to draw directly on part of the experience of its predecessors. At Matara, the archaeological evidence implies that there was a clear break between the two periods (Anfray and Annequin 1965), but this need not have been the case everywhere in the country. The solution to these questions can only await further clarification from archaeology.

The subsequent periods are those which represent the duration of the Aksumite kingdom proper. In the following table approximate dates for these periods, numbered 1-5, are indicated, together with the names of the known rulers, with notes about any references in texts or inscriptions, contemporary constructions (Ch. 16) at Aksum (using the terminology in Munro-Hay 1989), and significant international events with a bearing on Aksum.

2. Comparative Chronological Chart; Rulers, Sources and Sites

Period 1. Early Aksum until the reign of GDRT. 1st-2nd centuries AD.
Earliest platforms
Lower Stele Park levels
Zoskales? Periplus?
Platform A,
c150AD 1st platform extension
Period 2. GDRT-Endubis. Beginning of 3rd century AD to c270AD.
GDRT, BYGT South Arabian inscriptions
250AD Nefas Mawcha
Walls M2, M5, and M9
Gudit Tomb; Gudit, Southeastern, and GA stelae fields
Period 3. Endubis to Ezana before his conversion. c270AD to c330AD.
Endubis* Main Stele Field
Coinage begins
Dressed stelae
Wall M1
Decorated stelae
Tunnel Complex, East Tomb, Sh.T.B
Tomb of Brick Arches
Ezana* Inscriptions, Anza stele, Brick Vaulted Structure, Mausoleum, Matara stele.
Period 4. Ezana as a Christian to Kaleb. c330AD to c520AD.
Christian Inscriptions and coins.
2nd platform extension
Wall M'1, East wall, East steps
Anonymous Christian coins
Athanasius Apologia
Ouazebas*Walls M7 and M8
400AD Fall of Stele 1
Kaleb I-IA buildings
Eon*Tomb of the False Door
IW building
Ousas*/Ousana(s)*. Tazena
Period 5. Kaleb until the end of the coinage. c520AD to early C. 7th AD.
Kaleb* InscriptionKosmas Indikopleustes
Conquest of Yemen
Yusuf As`arTombs of Kaleb and Gabra Masqal
530ADSumyafa` Ashwa`
AbrehaAlla Amidas*
W'ZB Inscription, Ella Gabaz*
575AD Persians in Yemen
Hataz* = `Iathlia'Fill of ES building
614ADArmah*Jerusalem falls to Persia
619ADEgypt falls to Persia
End of Aksum as capital, Matara tomb
Period 6. After the end of the coinage.
630AD Death of Ashama ibn Abjar
640ADArab expedition in Red Sea, Egypt falls to Arabs
705-715ADReign of al-Walid, Qusayr Amra painting
The symbol * denotes issues of coins.

3. Period 1; Early Aksum until the Reign of Gadarat

The process of development of the Aksumite state is obscure. The earliest surviving literary references to Aksum, in the Periplus of the Erythraean Sea (Huntingford 1980) and Ptolemy's Geography (Stevenson 1932), together with some finds of early date from the site itself (Munro-Hay 1989), indicate that the city was probably established at the beginning of our era. The dates of both the above references have been disputed. It has been suggested that the Periplus could have been written in the mid-1st century (Bowersock 1983: 70; Casson 1989: 7) or even as late as the 3rd century AD (Pirenne 1961), whilst one scholar proposed that the earliest surviving versions of Ptolemy's work relay information which was continually updated until the 4th century AD (Mathew 1975: 152). However, if they are accepted as early documents, their references to Aksum do seem to be backed up by the excavation there of certain features which can be satisfactorily dated to the first and second centuries AD. These include stone-built platforms, perhaps originally laid out with some funerary purpose since they were found below the deposits later formed by the main cemetery (now the so-called Stele Park; see Ch. 16). They have been dated by radiocarbon tests on material found in associated contexts to the first two centuries AD (Chittick 1974; Chittick 1976i; Munro-Hay 1989). Among the finds from this region were fragments of glass vessels of particular types, dateable to the first century AD; and certain types of glass were actually included in the list of imports into Aksum provided by the Periplus.

If we are right in thinking of the Periplus as a probably mid-first century document, we can hope to find at Aksum evidence of the "city of the people called Auxumites" (Schoff 1912: 23) — translated by Huntingford as "the metropolis called the Axomite" (1980: 20) — which it mentions, together with a comprehensive selection of such goods as it describes as being imported into Aksum. Ptolemy, if we accept that his reference is not a later addition, leads us to expect a city with a king's palace at some time around the mid-second century AD. Archaeology has so far revealed little of this, but the early platforms and glass indicate that further evidence for the existence of the city by the first century AD may now be expected. With more archaeological excavation, other early remains apart from the platforms may be discovered. Much of the other material excavated is at the moment difficult to date reliably and so remains inconclusive.

Accepting, with the modern consensus of opinion, that the Periplus dates to the mid-first century AD, we find that at this stage Aksum, under the rule of king Zoskales, was already a substantial state with access to the sea at Adulis. Zoskales is the earliest king of the region known to us at the moment (though Cerulli 1960: 7; Huntingford 1980: 60, 149-50 and Chittick 1981: 186 suggested that he was not king of Aksum but a lesser tributary ruler). In his time there was a vigorous trading economy, and already a notable demand for the luxuries of foreign countries. The monarchy was established; and Ptolemy confirms that Aksum was the royal capital by the mid-second century AD. This period, then, saw the rise of the city into the governmental centre for a considerable area of the Ethiopian plateau and the coastal plain. Such a line of development is to be expected since by the time of king GDRT (Gadarat) Aksum had attained a position which allowed it to venture to send its armies on overseas expeditions and even establish garrisons in parts of Arabia. The fact that the Periplus does not mention Aksumites in connection with South Arabia is another feature which seems to date it before the period of Abyssinian intervention there, and indeed the Periplus notes that the power of king Kharibael of Himyar and Saba, and the tyrannos Kholaibos of the southern coastal Mopharitic region (al-Ma`afir), reaching from their capitals of Zafar and Saue, was sufficient to allow them to control Azania, the east African coast to Tanzania, and its rich trade in ivory and tortoise-shell (Casson 1989: 61, 69). Though, by the early third century, Aksum had come to dominate al-Ma`afir, and much weakened the Arabian trading system, at the time of the Periplus the Ethiopians were not in a position to reach so far, and the Arabian port of Muza seems to have been rather more important than Adulis. Gradually, during the second century, Aksum must have begun to interest itself in weakening Himyarite maritime control, culminating in its allying with Saba (see below) and seizing certain areas formerly under Himyarite rule (Bafaqih and Robin 1980; Bafaqih 1983: Ch. 3).

One uncertain but interesting hint that Aksumite power may have been increasing notably in the 1st century AD comes from accounts preserved by Seneca (ed. Corcoran 1972), the Roman writer who became the emperor Nero's tutor, and Pliny (ed. Rackham 1952: VI. 35, 184). These authors record details about an exploration (or two separate expeditions; Shinnie 1967: 21-22) in 61AD into the southern part of the Sudan. Certain Roman officers were able to penetrate as far as the great papyrus swamp region of the Sudan, the Sudd, it seems with a certain amount of help from the Meroitic king. Even in Augustus' time, according to Strabo (ed. Page 1930: 353) Aelius Gallus had been sent not only to explore Arabia, `but also in Aethiopia, since Caesar saw that the Troglodyte country which adjoins Aegypt neighbours upon Arabia, and also that the Arabian Gulf, which separates the Arabians from the Troglodytes, is extremely narrow' but this earlier effort came to nothing, it seems, since Gallus' expedition was a failure. It has been suggested that at the time of the Sudan expedition Rome, as Meroë's ally, was trying to assist in preventing the nascent Aksumite kingdom from seizing control of the routes formerly used by the Kushite monarchy's merchant caravans. Whilst we have no certain confirmation of this, there may have come a time when Meroë and Aksum clashed over their interests in the control of the Nile routes. Schur (1923) says that the emperor Nero intended to move against Aksum, and therefore sent an army to Ptolemais under Vespasian and Titus; but this can only be conjecture. Nevertheless, with the decline of Meroitic power and the fragmentation of authority in the region, Aksum would certainly have had a better opportunity for advancing its interests to the west and north than when Meroë was still a powerful state. The Meroitic relief at Jabal Qayli, close to the route leading to Kassala, where the king of Meroë is shown with slain enemies under the image of an Apollo-like deity, is the furthest actual trace of Meroitic influence to the east (Shinnie 1967: 50-51). This relief, bearing the name of king Sherkarer, is attributed to the early first century AD, but the Meroitic dates are not certainly fixed, and there is considerable leeway. It has been suggested that the distorted figures of the enemy represent slain Aksumites, but they could just as easily be depictions of any local group who had incurred Sherkarer's enmity.

4. Period 2; Gadarat to Endubis

This period may be characterised as Aksum's first `South Arabian' period, since most of the information available comes from inscriptions found in South Arabia (Beeston 1937; Jamme 1962; Robin 1981). The inscriptions name the Ethiopian kings as `nagashi of Habashat (Abyssinia) and of Aksum(an)' and are written in the old South Arabian script and language. Since there are no vowels marked, the royal names mentioned by these inscriptions actually read GDRT, `ADBH, ZQRNS and DTWNS, but for convenience here simple vowelling has been added, as for example, in the name `Gadarat'. The letters GDRT could represent a Ge`ez name such as Gedur, Gadura, Gedara or the like, but until a correctly vowelled spelling is found we remain unsure of the precise pronunciation.

The inscriptions which refer to Gadarat and `Adhebah (perhaps `Azba or `Azeba in Ge`ez), kings of Aksum and Habashat, come from the famous temple at Marib called by the Arabs `Mahram Bilqis', after the Arab name for the Queen of Sheba. Mahram Bilqis was in fact the great temple of the moon-god Ilmuqah at the ancient capital of the kingdom of Saba, now in north Yemen. Dated inscriptions, using an `era of Himyar' are now interpreted as providing a date for Gadarat around the beginning of the 3rd century AD. It was previously suggested (Munro-Hay 1984: 20) that these were fourth century rulers, on the strength of the reading of `ADBH as WDBH, identified with Wazeba (WZB), one of the earliest kings named on the coinage, but since new discoveries about the dating of the inscriptions this theory has been abandoned.

The inscriptions which mention the Aksumite rulers were written as official accounts of wars and victories by the kings of Saba and Himyar. Since these kings were usually the enemies of the Aksumites, they do not deal very often with Aksumite successes. Nevertheless, we find that the military forces of the Aksumites were in control of certain regions of the Arabian peninsula, a situation doubtless partly facilitated by the political situation in Arabia, where the rulers of both Saba and Himyar at different times called in the help of Aksumite armies against each other.

The situation is still not entirely clear, but it appears that Arabia at the end of the second century was dominated by four states, Himyar (a relatively new polity), Saba, Hadhramawt and Qataban. Somewhere between c160-210AD Qataban was annexed by Hadhramawt, while the Sabaean rulers tried to subjugate Himyar, then ruled by king Tharan Ya`ub Yuhan`im. The Sabaean king `Alhan Nahfan, son of Yarim Ayman I, and his sons Sha`ir Awtar and Yarim Ayman II allied themselves against Himyar with Gadarat, nagashi of Aksum. This latter power was probably a relatively recent arrival on the Arabian scene, interested in curtailing Himyarite trading control in the Red Sea and beyond, and its assistance helped the Sabaeans to achieve a favourable balance of power. But it also brought a new factor into South Arabian politics, not finally disposed of until the Persian conquest centuries later; the Abyssinian presence ultimately protracted the conflict between Saba and Himyar for eventual control of the entire region.

This realignment occurred in the early part of the third century. The three Sabaean kings had previously allied themselves with Yada`ub Gaylan, the king of Hadhramawt. An inscription celebrating their treaty with Aksum declares that

`they agreed together that their war and their peace should be in unison, against anyone that might rise up against them, and that in safety and in security there should be allied together Salhen and Zararan and `Alhan and Gadarat'.
In this inscription what seems to be Gadarat's castle or chief residence Zararan, is mentioned in parallel to the palace of `Alhan at Marib, capital of Saba, which was called Salhen; Zararan might even be one of the palaces whose ruins are still visible at Aksum (Ch. 5: 4).

After `Alhan Nahfan's death his son Sha`ir Awtar (whose reign seems to date from about 210 to 230AD, linked for a time in co-regency with his brother Hayu`athtar Yada`) abandoned this alliance. Frictions had doubtless begun to develop as Aksum grew more powerful in the region, and learned to play off the Arabian kingdoms and tribal allegiances against each other. By about 225AD Sha`ir Awtar had defeated and captured Il`azz Yalut, king of the Hadhramawt, and taken his capital, Shabwa. Il`azz was married to the sister of Sha`ir Awtar, and in 217-8AD the latter had helped put down a rebellion against the Hadhrami king; the enmity between Hadhramawt and Saba was a major change in policy. The Abyssinian position in these events is not clear. Sha`ir Awtar apparently used both Himyarite and Sabaean troops in this campaign, and the Himyarite ruler, Li`azz Yahnuf Yuhasdiq, whose reign may have overlapped with the end of Sha`ir's, also allied with the Sabaeans against Gadarat. Aksum suffered a defeat, and was expelled from the Himyarite capital, Zafar, which had been occupied and garrisoned under the command of a son of the nagashi, Beyga or Baygat (BYGT). However, Aksum still retained territory in Arabia in the reign of Sha`ir Awtar's successor Lahay`atat Yarkham, who had at least one clash with Habash troops. In any event, these activities, dating from perhaps the beginning of the third century to the 230's AD, are confirmation that Aksum had reached a new zenith in its power. Overseas wars, the occupation of territories in Arabia, military alliances, a fleet, and the extension of Aksumite political and military influence from the Hadhramawt to Najran in modern Saudi Arabia bespeak an important increase in the scope of the Aksumite state.

A peace may have been patched up between the contestants for a while, but it was only temporary. A little later, in the 240s, we find two rival dynasties calling themselves kings of Saba and Dhu-Raydan, one of which, represented by a certain Shamir of Dhu-Raydan and Himyar, turned for help to king `Azeba or `Adhebah (`ADBH) of Aksum, and his son Girma, Garima or Garmat (GRMT), with their allies from Sahartan and the tribe of Akk, against the Sabaean kings Ilsharah Yahdub and Yazzil Bayyin, sons of Fari`um Yanhub (who only called himself king of Saba, perhaps recognising that he was not in the same position of power as his two predecessors, who had employed the dual title `king of Saba and Dhu-Raydan'). These kings considered that Himyar, the Abyssinians, and Sahartan were in breach of a peace-treaty during the ensuing war. Shamir Dhu-Raydan was almost certainly the Himyarite king Shamir Yuhahmid, who became an ally of Aksum under `ADBH and the `son of the nagashi' GRMT. He sent for help to the nagashi, and, though one inscription claims that

"Shamir of Dhu-Raydan and Himyar had called in the help of the clans of Habashat for war against the kings of Saba; but Ilmuqah granted . . . the submission of Shamir of Dhu-Raydan and the clans of Habashat",
Shamir seems to have to some extent recovered Himyarite power. It may have been such a request for aid that eventually led the Aksumite kings to claim the much-used titles of `king of Saba and Himyar' in their own titulature, asserting some sort of theoretical suzerainty over the Arabian kings. Incidentally, it is unknown whether the two generals entitled `son of the nagashi' Baygat and Garmat were `crown princes' who succeeded to the throne in their turn, or whether they were merely military captains under the nagashis. Their names are unknown except for these inscriptions.

Around the end of the 240s until c260, the Himyarite king was Karibil Ayfa`, who fought with Yada`il Bayyin and his son Ilriyam Yadum of the new dynasty in Hadhramawt, with the Abyssinians, and with Ilsharah Yahdub and Yazzil Bayyin of Saba; all the main forces then in the Yemen. One of the al-Mis`al inscriptions (no. 3) mentions that a son of the nagashi, unfortunately unnamed, came to Zafar with the troops of al-Ma`afir and the Abyssinians, and that a sortie was made against them.

A new gap now occurs in the records. Possibly it may be filled by one of the most mysterious of the Aksumite kings, Sembrouthes (Littmann 1913: IV, 3). He is known only from his Greek inscription from Daqqi Mahari, well north of Aksum in present-day Eritrea. The inscription is on a roughly shaped building block, and, for so brief a text, is filled with expressions of the royal self-esteem;

"King of kings of Aksum, great Sembrouthes came (and) dedicated (this inscription) in the year 24 of Sembrouthes the Great King".
His substantial reign of at least 24 years, if correctly placed here, fills the period between the last mention of `Adhebah and the next known Aksumite rulers, DTWNS and ZQRNS. Himyarite power was growing stronger throughout this period, and perhaps to curb this Aksum decided to act; in c267-8 Yasir Yuhan`im of Himyar (c260-270) suffered an invasion led by two Aksumite kings. More South Arabian inscriptions, recently brought to notice by Christian Robin, (whose dating of the South Arabian eras, and general historical scheme (1981) we have followed here) come from the Yemeni site of al-Mis`al. One inscription (no. 5) deals with this war in which the two kings of Aksum, Datawnas (DTWNS) and Zaqarnas (ZQRNS), with their allies of al-Ma`afir, were involved. Whether these were co-rulers, or successive occupants of the Aksumite throne, is not certain, but they appear to have renewed or continued the Aksumite presence in South Arabia sometime during the years between 260 and 270 AD. The results of their efforts remain unclear; when the al-Mis`al inscriptions are fully published more may be known about the events of this period, but the fact that the Aksumite kings were still interfering in Arabian politics indicates that their interests in South Arabia were not lightly abandoned. An inscription of the last Sabaean king, Nashakarib Yuha`min Yuharib, also mentions Abyssinian incursions at this time, but it is notable that accounts of his wars in Sahartan do not mention the Abyssinians.

Illustration 10. The Greek inscription of the king of kings Sembrouthes of Aksum, from Daqqi Mahari, Eritrea (courtesy of G. Tringali).

