New light on the early Islamic West African gold trade: coin moulds from Tadmekka, Mali.

Nixon, Sam ; Rehren, Thilo ; Guerra, Maria Filomena 等

[ILLUSTRATION OMITTED]

Introduction

As South American gold enchanted the early modern world, West African gold enchanted the medieval world before it. Following the Islamic conquest of North Africa in the seventh century AD, Islamic traders crossing the Sahara to West Africa found abundant, high-quality gold there (Levtzion & Hopkins 2000). Over the next 300-400 years West Africa was incorporated into the Islamic trade network, leading it to become the world's major gold supplier until Columbus' era (Messier 1974; Devisse 1988; Spufford 1988; Gondonneau & Guerra 1999; Levtzion & Hopkins 2000; Mitchell 2005; Figure 1). The Islamicised West African trading towns, where traders obtained gold, took on huge significance in people's minds (Figure 2), and fantastic and mythical tales of their splendour and riches were told (e.g. Levtzion 1968; Benjaminsen & Berge 2004). Medieval texts have provided great insight into the reality of these towns and their trade (Levtzion & Hopkins 2000), but they also leave uncertainties. A particularly intriguing issue that has so far been impossible to resolve is whether references within these texts to local, pure gold coinages at the West African towns described myth or reality.

[FIGURE 1 OMITTED]

In AD 1526 Leo Africanus--a Muslim prisoner in Europe who became a protege of the Pope--recorded his knowledge of West Africa (Brown 1896). Within his description of Timbuktu, the trans-Saharan trading town which flourished from the fourteenth century, he stated that "the coin of Timbuctoo is of gold without any stamp or superscription" (Brown 1896: 825). He also described "pure gold coins without inscription" at another major town, Djenne (Brown 1896: 822). Some 500 years earlier (AD 1068) the respected Arabic geographer al-Bakri wrote a world geography using contemporary and historical sources. He noted of the West African trans-Saharan trading town of Tadmekka that "their dinars [Islamic gold coins] are called 'bald' because they are of pure gold without any stamp" (Levtzion & Hopkins 2000: 85). Other chroniclers also reported exchanges of "dinars" at Tadmekka and other important West African towns, including the early trading centre of Audaghust (Levtzion & Hopkins 2000: 45-9, 81, 90).

[FIGURE 2 OMITTED]

None of these unstamped gold 'coins' has been recovered, nor indeed is there concrete evidence of any coinage produced in West Africa before twentieth-century colonial issues. This stands in stark contrast to significant archaeological recovery of coinages produced elsewhere throughout the early Islamic world (e.g. Ehrenkreutz 1992; Gondonneau & Guerra 1999; Figure 3), including from areas where mints are not historically documented, such as the East African coast (Horton et al. 1986). Despite the lack of finds, early scholars believed the West African gold coins to be a reality (Mauny 1958; Messier 1974), suggesting that failure to find them in West Africa was because they were likely transported across the Sahara to be stamped on arrival. Subsequent prominent studies, however, made no reference to them (Garrard 1982; Devisse 1988; Devisse 1993), and they are also absent from wider texts drawing on these specialist studies (Spufford 1988; Blanchard 2001; Mitchell 2005). The consensus has clearly been voiced (Devisse 1988: 387) that while complex gold-working traditions existed in West Africa--producing jewellery and other display items (Joire 1943; McIntosh 1995: 267; Insoll 2003: 245)--commercial exchange of gold in the West African towns was restricted to gold dust and rudimentary processed forms; alongside other commonly encountered West African money such as cowrie shells, copper ingots, and cloth (see Eagleton et al. 2009). Meagre archaeological finds of gold trade in West Africa--wire and simple ingots (Devisse 1988: 404)--have not altered this view. Seemingly, then, the references to local gold money in an internationally recognised coin form in medieval West Africa had come to be seen as merely colourful elaborations of early Muslim authors.

