lable at ScienceDirect

Journal of Archaeological Science 37 (2010) 1898e1912

Contents lists avai

Journal of Archaeological Science

journal homepage: http: / /www.elsevier .com/locate/ jas

Southern African glass beads: chemistry, glass sources and patterns of trade

Peter Robertshawa,*, Marilee Wood b, Erik Melchiorre c, Rachel S. Popelka-Filcoff d, Michael D. Glascock e

aDepartment of Anthropology, California State University, 5500 University Parkway, San Bernardino, CA 92407-2397, USAb School of Geography, Archaeology and Environmental Studies, University of the Witwatersrand, Private Bag 3, Wits 2050, South AfricacDepartment of Geological Sciences, California State University, San Bernardino, CA 92407-2397, USAd School of Chemical and Physical Sciences, Flinders University, GPO Box 2100, Adelaide, South Australia 5001, AustraliaeResearch Reactor, University of Missouri-Columbia, Columbia, MO 65211, USA

a r t i c l e i n f o

Article history:Received 14 December 2009Received in revised form10 February 2010Accepted 12 February 2010

Keywords:GlassBeadsSouthern AfricaICP-MSIndian Ocean tradeGlass chemistry

* Corresponding author. Tel.: þ1 909 537 5551.E-mail address: proberts@csusb.edu (P. Robertsha

0305-4403/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.jas.2010.02.016

a b s t r a c t

Three-hundred-and-sixty glass beads from 19 archaeological sites in southern Africa dating betweenabout the 8th and 16th centuries AD were analyzed using LA-ICP-MS, determining 47 chemical elements.The eight different bead series, previously defined on morphological characteristics, possess differentglass chemistries. Some bead series were made from plant-ash glasses, others from soda-alumina glasses.Zhizo series beads of the late 1st millennium AD were probably made from Iranian glass. Later beadseries were made of glass probably manufactured in South Asia, though there are changes through timein both South Asian glass recipes and bead morphologies.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Glass beads, often in large quantities, have been discovered atarchaeological sites in southern Africa from the 8th century ADonwards. They are the most abundant artifacts attesting to inter-national trade with southern Africa and, at least until the late 13thcentury AD, they occurmost frequently in the area of the confluenceof the Shashe and Limpopo rivers where the modern countries ofSouth Africa, Zimbabwe and Botswana meet (Fig. 1). It was in thisregion that the first state-level society, the precursor to GreatZimbabwe, evolved in the early secondmillennium AD. Excavationsat the roughly 10th century site of Schroda, the largest Zhizo periodsite south of the Limpopo River, produced over 1000 glass beads. Farfewer such beads were discovered at contemporary smaller sites.The presence of numerous ivory artifacts at Schroda indicates it was“articulated directly with the Indian Ocean commercial network”(Huffman, 2000:19). Around AD 1000 immigration into the regionled to the development of the major political center known as K2,where tens of thousands of beads were recovered from a massivemiddennext to the central cattle enclosure and the court (the centerof political decision-making), as well as in residential areas.Although ivory exports probably continued, alluvial gold became

w).

All rights reserved.

the most important trade commodity. The K2 elite who controlledthe distribution of exotic goods were able to amass wealth andpower. Glass beads were traded or gifted on to communities inneighboring regions. “Garden Roller” beads (see below),whichweremade at K2 and possibly other related sites by reworking importedglass beads (and thus are evidence of interaction with K2), werewidely distributed as well; the most distant example coming fromthe southern Zambian site of Isamu Pati (Wood, 2005:49).

In the 13th century, the capital moved from K2 to nearbyMapungubwe,where for thefirst time in southernAfrica the head ofwhat was now a state-level society lived on a hilltop separate fromthe rest of the inhabitants and the cattle enclosurewasmoved awayfrom the center of the settlement. This spatial pattern was alsoevident at the later, more famous site of Great Zimbabwe.More than100,000 glass beads have been found at Mapungubwe, of which atleast 26,000 (Saitowitz, 1996:201), plus over 12,000 gold beads,were discovered in a royal burial, one of several at the site which isalso the first to have grave goods indicative of high status (Huffman,2007:58). Clearly, the rulers of the Mapungubwe state, which wassupplanted by Great Zimbabwe at the end of the 13th century,controlled the distribution of imported glass beads, as well as thecotton cloth that historical sources indicate was probably the mostsought-after import, despite its low visibility in the archaeologicalrecord. In the final pre-European period, after the decline of GreatZimbabwe in the mid 15th century, beads of the Khami Series are