The subsequent events, culminating in a Himyarite victory over Saba, are conjectured to be more or less contemporary with the Aksumite kings Endubis and Aphilas, and are detailed below (Ch. 4: 5).

Sadly, nothing is known of these Aksumite kings of the third century from the Ethiopian side except for the discovery at Atsbi Dera of a sceptre or wand in bronze, which mentions the name of `GDR negus of Aksum' (Caquot and Drewes 1955: 32-38; Doresse 1960). This appears in a short inscription which has been translated as either "GDR king of Aksum occupied the passages of `RG and LMQ", or "Gedara, King of Axum is humbled before the [gods] Arg and Almouqah" (Jamme 1957). GDR is very likely the same king called by the Arabian kings GDRT (Gadarat). In addition, some finds of Himyarite coins at Aksum may be attributed to this overseas intercourse (Munro-Hay 1978).

The stelae of two prominent Ethiopians of the late third century offer a little information about local matters (Drewes 1962: 67-8). One, the Matara stele, reads `This is the stele which Agaz has made for his ancestors . . .', but no information is given about Agaz himself. The other, the Anza (near Hawzien) stele, was erected by Bazat (BZT) negus of Agabo, perhaps a local king. His stele seems to celebrate a 15 day festival, and 520 containers of beer and 20,620 loaves are recorded as a donation.

The Aksumite state at this stage appears fully-fledged as a militaristic monarchy with wide-reaching foreign connexions. The interest in South Arabia may have been encouraged by the need to keep the Red Sea efficiently policed so that vessels of the Aksumites or their trading partners could come and go safely. Aksum may also have been concerned to be included in the enormously profitable trade in incense and other valuable goods along the routes which crossed Arabia to the markets of the Roman empire. Sembrouthes' inscription attests Aksumite power as far north as Daqqi Mahari, and confirms that he controlled subordinate kings, since he uses the title `king of kings'. His inscription is in Greek, the language Zoskales also knew. It remains possible that Sembrouthes should be situated at an earlier date, though the elevated title of `king of kings' does perhaps tend to support the dating proposed above.

5. Period 3; Endubis to Ezana

From the reign of king Endubis we are fortunate in having the newly issued coinage, in gold, silver, and bronze, to guide us in tracing out a framework for the history and chronology of Aksum (Munro-Hay, loc. var.; Hahn 1983). The issue of a coinage (Ch. 9) is of very great importance in itself, and for Aksum the issue of an independent gold-based currency was a move which announced that the state considered itself on a par with its great neighbours at least in so far as sovereignty was concerned. It further enabled the rulers to employ a powerful propaganda instrument, simplified trade, and, not to be forgotten, was profitable.

As far as publicising themselves and their state was concerned the Aksumite rulers were highly successful from our point-of-view; most of the Aksumite kings are known to us only from the legends on their coins, all other evidence for their existence having perished or disappeared among the ruins of Aksum. The main features and significance of the coinage are dealt with in Ch. 9 below. From the evidence presented through study of the coinage (Munro-Hay 1978) it can be inferred that Endubis employed the Roman monetary system as a model, but used his own selected designs to maximise the impact of his coinage as a vehicle to convey the official propaganda. The subsequent kings added or removed motifs and other elements of the design as the current situation recommended.

Illustration 10a. Drawing of a silver coin (d. 12mm) of king Wazeba with its alternative reverses, the right-hand example belonging to king Ousanas and perhaps indicating joint tenure of the throne.

A new title, not met with before in Aksumite records, first appears in the coin-legends of the pagan rulers. This consists of the word `Bisi', from `be`esya', `man of . . .' in Ge`ez, followed by a name. It could be perhaps a tribal or clan designation, or perhaps a military title, and it remained in use until the sixth century AD (see Ch. 7: 5), and possibly even on into the eleventh and later centuries (Conti Rossini 1901).

Endubis and his successors all included the pre-Christian disc and crescent symbol on their coins, until, with Ezana's conversion in c333AD it was replaced by the cross. This enables us to group the five kings Endubis, Aphilas, Wazeba, Ousanas and Ezana at the head of the coinage sequence. Although the first four of these pre-Christian kings are not mentioned anywhere else, the archaeological record, in so far as it can be interpreted, almost certainly leads to the conclusion that at least some of them were responsible for the erection of the series of large decorated stelae in the central necropolis of the capital (Ch. 5: 5). Some of the tombs marked by these stelae must also be theirs, but in most cases the tombs belonging to the various stelae have not yet been identified. Very little political information can be extracted from the coins for this period, but it may be that Wazeba and Ousanas ruled for a time conjointly (see Ch. 7: 3), since there is one issue which combines obverse dies of Wazeba with reverse dies of Ousanas. The scarcity of Wazeba's coins may hint at a short reign. His unique use (at this period) of Ge`ez for his coinage, instead of the usual Greek, may betray an interest in encouraging the use of the coinage in Ethiopia itself, rather than mainly for external trade.

It may have been during the reigns of Endubis or Aphilas that the last events we know of during the first Abyssinian involvement in Yemen occurred. By the 270s Yasir Yuhan`im of Himyar and his son and co-ruler Shamir Yuhar`ish seem to have ended the Abyssinian danger, and, in addition, to have triumphed to such an extent that they could annex Saba itself. About 290AD Hadhramawt fell in its turn, and Shamir Yuhar`ish adopted, by 295, the longer title of king of Saba, Dhu-Raydan, Hadhramawt, and Yamanat. If the Ethiopians retained territory on the east side of the Red Sea, it must have been at most some minor coastal districts; at any rate, the inscriptions of Shamir no longer mention them.

In the fourth century, after the reign of Shamir Yuhar`ish, another South Arabian inscription alludes to Karibil Watar Yuhan`im, king of Saba, Dhu-Raydan, Hadhramawt and Yamanat, sending ambassadors to the "land of Habashat and Aksuman, to the nagashi . . . and he (the nagashi?) sent with him as emissaries `HQM and ZLNS".

Illustration 11. A gold coin (diameter c. 18mm) of king Ousanas of Aksum with the pre-Christian disc and crescent symbol above his head.

Ousanas seems very likely to have been the king to whom the two captive Tyrian boys, Frumentius and Aedesius (see Ch. 10: 2), were brought after the killing of their shipboard companions. This king is called Ella Allada or Ella A'eda in the traditional account, and Budge (1928: 1164-5) interpreted this name as Alameda, Ella Amida; a reasonable enough suggestion, since from the numismatic point of view Ezana, the king who adopted Christianity, seems to follow Ousanas, while the tradition relating the circumstances of Ethiopia's conversion states that the converted king was the son and successor of Ella Allada/A'eda, though under the regency of his mother (but see also Dombrowski and Dombrowski 1984: 131-3). The name is testified later as Alla Amidas (Ch. 4: 7); Ousanas may have adopted it as his throne-name, and it is not impossible that one of the inscriptions published by the Deutsche Aksum-Expedition (Littmann 1913: IV, DAE 8) actually belongs to Ousanas rather than Ezana; its `Bisi' title certainly includes the letter `s' and whatever identity, such as `Ousanas Bisi Gisene', is accepted, `Ezana Bisi Alene' is definitely precluded (Munro-Hay 1984ii: 108).

Ezana is the most famous of the Aksumite kings before Kaleb. Several inscriptions of his are known, which tell a good deal about his military exploits and furnish many other details about fourth century Aksum. His most significant contribution to Ethiopian history was his official adoption of Christianity around 333AD, which he signalised by putting the cross on his coins (it also appears on one of his inscriptions; Schneider 1976ii: fig. 4), and by dropping the claim to be the son of the god Mahrem.

Illustration 11a. Drawings of two silver and three bronze issues (d. c. 10-16mm) of king Ezana of Aksum, some with the disc and crescent symbol, and some with no religious symbol at all.

In Ezana's time intercourse with the Roman empire continued, but even if the conversion to Christianity (Ch. 10: 2) was designed to bring Aksum closer to Rome or Constantinople, it was not a policy which he followed slavishly. There seems to have been little response to Constantius II's suggestion (c356AD) that Frumentius, by now bishop of Aksum (to whom Mommsen (1886: 284, n. 2) referred in his phrase `an Axomitic clergyman'), should be sent for examination for doctrinal errors to the emperor's bishop at Alexandria (Szymusiak 1958). Constantius, leaning towards the Arian heresy, was currently at loggerheads with patriarch Athanasius of Alexandria, who had consecrated Frumentius for his new see probably around 330AD. Athanasius had been sent into exile, and an Arian bishop installed in his stead. It was to this man, George of Cappadocia, that Constantius, declaring himself fearful for the Christian faith in Aksum, wanted Ezana and his brother to send Frumentius. But since Frumentius remains revered as the founder of the Ethiopian church, which does not follow Arianism, it may be assumed that the request was ignored, and that, as the Ethiopian Synaxarium says, he `died in peace' (Budge 1928). In any event, the Arian emperor and bishop did not last much longer, and delaying tactics might have avoided the necessity to give a definite response to the request before the emperor's death in 361.

Ezana's titles (see Ch. 11: 5) show that he considered himself to be at least theoretically the ruler of very large areas of present-day Yemen, Ethiopia and the Sudan. Interestingly enough, his use of the title `king of Saba (Salhen) and Himyar (Dhu-Raydan)' is similar to only the most modest of those used in the Yemen itself; around 300AD the title `king of Saba and Dhu-Raydan and Hadhramawt and Yamanat' came into existence, and was used by rulers such as Shamir Yuhar`ish and Karibil Watar Yuhan`im, whilst by the end of the fourth century, under Abukarib As`ad, it developed into `king of Saba and Dhu-Raydan and Hadhramawt and Yamanat and the Arabs in the Tawd (highlands) and the Tihamat (coastal plain)'. It seems certain that Ezana did not actually control any of the Arabian kingdoms, but his use of only the attenuated Arabian title and the apparent circulation of some of his coins in Yemen perhaps indicate that some sort of arrangement was reached between the two regions, or even that a coastal foothold was still retained by Aksum on the other side of the Red Sea. If predecessors of Ezana, like the `king of kings' Sembrouthes, had claimed the Arabian titles, they might simply have remained in the least expanded form by tradition; the Arabian kings themselves never used the parallelism Saba/Salhen, Himyar/Raydan in their own titles, though Shamir Yuhahmid was referred to as `of Dhu-Raydan and Himyar'.

In Africa, though most of Ezana's military expeditions were more or less tribute-gathering rounds in his own kingdom, pacifying any unrest in transit, he mounted at least one large-scale campaign against the Sudanese Noba and Kasu which his inscriptions (see Ch. 11: 5) claim as a major victory. Two fragmentary Aksumite inscriptions found at Meroë itself may be traced to this campaign, or perhaps to a similar one by a predecessor (Sayce 1909, 1912; Hägg 1984; Burstein 1980; Bersina 1984). It appears that Ezana's campaign was celebrated by Christian inscriptions, while some of the interpreters of the Meroë inscriptions believe that they were dedicated to the pagan Ares/Mahrem; if so, they probably belong either to an early campaign of Ezana, or to some predecessor.

At some uncertain point in our Periods 2 and 3, comes one of the best known of all Aksumite inscriptions; the `Monumentum Adulitanum' (Ch. 11: 5). The inscription itself has been lost, but its Greek text detailing the campaigns of an unnamed Aksumite king was preserved by the merchant Kosmas in the sixth century when he copied it for king Kaleb at the behest of Asbas, archon or governor of Adulis (Wolska-Conus 1968: 364ff). It was inscribed on a stone throne, behind which lay a fallen and broken inscription of king Ptolemy III of Egypt, who reigned in the third century BC. Unfortunately, Kosmas, copying the two inscriptions, simply carried on from the end of Ptolemy's inscription to the Aksumite one without including the section (if it still existed) with the Aksumite ruler's name and titles.

Certain details put the inscription broadly into context. It is of a pre-Christian ruler, whose campaigns took him from the Nabataean port of Leuke Kome (`White Village' — the exact position of which is still uncertain, Gatier and Salles 1988) at the limits of the Roman possessions on the east coast of the Red Sea, to the country of the Sabaeans in South Arabia, and to extensive African territories, apparently ranging from the lands bordering Egypt to the Danakil desert. Huntingford (1989) gives the latest of many attempts to outline the historical geography of the text. The author refers to himself as the first and only king of his line to subdue so many peoples, but this could be mere hyperbole. The gods he mentions, and the ritual of setting up a victory throne, are also known from Ezana's inscriptions. In short, he could be situated chronologically almost anywhere between Gadarat and Ezana. His inscription is of immense value, since it supplies a sort of gazetteer for the limits of the contemporary Aksumite empire; or at least the limits of the sphere of influence, since it is not very likely that some of the more far-flung areas could ever have been retained as Aksumite possessions. It is notable that this inscription has a year-date 27, while Sembrouthes' inscription has 24 years and one of the inscriptions from Meroë has the date of year 21 or 24. Sembrouthes or Ezana (whose reign spanned at least a quarter century) are therefore both candidates for the Monumentum Adulitanum and Ezana (who campaigned in the Meroitic region) may, as mentioned above, be responsible for the Meroë 1 inscription. But Sembrouthes, if he really fits as we have suggested in the mid-third century, would have reigned at a time when just such activities in Arabia as are detailed in the Monumentum Adulitanum are to be expected. He also gives himself the titles of Great King and king of kings, perhaps suitable to one who had campaigned, like the unknown author of the Adulis inscription, so vigorously to establish his kingdom's power in new regions. It now seems very unlikely that Ezana could have set up an inscription dedicated to pagan deities so late as his 27th year. It is not beyond hope that future excavations may actually find the famous Monumentum Adulitanum, which Kosmas saw set up outside the port-city on the Aksum road.

6. Period 4; Ezana after his Conversion, to Kaleb

Owing to the lack of inscriptions and other sources of information, this period is very inadequately documented. A regulation of Constantius II of the late 350s, in the legal text called the Codex Theodosianus (Pharr 1952: 380, 12.12.2) speaks of persons travelling (presumably on official business) to the Aksumites and Himyarites. It reads;

The same Augusti and Julianus Caesar to Musonianus, Praetorian Prefect; No person who has been instructed to go to the tribe of the Aksumites or the Homerites shall henceforth tarry at Alexandria beyond the space of the time-limit of one year, and after a year he shall not receive subsistence allowance.
This gives the impression that contact was relatively frequent between the empire and Aksum in the later fourth and early fifth century. The law was issued at about the same time as the letter of Constantius II to Ezana and Sazana was written.

Illustration 11b. Drawings of a gold and silver issue (d. c. 16mm and c. 14mm) of king Ebana. The obverse of the gold issue bears a legend which may read `king of the Habashat', and the reverse of the silver shows a cross-crosslet with a gold-inlaid diamond-shaped centre.

The Aksumite coins continue to provide a sequence of rulers, and from their designs and legends some conclusions can be drawn (see Ch. 9). One remarkable feature of these coins, is that the gold (in one case in a hoard with some late Roman solidi — Munro-Hay 1989ii) has often been found in South Arabia, and that much of the gold and some of the silver bears a legend which may read `king of Habashat' or `king of the Habashites'. Could this be a hint that the Aksumites still managed to maintain some kind of legal (or even actual?) footing there, as asserted by Ezana's title `king of Saba and Salhen, Himyar and Dhu-Raydan'? Only one silver Aksumite coin, probably of the fifth century, has so far been reported from South Arabia, from Shabwa, capital of Hadhramawt (Munro-Hay, Syria, forthcoming.)

Other interesting features of the coinage of this period are the appearance of the cross, often gilded, on the coins, and the issue of coins with no royal name in all three metals. One king only, Mehadeyis (MHDYS on the unvowelled coin legends), issued copper/bronze coins with a Ge`ez legend, interestingly enough almost an exact translation of the Roman emperor Constantine's famous motto `In hoc signo vinces' (by this sign (the cross) you will conquer). He is also the last king to specifically mention Aksum or the Aksumites in his coinage legends until the post-Kaleb period.

Illustration 11c. Drawing of a bronze coin (d. 14mm) of king MHDYS of Aksum, with a gold-inlaid cross on the reverse, and a legend which reads `By this cross he will conquer'.

Christian mottoes on the late fourth and fifth century coins abound, and it is evident that the climate was right, from the point-of-view of official sanction, for the missionary efforts of the Nine Saints whom the Ethiopian hagiographies attribute to the latter part of this period (Ch. 10: 4). There is no sign on the coinage of a lapse into paganism again (Pirenne 1975; Shahid 1979), but rather of an increasing emphasis on the rulers' Christian faith. The names of the kings whose coins situate them in this phase of Aksumite history are; Ezana, MHDYS, Ouazebas, Eon, Ebana, Nezana, Nezool, Ousas and/or Ousana(s). Specific coinage questions are dealt with in Ch. 9 below.

Illustration 11d. Drawings of Aksumite anonymous coin issues in gold (d. c. 17mm), two types of silver (d. c. 13mm), and two types of bronze (d. c. 13mm and 15mm). The cross on the reverse of the lower examples of the silver and bronze is inlaid with gold. The bronze issues are the commonest of all Aksumite coins.

Illustration 11e. Drawings of gold issues (d. c. 18mm) of the kings Ousas/Ousanas/Ousana, possibly different spellings for one name. The reverse dies all bear the Greek legend `Theou Eukharistia' — By the Grace of God — under a cross.

Since coins of king Ouazebas were found in the occupation debris of a room buried under which were some of the broken fragments of the largest of the stelae at Aksum (de Contenson 1959; Munro-Hay, forthcoming), this, Aksum's largest monolithic monument, could have fallen as early as the reign of Ouazebas himself, very likely in the late fourth or early fifth century. The stele seems to have been the last of such monumental funerary memorials and possibly they went out of favour as Christianity spread, bringing with it new ideas about burial (see Chs. 5: 4-5, and 14).

Illustration 11f. Drawing of a bronze issue (d. c. 17mm) of king Ouazebas of Aksum; the reverse bears the Greek legend `Touto aresē tē khōra' — May this please the people — and the area around the royal bust is gilded.