[FIGURE 3 OMITTED]

Scholarly caution is certainly necessary. However, continued investigation reveals further neglected medieval descriptions of West African coinages, this time of silver. Firstly, there is a fourteenth-century reference to a silver coinage of Kanem, near Lake Chad (Levtzion & Hopkins 2000: 260). This is mirrored by a fifteenth-century reference to the presence in Egypt of silver coinage from 'Bilad al-Takrur', a general name for West Africa (Bacharach 1967: 224). If we dismiss the references to the gold coinages, should we also dismiss these references? While certainly we have not found medieval West African coinages, equally we have not found the local gold coinages reported by nineteenth-century European explorers in West Africa (Johnson 1968). It is also important to note that while archaeology has significantly advanced understanding of medieval West Africa (e.g. Devisse 1983; McIntosh 1995; Insoll 2003), little investigation has actually been undertaken at the key locations where coinages were described (cf. Insoll 1998).

Here we present a group of artefacts identified as moulds to make gold coins. They were excavated in a stratigraphic context datable to the ninth to tenth century AD, at the town of Tadmekka where al-Bakri described a pure gold coinage in AD 1068.

Tadmekka: history and archaeological background

Seemingly functioning by the eighth century AD, Tadmekka (or 'Tadmakka') (Figure 1) was one of the earliest towns established at the Sahara's southern edge where cross-Saharan camel caravans traded (Nixon 2009). It was located in territory governed by the Tuareg, the Berbers associated with Saharan camel caravans. Tadmekka's merchants brought gold, slaves and ivory from further south, exchanging these for North African goods, including copper and cloth. In AD 1068, three centuries before Timbuktu is first recorded, al- Bakri described Tadmekka as "a large town ... better built than Ghana or Gao [the two earliest gold trading states further south]", with a richly attired king and pure gold coinage; it was, he said, "of all the towns in the world ... the one that resembles Mecca the most" (Levtzion & Hopkins 2000: 85).

[FIGURE 4 OMITTED]

By the tenth century Tadmekka was supplying substantial quantities of gold (Devisse 1988; Levtzion & Hopkins 2000: 90). Amongst the early purchasers were the North African Fatimids, who had broken from the central Muslim authority of the Baghdad Caliphate and required gold for their coinage (ah example is shown in Figure 3). Tadmekka likely also supplied the Almoravids, the dominant North African trading power in the eleventh and twelfth centuries who conquered Morocco and southern Spain, and whose prestigous coinage was even known in China (Messier 1974). Tadmekka was still trading in the fourteenth century, when Timbuktu started to became famous.

Tadmekka's extensive stone ruins, situated in the desert in northern Mali, further underline its significance. These ruins (Figure 4)--today called Essouk--cover some 75ha. The site's Arabic epigraphy identifies it as ancient Tadmekka, and includes the earliest known writing in West Africa, from the early eleventh century AD (Moraes Farias 2003: 87-9). In 2005 excavations were undertaken to better understand this previously unexcavated town and its trade (Nixon 2009, 2010). Three interventions explored a 6m-deep archaeological sequence, of which 5m were associated with permanent architecture radiometrically dated to c. AD 750-1400 (for details on extensive radiometric dating and dated trade goods see Nixon 2009). The results firmly establish Tadmekka as one of the most important early trans-Saharan trading towns, and crucially provide physical evidence for the gold coinage described by al-Bakri.

The coin moulds and related finds

Three fragmentary ceramic objects identified as coin moulds were excavated within a stratigraphically sealed room 4m below the surface, radiometrically dated to c. AD 850-950 (Figure 5; Nixon 2009). The 7.5m2 room had stone and rammed earth walls and was part of a continuous sequence of similar rooms built one on top of another. The moulds were found at different levels within c. 700rum of coarse sand, likely representing multiple sand floors. The room--of unknown function but seemingly part of a larger structure--also contained other metalworking remains, glass beads and animal bones.

[FIGURE 5 OMITTED]

The moulds were identified by the roughly circular cup-shaped depressions in their surface (Figures 6 & 7). These have curved interiors, c. 20rum internal rim diameters, and are c. 10mm deep. They were formed as impressions in soft clay. Ali fragments feature surface vitrification. They are fairly porous and low fired, with chaff and sand temper, matching Tadmekka's domestic pottery. Two fragments clearly originally formed parts of plate-like objects, one roughly circular, one rectilinear. The circular mould was likely c. 180mm in diameter, featuring about 30 cups.