Fig. 1. Map of sites from which the analyzed beads were recovered.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1899

widely distributed at settlements across the eastern part ofsouthern Africa, though as in earlier times theyaremore common atthe larger political centers. In summary, “control of the supply ofimported glass beads and cotton cloth, and of their redistribution,appears to have been a central mechanism in the development ofcomplex societies in this region [the Limpopo Valley] (Hall,1990:88e90; Huffman, 2000)” (Killick, 2009:188). Easily storedand transported, beads provided an alternative expression ofwealthand power that was not prone to the vicissitudes of cattle-keeping.

Two questions have dominated studies of the southern Africanglass beads: 1) How old are they? 2) Where did the glass comefrom? This first question has recently been definitively answered byWood (2000, 2005), who succeeded both in defining bead seriesbased on morphological and technological attributes and in datingthese series by reference to established radiocarbon chronologies.Answering the second question involves comparative studies ofbeads from various parts of the world and chemical analyses. Thisreport presents the results of the chemical analysis of 360 glassbeads, belonging to 8 different bead series, recovered from 19 sitesin southern Africa. We then compare our results with publishedglass compositions to identify the probable sources of the differentbead series found in southern Africa.

This study is part of a larger project on the chemistry of glassbeads from African sites dated between about AD 800 and 1500, i.e.preceding most European contact (see Robertshaw et al., 2003 fora longer introduction to the project). We have analyzed more than1000 glass beads recovered from relatively well-dated contexts onAfrican archaeological sites. The major goal of the project is to usethe results of the chemical analysis to identify the regions wherethe glass was manufactured and thereby to reconstruct changingpatterns of trade between various regions of Africa and the widerworld. Our southern African research builds upon pioneeringchemical studies of glass beads by Davison (1972, 1973; Davisonand Clark, 1974, 1976), who analyzed about 130 beads fromsouthern African sites using destructive neutron activation analysis(NAA) and a smaller number of beads using non-destructive x-rayfluorescence analysis (XRF). Both techniques provided quantitativedata on a limited number of elements, but a dearth of comparativedata from other parts of the world hampered Davison's ability tointerpret her results in terms of provenance.

More recently, Saitowitz (1996; Saitowitz et al., 1996), employ-ing mostly laser-ablation inductively coupled plasma mass spec-trometry (LA-ICP-MS), concluded that the results from the rareearth elements (REE) “showed positively that some beads

Fig. 2. Examples of the different bead series.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121900

excavated in the northern and eastern Transvaal are [chemically]identical to beads that were produced in Fustat a thousand yearsago” (Saitowitz, 1996:2). This conclusion, based on the presence ofa negative Ce anomaly, is incorrect. Among other problems, theSouth African beads with a negative Ce anomaly include membersof four different bead series spanning almost a millennium (seebelow), representing four different glass types that used differentglass-making recipes and in some cases different raw materials.Moreover, beads with very similar major and minor elementchemistries exhibit widely varied Ce anomalies. In essence a nega-tive Ce anomaly merely indicates a fractionated source material orprocess inherent to a specific glassmaking recipe that preferentiallyremoved Ce from the glass melt.

2. The bead series (Fig. 2)

The earliest bead series, the Zhizo, began arriving in southernAfrica in the 8th century AD. These beads are cut from drawn tubesand are usually left with untreated ends. Many are 2.5e4.5 mm indiameter but larger ones are common, occasionally up to 13 mm indiameter and 20 mm long. Cobalt blue is the most common color,followed by yellow, blue-green and green. The glass is usuallytranslucent (or even transparent in blue samples) but normallyappears to be opaque because of the condition of the glass. Zhizoglass corrodes easily, resulting in a patina that is often powderywhite. The beads often appear striated because of numerous rowsof tiny bubbles just under the surface. This series disappears rathersuddenly in the mid-10th century.

The K2 series is made up of drawn, small (2e3.5 mm diameter;1.2e4 mm long) transparent to translucent blue-green to lightgreen beads that are characteristically tubular in shape with faintlyrounded ends, but cylindrical examples also occur. The glass is verydurable and often very shiny. K2 beads first arrived in southernAfrica during the 10th century and were no longer traded into the

area after about AD 1200. They were, however, often curated and sooccur occasionally in later deposits.