The story of a skholastikos, or lawyer, of Thebes in Egypt who travelled to India after a stay at Aksum is preserved in a letter written by a certain Palladius, probably the bishop of Helenopolis by that name who lived from 368-431AD (Derrett 1960; Desanges 1969). Palladius travelled to India to investigate Brahmin philosophy in the company of Moses, bishop of Adulis. Palladius' journey seems to have been undertaken sometime in the first quarter of the fifth century, and it was after that that he wrote his letter to some personage of high rank to inform him about the Brahmins, using the skholastikos' information. The latter had spent some time in Ethiopia, entering the country at Adulis and going on to Aksum. He eventually was able to go on an `Indian' ship to India. His comments have been thought to indicate that the reputation of Aksum had declined rather in this period: the king of Aksum is apparently referred to rather scathingly as an Indian minor kinglet (basiliskos mikros). Although this title could simply result from the attitude of a Roman subject to almost any ruler in comparison to the Roman and Persian emperors, it may also reflect a certain dimunition of Aksumite power at the time. The exact meaning of the title `basiliskos', however, is still the subject of discussion (Donadoni 1959; Hansen 1986), and may actually be a superior title to basileus in certain circumstances. A significant point is that basiliskos appears to be used otherwise particularly for Nubian rulers. It is the title used by the Blemmye king Kharakhin, a certain Pakutimne refers to himself as `epiph(ylarkhos) of the basiliskos', the six Blemmye/Beja kings captured by Ezana's brothers (DAE 4, 6 & 7) are called basiliskoi, and king Silko of the Nobatae is also basiliskos (Munro-Hay 1982-3: 93, n. 23). It may be that Palladius or the skholastikos confused the titles of the Aksumite king and a Nubian ruler; but mikros is still not very complimentary.

Illustration 11g. Drawing of a gold coin (d. 16mm) of king Kaleb of Aksum, with, on the reverse, the king's monogram followed by the Greek words `Uios Thezena' — Son of Thezena.

Kaleb, first ruler in our next period (see below), calls himself the `son of Tazena' on his inscription found at Aksum (Schneider 1972) and on his gold coins (where it is written in Greek as Thezena or variants; Munro-Hay 1984: 116-123). Tazena is also known from the Ge`ez stories about the Nine Saints, where he is identified as a king (Sergew Hable Sellassie, 1972: 115ff). Kaleb's own inscription from Aksum refers to `the throne of my fathers' which may not actually confirm that his father Tazena was king but at least means that Kaleb regarded himself as belonging to the legitimate dynasty. Tazena, if he actually ruled as a king of Aksum, may be identified with one or other of the names known from the coinage, since Aksumite rulers used several names; Ousas/Ousana(s) being perhaps the most likely identification from the numismatic point-of-view (Munro-Hay 1987). The sequence given by the king-lists (Conti Rossini 1909) and hagiographies is usually Sa`aldoba, Ella Amida, Tazena, Kaleb, but apart from Tazena's name on Kaleb's coinage, so far only Kaleb himself can be accurately identified from other sources. There exists a Syriac work, the Book of the Himyarites (Moberg 1928), about the war in Himyar which Kaleb waged against the Jewish king Yusuf. The surviving leaves of this book were found, remarkably enough, acting as padding for the covers of another much later book and through this safe concealment survived to our day. In the Book of the Himyarites mention is made of a previous expedition conducted by the Aksumites to Arabia, led by a certain Hiuna. This name unquestionably resembles the royal name Eon (EWN) as it is written in Greek on the coins of Eon Bisi Anaaph (Munro-Hay 1984: 88-9). The difference in the spelling is no more than would result from transposing the name into the two languages concerned. However, the coins in question, apparently of the beginning or early part of the fifth century, seem to be of too early a style to admit this identification. Accordingly, it has to be assumed that Eon and Hiuna are different people, unless, as is also possible, there is some distortion in the apparent chronology of the Book of the Himyarites at this point. The book only mentions Hiuna in its contents list, the particular chapter concerned having disappeared; `On the first coming of Hiuna and the Abyssinians'. The chapter, then, could refer to a previous expedition at some remove in time. However, the possibility that Hiuna might be a contemporary of Kaleb is enhanced by the latter's inscription (Schneider 1974: Drewes 1978), which apparently declares `I sent HYN ..BN ZSMR with my troops and I founded a church in Himyar'.

Illustration 12. Drawing of a gold coin (d. 17mm) of king Eon, with, on the obverse, the legend + CAC + ACA + XAC + CAC, which, with its variants, may mean `King of Habashat'.

Eon, interestingly enough, appears to be the first of the Aksumite rulers to use the mysterious title + BAC + CIN + BAX + ABA, on his gold coins. This is of uncertain meaning but has been interpreted to include the phrase `Basileus Habasinōn', or king of Habashat/the Habash, one of the titles used by the South Arabians in their inscriptions when referring to the Aksumite rulers (Doresse 1957: I, 278ff). Further, Eon's gold coins have been found in South Arabia, as have those of almost all his successors until the reign of Kaleb (Anzani 1928, Munro-Hay 1978, Munro-Hay forthcoming). It is possible, then, that the Aksumites continued to struggle to preserve some sort of foothold or official presence in South Arabia during the fifth century, in spite of the consolidation and expansion there of the power of local rulers such as Abukarib As`ad. We cannot know for certain how much truth there might be in this suggestion until inscriptions of one or other of the Ethiopian kings of the period or their Arabian contemporaries come to light.

7. Period 5; Kaleb to the End of the Coinage

The events of the time of Kaleb swell the sources available for Aksumite history to a disproportionate degree, and have resulted in the assumption that Kaleb was Aksum's greatest and most powerful ruler. However, the majority of the sources merely consist of more or less repetitious accounts by ecclesiastical historians full of praise for Kaleb's incursion into the Yemen to crush the anti-Christian persecutor, and often add little or nothing to the information of more reliable sources.

Illustration 12a. Drawings of two gold coins (d. c. 18mm) and a bronze coin (d. 15mm) of king Kaleb.

Kaleb invaded the Yemen around 520, in order to oust the Jewish Himyarite king Yusuf Asar Yathar, who was persecuting the Christian population. This ostensible reason for mounting the expedition across the Red Sea probably covers a number of other causes, since it seems that Yusuf may have also acted against Aksumite interests, and those of her Roman allies, in the political and commercial spheres.

One later source (the historian known as Pseudo-Zacharias of Mitylene; Brooks, ed. 1953) claims that Yusuf had acceded to power in Himyar because an Aksumite appointee to the throne had died, and, it being winter, the Aksumites could not cross the Red Sea to install another king. However, the source is not a contemporary one, though the very fact that he makes such a statement is interesting. King Ma`adkarib Ya`fur, who left an inscription of year 631 of the Himyarite era (c516AD), may have been the deceased ruler, but his titulature is the very long one including `Hadhramawt, Yamanat, and their bedouins of the high plateau and coastal plain', and it seems very unlikely that he was actually an Aksumite appointee (Rodinson 1969: 28, 31).

The various Latin, Greek, Syriac and Ge`ez sources (admirably summarised in Shahid 1971) have left a complicated set of names for the two protagonists in this war. Kaleb Ella Atsbeha is usually referred to by variants of his throne-name (such as Ellesbaas, Hellesthaeos, etc.), though John of Ephesus calls him Aidog. Yusuf is called Dhu-Nuwas in the Arab accounts, and a variety of names in other sources (Damianus, Dunaas, Dimnus, Masruq, Finehas, etc.) which seem to be based on this epithet or nickname or are derived from other epithets.

The chronology of the rulers of the Yemen in Kaleb's time is tentative, and one inscription which refers to the death of a king of Himyar, dated to 640HE/c525AD, has been taken as announcing the death of king Yusuf, thus situating Kaleb's invasion in 525. But it may well refer to the death of his successor, the viceroy Sumyafa` Ashwa`. With the useful assistance of dated inscriptions the chronology can be reconstructed as follows;
Ma`adkarib Ya`fur - c517/8AD;
Yusuf Asar Yathar517/8 - 520;
Sumyafa` Ashwa` 520 - c525;
Abreha 525 - at least 547.

The Arab historians Ibn Hisham, Ibn Ishaq and Tabari, each of whom has a slightly different version of events (Guillaume 1955; Zotenberg 1958), tell of rivalry between Abreha and another of the najashi's generals in the Yemen, Aryat. This would seem to have occurred around 525.

When Kaleb's forces arrived in Yemen, there was a certain amount of fighting, celebrated by various inscriptions. One (Rodinson 1969), of some qayls or princes of `Yusuf Asar Yathar, king of all the Tribes', is dated to year 633 of the Himyarite era/c518AD, and mentions that the king destroyed the church and killed the Abyssinians at Zafar, the Himyarite capital, demonstrating clearly that there were already Abyssinians in the country at the time; this rather speaks in favour of Pseudo-Zacharias' statement noted above. In the end, Kaleb's invading force was able to rout and eventually kill Yusuf. Another, Christian, ruler, Sumyafa` Ashwa`, was appointed, whose inscription (Philby 1950; Ryckmans 1946; Ryckmans 1976) refers to him by the title of king, but also as viceroy for the kings (in the plural) of Aksum; this inscription actually names Kaleb by his `Ella'-title, as Ella A(ts)bahah. Another inscription, possibly part of this one or a close parallel, appears to name the town of Aksum itself (Beeston 1980ii).

The Ethiopian viceroyalty lasted for perhaps four or five years, until c525, when the viceroy, Sumyafa` Ashwa`, was deposed, and Abreha became king in his stead. The contemporary Byzantine historian Procopius mentions that Sumyafa`, whose name he graecises as Esimiphaios, was a Himyarite by birth. The deposition of Sumyafa` was, apparently, accomplished with the support of Ethiopians who had remained in the Yemen, and Kaleb attempted to punish them and Abreha by sending a force of three thousand men under a relative of his. But this force defected, killing their leader and joining Abreha. The infuriated Kaleb sent yet another army, but this was defeated and accordingly Abreha was left on his throne (Procopius; ed. Dewing 1914: 189-191).

Abreha in later years used the titles of `king of Saba, Himyar, Hadhramawt, Yamanat, and all their Arabs of the Coastal Plain and the Highlands' (Ryckmans 1966; Smith 1964; Sergew Hable Sellassie 1972: 148, 153). This title, apparently last used by Ma`adkarib Ya`fur, seems to have lapsed during Yusuf's usurpation (Rodinson 1969) and Sumyafa` Ashwa`'s viceroyalty, and it is interesting to observe that although Kaleb of Aksum used these same titles, his son Wa`zeb abandoned the longer title and contented himself with the old claim of Ezana's time to overlordship of Saba and Himyar only (Schneider 1972). This seems to reflect exactly the position as Procopius, writing after the death of Kaleb, related it, and seems to indicate that the titulature of the Aksumite kings did have some real significance in relation to events, rather than consisting of a merely traditional listing of both actual possessions and former claims (Ch. 7: 5). The essential of the situation is, that while the Aksumites may have been palliated by a formal submission and tribute, in actual terms they had permanently lost the control of the Yemen. A poem recorded by one of the Arab authors (Guillaume 1955: 34) sums up the history of Dhimar (Yemen) in this period, (as seen through anti-Ethiopian eyes);

"To whom belongs the kingdom of Dhimar? to Himyar the righteous;
to whom belongs the kingdom of Dhimar? to the wicked Abyssinians;
to whom belongs the kingdom of Dhimar? to the noble Persians;
to whom belongs the kingdom of Dhimar? to the Qoreysh, the merchants".

Glorious though Kaleb's re-establishment of the Christian faith in the Yemen seemed to contemporary (and later) ecclesiastical historians, it was Aksum's swan-song as a great power in the region. The real result may well have been quite the opposite; a weakening of Aksumite authority, over-expenditure in money and man-power, and a loss of prestige. The venture was, it seems, too ambitious for the times, and did Aksum nothing but harm in the long run. Nevertheless, for a while we still hear of embassies arriving from the Byzantine empire with trade proposals, and others going to Abreha in the Yemen (recorded on his Marib Dam inscription of 658HE/c543AD) and to the Persian king to persuade him to release certain bishops jailed at Nisibis (according to John of Ephesus' life of Simeon, bishop, of Beth Arsham — Brooks 1923; Doresse 1971: 102). Evidently Aksum still remained in the main stream of international affairs for a while.

Kaleb's inscription and coins, the hagiographical tales (and the king-lists, rather surprisingly, as well), confirm that Kaleb Ella Atsbeha's father was a certain Tazena. He is not represented on the coins, but could be included under some other name; one of the great difficulties of Aksumite numismatics. The gold coins which most closely resemble those of Kaleb come from a group bearing the royal names Ousas/Ousana/Ousanas, which may all belong to only one king. Perhaps he is to be identified with Tazena, as noted above (Ch. 4: 6). On the other hand, the emphasis on naming Kaleb's father on all his son's surviving official documents, and his use of the phrase `throne of my fathers' in his inscription (see above) might lead us to suppose that some political need was felt for this assertion of legitimacy; perhaps Tazena's kingship was somehow in dispute, or perhaps he was not an Aksumite king, but a claimant in the line of succession? The late compilations of the tales of the Nine Saints mention a king Tazena as the father of Kaleb, but their evidence is not necessarily reliable.

Illustration 13. Drawings of coins (d. 17mm) of king Alla Amidas of Aksum.

The Ethiopian traditions say that Kaleb eventually abdicated the throne, sent his crown to be hung on the Holy Sepulchre at Jerusalem, and retired to a monastery. Since there are actual die-links between the coins of Kaleb and a king Alla Amidas (Munro-Hay, Oddy and Cowell, 1988), it is possible that he and Kaleb may have ruled together, Kaleb perhaps later retiring, thus explaining the plural term `the neguses (nagast) of Aksum' in the inscription of Sumyafa` Ashwa` (Ryckmans 1946: line 3). Another line (8) in this inscription mentions that `they submitted to the kings (amlak) of Aksum'; however, these locutions might apply to a concept of `the crown of Aksum', since a further phrase (line 14) alludes to the ngsy in the singular form.

Kaleb's son Wa`zeb (W`ZB) is known from an inscription (Schneider 1974), where he is actually called `son of Ella Atsbeha' using his father's throne-name. He is presumably represented on the coins by another name, possibly Ella Gabaz (Munro-Hay 1984ii). Wa`zeb has left only this inscription, in Ge`ez, written in South Arabian script like Kaleb's, but so damaged that it is very difficult to decipher (but see Ch. 11: 5). The story of Abba Libanos, the `Apostle of Eritrea', mentions a king called `Za-Gabaza Aksum', perhaps another version of the name Ella Gabaz (Dictionary of Ethiopian Biography 1975: I, 103); a suggestion confirmed very recently by Sergew Hable Sellassie (1989) who notes a homily of the Metropolitan Elias of Aksum about Abba Mat'e, Libanos, in which it is stated that the contemporary king was Ella Gabaz. Ella Gabaz and Za-Gabaza Aksum may be epithets indicating that Wa`zeb (if the identity is admitted) did some important building work at Maryam Tseyon cathedral.

Illustration 14. Drawings of a gold coin (d. 18mm) of king Iathlia/Hataz, with two issues each of Hataz in silver and bronze.

The coinage of this period is extremely difficult to put in order. There are often only single surviving specimens of issues, or a bewildering array of mutually exclusive factors to take into account when attempting to classify them into a sequence. However, among names known from the coinage, apart from those already noted above, are Wazena (tentatively identified with Alla Amidas), Ioel, Iathlia/Hataz, and Israel. If the identity of Alla Amidas with Wazena is correct (Munro-Hay 1984ii), and this ruler was a colleague or immediate successor of Kaleb, it may be that Wazena is the name found in the opening phrases of Abreha's 543AD Marib inscription; `viceroy of the king Ella `ZYN' (see Schneider 1984: 162-3). This would be a somewhat bizarre rendering of a supposed name `Wazena Ella Amida' with, in addition, the waw written as `ain.

Illustration 15. The obverse of a gold coin (d. 18mm) of king Israel of Aksum. Photo British Museum.

King Israel bears the name of one of Kaleb's sons in the legendary histories (Conti Rossini 1909; Sergew Hable Sellassie 1972: 161), but seems too far removed from him from a numismatic point-of-view to be so identified (Munro-Hay 1984ii). Two more kings, Armah and Gersem, close the sequence of the coinage. The coins of the later kings are very degenerate in appearance in comparison to the earlier issues, and their gold content is much debased. It would seem that Aksumite prosperity was on the decline through a combination of reasons, but that coins continued to be issued until the disruption of the Red Sea trade which had brought the experiment with a monetary economy into being finally removed the need for it. Some features of the design of the later coins are extremely reminiscent of Byzantine models.

Illustration 16. Drawing of a silver coin (d. 16mm) of king Armah, depicting, on the reverse, a gateway adorned with three crosses, possibly representing the Holy Sepulchre in Jerusalem.

There are certain factors which favour suggesting that king Gersem belongs at the end of the coinage sequence, but a number of alternative points might favour Armah as the last king of Aksum to issue coins (Munro-Hay 1984ii; Hahn 1983). Armah's name could conceivably be related to that of Ashama ibn Abjar, or his son Arha, known from the accounts of some early Islamic historians, through the common process of Arab mis-spellings or inversions by copyists (Hartmann 1895). Armah produced (so far as we know to date) no gold coins, which might suggest that he had accepted that there was no purpose in producing them, as his kingdom was by now at least in part cut off from the Byzantine trade network. His silver has an unusual reverse, depicting apparently a structure with three crosses, the central one gilded; Hahn (1983) suggested that this could represent the Holy Sepulchre, and have been placed on the coins in reference to the Persians' seizure of Jerusalem and the holy places in 614. On the obverse the king's crown is gilded as well. Armah's bronze coins are the largest produced by the Aksumites, and unusually show the king full-length, seated on a throne; these and the elaborate gilded silver coins may have been designed to compensate for the lack of gold. Unfortunately there are neither archaeological contexts, nor overstrikings among the coinage, which can confirm which of these kings actually issued the last coins of the Aksumite series.