[FIGURE 6 OMITTED]

The artefacts closely resemble common European Iron Age coin moulds (Tournaire et al. 1982; Gebhard et al. 1995; Heinrichs & Rehren 1996; Robbins & Bayley 1997). In these, metal dust and nuggets were heated in the cups, forming irregular discs that were then struck to produce coins. Similar coin moulds are documented from Hellenistic Cyprus (Nicolaou 1990), Aksumite Ethiopia (Wilding 1989), and early Islamic Pakistan (Khan 1990). Forty years ago excavations at Audaghust (Figure 1) did in fact recover similar artefacts, but their formal similarity to coin moulds was not discussed and they were interpreted, without technical analysis, as moulds for glass beads (Vanacker 1984: 48). These moulds are unfortunately not available for us to reassess, but this interpretation is certainly not appropriate for the Tadmekka moulds. The cups are far too large to correspond with the well-documented small size of medieval West African glass beads, and glass- working leaves much more heavily vitrified residues. Other similar West African finds (Griaule & Lebeuf 1948: 54), while observed to have been subjected to heating, remain functionally uninterpreted.

Examination of the Tadmekka moulds' surfaces and firing characteristics showed very strong resemblances to known coin moulds. Crucially, through chemical analysis we identified three prills (droplets) of gold within the cups of a mould (see Figure 8).

Within the same contexts as the moulds were fragments of two crucibles, but featuring no metal prills. A goldworking crucible was, however, found in a slightly younger (eleventh-century) context within the same excavation unit (see below). Contemporary evidence for copper smelting and iron smelting and smithing shows the diversity of local metallurgy. Six copper/silver coins were found in later contexts (c. AD 1100-1400), but corrosion layers have so far prevented their identification.

[FIGURE 7 OMITTED]

Technical and chemical analysis

Polished sections were prepared from the moulds and crucible for optical microscopy and SEM-EDS analyses. We aimed to locate further metal inclusions, and to better understand the technical ceramic. Additionally, gold prills were analysed using Proton-Induced X-ray Emission (PIXE) at the AGLAE accelerator of the Centre for Research and Restoration of the Museums of France, Paris.

EKA-96

The intact upper surface features vitrification, about 50-100[micro]m thick and incorporating quartz grains from the ceramic. Extensive shattered quartz within the ceramic (Figure 9D) suggests high, relatively even, firing temperature. The vitrified layer is chemically similar to the ceramic, except for increased lime and potash. No lead, copper or other base or noble metal was found (detection limit c. 0.2%). Vitrification formed through fuel ash, rich in lime and potash, fluxing the ceramic, rather than through reaction with metal oxides as crucible slag. Minute iron phosphide prills observed within the vitrified layer likely formed from reaction between fuel-ash phosphate and iron oxide in the ceramic. The fragment features an impression of charcoal used during firing (Figure 9A).

EKA-95

EKA-95 appears less highly fired than EKA-96, and few shattered quartz grains were seen. Vitrification is seen on all surfaces but is patchy due to erosion. No vitrification was seen in section. The base of one cup features a circular patterning in the vitrification (Figures 7 & 9B), almost certainly the 'ghost' of a molten coin disc. No metal inclusions were identified.

EKA-93

All surfaces are vitrified but the cross-section shows vitrification is very thin. Clear evidence of localised 'bloating' (Figure 9C) suggests the mould was exposed to significant heat, above the ceramic's thermal stability threshold. Shattered quartz is restricted to immediately underneath the bloated ceramic, pointing to somewhat uneven heating. We again see a circular vitrification pattern in one of the cups. No further metal inclusions were identified in section. Three gold prills (referred to above: Figure 8) are visible using a hand lens.

[FIGURE 8 OMITTED]

The two larger gold prills on the sample's surface were analysed using a 3 MeV proton beam of 30[micro]m diameter, using the PIXE set-up explained elsewhere (Guerra & Calligaro 2004; Guerra 2004). Both were found to be very high-purity gold (98%: Table 1). Relatively high traces of zinc, lead and tin are well above usual background levels in ceramic, and reliably attributable to the gold.