The K2 Garden Roller series consists of large (10e14 mmdiameter; 7e15 mm long), mainly barrel-shaped beads that weremade at K2 and perhaps elsewhere by melting K2 series beads andforming new large ones in single-use clay molds (Wood,2005:45e49). Like K2 beads, K2 Garden Rollers appear mainly inshades of transparent to translucent blue-green to soft green. Thetemporal span of this series is similar to that of the K2 series.

The Indo-Pacific series first arrived in southern Africa in theearly 11th century AD. Characteristic are drawn beads that varyconsiderably in size and shape, with ends that have been roundedthrough reheating. Most are 2.5e4.5 mm in diameter and cylin-drical in shape. Black and brownish-red beads are opaque; yellow,soft orange, green and blue-green ones are translucent. Cobalt bluebeads do not appear in this series, which was displaced by theMapungubwe Oblate series in the early to mid-13th century.

Beads belonging to the Islamic series were made at Islamiccenters in the Middle East or north Africa. The series is not welldefined here since so few of these beads are found in southernAfrica. Those that are, are large and decorated with patterns madeup of glasses of several colors.

The Mapungubwe Oblate series is characterized by drawn, heatrounded beads that are often remarkably uniform in size (2e3.5mmdiameter) and shape. Many are oblate, others are cylindrical withwell-rounded ends. Opaque black is the most common color fol-lowed by a range in translucent-opaque glass including blue-green,green, yellow and butterscotch orange. Transparent cobalt blue andplum (burgundy) are rare. This series began arriving in the Shashe-Limpopo region before AD 1250, by which time they appeared inenormous quantities. Theywere supplanted by the Zimbabwe seriesat the end of the 13th century.

Subtle changes distinguish the Zimbabwe series from theMapungubwe Oblate series. The glass used to make most colors,

Table 1Breakdown of the analyzed beads by series. Associated dates are approximate.

Bead series Site N of beads

Zhizo 8th century e ca. AD 950 Bosutswe-Kgotla 3 3Nqoma 7Mmadipudi 1Matlapaneng 1Pont Drift 2Schroda 10Diamant 2Total 26

K2 ca. AD 980e1200 K2 16Mapungubwe 3Kgaswe 2Pont Drift 1Schroda 2Skutwater 2Hlamba Mlonga 5Total 31

K2 Garden Roller ca. AD 980e1200 K2 7Mapungubwe 2Bosutswe 1Total 10

Indo-Pacific ca. AD 1000e1250 K2 15Mapungubwe 11Bosutswe-Kgotla 3 1Kgaswe 3Pont Drift 2Schroda 5Skutwater 2Total 39

Islamic ca. AD 1250e1300 Mapungubwe 2Total 2

Mapungubwe Oblate ca. AD 1240e1300 K2 8Mapungubwe 21Bosutswe 15Skutwater 20Kgaswe 3Total 67

Zimbabwe ca. AD 1300e1430 Skutwater 1Thulamela 7Great Zimbabwe 59Hlamba Mlonga 28Total 95

Khami ca. AD 1430e1650 Faure 16Bosutswe 19Tora Nju 9Toutswe 1Thulamela 22Sibudu Cave 3Great Zimbabwe 5Melora 14Total 89

(Unasssigned) Schroda 1Total 1

TOTAL 360

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1901

excluding black, is less opaque and ranges from almost transparentto translucent. New colors appear, including a translucent limegreen and a transparent dark green. Opaque brownish-red beadsalso belong to this series, which was imported into southern Africaup to the early 15th century.

Khami series beads, which span the 15th through the mid 17thcenturies, tend to be somewhat irregular in shape, but most arecylinders with rounded ends. Sizes vary considerably (2e7 mmdiameter), many being significantly larger than beads found inearlier series. Opaque colors include black and brownish-red whiletranslucent-opaque glass is found in blue, blue-green, yellow, dullorange, and green. This series includes the earliest white beadsfound in the region; these are easily distinguished from laterEuropean-made white beads because they are slightly translucentand not pure white.