Structures have been found at Adulis, Aksum and Matara which contain the coins of all these later kings, showing that these towns were probably functioning at least until the end of the coinage, and were still under Aksumite control. Considerable quantities of pottery, often decorated with elaborate crosses, and other material including imported amphorae, provide evidence of still flourishing trade and local industries during the time of at least some of the later rulers. It is interesting to note, however, that the mottoes on the coins seem to grow increasingly less interested in royal and religious themes, but in the reigns of the later kings begin to ask for `Mercy and Peace to the Peoples'. Perhaps it is pure imagination to see in this a response to the current situation of Aksum, but it can be suggested that gradually things were getting worse in both the condition of the kingdom in general and in the capital itself. Aksum may possibly have suffered from the plague which reached Egypt in 541 (Procopius, ed. Dewing 1914: 451ff; see also Pankhurst 1961 for more notes on diseases in Ethiopia), and had spread all over the eastern part of the Roman empire a year or two later; some claimed that it had originated in `Ethiopia', but the term could refer to the Sudan or even other parts of Africa. If, however, Aksum was the victim of an epidemic, it might be an additional reason for the failure of the attempt to control king Abreha in the Yemen, and for the gradual decline of Aksum from that time onwards.

An interesting, but very tentative, source for Aksumite history in the later sixth century takes us much further afield for sources, to China. In 1779 the T'ien-fang Chih-sheng shih-lu, written between 1721 and 1724 by Liu Chih, was published. This was a life of Muhammad, the `True Annals of the Prophet of Arabia', written using an older book of records about the prophet in Arabic found at Ts'eng Liu. It has been translated into English by Mason (1921). According to Leslie (1989) the original used by Liu Chih is very likely to have been a Persian translation of a biography of the prophet written by Sa`id al-Din Muhammad b. Mas`ud b. Muhammad al-Kazaruni, who died in AH758/1357AD, but at the time of writing the sources used by the latter have not yet been traced. However, it is interesting to note that the book contains a number of mentions of Abyssinia. The reigning najashi was said (Mason 1921: 35) to have sent an ambassador with gifts on sighting a star which marked the birth of the prophet (c570), and later (c577) when Muhammad was seven, the text (Mason 1921: 47) tells that the najashi Saifu ascended the throne. Abd al-Muttalib went to congratulate Saifu on his accession, and a speech of the former is quoted. He declared that

"The great king, your grandfather, was a benevolent king, and his grandson is a holy sovereign, who breaks off with flatterers and follows what is right, avenges the oppressed and, acting upon right principles, administers the law equitably. Your servant is the superintendent of the sacrifices in the sacred precincts of the True God, a son of the Koreish, who, hearing that your Majesty has newly received the great precious throne, has come to present congratulations".
Saifu recognised Muhammad's greatness from portents he had found in the books, but prophesied that he would have troubles.

However much the Chinese translation may colour the narrative, there remain very interesting points in this account. If a king `Saifu' came to the Aksumite throne in c577, and was the grandson of a particularly eminent Aksumite monarch, could that monarch have been Kaleb, so well-known to the Arabs as the conqueror of the Yemen? The text also adds (Mason 1921: 102) that Saifu's own grandson was the najashi who received the Muslim emigrants in the fifth year of Muhammad's prophetship, 615-6AD (Ch. 15: 4). If this very late and very much second-hand account deserves any credit we may postulate that Kaleb's son(s) ruled until c577, to be succeeded then by his grandson, followed by other members of the same family until Ashama ibn Abjar. Ibn Ishaq (Guillaume 1955: 153ff) says that Ashama only succeeded after the reign of an uncle who had usurped the throne from Ashama's father (Ch. 7: 5). The coins and inscriptions offer, as we have seen, Alla Amidas/Wazena, and Ella Gabaz/W'ZB (Kaleb's sons?), followed by Ioel, Hataz, Israel, Armah and Gersem, for this period. It seems that this comes close to the number of rulers recorded in the later sources; two sons of Kaleb according to the Ge`ez texts, a grandson Saifu succeeding in c577 after the Chinese text, another ruler and his usurping brother according to Ibn Ishaq, and finally Ashama, who died in 630, according to the Muslim chroniclers.

The situation at the ancient capital at this period was not what it had been in former times. Overuse of the land around a city which had supported a substantial population for some six hundred years was doubtless beginning to result in food-supply deficiencies, and in addition the Nile-levels recorded in Egypt indicate that the rainfall was not so constant as before (Ch. 3: 1-2). The clearance of the wooded hills around the city, whether for charcoal or planting of crops, allowed the rains to carry off the topsoil, exacerbating the agricultural problems (Butzer 1981). The very grave dimunition of the Red Sea trade, and the loss of revenue from that source may have combined with internal troubles, such as a resurgence of independence among the northern Beja tribes, or even dynastic difficulties, to hasten the abandonment of Aksum as the capital of the kingdom. Whatever the case (for the situation under Ashama ibn Abjar and the wars of the hatseni Danael see Ch. 15: 4-5) it seems that possibly before the middle of the seventh century AD, Aksum, though it continued to exist in a reduced way until the present as an ecclesiastical centre, and even ritual centre for the kingship, had ceased to be viable as Ethiopia's capital city.

A famous Ethiopian of this last period of Aksum was one of the early converts to Islam, Bilal ibn Rabah. He was a freed slave of Ethiopian origin born in Mecca who became the first muadhdhin — muezzin — or chanter of the call to prayer. He also bore the prophet's spear, which was a gift of the najashi Ashama to the prophet's cousin al-Zubayr, and was used from 624AD to point the direction of prayer. Bilal died about 640AD (Dictionary of Ethiopian Biography 1975: I, 41).

8. The Post-Aksumite Period

The period from the seventh to the twelfth century, though recognised here as post-Aksumite in the sense that Aksum was no longer the political centre of the kingdom, has generally been included by previous writers on Ethiopian history as Aksumite; accordingly, a brief sketch of the few known events occurring during this time may be useful. Aksum's name seems to have no longer been applied to the Ethiopian people, but `Habash' remained, as usual, the Arab name for them, and the country was called `Habashat' (Irvine 1965; but see Beeston 1987). The period concerned includes the greater part of the seventh century, and terminates with the advent of the Zagwé dynasty in about 1137 (another long-disputed date, see Munro-Hay, The Metropolitan Episcopate of Ethiopia and the Patriarchate of Alexandria 4th-14th centuries, forthcoming).

1. Realignment.

The reign of Ashama ibn Abjar, and the inscriptions of the hatsani Danael, are discussed in Ch. 15: 4-5, and are suggested to represent the end of the Aksumite period in Ethiopia. An entirely different picture of the kingdom now emerges. The coinage, and with it the use of Greek and the trade connections into the Red Sea and the Roman Christian world gave way to a different economic and political orientation. Such commodities as cloth (al-Muqaddasi, in Vantini 1975: 176) and probably the ever-needed salt were used in barter, but now trade was, it seems, limited only to neighbouring countries in Africa and Arabia. The kingdom, though almost always regarded by Arab writers during this period as a powerful and extensive state, eventually lost the use of the coast, and other areas formerly under a tribute relationship to the Ethiopian state became completely independent.

As Connah (1987: 71-2, 95) has pointed out, the new condition of relative isolation had both advantages and disadvantages, the greatest benefit being that Ethiopia could safely withdraw itself into the mountains and take up a strong defensive stance. This was, of course, offset by the loss of its international position, but, as we have said, the decline of the Red Sea trade was the result of great events in the outside world which Ethiopia could only accept by its own realignment. The country's political connexions were nearly all with Muslim states from this time onwards, though a few brief mentions are made of the Christian Nubian kingdoms (see below). Al-Mas`udi speaks of a treaty between Abyssinia and Ibrahim, ruler of Zabid in the Yemen, by which the latter's ships continually moved between the two countries with merchants and merchandise (de Meynard and de Courteille 1864: 34), and relations apparently remained intact with the Yemeni rulers after the country's conquest by a neighbouring queen (see below). Ethiopia's metropolitan bishops still came from the patriarchate of Alexandria, but were now obtained by application to the Muslim governor of Egypt (Munro-Hay, The Metropolitan Episcopate of Ethiopia and the Patriarchate of Alexandria, 4th-14th centuries, forthcoming). Muslim states arose in the Dahlak islands and on the coasts, and in later times became a grave danger to the Christian state.

But the Ethiopian kingdom itself did not remain static; as it lost in the north and east, it gained in the south, and the dynastic capitals of the later Zagwé (c1137) and Solomonic (c1270) dynasties were successively situated further in that direction. Trading objectives changed too; but still Arab traders continued to come to Ethiopia. The country still possessed excellent agricultural resources, gold, ivory, hides, and many other products, and doubtless the expansion southwards allowed the Ethiopians of the post-Aksumite era to develop certain aspects of their export trade in luxuries as much in demand in the Arab world as they had been in the Roman; we hear of merchants of Oman, Hejaz, Bahrein, and the Yemen trading there (see below). Ethiopia may have found itself increasingly outside the main stream, but was certainly not finished as a polity. The brief reports of the next few hundred years speak of a large and powerful Habash realm, and, as far as we can tell, only in the later tenth century did disaster strike.

2. Successor Capitals.

The immediate successor capitals to Aksum are mentioned mainly by Arab authors, illustrating well the new alignment of trade and therefore knowledge about the country. In many cases these repeat the information of preceding writers, sometimes anachronistically. The astronomer Al-Battani, for example, who died in 929AD, repeats information from Ptolemy, and names `Ksumi, the town (or, land) of the king of Kush' (Nallino 1907: II, 47). This obviously refers to Aksum; but most other Arab writers provide completely different names. The earliest is Jarmi or Jarma, followed by Ku`bar or Ka`bar.

Abu Ja`far al-Khuwarizmi, writing before 833AD, seems to be the first to cite a town called `Jarma, the town of the kingdom of Habash', as well as `Jarmi, the great town', in company with such other towns as Dunqula (Dongola), capital of the Nubian kingdom of Muqurra (Vantini 1975: 50). Al-Farghani, writing before 861AD, mentions `the towns of the kingdom of the Habasha, which are called Jarmi (Jarma), Dunqula and the town of the Nuba' (Vantini 1975: 53). Ibn Rusta, who died before 913AD, wrote of `Jarmi, the capital of the Habasha and Dunqula, the capital of the Nuba' (Vantini 1975: 87), and around 950AD Ishaq ibn al-Husain wrote that "The main town in the country of the Habashat is (the town of) Jarmi, (which is) the capital (dar) of the kingdom of the Habasha. This kingdom is ruled by the najashi". Interestingly, this author repeats Kosmas' old tale about silent trade in the lands of the Habasha (Vantini 1975: 122-3), as do others even later (e.g. al-Zuhri, Vantini 1975: 262).

About 966AD al-Maqdisi simply confirms the previous information, writing that "One of the towns of the First Climate is . . . Jarmi, a town of the king of the Habasha, another is Dunqula, the town of the Nuba" (Vantini 1975: 147), and a little later Ibn Yunus (d. 1009AD) listed both Jarmi and Madinat al-Habash (here Dongola/Dunqala?) with their latitudes and longitudes (Vantini 1975: 223). Al-Tusi, an astronomer who died in 1273AD, also mentions these two towns with their latitude and longitude (Vantini 1975: 380).

Al-Biruni (d. 1048), though a pupil of al-Mas`udi, does not mention Ku`bar (see below), but in his list of towns, he includes "Jarma (Jarmi), a town of the Habasha" and Aydhab as a Habasha town, the frontier between Beja and Habasha (Vantini 1975: 231-2), while al-Marwazi (d. after 1120AD) mentions that the First Climate "passes through a country called Jarma, which is the residence of the king of the Habasha, and through Dunqala, which is the capital of the Nuba" (Vantini 1975: 250). Yaqut (d. 1229) mentions among the towns of the First Climate `Jarma, the town of the king of the Habasha; Dumqula, the town of the Nuba' (Vantini 1975: 341). Abu'-l-Fida` (1273-1331AD) wrote that "Jarmi is the capital of the Habasha. It is mentioned by the majority of the travellers in the books of routes" (Vantini 1975: 463), and Ibn al-Shatir (1304-1379AD) still notes the latitudes and longitudes of "Jarmi of the Habasha" and "Dongola of the Nuba" (Vantini 1975: 525). Finally, it may be that the name Tambra or Tarma, used by al-Wardi (d. 1457AD) is a last memory of Jarmi or Jarma; he describes the town as "a big town on a lake in which the Nile waters collect" (Vantini 1975: 724).

Al-Ya`qubi (fl. c872-891) is the first of the Arab authors to mention Ku`bar or Ka`bar. His report is interesting in that it names five Beja kingdoms bordering on the najashi's realms and on Alwa (the Arab name for the Nubian Christian state of Alodia, with its capital at Soba). They each had their own king, and there is no mention of dependence on the najashi's kingdom; the last known Aksumite claim to control Beja and Noba is in W`ZB's titulature (Schneider 1974). Al-Ya`qubi describes

"a vast and powerful country. Its royal town is Ku`bar. The Arabs go thither to trade. They have big towns and their sea coast is called Dahlak. All the kings of the habasha country are subject to the Great King (al-malik al-a`zam) and are careful to obey him and pay tribute" (Vantini 1975: 73).
This information, well over two hundred years after the time of Ashama ibn Abjar, indicates that the Ethiopian kingdom had maintained itself relatively well, and was still in control of some of the coastal area.

Al-Mas`udi, who died in 956AD, gives rather similar information in his geographical work Muruj al-Dhahab, the `Meadows of Gold'.

"The chief town of the Habasha is called Ku`bar, which is a large town and the residence of the najashi, whose empire extends to the coasts opposite the Yemen, and possesses such towns as Zayla`, Dahlak and Nasi" (Vantini 1975: 131).
He repeated this in his Akhbar al-Zaman;
". . . the Habasha are the descendants of Habash b. Kush. b. Ham. The largest of their kingdoms is the kingdom of the najashi, who follows Christianity; their capital is called Kafar (Ka`bar). The Arabs used since the earliest times to come to this kingdom for trade" (Vantini 1975: 143).
Al-Harrani, writing about 1295AD, mentions that
"one of the greatest and best-known towns is Ka`bar, which is the royal town of the najashi . . . Zayla`, a town on the coast of the Red Sea, is a very populous commercial centre. . . . Opposite al-Yaman there is also a big town, which is the sea-port from which the Habasha crossed the sea to al-Yaman, and nearby is the island of `Aql" (Vantini 1975: 448).
Ibn Khaldun (1332-1406) also knew that Ka`bar was formerly the capital of Abyssinia (Trimingham 1952: 59).

Conti Rossini proposed that Ku`bar originated in a mistaken rendering of [A]Ksum in Arabic (1909: 263, n. 1). He says that "Aksum was the political capital still in the tenth century . . . I call . . . kings of Aksum the kings anterior to the Zagwé, who went down into Lasta". Vantini (1975: 131), whose extremely useful compendium of translations from Arab authors we have used extensively in this section, also thought Ku`bar was Aksum, suggesting a distortion from an epithet, `kabur', The Noble, or the name of some place near Aksum. Paul (1954: 71) identified Ku`bar with Adulis. The editors of Mas`udi's book The Meadows of Gold, where he mentions Ku`bar, identified it with Ankober in Shewa, influenced by the current name for the Shewan capital (de Meynard and de Courteille 1844: 34). Taddesse Tamrat (1970: 87-8; 1972: 37) thought Ku`bar might be in southern Tigray or Angot; an Ethiopian legendary account (Kur 1965: 18) says that the (ninth-century?) king Dil Na`od moved the capital "from Aksum to the country of the east" in the seventh year of his reign, and since the same phrase elsewhere in the text describes the Lake Hayq region, Tamrat suggested Ku`bar was in that general direction. In fact, little reliance can be placed on the names and dates (Tamrat 1972: 36, n. 3) in these legends, and it seems that Ku`bar's position must remain a mystery for the time being. It is almost certainly not Aksum, but could conceivably be some other major Aksumite town which, for various reasons, was considered to be a more suitable spot for the capital; possibly in the eastern region where many towns were built along the north-south route west of the escarpment.

We thus have Arab authors writing of Jarmi from before 833AD to well into the fourteenth or even fifteenth century, and of Ku`bar from the later ninth century until c1259AD. There seem to be three possibilities here.

One is that the two cities were in fact one and the same. It is very unlikely that any of the later writers did anything more significant than simply copying the older works, and one might propose that this town, wherever it was, lasted as capital of Ethiopia only from the transfer from Aksum (between c616-630?), or later if there was another capital between this time and the first mention of the new one in Arab sources, until the fall of the kingdom to the Queen of the Bani al-Hamwiyya in the mid-tenth century (see below). This would credit the work of al-Ya`qubi and al-Mas`udi, who generally seem well informed.

The second possibility is that there are two cities in question. If the name Jarma were an epithet for the capital, perhaps from the Ethiopic word girma, which means something like venerable or revered, it would seem likely that Jarmi was Aksum, to which this epithet was well-suited. Some authors would have simply repeated the outdated information that it was still the capital, and continued to do so even after Ku`bar had taken over that status, in spite of the more up-to-date reports of the better informed writers such as al-Ya`qubi and al-Mas`udi. In support of this theory, according to Vantini (1975: 380, n. 1) it appears from al-Tusi's own map that Jarmi corresponds to Aksum. However, the version of the map in Yusuf Kamal (1930-35: III, 1045r) as reconstructed by Lelewel, shows `Dziarmi' considerably south of Aden following al-Tusi's longitude, and Kammerer (1929: 48) noted that "la carte de Ahmad el-Tusi est un croquis élémentaire".

The third possibility seems the most likely; that Jarmi has nothing to do with Abyssinian Ethiopia, but is in fact Ptolemy's town of Garami, metropolis of the Garamantes in Libya. Arab geographers would have taken Garami in `Ethiopia' (the broad term used of all the area occupied by dark-skinned peoples) as the capital of the `Habash', used as a similarly general term. Jarmi is thus no longer a rival to Ku`bar as the post-Aksumite capital.