EKA-86A

This small ceramic fragment was tentatively identified as from a crucible. The fabric differs from the moulds, being dominated by badly-sorted quartz grains, probably added as temper. Covering its concave surface is glassy slag containing numerous minute gold prills and heavy mineral inclusions (Figure 10). The majority of the prills tested were around 97% gold, although a few averaged 94% (Table 2). This fragment suggests separation of panned gold dust from heavy mineral impurities using a slag-forming crucible process (to be detailed in a forthcoming publication). Its presence further supports the notion of established, competent goldworking at Tadmekka.

[FIGURE 9 OMITTED]

Discussion

Gebhard et al. (1995) discuss European gold coin moulds closely resembling the Tadmekka artefacts, and their technical evidence is virtually identical. Absence of base- metal remains (e.g. lead, copper or zinc) on, or in, the moulds rules out a copper alloy being processed. Similarly, absence of significant amounts of glass-indicating oxides such as soda or potash excludes glass processing. Only melting of pure gold would leave no chemical traces except superficial fuel-ash vitrification, and occasional gold prills, as seen here. Combined with al-Bakri's description of Tadmekka's coinage, the evidence overwhelmingly indicates gold coin production at medieval Tadmekka. According to the archaeology, this coinage goes back to ar least the tenth century AD, and likely the ninth century. This is the first concrete evidence of locally-made coinage anywhere in pre-colonial West Africa.

[FIGURE 10 OMITTED]

On reflection, the decision to produce gold coins ar this flourishing early medieval trading town is unsurprising. Ir was fairly easy to process highly pure West African gold, and trans-Saharan merchants would prefer clean, processed gold in a standardised, exchangeable form. Also, other goods were exchanged here, including slaves and ivory, and efficient gold exchange would facilitate this. Ingots would obviously be useful for large transactions. However, for such a valuable commodity smaller exchangeable units would be required. As Islamic merchants familiar with coinage were operating ar Tadmekka, and as Tadmekka traded with North African coin-based economies, smaller units were produced in coin form.

The purity of the gold droplets (>98%) significantly exceeds previous findings for medieval West African gold (92-94%: Messier 1974; Roux & Guerra 2000). This high purity is consistent with supergene gold, where primary gold, typically containing 5-35% silver, naturally dissolves over geological time-spans and is re-precipitated in a highly pure form (Guerra & Rehren 2009, and references therein). Such deposits are typically accessible through panning or surface mining. This fits with analyses of modern West African nuggets (> 97%: Gondonneau et al. 2001), and medieval descriptions of sourcing high- quality gold from soils and streams in West Africa (Levtzion & Hopkins 2000). Alternatively, high purity could indicate refining of primary gold, a process known at least since the sixth century BC when the first known gold coinage emerged in Lydia (Turkey) (Ramage & Craddock 2000). However, the levels of silver and copper still present argue against this, as does the Tadmekka evidence for extraction of high-purity gold from heavy mineral concentrates. The exact geological origin of this Tadmekka gold cannot currently be pinpointed. Bambuk and Bure (Figure 1) are the most commonly discussed goldfields (Devisse 1988), yet we should also consider the Sirba Valley.

Interestingly, the mould technology used to make the Tadmekka coinage does not feature in medieval Islamic treatises, which mainly record cutting coinage from sheets or ingots (Ehrenkreutz 1953; Ramage & Craddock 2000). While it is better known from pre-Islamic contexts, an excavated example is known from a very early Islamic site in Pakistan (Khan 1990). Potentially this was actually an important technology in the earliest centuries of Islam when no treatises on coin production exist. Tadmekka's coinage was possibly also produced using other techniques. The moulds themselves would quite likely be broken up after use to extract gold inclusions, as even small amounts were hugely valuable.

Tadmakka's coins likely corresponded to Islamic weight and size standards. Al- Bakri's reference to them as "dinars" supports this, as does the presence at contemporary Audaghust of Islamic standard weights (Launois & Devisse 1983). The moulds would initially produce irregular blanks (e.g. Ramage & Craddock 2000). However these could be easily hammered into a dinar form to meet the expectations of Islamic traders. Al-Bakri also states Tadmekka's coins were "bald", i.e. blank. We have to assume this is accurate, or that they featured only very superficial identification marks. Objects considered 'coin blanks' north of the Sahara were seemingly then in this trading town a fully functioning money. A comparable example are blank 'coins' from medieval Turkey having very superficial identification marks (Hinrichs 1997), in some cases single indented dots or lines of dots, in others small marks or symbols covering only c. 10% of the coin surface. These would be largely obliterated when struck with a die.