Other names for capitals are mentioned in the sources. An anonymous treatise of 982/3AD, written in Persian, the Hudud al-Alam, notes about the country of the Habash that

"This country has a very mild climate. The inhabitants are of a black complexion. They are very lazy and possess many resources. They obey their own king. Merchants from Oman, Hejaz and Bahrain often go to that country for trade purposes. Rasun, a town on the sea coast, is the residence of their king, while the army dwell in the town of Suwar: the Commandant-in-Chief resides at Rin, with (another?) army. In this province gold is abundant".
Rasun, it has been proposed, could stand for Jarmi through miscopied Arabic, though the coastal setting is unusual, and Minorsky proposed Aydhab and Zayla` for the other two (Minorsky 1937: 164, 473-4; Vantini 1975: 173).

An Ethiopian legend mentions that king Degnajan, who is supposed to have lived at the period just before Gudit's attack, left Tigray and made Weyna Dega his capital, apparently a place in Begemder east of Gondar (Sergew Hable Sellassie 1972: 203, n. 115; 231, n. 98). Al-Idrisi, the famous geographer at the court of the Norman king Roger II of Sicily, writing before 1150AD, called the greatest of all the towns of the Habasha `Junbaitah' (and variants), which seems likely to come from the Ethiopic phrase `jan-biet', or king's house (Vantini 1975: 278; Conti Rossini 1928: 324). Another interpretation identified the three Ethiopian towns mentioned by al-Idrisi (which Kammerer (1929: 54) thought were `toutes trois fantaisistes') with present-day villages; Miller (1927) decided that Gunbaita (Junbaitah) was Genbita near Kassala, Markata was Hanhita near Gondar, and Kalgun was Aganiti about half-way between Aksum and Ankober.

Al-`Umari (1300-1348AD) was aware of the antiquity, and even the (more or less) correct name of Aksum. He was employed in the chancery of the Egyptian sultan, and knew the protocol to observe when writing to the haty or king (Munro-Hay, The Metropolitan Episcopate of Ethiopia and the Patriarchate of Alexandria 4th-14th centuries, forthcoming), and perhaps had access to rather better archival material than many other writers. He states that a certain wadi

"leads to a region called Sahart, formerly called Tigray. Here there was the ancient capital of the kingdom, called Akshum, in one of their languages, or Zarfarta, which was another name for it. It was the residence of the earliest najashi, who was the king of the entire country. Next is the kingdom of Amhara, where is the capital of the kingdom nowadays, called Mar`adi; next is the territory of Shewa . . ." (Vantini 1975: 509).
Amda-Tseyon's (1314-1344) chronicler confirms that his court was located at Mar'adé, apparently in Shewa (Taddesse Tamrat 1972: 274).

Finally, we have the town of `Arafa or Adafa. This is mentioned in the History of the Patriarchs as the `city of the king' of Ethiopia in 1210, with the additional note that the king was called Lalibela (Atiya et al: III, iii, 192). The town is called Adafa in the Gedle Yimrha-Kristos (Taddesse Tamrat 1972: 59), and was the capital of the Zagwé dynasty. This presumably means that those Arab writers who continue to mention Ku`bar after about 1200AD are simply repeating earlier information without updating.

3. The History of the Patriarchs and Ethiopia.

Beyond the brief descriptions of the country in Arabic historical or geographical works, we have almost no other information about events in Ethiopia except from the History of the Patriarchs of Alexandria (Evetts 1904 and Atiya 1948) and the Ethiopian Synaxarium (Budge 1928). The biographies of the Coptic patriarchs of Alexandria, spiritual heads of the Ethiopian Orthodox church, were careful to include references to the patriarchate's dealings with the kingdoms of the south, Nubia and the Habash, since the authority of the patriarchs there was a useful counter to their subordination to the Muslim government in Egypt. From this source we have a few glimpses of conditions in Ethiopia, though the majority of references concern purely church matters and are not very informative on other questions. A detailed analysis of all the notes about Ethiopia in the ecclesiastical records can be found in Munro-Hay (The Metropolitan Episcopate of Ethiopia and the Patriarchate at Alexandria, 4th-14th centuries, forthcoming).

During the patriarchate of James (819-830) the metropolitan John was ordained for Ethiopia (Evetts 1904: 508ff). During his metropolitanate, military defeats, compounded by plague which killed both men and cattle, and lack of rain, are mentioned. The metropolitan had been driven to return to Egypt by the opposition of the queen and people while the king was occupied at war, and he was only able to return in the patriarchate of Joseph (831-849). Dates around c820-30 have been suggested for these events, and it is interesting to note that during that decade the king of Abyssinia was involved in unsuccessful military engagements. At just this time, in northern Abyssinia, the Beja tribes were growing stronger, and in 831 a treaty between the caliph al-Mu`tasim and Kannun ibn Abd al-Aziz, `king' of the Beja (Hasan 1973: 49-51), seems to recognise his power even as far south as Dahlak. This may well indicate that there had been disputes between the Beja and the Ethiopians, with the Ethiopians, at least temporarily, coming off rather the worse in the conflict.

It is also noted in his biography that Joseph had problems over his Abyssinian pages, presented to the church as gifts by the Ethiopian king (Evetts 1904: 528-31).

Nothing more is noted until the patriarchate of Kosmas III (923-934), when Ethiopian questions again came to the fore (Atiya 1948: 118-21; Budge 1928: III, 666-8). Interestingly, the eleventh century biographer, when he introduces al-Habasha (Abyssinia), adds the gloss that it is "the kingdom of Saba, from which the queen of the South came to Solomon, the son of David the king", showing that this identification was current even then. The Ethiopian king, Tabtahadj (or Yabtahadj, Babtahadj — a name not recorded elsewhere — Perruchon 1894: 84; it is, in fact, a misreading of bi-ibtihaj `with joy'), received a new metropolitan, Peter, and is said to have given him authority to choose his successor on his death-bed. Peter selected the youngest of the late king's two sons, but soon a monk, Minas, arrived with forged letters which declared that he himself was the rightful metropolitan, and Peter an imposter. This naturally found favour with the rejected brother, and Peter and the king he had chosen were deposed. Eventually the new king learned from Kosmas that Minas was actually the imposter, and he was executed, but meanwhile Peter had died in exile, and patriarch Kosmas refused to consecrate another metropolitan. The king therefore forced Peter's assistant to take up the post, uncanonically, thereby instituting a quarrel between monarchy and patriarchate which lasted for the unconsecrated metropolitan's lifetime, through four succeeding patriarchates and into a fifth. Eventually, during the patriarchate of Philotheos (979-1003), the quarrel was resolved, but only after Ethiopia had suffered terrible devastation.

The story is preserved in the History of the Patriarchs (Atiya 1948: 171-2) and in the Ethiopian Synaxarium (Budge 1928: I, 233-4), as well as in the accounts of certain Arab historians (see below). The first two refer to enemies who attacked Ethiopia, driving the king out and destroying his cities and many churches. In the History of the Patriarchs the enemy is named as the queen of the Bani al-Hamwiyya; a title which has not much assisted in identifying her. These ecclesiastical sources claim that the troubles were all due to the fact that metropolitan Peter had been deposed illegally, and when patriarch Philotheos, responding to the pleas of the king of Ethiopia sent through the agency of king Girgis II of Nubia, appointed a new metropolitan, Daniel, to Ethiopia, the troubles ceased.

The Arab historians add considerably to the history of Ethiopia at this point. Ibn Hawqal (Kramer and Wiet 1964: 16, 56) mentions that the country had been ruled for thirty years by a woman, who had killed the king, or hadani, and now ruled the hadani's territory as well as her own lands in the south. This was written in the 970s or 980s, and so the queen's advent was probably in the 950s. A confirmatory note occurs in a reference which states that the king of the Yemen sent a zebra, received as a gift from the female ruler of al-Habasha, to the ruler of Iraq in 969-70 (el-Chennafi 1976). It seems more than likely that this queen is identical with the queen enshrined in Ethiopian legend as the destructive Gudit, Yodit, or Esato, who invaded the kingdom and drove the legitimate kings into hiding, in spite of her legendary association with the establishment of the Zagwé kings. In fact, it seems likely that a period of well over a hundred years yet had to elapse before this new dynasty came to the throne, and that the Ge`ez chronicles have become hopelessly muddled at this stage.

Between 1073 and 1077 more trouble occurred because of a false metropolitan (Atiya 1948: 328-330; Budge 1928: IV, 995). A certain Cyril arrived with forged letters to claim the metropolitanate, and actually managed to bribe the ruling amir of Egypt to force the patriarch to ratify the appointment. The problem was only solved in the patriarchate of Cyril II (1078-92), who consecrated a new metropolitan, Severus; Cyril fled to Dahlak, where he was arrested by the sultan, and sent to Egypt, where he was duly executed.

Metropolitan Severus wrote to the patriarch in Egypt that the country was in good order but for the practice of polygamy by the Ethiopian ruler and some of his subjects. He was, however, soon in trouble. He apparently tried to please the Egyptian amir by building mosques in Ethiopia, for which he was arrested. Letters with threats from the amir to demolish churches in Egypt were despatched to Ethiopia, but the king's reply was uncompromising and unimpressed; "If you demolish a single stone of the churches, I will carry to you all the bricks and stones of Mecca . . . and if a single stone is missing I will send its weight to you in gold" (Atiya 1948: 347-51).

The patriarchal biographies continue, without much useful information about Ethiopian affairs, to relate the history of the metropolitan George (Atiya 1948: 394-5), consecrated by patriarch Michael IV (1092-1102). This prelate behaved so badly that he was arrested and sent back to Egypt. A successor, consecrated by Macarius II (1102-28), was the metropolitan Michael (Sergew Hable Sellassie 1972: 203). He was at the centre of a dispute in the time of patriarch Gabriel II (1131-45) with the reigning king of Ethiopia, who wanted him to consecrate more bishops than was allowed by custom (Atiya 1950: 56-7; Budge 1928: 800-1). This would have allowed the king to elect his own metropolitan, since ten bishops could constitute the synod necessary to do so, and Ethiopia and Nubia were therefore limited to seven. The Egyptian caliph was at first supportive of the king, but when the patriarch, anxious not to lose his influence in Ethiopia, pointed out that the Ethiopian king, absolved from his obedience to the patriarch, could attack Muslim lands with impunity, he seems to have changed his mind. In any event, famine in Ethiopia is supposed to have persuaded the king that he was wrong, and the attempt to gain independence for the Ethiopian church was over.

Michael's metropolitanate had one more trial to go through (Atiya 1950: 90-1). In 1152, when he was very old, a messenger arrived in Egypt from an Ethiopian king to the vizir with a request that patriarch John V (1146-67) replace Michael with a new metropolitan. Michael had apparently quarrelled with this king, who was a usurper. Possibly this usurping ruler may be identified with an early king of the Zagwé dynasty of Lasta, who seem to have come to power around 1137. The fact that he did not write to the patriarch until 1152 may mean that the king had not found Michael too troublesome until then, or that the affair took some time to build up, and does not really militate against his identification as a Zagwé ruler; alternatively, the 1152 incident might have been the culmination of the old quarrel of the time of patriarch Gabriel, with the Ethiopian ruler taking revenge against Michael for thwarting him previously. In any event, John V declined to replace the metropolitan, who had done no wrong and could not therefore be legally deposed, and for this refusal was himself imprisoned; he was eventually released on the death of the vizir.

While limited in scope, the biographies of the patriarchs do allow glimpses of Ethiopia, which, combined with those from the Arab historians and geographers, are not uninformative. We learn that the kingdom was, if erratically, in touch with Egypt, and that the monarchy and ecclesiastical structure remained intact until around the 950's. At some time before 1003 the foreign queen's rule was terminated, the Ethiopian kingdom restored and the church hierarchy reinstated with metropolitan Daniel; but one might imagine that the destruction of cities and churches, and the death or captivity of part of the population, left the country in a weaker condition than before. In the reconstruction, foreign influences were not lacking; in patriarch Zacharias' time (1004-32), Coptic Christians were allowed to emigrate to Abyssinia, Nubia, and Byzantine lands (Atiya 1948: 196), while near the town of Qwiha, in Enderta, funerary inscriptions of Muslims, perhaps a trading community, have been found dating from 1001-1154AD (Schneider, M. 1967); it may have been for such communities that some of Severus' mosques were built.

From the History of the Patriarchs we learn that in the ninth century the country suffered from war, plague, and insufficient rainfall, that it suffered a major disaster in the tenth century with the depredations of the queen of the Bani al-Hamwiyya. The remainder of the information from the eleventh and early twelfth centuries deals mainly with ecclesiastical questions, or in the case of the Arab geographers of repetitive and more or less inaccurate travellers tales. But the story about the attempt to increase the number of bishops in metropolitan Michael's time has some interesting facets. It mentions, unless this is mere rhetoric to emphasise the contrast with the later state of affairs after the patriarch blessed the land, that between 1131 and 1145 Ethiopia suffered several disasters. The king's palace was struck by lightning, and the land suffered from pestilence, famine and drought. Although the country recovered in due course, it must have been at about this time that the change of dynasty actually occurred, and Zagwé rule was established until c1270.

5.   The Capital City

1. The Site

Aksum was built on gently sloping land which rose, north and east of the city, to two flat-topped hills, now called Beta Giyorgis and Mai Qoho respectively. The hills around the town are formed from a granitic rock, nepheline syenite (Littmann 1913: II, 6; Butzer 1981). Between Beta Giyorgis and Mai Qoho runs the course of a stream, the Mai Hejja or Mai Malahso in its upper reaches, which rises on the eastern slopes of Beta Giyorgis. Further west another stream bed, that of the Mai Lahlaha, also descends from the top of Beta Giyorgis. Run-off from the Mai Hejja and down the flanks of Mai Qoho above the town is caught in a large excavated basin, officially called Mai Shum, but often referred locally to as the `Queen of Sheba's bath'. This is said to have been dug by one of the later metropolitan bishops of Aksum, Samuel, in the reign of king Yeshaq in about 1473 (Salt 1812; Monneret de Villard 1938: 49), but may very well be of Aksumite origin, enlarged or cleared by Samuel, just as it has been again enlarged and cleared recently. The basin lies directly below the north-west side of Mai Qoho, and access to it from above is aided by a series of steps cut into the rock, which may also date back to Aksumite times. Butzer thought that there was evidence for an earth dam some 50 m. below the Mai Shum reservoir (Butzer 1981: 479), and it is possible that much more water was caught or diverted in Aksumite times than today. Water was probably a very important element in the development of Aksum as the capital city of ancient Ethiopia. The name of the town itself is thought to be composed of two words, ak and shum, the first of Cushitic and the second of Semitic origin, meaning water and chieftain respectively (Sergew Hable Sellassie 1972: 68; Tubiana 1958). This name `Chieftain's Water' seems to suggest that Aksum could have been the site of a spring or at least a good water supply, and perhaps it early became the seat of an important local ruler.

Illustration 17. A view of Aksum taken from below Mai Qoho hill looking over the Stele Park towards Beta Giyorgis hill.

Illustration 18. The reservoir Mai Shum in its present enlarged state. Photos BIEA.

The two streams run south, that of the Mai Hejja skirting Mai Qoho hill (in its lower course it takes on different names, such as the Mai Barea, or the Mai Matare), and the Mai Lahlaha running directly through the town. Both lose themselves in the broad plain of Hasabo facing the town to the south and east. The two hills, both reaching to around the 2200 m mark above sea level, rise about 100 m above the town area, enclosing and sheltering it on two sides. Nowadays they are almost bare of trees, except where recent eucalyptus planting has occurred, but in Aksumite times there is reason to believe that they were probably forested to some extent. The geomorphologist Karl Butzer, exploring the area around Aksum, found that in the plateau of Shire, of which the Aksum region is part, were remnant stands of trees favouring a more moist climate, whilst the present montane savanna vegetation is the result of intensive human activity (Butzer 1981: 474-6). Only a few great sycomores still stand, some of which appear on Salt's aquatint of 1809, which also shows a fair scattering of other trees around the stelae and on the slopes of Beta Giyorgis.

The streams, which are seasonal, may either have run more continually in ancient times, when the rainfall was more constant, or have been supplemented by permanent springs. Possibly such springs helped to keep the Aksumite Mai Shum filled. In any event, travellers like Bruce (1790: III, 460-1) noted that there were springs functioning relatively recently and that the town was able to maintain gardens, though this was of course in its less populous days. Nathaniel Pearce, who lived in Ethiopia from 1810-1819, declared

"There is no river within two miles of Axum, but the inhabitants have good well water; there are many wells hidden, and even in the plain have been found, but the people are too lazy to clear them from rubbish. It appears probable that, in ancient times, almost every house had its well, as I have been at the clearing of four, situated at not more than ten yards from each other. The stone of which they are constructed is the same kind of granite of which the obelisks are formed" (Pearce 1831: 162-3).
A well was found near the Tomb of the False Door, probably sunk to serve one of the houses built over the Stele Park in later times (Chittick 1974: fig. 2). Such wells would have been essential for those who lived at a distance from the streams, and also would have helped to make the inhabitants more independent of the behaviour of the natural springs and streams available.

Alvares (Beckingham and Huntingford 1961: 155) mentions a "very handsome tank [or lake of spring water] of masonry [at the foot of a hillock where is now a market]" behind the cathedral, "and upon this masonry are as many other chairs of stone such as those in the enclosure of the church". Since there are thrones along the rock wall (thought by the DAE, who called it Mehsab Dejazmach Wolde Gabre'el, to be a natural formation (Littmann 1913: I, 31), but illustrated by Kobishchanov (1979: 118) as `cross-section of fortification embankment') on the west side of Mai Qoho, Alvares' description could possibly refer to this. The rock-wall, whether natural or man-made, could have acted as a retaining wall to waters overflowing from the Mai Hejja, or down the slopes of Mai Qoho, (or even from Mai Shum itself), and thus formed a lake of sorts along the foot of Mai Qoho. The word `mehsab' means something like `washing-place', which seems to confirm this idea.