Stamping coins in the early centuries of Islam was the prerogative of the Caliphs. A minted Tadmekka coinage would certainly have projected Tadmekka's power, and indeed would have communicated the ruler's guarantee of the coinage purity, but it is unlikely its Islamic authorities wanted to make this overtly political gesture (Tadmekka's tenth-century rulers are named as Fusahr bin Alfara and Inaw bin Sabanzak--Levtzion & Hopkins 2000: 51). A potential additional factor keeping coinage blank is that high-quality blank coins had huge commercial value: they could simply be stamped with no reworking by the buyer to produce high-quality coinage. While intriguing to reflect on the possibility of the 'king' of Tadmekka described by al-Bakri stamping coins in his name, these coins were seemingly not a political tool but simply a trading mechanism.

That Tadmekka's coinage featured in al-Bakri's world geography suggests it was fairly significant--a very small-scale industry is unlikely to have been heard of across the Saharan routes and in Spain where al-Bakri lived, let alone considered worthy of mention. Understanding how much Tadmekka gold was converted to coins is currently impossible. However, even a small percentage would have been significant as substantial quantities of gold moved through Tadmekka. A tantalising insight is perhaps provided by a report that a single Tadmekka merchant annually sent across the Sahara 16 bags containing 500 dinars each (8000 in total, c. 34kg of gold) (Levtzion & Hopkins 2000: 90).

Amongst the gold coinages of the North African dynasties that traded with Tadmekka--such as the Aghlabids and the Fatimids--we find many coins of very high (98-99%) purity (Ehrenkreutz 1992; Gondonneau & Guerra 1999). Given our analytical results (> 98% purity), it is likely much of this high purity coinage was made directly from high-quality supergene gold arriving from Tadmekka, rather than from gold artificially refined in the mint. Indeed, highly pure blank Tadmekka coins likely became a trans-Saharan trade item in their own right, struck on arrival into inscribed coinage in North African mints with little or no reworking. Results of 94% gold from the crucible suggests some Tadmekka gold may also have corresponded with Almoravid coins (average 94%: Roux & Guerra 2000). The non-destructive analysis we conducted did not test for certain diagnostic trace elements, but the Tadmekka trace contents differ from Almoravid coins, Gao's bead, and many West African gold nuggets, while equivalent trace contents were found in Audaghust gold wire and a modern nugget from Ivory Coast (Roux & Guerra 2000; Gondonneau et al. 2001).

The fact that al-Bakri describes Tadmekka as the best built town in West Africa, controlled by a rich king, would have added further allure to this place. Tadmekka and its coinage would have been icons of the riches to be gained across the Sahara. In time more famous icons developed, such as the Malian king Mansa Musa (Figure 2), and the legend of Timbuktu and its gold. However, Tadmekka would have been a crucial element in the mental map of West Africa in the early phase of this region's integration into the medieval world trade network.

It is unclear what happened to Tadmekka's coinage after al-Bakri. Possibly social or economic changes caused it to decline quickly. Specifically, we should consider it being affected by the Almoravids, from the late eleventh century the dominant power within trans-Saharan trade (Nixon 2009). It is possible the Almoravids attempted to eradicate coinages they saw as competitors to their own. We do have limited evidence for later silver and base metal coinage at Tadmekka, but it is not clear if these are locally produced. Certainly this at least suggests a continued coinage function within the economy.

Having established the existence of Tadmekka's coinage, we should now consider more seriously other evidence for West African gold coinages. Firstly, "dinars" are mentioned within the trade network of Audaghust (Levtzion & Hopkins 2000: 45-9, 81), the contemporary gold trading town at the southern edge of more westerly Saharan routes, with a similar socio-economic role and Islamicised population to Tadmekka. As mould fragments resembling Tadmekka's were found here, there is a strong case for gold coin production at Audaghust. While the reference we have to Timbuktu's 'unstamped' gold coinage is some 500 years later, this town also fulfilled the same function as Tadmekka and Audaghust, meeting cross-Saharan traders and supporting cosmopolitan Islamicised trading populations. In light of the Tadmekka evidence, it would not be surprising to find archaeological evidence of Timbuktu's 'unstamped gold coinage' ir its medieval layers were excavated.