2. The Town Plan

By the time that the Periplus Maris Erythraei was written, the town of Aksum, together with Meroë, capital of the Kushitic kingdom (though here the text is corrupt — Huntingford 1980: 19), was prominent enough to be called by the anonymous author of that work a `metropolis', a word reserved for relatively few places. It seems as if the main part of the town lay on either side of the Mai Lahlaha in the areas now known as Dungur and Addi Kilte. Here were all the élite dwellings found by the various archaeological expeditions (Littmann 1913; Puglisi 1941; Anfray 1972; Munro-Hay 1989). The Deutsche Aksum-Expedition traced the approximate ground plans of three very substantial buildings, which they called Ta`akha Maryam, Enda Sem`on and Enda Mikael after local identifications based on the Book of Aksum (Conti Rossini 1910), and they found traces of many others in the immediate neighbourhood. Subsequently, Puglisi, an Italian archaeologist, and Anfray, working on behalf of the Ethiopian Department of Antiquities, established that to the west were many more such structures. The whole area is scattered with the debris of the ruined buildings of this ancient quarter of the town. These large residences were basically, it seems, of one plan; a central lodge or pavilion, raised on a high podium approached by broad staircases, surrounded and enclosed by ranges of buildings on all four sides. The central pavilion was thus flanked by open courtyards. The plan shows a taste for the symmetrical, and the buildings are square or rectangular, with a strong central focus on the main pavilion. Ta`akha Maryam was furnished with an extra wing, and is the largest of such structures to have been excavated and planned so far.

How widespread the central part of the town was formerly is not yet known, but it may be assumed that the less permanent habitations of the poorer sections of the population were constructed all round the more substantial dwellings, and on the slopes of the Beta Giyorgis hill. Nothing of these has survived, but in time archaeologists may find evidence for the sort of dwellings we would expect; rough stone and mud, or wood, matting and thatch. One or two house models in clay found during the excavations give an impression of the smaller houses of Aksumite times (Ch. 5: 4).

Interest in fortification seems to have been minimal. The country itself was a natural fortress, enclosed within its tremendous rock walls and defended by its mountainous and remote position, as well as by the military superiority of its armies. Within the town, the pavilion style of dwelling, enclosed by inner courts and outer ranges of buildings, were in some measure given privacy, and if necessary defence, by their very layout. Kobishchanov writes of fortified bastions around the sacred area, but these were in fact only the outer walls of the large structure of typical Aksumite plan which now lies beneath the cathedral, not walls for specific defence reasons (Kobishchanov 1979: 141; de Contenson 1963). Nothing is known about the street plan of the suburbs where these mansions lay, and whether they too partook of the prevailing liking for symmetry by using a grid-pattern. The outer parts of the town very likely developed organically in a piecemeal fashion, and were in a constant state of alteration, enlargement, or rebuilding as structures decayed or developed. Such a development can be seen on the plan (Anfray 1974) of the excavated structures at Matara.

Map C. Plan of the ancient city of Aksum.

Flanking the town in various directions were the necropoleis or cemeteries. These are marked today by the numerous granite stelae, standing stone monuments which vary from rough and simple marker-stones to some of the largest single stones ever employed in human constructions. Fields of such stelae are found in the following locations;
  to the south of Dungur (called the `Gudit' field, after the legendary queen who sacked Aksum in the tenth century);
  to the east of the town below the south side of Mai Qoho;
  on the top of Beta Giyorgis to the north of the two church buildings there excavated by Lanfranco Ricci (1976);
  and above all in the main stelae field running along the north side of the Mai Hejja. This latter can be divided into two, a northern group in the area known as Geza `Agmai, and the main, southern group, ending almost opposite the cathedral and embracing the recently-made `Stele Park', which includes all the decorated monoliths, and some enormous granite-built tombs (Littmann 1913; Munro-Hay 1989).

Rock-cut or built tombs are also to be found on the slopes of Mai Qoho hill, in the courtyard of the church of the Four Animals (Arbate Ensessa) in Aksum, and in the region of the so-called tombs of Kaleb and Gabra Masqal some 2 km. north of the present town (Littmann 1913; Munro-Hay 1989). It is evident that there are many other structures in this region and to the north; an idea of the density of occupation can be gained from the survey work of Michels quoted in Kobishchanov (1979: Map 4).

The southern group of stelae in the main stelae field marks what was evidently the chief necropolis of the city, and the royal burial place. Directly facing it was a religious and ceremonial area now occupied by the sacred enclosure called Dabtera where the two cathedrals, old and new Maryam Tseyon, stand. The base of the podium of the old cathedral is an Aksumite structure, and other very substantial buildings have been traced in the area (de Contenson 1963i); if the custom of building churches on former sacred spots was followed here also, these may be traces of earlier cathedrals or pre-Christian temples. The architectural ensemble of this part of the town, of cathedral/temple, thrones, and stelae, shows an arrangement which may owe something to intentional design, but which was evidently also an extended process. The earlier examples among the great decorated stelae may well have been situated one after the other following a deliberate design; this impression is much stronger for the last three, which, dominating the terrace of lesser stelae, must have offered a sight which for dramatic quality was rarely equalled in the ancient world.

Near to and facing the old cathedral, is a cluster of granite thrones (Littmann 1913: II, 45ff) of which now only pedestals remain. Most of them are in a row running approximately north-north-east to south-south-west. There are eleven in the row, two being double, with another two immediately in front of the main row. At least six of the thrones had some sort of pillared canopy, as emplacements for pillars can be seen in the stonework of the pedestals. Slots for their backs and sides show that the original design was for closed chairs like the picture (Wolska-Conus 1968) of the Adulis throne in Kosmas' book, and very likely at least some of these now-missing slabs bore inscriptions as did the Adulis monument. Some way to the southeast, between the row of thrones and the inner enclosure of the church, stand two other throne-pedestals, one with four columns still erect, and another set on a massive plinth. The throne-bases are noted in the Book of Aksum (Conti Rossini 1910) as the thrones of the Nine Saints, with others for Kaleb, Gabra Masqal and so on, or are attributed to the twelve judges of Aksum. In later times they served in the ceremony of the coronation. They may well have been the thrones which inscriptions tell us were set up as memorials of victories or other great events, like the one which still existed at Adulis in the sixth century, when Kosmas copied its anonymous Aksumite inscription. One of Ezana's inscriptions, DAE 10, (see Ch. 11: 5) mentions a throne set up `here in Shado', possibly the ancient name of one of the two places at Aksum still marked by rows of thrones.

The second set of pedestals led in a row from beside the eastern stelae field towards the ceremonial centre of the town, and some of these still show traces which indicate that they once held statues. Plinths for statues are known from other parts of the town also, one having sockets for feet 92 cm long (Littmann 1913: II, 44, provided a photograph of this now-vanished monument). Perhaps this sort of monument gave rise to the legend that when Christ descended to earth to perform the miracle of filling up the lake where Abreha and Atsbeha later built the cathedral, he left his footprints in the rock; they were, according to the Book of Aksum, still visible in the fifteenth century. Some of the Aksumite inscriptions mention the erection of metal statues as victory memorials, but as yet only stone statues have been found in Ethiopia. These, of which the finest examples come from Hawelti, near Aksum, date to some centuries before Aksumite times, but there may have been a continuity of tradition from one period to the next (de Contenson 1963).

Illustration 19. One of the stone thrones or statue-bases which lined the entrance avenue leading to Aksum's main Stele Field. Photo BIEA.

The town-plan of Aksum is thus fairly simple; it may be envisaged as commencing with a ceremonial approach from the east, lined with granite victory-thrones and statues of bronze and precious metals dedicated to the gods, leading to the religious centre with the royal cemetery lying to the north and east. The focus for this region seems to have been the temple/cathedral area, with another row of thrones. The main residential suburb with its huge palaces was situated to the west; and the whole was flanked with lesser cemeteries and more humble residential suburbs. It is probable that there was at least one open square, a market-place perhaps, somewhere in the town centre. Since inscriptions and a statue base are reported to have come from the area between Ta`akha Maryam and Enda Sem`on (Littmann 1913; Schneider 1974), it may have been situated there, as such monuments may well have been set up in a public place. Civic building has not been identified; nothing has yet been excavated which can be categorised as public architecture, such as the structures housing town administrations, law-courts, covered markets or shopping arcades, baths, and the like so common in Roman town centres. As we have noted above, there is no hint that the Aksumite rulers needed to dominate their towns with citadels, or surround them with defensive walls, and the town must have simply petered out in the plain and on the slopes of the hills.

3. Portuguese Records of Aksum

The most significant of the Portuguese accounts of Aksum is undoubtedly that of Francisco Alvares (or Alvarez), who came to Aksum in 1520. He was a careful and sympathetic observer who noted a good deal about the town, including details of many of the then extant monuments. His account has the special significance that it was the only one made before the sack of the city by the Muslim leader Ahmad Gragn. Several of the buildings which he mentions now do not exist (at least on the surface), but from the accuracy of those descriptions which can be checked, it is evident that his statements are worthy of respect. His description of Aksum was as follows (from Beckingham and Huntingford 1961); the square brackets indicate additions in Ramusio's Italian edition, apparently made from a different manuscript than that published in Lisbon in 1540.

Chapter XXXVIII.

(After the description of the church of Maryam Tseyon; see Ch. 10: 5);

Inside the large enclosure (the outer enclosure around Maryam Tseyon) there is a large ruin built in a square, which in other times was a house and has at each corner a big stone pillar, square and worked [very tall with various carvings. Letters can be seen cut in them but they are not understood and it is not known in what language they are. Many such epitaphs are found.] This house is called Ambaçabet, which means house of lions.
It is not known what this structure, which has now disappeared, was.
Before the gate of the great enclosure there is a large court, and in it a large tree which they call Pharaoh's fig tree, and at the end of it there are some very new-looking pedestals of masonry, well worked, laid down. Only when they reach near the foot of the fig tree they are injured by the roots which raise them up. There are on the top of these pedestals twelve stone chairs [arranged in order one after the other] as well made with stone as though they were of wood. They are not made out of a block, but each one from its own stone and separate piece. They say these belong to the twelve judges who at this time serve in the court of Prester John.
These are the thrones, of which only the pedestals now exist.
Outside this enclosure there is a large town with very good houses . . . and very good wells of water of [very beautiful] worked masonry, and also in most of the houses . . . ancient figures of lions and dogs and birds, all well made in [very hard, fine] stone. At the back of this great church is a very handsome tank [or lake of spring water] of masonry, [at the foot of a hillock where is now a market] and upon this masonry are as many other chairs of stone such as those in the enclosure of the church.
This seems to refer to the row of thrones set on what seems to be a natural rock wall at the base of Mai Qoho; possibly in the sixteenth century this wall acted as a retaining wall for water, perhaps overspill from Mai Shum?
This town is situated at the head of a beautiful plain, and almost between two hills, and the rest of this plain is almost as full of these old buildings, and among them many of these chairs, and high pillars of inscriptions [; it is not known in what language, but they are very well carved].

Above this town, there are very many stones standing up, and others on the ground, very large and beautiful, and worked with beautiful designs, among which is one raised upon another, and worked like an altar stone, except that it is of very great size and it is set in the other as if enchased (the standing stele and its base-plate). This raised stone is 64 covados in length, and six wide; and the sides are 3 covados wide. It is very straight and well worked, made with arcades below, as far as a head made like a half moon; and the side which has this half moon is towards the south. There appear in it five nails which do not show more on account of the rust; and they are arranged like a quinas (the five dots on dice). And that it may not be asked how so high a stone could be measured, I have already said how it was all in arcades as far as the foot of the half moon, and these are all of one size; and we measured those we could reach to, and by those reckoned up the others, and we found 60 covados, and we gave 4 to the half moon, although it would be more, and so made 64 covados (thus making it about twice too great, the covado being apparently 27 inches). This very long stone, on its south side, where the nails in the half moon are, has, at the height of man, the form of a portal carved in the stone itself, with a bolt and a lock, as if it were shut up. The stone on which it is set up is a covado thick and is well worked; it is placed on other large stones, and surrounded by other smaller stones, and no man can tell how much of it enters the other stone, or if it reaches to the ground. [Near these] there are endless other stones raised above the ground [, very beautiful] and very well worked [; it seemed as if they had been brought there to be put to use, like the others that are so big and are standing up]; some of them will be quite forty covados long, and others thirty. There are more than thirty of these stones, and they have no patterns on them; most of them have large inscriptions, which the people of the country cannot read, neither could we read them; according to their appearance, these characters must be Hebrew (perhaps actually Epigraphic South Arabian; does this mean that originally there were some inscriptions in the Stele Park?). There are two of these stones, very large and beautiful, with designs of large arcades, and tracery of good size, which are lying on the ground entire, and one of them is broken into three pieces, and each of these equal eighty covados, and is ten covados in width. Close to them are stones, in which these had been intended to be, or had been enchased, which were bored and very well worked.

These are the two largest stelae, but now the base-plate of the largest is missing.

Chapter XXXIX.

Above this town which overlooks much distant country [on every side], and which is about a mile, that is the third of a league, from the town, there are two houses under the ground into which men do not enter without a lamp (the `Tombs of Kaleb and Gabra Masqal'). These houses are not vaulted, but of a very good straight masonry, both the walls and the roof. The blocks are free on the outside. The walls may be 12 covados high; the blocks are set in the walls so close one to the other, that it all looks like one stone [for the joins are not seen]. One of these houses is much divided into chambers and granaries. In the doorways are holes for the bars and for the sockets for the doors. In one of these chambers are two very large chests (the sarcophagi), each one 4 covados in length, and one and a half broad, and as much in overall height, and in the upper part of the inner side they are hollowed at the edge, as though they had lids of stone, as the chests are also of stone (they say that these were the treasure chests of the Queen Saba). The other house, which is broader, has only got a portico and one room. From the entrance of one house to that of the other will be a distance of a game of Manqual (a type of skittles) and above them is a field. . . . In this town and in its countryside . . . when there come thunderstorms . . . there are no women or men, boys or children . . . left in the town who do not come out to look for gold among the tillage, for they say the rains lay it bare, and that they find a good deal.
This remark was taken by the editors of Alvares' book to mean that gold was actually washed out of the soil; it is, however, much more likely to refer to the finding of gold coins and other items in the earth, something still not infrequent at Aksum after the rains. Alvares also describes the now-disappeared western church of St. Michael with a `tower of very fine masonry' and the two shrines of Abba Liqanos and Abba Pantelewon.

Illustration 19a. The title page of Telles' 1660 Historia de Ethiopia a alta after d'Almeida; some attempt has been made to `Ethiopianise' the figure of the king and his courtiers.

About one hundred years later, after Ethiopia had passed through the great convulsions of Gragn's wars, Manoel de Almeida described the town in his chapter XXI, Acçum and its Antiquities (Beckingham and Huntingford 1954: 90ff);

It is situated on the edge of very broad meadows in a gap where they come in between two hills. Today it is a place of about a hundred inhabitants. Everywhere there ruins are to be seen, not of walls, towers, and splendid palaces, but of many houses of stone and mud which show that the town was formerly very large.
Much of this was presumably the remains of the early sixteenth century town.
A church of stone and mud, thatched, is to be seen there built among the ruins and walls of another, ancient, one, the walls of which are still visible and were of stone and mud too (for in no part of Ethiopia is there any sign or trace that lime has ever at any time been seen there, or any building, large or small, constructed with it) but very wide apart. From what is visible, the church seems to have had five aisles. It was 220 spans long and 100 wide, it has a big enclosure wall of stone and mud and inside it a very handsome courtyard paved with large, well cut stones, ending, on the side the church is, in a flight of 8 or 9 steps, also made of well cut stones. At the top is a platform of 10 or 12 covados in the space before the façade and principal door of the church.

Outside this church's enclosure is another in which five or six big pedestals of black stone are to be seen. Near at hand are four columns of the same stone 10 or 12 spans high. Among them is a seat on which the Emperors sit to be crowned after first having taken his seat on the pedestals I mentioned and after various ceremonies have been performed on them

(see Ch. 7: 6 for an account of the coronation).
What is most worth seeing here, a display of presumptuous grandeur, is many tall stones like obelisks, needles and pyramids. They are in a meadow lying behind the church. I counted some twenty that were standing and seven or eight that have been thrown to the ground and broken in many fragments. The tallest of those standing, if measured by its shadow, is 104 spans. Its width at the base is ten spans, it becomes thinner as it goes up, like a pyramid, but it is not square; it has two sides broader and two narrower than the other two. It is carved as though in small panels each of which is like a square of two spans. This is the style of all those which have this carving, which are the taller ones. The rest are rough and unshaped slabs without any carving at all. The shortest are from 30 to 40 spans; the rest are all taller. It can be seen from the fragments of three or four of those that have been overthrown, that they were much bigger than the tallest of those now standing, which I said was a hundred and four spans, and some can be seen to have been over two hundred. The old men of this country say that a few years ago, in the time of King Malaac Cegued, and the Viceroy Isaac who rebelled and brought in the Turks to help him against the Emperor, they overthrew the six or seven that lie on the ground in fragments.

No one can say what was the object of the former kings who raised them up. It may well be thought that they were like mausoleums erected near their tombs, since this was the object of the Egyptians. It was no doubt from them, through their proximity and the constant communication there was between them, that they learnt about, and that the workmen came to make, these barbarous and monstrous structures. A bombard shot away from this spot is a broad stone not much higher than a man on which a long inscription can be seen. Many Greek and some Latin letters are recognisable, but when joined together, they do not make words in Greek, Latin, Hebrew, or any other known language, and so the meaning of the writing is not discoverable.