Further south of the desert's edge, the concept of coinage would certainly have been less common as links with North African coin-based economies diminished. Texts and archaeology indicate strong local currencies in these regions, such as cowries (Levtzion & Hopkins 2000; Insoll 2003; MacDonald in press). While it would be folly to anticipate widespread evidence for local coinages throughout medieval West Africa, one cannot ignore the description of local coinage further south at Djenne, an important gold trading centre in the sixteenth century. The fourteenth-century reference to a silver Kanem coinage is also interesting, describing how its value was calculated against cloth, an example of coinage interfacing with local exchange. Additionally, the fifteenth-century reference to a 'Takrur' silver coinage mentions that these coins were illegal in Mamluk Egypt, and so brought to the mint for melting down--a very precise and realistic description. While West African silver sources are unknown, it is possible that rulers imported silver as a monetary unit clearly distinguishable from gold. No significant archaeology has been conducted of medieval Kanem or late medieval Djenne.

Conclusion

The moulds for making gold coins excavated at medieval Tadmekka may be identified with the local 'unstamped' gold coinage mentioned in historical sources, but previously treated with some scepticism. The finds show that gold exchange employed formally manufactured and regulated pieces, rather than simply gold dust. The Tadmakka gold was very pure (> at least 98%), sufficient for re-use in the coinages of Tadmekka's North African trading partners without further refining. Amongst the coinages we suggest used Tadmekka gold--and possibly blank Tadmekka coins as 'coin blanks'--was the prestigious Fatimid coinage.

There is suggestive evidence that these gold pieces may also have been produced elsewhere in West Africa. Audaghust and Timbuktu, like Tadmekka, hosted large groups of Islamic traders, and engaged in continual exchange with coin-based economies across the Sahara--the evidence we have from these towns is therefore particularly compelling. While traditional West African monetary systems dominated further south, and coinage would have been a much less familiar concept there, we should note the historical description of a gold coinage ar Djenne. Additional intriguing evidence for silver coinage from Kanem and 'Takrur' also becomes compelling.

The new evidence firmly establishes Tadmekka as a major conduit for the gold trade with the Mediterranean. In early medieval times, Tadmekka's pure gold coinage would have been one of the first icons of the riches to be gained from the West African gold trade, and it would have led to the elevation of Tadmekka's reputation throughout the Islamic world, and even into Europe.

Acknowledgements

We are indebted to those who helped excavate Tadmekka, and to Mali's Institut des Sciences Humaines and Direction Nationale du Patrimoine Culturel. Funding carne from the Arts and Humanities Research Council, UCL, the University of London, and the Natural Environment Research Council. Analysis was undertaken at UCL's Wolfson Archaeological Science Laboratories and on C2RMF's AGLAE accelerator. Particular thanks go to Kevin MacDonald and Paulo Farias for input into the project, Thierry Borel and Dominique Bagault for photographs and radiographs, and James Lankton, Jack Hiscock and Jacke Phillips for their insights. Thanks are also due to the anonymous referees who provided many useful comments on the text.

References

BACHARACH, J. 1967. A study of the correlation between textual sources and numismatic evidence for Mamluk Egypt and Syria. Unpublished PhD dissertation, University of Michigan.

BENJAMINSEN, T. & G. BERGE. 2004. Myths of Timbuktu: from African El Dorado to desertification. International Journal of Political Economy 34:31-59.

BLANCHARD, I. 2001. Mining, metallurgy and minting in the Middle Ages. Stuttgart: Franz Steiner.

BROWN, R. (ed.) 1896. The history and description of Africa and the notable things therein contained. By Leo Africanus, done into English in the year 1600 by John Pory. London: Hakluyt.

DEVISSE, J. (ed.) 1983. Tegdaoust III. Recherches sur Aoudaghost. Paris: Editions Recherche sur les Civilisations.

--1988. Trade and trade routes in West Africa, in M. el Fasi (ed.) General history of Africa, Volume 3: Africa from the seventh to the eleventh century: 367-435. Berkeley (CA): University of California.

--1993. L'or, in J. Devisse (ed.) Vallees du Niger. 344-57. Paris: Editions de la Reunions des Musees Nationaux.

EAGLETON, C., H. FULLER & J. PERKINS. 2009. Money in Africa. London: British Museum.

EHRENKREUTZ, A. 1953. Extracts from the technical manual on the Ayyubid mint in Cairo. Bulletin of the School of Oriental and African Studies 15: 423-47.

--1992. Monetary change and economic history in the medieval Muslim world (Variorum collected studies series CS371). Aldershot: Ashgate.

GARRARD, T. 1982. Myth and metrology: the early trans-Saharan gold trade. Journal of African History 23: 443-61.

GEBHARD, R., G. LEHRBERGER, G. MORTEANI, C. RAUB, F. WAGNER & U. WAGNER. 1995. Coin moulds and other ceramic material: a key to Celtic precious metal working, in G. Morteani & P. Northover (ed.) Prehistoric gold in Europe: 273-301. Dordrecht: Kluwer.

GONDONNEAU, A. & M.F. GUERRA. 1999. The gold from Ghana and the Muslim expansion. A scientific enquiry into the Middle Ages using ICP-MS combined with a UV laser, in S. Young, A. Pollard, P. Budd & R. Ixer (ed.) Metals in antiquity (British Archaeological Reports international series 792): 262-70. Oxford: Archaeopress.

GONDONNEAU, A., M.E GUERRA & M.R. COWELL. 2001. Searching for the provenance of gold. The methodology of gold analysis by ICP-MS: first developments, in L. Barba (ed.) Proceedings of the 32nd International Symposium on Archaeometry, 15-19 May 2000, Mexico City. Mexico City: University of Mexico (CD-ROM).

GRIAULE, M. & J-P. LEBEUF. 1948. Fouilles dans la region du Tchad. Journal de la Societe des Africanistes 18:1-116.

GUERRA, M.F. 2004. Fingerprinting ancient gold with proton beams of different energies. Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms 226: 185-98.

GUERRA, M.F & T. CALLIGARO. 2004. Gold traces to trace gold. Journal of Archaeological Science 31: 1199-208.

GUERRA, M.F. & Th.REHREN. 2009. In-situ examination and analysis of the gold jewellery from the Phoenician tomb of Kition (Cyprus). ArcheoSciences--Revue d'Areheometrie 33:151-8.

HEINRICHS, J. & Th.REHREN. 1996. Keltische Munzschrotlingsformen aus Bonn. Metalla (Bochum) 3: 83-100.

HINRICHS, J-C. 1997. Blank 'coins' from Anatolia. Oriental Numismatic Society Newsletter 154: 13-14.

HORTON, M.C., H.M. BROWN & W.A. ODDY. 1986. The Mtambwe hoard. Azania: Archaeological Research in Africa 21: 115-23.

INSOLL, T. 1998. Archaeological research in Timbuktu, Mali. Antiquity 72:413-17.

--2003. The archaeology of Islam in sub-Saharan Africa. Cambridge: Cambridge University Press.

JOHNSON, M. 1968. The nineteenth-century gold 'Mithqal' in West and North Africa. The Journal of African History 9: 547-69.

JOIRE, J. 1943. Archaeological discoveries in Senegal. Man 43: 49-52.

KHAN, A. 1990. Al-Mansurah: a forgotten Arab metropolis in Pakistan. Karachi: Kifayat.

LAUNOIS, A. & J. DEVISSE. 1983. Poids de verte decouverts a Tegdaoust, chronologie du site et histoire des etalons de poids en Afrique occidentale, in J. Devisse (ed) Tegdaoust III. Recherches sur Aoudaghost: 399-419. Paris: Editions Recherche sur les Civilisations.

LEVTZION, H. 1968. Ibn-Hawqal, the cheque, and Awdaghost. The Journal of African History 9(2): 223-33.

LEVTZION, N. & J.F. HOPKINS (ed.). 2000. Corpus of early Arabic sources far West African history. Princeton (NJ): Markus Wiener.