Balthasar Tellez also added a little to the picture of Aksum in the seventeenth century, in his book published in Portugal in 1660, and later translated into English (1710); much of his information repeats de Almeida's.
At this time there is no settled city in all Ethiopia; formerly the town of Aczum was very famous among the Abyssinians, and still preserves somewhat of its renown; and this place seems to have been a city, at least they look upon it as most certain, that the Queen of Sheba kept her court there, and that it was the residence of the emperors for many ages after, and they are crown'd there to this day . . . this is the city Aczum, or Auxum . . . at present it is only a village of about 100 houses. There are to be seen many ancient ruins, particularly those of a spacious church. . . . The most magnificent thing that appears here, are certain very tall stones, in the nature of obelisks, or pyramids, the biggest of them 78 foot in length, the breadth at the foot seven foot six inches. It is cut as it were in small cushions, each of them about half a yard square; the smallest of them being between 25 and 30 foot high are rude misshapen stones. Some of those which seem to have been tallest are thrown down, and they say, the Turks entering Ethiopia overthrew them. The end of erecting them may reasonably suppos'd to have been for monuments, near their graves; which was the design of the Egyptians in their so famous pyramids. Here is also a stone set up with a large inscription, in Greek and Latin characters, but they do not make any sense.

Illustration 20. View of part of the central pavilion at Dungur, showing the granite corner-blocks, one of the re-entrants, and the rebates in the walls.

4. Aksumite Domestic Architecture

The general style of the élite domestic buildings of Aksum has been described above (Ch. 5: 2), and reconstructions have been attempted (in Littmann 1913; and, more modestly, by Buxton and Matthews 1974). The pavilions in their domestic enclosures are the most typical examples of the unique Aksumite form of construction, and embody most of the characteristics of Aksumite architecture. The podia (the only parts of these buildings which survive, except in very rare cases) were built according to a style whereby the walls exhibited no long straight stretches, but instead were indented, so that any long walls were formed by a series of recesses or re-entrants and salients. These, considering that the building material was mostly random rubble (coursed stonework is rarely found), bound only with mud mortar, must have been designed to strengthen walls with low cohesion inclined to sag, and to deal with expansion and contraction caused by widely-ranging temperatures. Later structures, such as the Lalibela churches, show that the indentations were not used only on the podia, but extended to roof level.

At intervals of a little less than 50 centimeters as the podium walls rose they were rebated, each rebate setting the wall back about 5 centimetres. These rebates or gradins, sometimes up to seven, were often topped with flat slate-like stones forming a shelf or string-course. This design was perhaps inspired by the same architectural tradition illustrated in the pre-Aksumite period at the Grat-Beal-Guebri and the temple at Yeha (Anfray 1972ii: 58). It not only narrowed the walls as they rose higher, but assisted the run-off of rain from the surfaces of the walls, and thus protected the mud-mortar to some extent. The rebated walls with re-entrants and salients are one of the chief distinguishing marks of Aksumite architecture, and apart from their reinforcing function they would have enhanced the appearance of the massive podia of the palaces and mansions by breaking up the solidity of their mass with light and shadow. A wall at the structure called Enda Sem`on was remarkable in that it had not only five surviving rebates, but the wall above was preserved for a further 1.80m, including a blocked-up window with a stone lintel (Munro-Hay 1989).

Most of the surviving podium walls of these Aksumite structures were furnished at all corners with large and carefully cut granite corner blocks, which protected, linked, and supported the weaker parts of the walls. Occasionally a podium might be strengthened by a complete row of cut granite blocks, as still visible at Dungur and in the Aksumite part of the Maryam Tseyon cathedral podium. Granite was also used for architectural features such as columns, bases and capitals, doors, windows, paving, and the like, and particularly for the massive flights of steps which sometimes flanked two or three sides of the pavilions. A good deal of this stonework consisted of undecorated but well-dressed blocks, but some of the doorway blocks, or the columns and their bases and capitals, were decorated with a variety of designs.

A thick lime plaster was noted on the walls of one room in the large tomb called the Mausoleum at Aksum (Munro-Hay 1989), and similarly appears on a chamfered column at Maryam Nazret. Lime mortar has also been observed fixing stones on the podium of the Aksumite church at Agula, but it does not appear to have been regularly used.

At Adulis the main construction material was porous basalt (the same material was used for the stele at Adulis which Kosmas saw behind the marble throne there, lying broken into two pieces — Wolska-Conus 1968: 364) or sandstone. Polygonal blocks of basalt were used for walls, and cut cubes were assembled to form square columns. During the excavations at Adulis, Paribeni (1907: 464) was puzzled by the lack of doors and windows in walls he was able to clear up to a height of 3.40m. He concluded that they must only represent foundations, and this was in a measure true, since the buildings rose on podia as indicated by the excavations at Matara and Aksum. Among the structures which Paribeni cleared was one which he called the Ara del Sole, Altar of the Sun, because of a number of designs which he interpreted as hills and sun-discs carved on bluish marble plaques destined for fixing to the walls. A church had been built on top of the original podium, but the latter conformed in all other ways to the usual Aksumite style.

A second type of architecture, though similar in most essentials to that already described, employed wooden beams as a strengthening element within the walls. A square horizontal beam set in the wall supported rounded cross-members embedded in the stonework and forming ties across the width of the walls. The ends of these cross-members projected from the external wall (and were sometimes visible internally as well) in rows, forming the characteristic `monkey-heads' often seen fossilised in stone in other examples of Ethiopian architecture. Doors and windows were constructed by a similar method, the openings being framed on all four sides and linked by cross members through the thickness of the walls; but the `monkey-heads' are square. Most of these features were carved in granite on the decorated stelae, and some can still be seen in the surviving ancient rock-cut or built churches of Tigray and Lalibela (Buxton and Matthews 1974; Plant 1985; Gerster 1970). The framed doors and windows appear as a repeated motif in some of these later structures, and have been dubbed the `Aksumite frieze', since they appear on the decorated stelae there (Plant 1985: 17, 20). However, the simple lintel was also known and employed, for example over the doorway of the East Tomb at Aksum and over a window at Enda Sem`on (Munro-Hay 1989).

Illustration 21. The lower part of the still-standing decorated stele at Aksum, showing the imitation in stone of wooden architectural elements.

Illustration 22. Apparently created long after the Aksumite period, the churches of Lalibela still employed the same general style of architecture, including imitation beam-ends in stone.

It is possible that the original inspiration for the design of the decorated stelae came from the South Arabian mud-brick multi-storey palaces familiar to the Aksumites from their involvements in that country, rather than from Ethiopian examples. On some of the Aksumite podia there could conceivably have been erected high tower-like structures of mud-brick around a wooden frame, such as that found at Mashgha in the Hadhramawt (Breton et al. 1980: pls. VIII, X) looking rather like the great stelae. But no evidence for such Yemeni-style buildings actually survives in Ethiopia, nor is there any archaeological indication there for mud-brick architecture. Alternatively, and more probably, the stelae could have been exaggerated designs based on the Aksumite palaces; and here there is archaeological support, since the structure called the `IW Building' partly cleared by the excavations of Neville Chittick (Munro-Hay 1989), included just such wood-reinforced walls. Though the load-bearing strength of the rough stone and mud mortared walls is apparently very considerable, it seems most likely that these buildings would have been in reality limited to only two storeys above the podia. Evidence that the pavilions and some of these outer ranges were more than one storey high is provided by the occasional staircases which have been found. The central pavilions of Aksumite palaces were completely surrounded by ranges of subsidiary structures, pierced here and there by gateways and doors. Each ensemble must have formed very much the sort of thing mentioned by the sixth-century merchant Kosmas, who speaks of the `four-towered palace of the king of Ethiopia' (Wolska-Conus 1973). The recessed central parts of each facade may have reached a storey less in height than the corner salients, giving the impression of towers (as shown on the reconstructions). Kobishchanov's eight-storied palace (1979: 141) of Ta`akha Maryam is probably an error for the number of stepped shelves which constituted the building's podium (Schneider 1984: 165).

Some of these structures were of very considerable size; Ta`akha Maryam measured 120 × 80 m, and its pavilion, at c. 24 × 24 m was the smallest of the three the German expedition cleared; if the other two were in proportion their overall size must have been very large indeed. Ta`akha Maryam thus covered around six times the total area of the more-or-less contemporary palace and portico of the kings of the Hadhramawt recently excavated (Breton 1987) at their capital of Shabwa, and, as a single architectural concept rather than an agglomeration of buildings, was larger than many European palaces (excluding such monumental constructions as the Roman and Byzantine Great Palaces) until the erection of such buildings as Hampton Court. Kosmas, when speaking of Kaleb's palace, sometimes simply refers to it as the royal dwelling, but on other occasions uses the latin word palatium, surely in recognition of its particular splendour (Wolska-Conus 1973: 321, n. 4.3).

Illustration 23. A column base from Aksum, possibly originally from the peristyle in the centre of the south wing of Ta`akha Maryam palace.

Illustration 24. At the mansion of Dungur, some of the basement rooms contained rough stone supports, perhaps for wooden columns.

The central pavilion at Ta`akha Maryam contained nine rooms, two of which were probably simply staircase-wells for access to the upper storey. The largest room was 7 × 6 m, and others measured 5 × 5 m, 7 × 4 m and 6 × 5 m. All had their roofs supported by two, three or four columns, and some had carefully flagged floors. In the south wing was a central peristyle with octagonal column-bases, and leading to the north corner-buildings were four-columned porticoes with elaborate floral column bases. The central pavilion of Enda Mikael measured 27 × 27 m with 10 rooms, following the same pattern as Ta`akha Maryam but with the central room divided into two. Room sizes were 6 × 6 m, 4 × 10 m, 5 × 9 m, and 3 × 9 m, with emplacements for four, eight or nine columns. The most substantial pavilion found to date was that in Enda Sem`on, 35 m square, with two enormous halls, each with twenty-eight column emplacements and measuring some 19 × 10 m; impressive dimensions, but needing more and more roof-support as the room sizes grew more ambitious, which must have resulted in a rather crowded effect. The lack of stone columns, commoner in the eastern Aksumite sites (Anfray 1974: 747), suggests that carved wooden ones were used, perhaps resting on rough stone pedestals as at Dungur, in some of the Matara buildings, and in some rooms excavated at Adulis (where they were capped with discs of basalt). These descriptions rest mainly on the published plans of the Deutsche Aksum-Expedition, which depend in parts on their assumption, probably correct as far as subsequent excavation has shown, that most buildings were more or less symmetrically arranged (see for example the plans in Littmann 1913: II, taf. XVII-XIX).

Anfray (1974: 762) suggested that the idea for such buildings ultimately derived from north Syria, and thought that `un certain caractère de sobriété, de rigidité, de rationalité même dans cette architecture axoumite . . . paraît d'inspiration romaine'. Whilst this may be partly true, a good deal of the inspiration might equally be derived from earlier local examples; the Yeha temple could hardly be plainer or more simple.

The architecture of the Aksumite élite residences should tell us something about the intentions and the character of the people who had them built, but this is in reality hard to interpret. The massiveness and solidity of the structures, and their simplicity and plainness, do indeed impress at first, but here we may well be missing such decorative elements as carved wooden columns, capitals and screens, and interior painting on plaster. Though in most Aksumite sites very few fragments of anything like elaborate carved stone or plaster-work have come to light as yet, churches in Tigray and Lalibela exhibit a rich selection of (albeit somewhat later) decorative elements (Plant 1985; Gerster 1970). These include painting on walls and ceilings, imaginative designs for windows, carved friezes, and carved wooden roof panels — some decorative woodwork survives at Dabra Damo, of uncertain date, and conceivably coming originally from a palace (Gerster 1970: 73). Some of this may well have been of Aksumite origin.

At Adulis, where perhaps more foreign influences might be expected, Paribeni (1907, loc. var.) found several examples of carved marble or basalt panels and decorative elements for affixing to walls, and carved architectural features such as acanthus or lotus capitals, or alabaster or limestone reliefs with formal floral designs or intertwined (vine?) leaves and branches; one also depicted a bird, possibly a peacock. Paribeni also found decorative marble colonnettes for framing screens — though these seem to have been imported ready-made from the eastern Mediterranean region, like those from an Adulite church excavated by the British in 1868 (Munro-Hay 1989i). He even found traces of lines, bands and leaves painted in red and brown on plaster in one house at Adulis. Most unusual among Paribeni's discoveries were plaques of a black schist, carved with shapes resembling oak leaves, which were cut in such a way as to accommodate metal inlay. At Aksum, the largest stele makes one concession to decoration with its filigree window-screens of superimposed stepped crosses under arches on the top storeys, perhaps modelled on something like the alabaster screens found until today in Yemeni houses. Constructional details like the beam-ends seem to have been left plain and visible. The churches of Lalibela, and others in Tigray, are nevertheless quite restrained where the architecture is concerned, however elaborate their interior paintings might be. As noted above, in the pavilions, smallish rooms, or larger halls thronged with columns, were a necessity given the limited means of spanning spaces, but the inevitable rather cramped feeling may have been largely offset by the use of open porticoes and wide courtyards.

A taste for the dramatic and the exclusive can perhaps be read into the appearance of the central pavilions in these courts, raised high on their podia, isolated by the courtyards surrounding them, and approached by massive flights of usually about seven steps. Such a design may be an expression of the special position of the rulers translated into architectural terms. There may have been an intention to isolate the pavilions as a convenience for security, but although the whole ensemble of pavilion, courtyards, and outer ranges was evidently to some extent defensible, that does not seem to have been a primary consideration. The Aksumites could surely, had they wished, have made stronger fortresses than these.

Some clay models of houses survive which illustrate the architectural style of the smaller Aksumite dwellings. A round hut, with a conical roof thatched in layers, and a rectangular doorway, is one type from Hawelti (de Contenson 1963ii: pl. XXXVII, b-c). A second type from Hawelti is rectangular, the doors and windows also rectangular, with a roof supported by beams whose `monkey-head' ends can be seen below the eaves. The roof has a small parapet and there is a waterspout to drain it (de Contenson 1963ii: pl. XXXVIII-XXXIX). A third type (de Contenson 1959: fig. 8) from Aksum shows only the remains of the bases of typical Aksumite window-apertures with square beam-ends at the corners; this may also represent a rectangular or square dwelling. In the BIEA Aksum excavations fragments of a fourth type of house, also rectangular, but with a roof consisting of sloping layers of what appears to be thatch of some sort, were found (Chittick 1974: fig. 21a). Its doorway seems to be surmounted by a dentilled lintel, like that of the largest stelae and the Tomb of the False Door (see below).

The last of these house-types is particularly interesting, in that it shows a pitched roof on a rectangular building. Such a roof was evidently of advantage in the rains, and may have been used on larger structures as well. Possibly the palaces themselves were roofed with thatch; the columns would have supported cross beams, perhaps with carved panels like those from Dabra Damo inset between them, and above some sort of layered thatch could have completed the weather-proofing (for some discussion of roofing in Ethiopian structures, see Buxton and Matthews, 1974). No trace of Roman-style tiles has yet been reported from Aksumite sites, not even from Adulis, nor do the brick vaults known to have been used in tombs (see below) appear to have been employed in domestic architecture as far as present evidence reveals. Paribeni does note (1907: 545) a report that some buildings at Tekondo were roofed with slabs of slate. He also made a few comments about domestic housing in Adulis, noting that open areas, perhaps for sleeping, would be useful in the hot climate of Adulis, and suggesting that some of the structures found without doorways could perhaps represent partially underground dwellings with wood or straw upper parts. These would have been entered from above by ladders, and perhaps were occupied by some of the `Cave-dwellers' mentioned by the texts.

5. The Funerary Architecture

Apart from rock-cut tombs of various types, and others constructed by walling excavated pits, the Aksumites built some much more elaborate tombs. The chronology of these is uncertain, but some idea has been gained from the stratigraphical evidence provided by recent excavations (Munro-Hay 1989). The tombs show that the Aksumites were deeply concerned with the well-being of their kings and other citizens after death, and from the finds in one partially-cleared tomb, called the Tomb of the Brick Arches, we can see that rich funerary goods were buried with them.

Illustration 25. Part of the great top-stone of the tomb called Nefas Mawcha.

Illustration 26. View of the great stele and the Nefas Mawcha, showing how the stele, in falling, struck the corner of the roof and destroyed the equilibrium of the tomb. Photos BIEA.

Perhaps the most extraordinary of all the funerary structures at Aksum is the tomb called (at least since the time of the German visit in 1906) Nefas Mawcha, or `the place of the going forth of the winds'. This name may be derived from a legend, related in the Book of Aksum (Conti Rossini 1910), that at the foot of the largest stele lay tunnels where winds blew out any lights. The tomb is (cautiously) dated to the third century AD (Munro-Hay 1989). It consists of a gigantic single roofing block, measuring about 17 × 7 × 1 m, placed over a paved chamber surrounded by ambulatories on all four sides. These passages were also roofed with granite blocks, those of the inner ambulatory (perhaps actually a rubble-filled supporting wall) fitting under both the great roofing block and the outer ambulatory roof blocks. The stones were trimmed to fit at either end, and linked together with metal clamps, the holes for which are still visible. This huge structure was almost certainly intended to be covered over by earth as an underground tomb. No entrance survives since the largest of the stelae, in falling, destroyed the west end of the building and caused the rest to settle as a result of the shock.

An unexpected find was the Tomb of the Brick Arches. The tomb itself lay beneath a rough-stone and mud-mortared superstructure, most of which has now disappeared, and whose original form cannot be reconstructed. Between two parallel walls a staircase roofed with rough granite slabs descended until the tomb's entrance was reached. The first sign of anything unusual was the discovery of a granite lintel, and then, underneath it, the upper part of an arch of baked bricks. As the excavations progressed, it became apparent that this was a horse-shoe shaped arch, forming about three-quarters of a perfect circle, which rested at each side on slate-like stones forming plinths supported by the usual Aksumite rough stone and mud-mortared walls. The entrance led to an antechamber, from which two further horse-shoe shaped arches led into the tomb-chambers proper. All had been blocked with stones, and all had been broken open in ancient times when the tomb was partially robbed.

Illustration 27. The Tomb of the Brick Arches. View from inside the vestibule, looking through the horseshoe arch towards the staircase. Photo BIEA.

The entrance-arch had an internal measurement of 1.3 m across the widest point, and the bricks were square. One of the internal arches resembled this, but the second was rather different, with oblong bricks arranged so that the long and short sides followed each other alternately. The square bricks measured 27 × 28 × 7 cm.