MACDONALD, K.C. In press. A view from the south: sub-Saharan evidence for contacts between North Africa, Mauritania and the Niger, 1000 BC-AD 700, in A. Dowler (ed.) Money, trade and trade routes in pre-Islamic North Africa. London: British Museum Press.

MCINTOSH, S.K. (ed.) i995. Excavations at Jenne-Jeno, Hambarketolo, and Kaniana (Inland Niger Delta, Mali), the 1981 season (University of California Publications in Anthropology 20). Berkeley (CA):

University of California. MAUNY, R. 1958. Anciens ateliers monetaires Ouest-Africains. Notes Africaines 78: 34-5.

MESSIER, R. 1974. The Almoravids. West African gold and the gold currency of the Mediterranean basin. Journal of the Economic and Social History of the Orient 17: 31-47.

MITCHELL, P. J. 2005. African connections. Archaeological perspectives on Africa and the wider world. Walnut Creek (CA): Altamira Press.

MORAES FARIAS, P.F. de. 2003. Arabic medieval inscriptions from the Republic of Mali. Epigraphy, chronicles and Songhay-Tuareg history. Oxford: Oxford University Press for The British Academy.

NICOLAOU, I. 1990. Paphos II. The coins from the House of Dionysos. Nicosia: Department of Antiquities.

NIXON, S. 2009. Excavating Essouk-Tadmakka (Mali): new archaeological investigations of early Islamic trans-Saharan trade. Azania: Archaeological Research in Africa 44:217-55.

--2010. Before Timbuktu: the great trading centre of Tadmakka. Current World Archaeology 39:40-51.

RAMAGE, A. & P. CRADDOCK. 2000. King Croesus' gold: excavations at Sardis and the history of gold refining. London: British Museum.

ROBBINS, K. & J. BAYLEY. 1997. Metallurgical analysis of coin pellet moulds and crucible fragments, in S. M. Elsdon (ed.) Old Sleaford revealed. A Lincolnshire settlement in Iron Age, Roman, Saxon and medieval times: excavations 1882-1995: 59-64. Oxford: Oxbow.

Roux, C. & M.E GUERRA. 2000. La monnaie almoravide: de l'Afrique a l'Espagne. Revue d'archeometrie 24: 39-52.

SPUFFORD, P. 1988. European silver and African gold, in P. Spufford (ed.) Money and its use in medieval Europe: 163-86. Cambridge University Press.

TOURNAIRE, J., O. BUCHSENSCHUTZ, J. HENDERSON & J. COLLIS. 1982. Iron Age coin moulds from France. Proceedings of the Prehistoric Society 48:417-35.

VANACKER, C. 1984. Perles de verte decouvertes sur le site de Tegdaoust. Journal des Africanistes 54:31-52.

WILDING, R. 1989. The pottery, in S. Munro-Hay (ed.) Excavations at Aksum: 235-316. London: British Academy.

(1) UCL Institute of Archaeology, 31-34 Gordon Square, London WC1H OPY, UK (Email: [email protected])

(2) UCL-Qatar, P.O. Box 23689, Georgetown Building, Education City Doha, Qatar (Email: [email protected])

(3) UMR 171 CNRS, Laboratoire du Centre de Recherche et de Restauration des Musees de France, Palau du Louvre, 14 quai Francois Mitterrand, 75001 Paris, France (Email: [email protected])

Sam Nixon (1), Thilo Rehren (2) & Maria Filomena Guerra (3)

Received: 7 October 2010; Accepted: 10 December 2010; Revised: 14 February 2011

Table 1. PIXE analysis of EKA-93 gold prills.

 Au% Ag% Cu% Zn Pb Pd Sn Sb
 PPM PPM PPM PPM PPM

Prill 1 98.8 0.9 0.2 667 72 23 162 26
 98.7 0.8 0.3 1323 238 15 254 25

Prill 2 98.1 1.8 0.1 212 171 -- 63 16
 98.3 1.4 0.1 484 229 21 490 19

Table 2. Selected SEM-EDS analyses of EKA-86A gold prills.

 Au% Ag% Cu%

Sol 3 large 98.4 1.6 nd
Sol 2 large 96.0 3.8 0.2
Sol 7 small a 94.1 5.4 0.5
Sol 7 small b 93.9 5.6 0.5