The contents of the tomb have been tentatively dated by various methods to the early/mid fourth century AD. To find horse-shoe shaped baked brick arches of this early date in Ethiopia was very surprising, and of great interest for the history of architecture. Horse-shoe shaped arches are known from an earlier period in India, a country with which Aksum had vigorous trading relations from probably the first century AD, but these arches were carved from the rock and not built. More or less contemporary built examples are reported from Syria, and so the Ethiopian examples have a pedigree as old as any others, at least for the time being (Munro-Hay, Rassegna di Studi Etiopici, forthcoming).

Some distance away was found the so-called Brick Vaulted Structure, presumed to be a tomb of the same date as the Tomb of the Brick Arches, since it was also situated in the main necropolis and similarly employed brick horse-shoe shaped arches. But it also included relieving arches and lintels, and the rooms were barrel-vaulted with brick. These bricks were mortared together, and it is evident that the Aksumites knew the use of mortar (nb. de Almeida's statement above, Ch. 5: 3), but rarely felt the need to employ it, preferring their drystone walling with simple mud-bonding.

The Brick Vaulted Structure first appeared during the excavations (Munro-Hay 1989) as a stone wall of Aksumite style, built parallel to the courtyard in front of the Tomb of the False Door to the west. In due course, a number of bricks began to appear, soon proving to be the remains of collapsed brick vaults. These consisted of double rows of square baked bricks forming radial barrel vaults resting on string-courses of slate-like stone on top of the usual Aksumite stone and mud-mortared walls. The chambers covered by the vaulting seem to have been approximately 1 × 2 m in size, and one retained traces of the stone-paved floor of a superstructure over the barrel vaulting. The height of the vaulted rooms was about 4 m, and a tentative reconstruction seems to indicate that they flanked a central passage.

The vaults themselves were not horse-shoe shaped. But the entrance to one of the vaults (the only entrance found) was formed by a horse-shoe shaped arch, also 1.3 m wide across the centre, sealed with a stone blocking, and surmounted by a granite relieving lintel above which the bricks of the vault rose. This revealed a new and more complex combination of architectural features, which, as far as our present knowledge goes, is entirely unique. It seems as if the structure originally had a number of these vaulted rooms opening off a central corridor, but the complete plan has not yet been completely recovered.

Illustration 28. Drawing of the granite entrance doorway to the tomb called the `Mausoleum'. Drawing BIEA.

A further tomb, probably the largest yet known at Aksum, was entered by a monumental granite doorway in typical Aksumite style, with carved granite square-headed beam ends protruding at the corners. This tomb was dubbed the `Mausoleum', as a testimony to its size and elaborate construction, both totally unexpected by the excavators. Its plan consists of a long corridor behind the stone doorway, also entered from above by three shafts, and flanked by ten rooms, five on each side. It has not yet been cleared, only planned by crawling through the narrow gap left between the mud fill and the roof. The tomb is about 15 m square and lies to the west of the foot of the largest stele. The entrance to another tomb was found on the east side of the stele with a simpler doorway of rough stone topped by a granite lintel. Both of these tombs opened onto a courtyard at the foot of the stele, which must have been filled in before the collapse of the stele. The `Mausoleum' was built largely of rough stone walling roofed with granite blocks, and was covered with huge quantities of dry stone fill. It may belong to the person for whom the giant stele was raised. At the west end of its central corridor can be seen the top of another brick arch, leading into a passage not yet entered; but whether it was of the horse-shoe type is as yet unknown, as it was never cleared. It is possible that the arch gives access to further chambers, but it seems unlikely that there will be any connection with the Brick Vaulted Structure to the west, since over 20 metres lie between them.

By the time this arch was found it was scarcely a notable discovery (see above). But earlier in the same season (1974), the very appearance of baked brick in Aksumite Ethiopian architecture would have been remarkable, since it had been previously noted only in a few special circumstances (Anfray 1974), and an arch in the same material was completely unheard of. It is certain that our ideas about the architectural limitations of the ancient Aksumites will require yet more revision when excavations can be resumed.

These baked brick features, horseshoe shaped arches and vaults, in Aksumite buildings of the fourth century AD, may mean that our ideas about the routes of dissemination of architectural ideas in Africa, the Near East, and Spain (where the horse-shoe arch was later familiar) also need some revision. Wherever the style originated, it was certainly not expected to turn up in Aksumite Ethiopia. Without being able to assert the idea too strongly until we have more evidence, there may even be a case for proposing the brick horse-shoe arch as another Aksumite innovation, perhaps based on ideas which arrived through the trade-routes with India.

It may be presumed that attached to all the stelae are as yet unrevealed tombs like those just described. The latest excavations confirmed that the whole area of the central `Stele Park', apart from the tombs already mentioned, was honeycombed with shaft tombs and tunnels (Chittick 1974; Munro-Hay 1989). These consisted of chambers and passages cut into the rock, sometimes irregularly, sometimes following a more orderly plan. Some may be a combination of smaller tombs linked by robber tunnels cut later. Very little could be done to clear them and investigate their plan and content in the short time available, but the remains of tomb-furniture were found in some.

Illustration 29. The Tomb of the False Door; the door block during excavation. It is very similar in detail to the doorways carved on the three largest of the decorated stelae at Aksum. Photo BIEA.

The most westerly of all the tombs found so far, excluding one small shaft tomb, was the Tomb of the False Door. This possibly late-fourth or fifth century tomb (Munro-Hay 1989) has a false-door facade with a dentilled lintel exactly like those on some of the decorated stelae, but instead set into a granite-built square structure exhibiting the typical Aksumite plan of symmetrical recessed facades. It faced onto a carefully paved court and was doubtless open to view. Below, however, was an underground tomb-chamber with a vestibule and a surrounding corridor, with two staircases descending from the court to the substructure. The staircases had been blocked by massive capping stones, only one of which now survives. The tomb now contains nothing but a smashed granite coffin.

Illustration 30. The ruined superstructure of the Tombs of Kaleb and Gabra Masqal, just outside Aksum. Photo BIEA.

Illustration 31. Detail showing the entrance to the Tomb of Kaleb.

The tombs customarily attributed to the sixth century Aksumite kings Kaleb and Gabra Masqal could conceivably be of that period or of the fifth century. The building-complex consisted of two underground granite-built tombs with a double superstructure which seems to have consisted chiefly of two columned halls set on a platform approached by a staircase. The façade was about 40 m long, and the two halls were not exactly the same size, that attributed to Gabra Masqal being a little larger. The eight- or ten-columned hall above the `Kaleb' tomb measured c. 10 × 11 m, and that above the `Gabra Masqal' tomb c. 10 × 13 m, though both also had niches to the east adding an extra 2 m. The Gabra Masqal hall contained some sort of architectural feature, possibly a cupola or baldaquin, perhaps for a statue of the deceased (see the illustration in Littmann 1913: II, 133). Each side of the superstructure also contained entrance halls and staircase-wells, and these side-buildings were linked by a broad main entrance-stair 23 m wide. This was surmounted by a terrace with two porticoes, each with a column in the centre supporting the roof; the base of the column on the Kaleb side was square and stepped, while the column (which survived) and base on the Gabra Masqal side were octagonal. The building may represent a memorial chapel or shrine to the deceased, perhaps a `Christianised' development of the principle which gave rise to the decorated stelae and the Tomb of the False Door. The Book of Aksum claims that these tombs were filled with gold and pearls.

Illustration 32. The white capping over a fill of stones on one of the platforms at Aksum. Photo BIEA.

The main southern stele field, particularly the area set aside in modern times as a `Stele Park', is also characterised by the terracing achieved by erecting walls or platforms. The earliest of these platforms are the oldest architectural works yet found at Aksum (Chittick 1974; Munro-Hay 1989), and may date to the first century AD or possibly even a little earlier in some cases. Platform-building seems to have continued for some time, the typical examples being simple stepped or rebated revetting walls acting as facing to enormous quantities of freshly-quarried rock fill. They appear to have been carefully topped with layers of white and red soils, doubtless specially chosen for some religious purpose, and there are signs that sacrifices or sacrificial meals took place on or around them. They seem to precede some of the stelae and to be contemporary with others. At some time, possibly in the fourth century, major work was undertaken to raise the height of the stele field, the whole being faced with a long rebated terrace wall at least three metres high. It is on this terrace that the largest of all the stelae were raised.

6. The Stelae

These are the most famous of all the monuments of Aksum. They range from very rough and simple stones erected to mark grave-pits, to massive sculpted towers which represent soaring multi-storeyed palaces. Such huge monuments represent an enormous outlay of labour and skill, particularly in the most elaborate specimens. There are six carved and decorated monuments, the largest, now fallen and broken, formerly exceeding 33 m in height, with a measurement at the base of about 3 × 2 m. It is carved on all four sides and shows 12 storeys. It seems that the kings each tried to outdo the achievements of their predecessors; this largest of all the stelae exceeds by far even its nearest companion, which was only about 24 m high, and c. 2 × 1 m at the base. It was only ten storeys high and, though also carved on all four faces, was not so elaborate. One wonders if the giant stele ever actually stood, or whether it immediately plunged down to smash its (variously estimated) 400-750 tons to pieces on the terrace wall below, destroying the great tomb, Nefas Mawcha, as it fell. The stelae very often have only about one-twelfth of their length buried, very inadequate support when the total height began to grow, let alone when it exceeded 33 m; the third tallest, also probably about 24 m long but only 21 m above the ground, still (rather surprisingly) stands dominating the terrace of stelae, in spite of the abraded surface of the Stele Park, which leaves it somewhat less support than it originally had. Only very few stones in the ancient world exceeded the Aksum stelae in bulk, notably those forming the trilithon at Baalbek, constructed in the first half of the first century AD, where the largest stone measures something in the order of 20 × 4 × 4 m.

Illustration 33. View from the Nefas Mawcha looking towards the restored terrace wall, showing the still standing decorated stele; to the left of it originally stood the two largest stelae.

Illustration 34. The largest of all the stelae, lying as it fell over the terrace wall. (Photo D. Phillipson).

The great monolithic towers unquestionably mark the sites of the tombs of the Aksumite kings, although only two tombs which can be directly associated with them have so far come to light. All six of the carved stelae are embellished with the elaborate doors, windows, beam ends and other features typical of Aksumite architecture. At their summits are emplacements for what seem to have been either one or two metal plaques, which we may perhaps imagine as gilded bronze embossed with the sign of the disc and crescent or some other emblem of the kings or gods. All that now remains are the traces of the fixing nails, arranged in positions which could even represent the cross; conceivably those which bore pagan symbols were later `Christianised' (van Beek 1967). At the base of the decorated stelae were granite base-plates, carved in the case of all but the second largest (those of the largest have never been found) with the kylix, or Greek-style wine-cup with two handles. Some show several of these carved cups, with surrounds of decorative carving to the base-plates. The missing front base-plate for Stele 6 was found by Chittick (1974: pl. VIa). It is supposed that the cups were for offerings; since there is a similar cup carved in the base of a fruit-press cut from the solid rock at Atshafi near Aksum, it may indeed have been for wine-offerings that these altar-like base-plates were prepared. Perhaps the wine-offerings were poured out during memorial ceremonies for the deceased.

Illustration 35. The top of one of the decorated stelae (no. 4) showing the emplacements presumably for fixing decorative plaques.

Illustration 36. The base-plate of one of the decorated stelae at Aksum (no. 4), showing the kylix carved in the raised central portion, and three others around it; in the background can be seen the fallen stele.

The stelae appear to have been extracted several kilometres away at the quarries of Wuchate Golo to the west of Aksum, and dragged from there into the city. At the quarry the rows of holes cut into the rock to delineate the line of the desired cut can still be seen; wooden wedges, rammed into these holes and swollen with water would have eventually caused the rock to split. In some cases, traces of these wedge-holes remain on the stelae themselves.

The stelae were probably erected with the aid of earthen ramps and tremendous human effort. Very likely each stele was hauled into position, perhaps using wooden rollers, until its base lay beside a pit dug ready to receive it (or possibly sometimes a tomb shaft, only partly refilled, was utilised). Then it was slowly levered up, and as it rose, stones and earth were placed as a ramp beneath it. Eventually, when it had slid into the pit and been levered completely upright, the pit, after being sometimes lined with larger stones, was packed with rubble and the base-plates installed. It must have been a tense moment when the stele reached the point of perpendicularity, and also when, once erect, the first steps were taken to remove the supporting ramp. Whether the rather unreliable African elephant could have been utilised in helping to manoeuvre these giant stones is not known, but makes an interesting speculation.

No Egyptian obelisks could equal the size of the largest of the Aksumite stelae (though the unfinished one in the quarry at Aswan comes quite close). This enormous stone must surely represent the apogee of the manifestation of personal power in monumental structures in the ancient world. The five others are more modest in size, but are still very large. We have no accounts of these stelae from ancient visitors to the city, but they must have been an awe-inspiring sight rising in a row on their terraces overlooking the capital. Now all but one lie smashed or partly buried in the stream-bed, save for the second largest, which, transported to Italy as loot during the period of the Italian occupation of Ethiopia, now rises, repaired and restored but deteriorating in the polluted air, near the site of the Circus Maximus in Rome.

Illustration 37. View (from below) of part of the Ionic capital and shrine on stele no. 7; the plain stele no. 36 now lies beneath it.

Prior to these very few stelae were decorated. There is one roughly-shaped specimen with very rudimentary representations of beams and `monkey-heads', and another with an Ionic pillar and some sort of shrine rather elegantly depicted on it. A third has a circular object, perhaps a shield, topped with an angled and pointed line, carved on it. Finally, one has, carved near its base, the ancient Egyptian symbol of life, the `ankh' (Anfray 1974).

Illustration 38. Two of the plain round-topped stelae at Aksum. (Photo R. Brereton).

A common type of stelae comprises those plain dressed ones with rounded tops. In some cases these preceded the great carved monuments, but in others the differentiation was probably caused by respective wealth; they were probably erected by contemporaries who could not hope to raise anything of the order of the decorated monoliths. These plain dressed stelae range in size from very large (more or less equal in height to all the decorated ones save the very largest examples) to comparatively modest dimensions. From the archaeological evidence, it can be said that in broad terms the development was from the smaller and rougher completely undressed type to the dressed and then decorated types. Some allowance, however, must be made for the relative prosperity of those who erected stelae, and their social grouping. Even after the development of the dressed stelae, some people could doubtless still only aspire to a rough-hewn memorial for their tombs; very likely the larger plain dressed stelae belonged to persons of very high rank, and the decorated ones only to the rulers.

Illustration 39. In the eucalyptus groves now planted over the northern Stele Field at Aksum, two of the smaller unworked stelae. Photo BIEA.

However, we remain completely uninformed as to who was buried beneath or near these memorials, though it is a natural inference that only the kings could have mobilised the necessary labour and skill to quarry, carve, decorate, and erect the giant stelae or build the larger tombs. At the other end of the scale a simple rough-cut tomb marked by a rough stele proved to contain sets of glass goblets and beakers, large numbers of iron tools, around eighty pottery vessels of excellent quality and many shapes (some certainly not for basic domestic use). This tomb, in the `Gudit' stelae field, probably dates to the third century AD and in spite of its unimpressive appearance it clearly belonged to a person of some affluence (Chittick 1974: 192 and pls. XII, XIIIa, XIVa).

Some of the burials found in the Stele Park were of a different nature. In one shaft (Chittick 1974: pl. VIIIa) bodies were found in layers, with few grave goods beside personal ornaments. These, from coins found among them, seem to have been later Christian-period burials of simple type, and there seems to be no stele associated with the tomb.

One or two bodies found may have been human sacrifices, offered as dedications or on celebratory occasions. King Ezana, for example, mentions offering 100 bulls and 50 captives in an inscription (see Ch. 11: 5); these captives may, of course, have only been dedicated to the gods as slaves. The stelae base-plates furnished with carved goblets resembling the Greek kylix could have acted as offering bowls; but these may rather have been used for wine or the like instead of blood, as noted above. A Yeha stele has a similar offering-cup in its base-plate (Littmann 1913: II, 2), and perhaps the custom was an old one. However, one rough stele (no. 137) was found to have the bones of perhaps two individuals seemingly thrown down into the pit in which it was erected, and there are other examples of human and animal bones, often burnt, among the fill and capping material of the platform complex (Munro-Hay 1989). Such rituals in pre-Christian Aksum call to mind contemporary sacrifices in the neighbouring Meroitic kingdom, where the kings are often depicted slaughtering captives en masse in an imagery which descends from the pharaonic art of very early times. In the vast and rich graves found under the mounds at Ballana and Qustul in Nubia, the so-called X-Group rulers contemporary with the later Aksumite kings were buried with human and animal sacrifices, and, though not in Africa, there is also the example of the Romans, who sometimes sacrificed victims at their triumphs.

A few stelae have been noted placed beside structures which have been firmly dated to the Christian period, such as the Dungur villa, the building at Wuchate Golo, and Matara, Tertre D; at the latter place Anfray suggests the stele might denote a sacred edifice (Anfray and Annequin 1965: p. 68 and pl. XLIX, fig. 4). This might well apply also to Wuchate Golo, but does not seem to be the case at Dungur.

The origins of the stelae are very difficult to disentangle. Attributions of stelae in Ethiopia to the pre-Aksumite period, though customarily accepted (Munro-Hay 1989: 150), are not necessarily correct (Fattovich 1987: 47-8). A stele tradition appears nevertheless to have existed in the Sudanese-Ethiopian borderlands, and in parts of northern Ethiopia and Eritrea in pre-Aksumite times. Fattovich suggests, plausibly enough, that stelae belong to an ancient African tradition. In the case of the stelae at Kassala and at Aksum — despite the difference in time and the difference in the societies which erected them — he sees a similarity in several features. These include the suggestion that `the monoliths are not directly connected with specific burials' (Fattovich 1987: 63). However, this is questionable as far as the Aksum stelae are concerned, now that it has been possible to analyse the results of Chittick's work. Though it is not yet easy to identify tombs for all the stelae, it does seem that, at the Aksum cemeteries, wherever archaeological investigations have been possible there is a case for suggesting that stelae and tombs are directly associated.

End of Chapters 4-5.
Continue with Chapters 6-10.
Back to Table of Contents.