3. Sample selection

We selected 360 beads but also analyzed separately the differentcolored glasses used to produce the few multicolored beads,resulting in a total of 373 analyses. Table 1 summarizes these beadsby site and bead series, while the locations of the sites are shown onFig. 1. Wood chose non-corroded beads of a wide variety of typesand colors for analysis. Where numerous beads of the same typeand color occurred within an assemblage, several were randomlychosen. We had ample quantities of beads from all series with theexception of the Zhizo series, beads of which are both relatively rareand prone to corrosion.

4. Analytical method

The bead samples were analyzed at MURR by Glascock andPopelka-Filcoff who used a high-resolution Axiom ICP-MS witha New Wave 213 nm laser-ablation system for the sample intro-duction system. As described in greater detail elsewhere(Robertshaw et al., 2006; Speakman and Neff, 2005), each samplewasmeasured for elemental concentrations using a continuous linescan of approximately 2mm length on the bead to provide themostaccurate bulk characterization. To eliminate the effects of surfacecontamination, each line was pre-ablated twice before the start ofdata collection.

The analytical menu consists of 47 elements ranging fromlithium to uranium measured sequentially. Calibration lines foreach element are established by analyzing SRM612 and SRM610glass standards from the National Institute of Standards andTechnology (NIST) along with the Brill (1999) glasses (II, 527e544)and obsidian glass from the Pachuca and Glass Buttes sources(Glascock, 1999). Data for beads and standards are acquired bythree or five runs through the analytical menu followed by blankruns to avoid memory effect. This method allows for averaging theresults to account for variation occurring within the system duringthe run as well as between runs throughout the day.

Elemental concentrations for the beads were calculated usinga normalization procedure described by Gratuze et al. (2001) thatsums the total concentration of oxides to 100%. Precisions reportedas relative standard deviation for the elements range from 2% to20% depending upon the strength of the signal for each element. Acomparison of the accuracy between our results for the calibrationstandards and published values for the standards is in the range of5e10% for most of the elements.

While the concentrations of the rare earth elements (REE) arereported in the tables of results, for the purposes of analyzing theREE data we computed chrondrite normalized values for theseconcentrations, consistent with standard reporting methodologyfor geological materials. We then quantified the Ce and Eu

anomalies for the beads, which we compared with geologicalreference materials (derived from Turekian and Wedepohl, 1961;Taylor and McLennan, 1985).

5. Results

The data are presented in Tables S1 and S2. A common first stepin the examination of glass compositional data is to calculate thereduced compositions of the glass by normalizing the seven majorand minor oxides to 100%. This normative reporting removes mostof the compositional effects of additives, such as colorants, so thatone can examine the main components of the glasses (Brill, 1999:II,8e11). The reduced compositions of the glasses are presented inTable S3, where the oxides are marked with an asterisk to indicatethat these represent reduced compositions.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121902

All the beads except one (PR226) are made of soda-lime-silica(Na2O:CaO:SiO2) glass. PR226, which is a unique bead and does notfit any of our series chemically or morphologically, is made froma mixed alkali (potash and soda) glass. Some analyzed beads arecorroded as a result of complex interactions between the surface ofthe bead and water. Corroded glass is generally depleted in K, Na,Ca, and Mg, but enriched in Si, Al, Ti, and Fe, with the alkalineelements being always heavily leached out (Dussubieux et al.,2009b:157e8). For the purposes of this study we have generallyexcluded any beads with less than 10% Na2O from any summaryreporting or statistical analyses on the grounds that the originalcompositions of these beads have been drastically altered bycorrosion. Some beads with more than 10% Na2O may also beslightly corroded, but the damage is probably not so serious as topreclude use of the data.

6. Glass chemical types

Identification of all the beads (except PR226) as soda-lime-silicaglass is no surprise since almost all Roman, Byzantine, Islamic andIndian glasses, as well as some European glasses, are of this type(Brill, 1999:II; Dussubieux, 2001). However, the type of soda used tomake this glass is sometimes diagnostic of a broad region and/orchronological period. Twomajor sources of alkali were used in glassmanufacture: mineral soda, often in the form of naturally occur-ring natron (or trona), or soda derived from plant-ash that wasobtained by burning alkali-tolerant, halophytic plants found incoastal, salt marsh or desert regions (Barkoudah and Henderson,2006; Brill, 2001a; Tite et al., 2006:1285). These soda types canbe differentiated chemically by the amount of MgO in the glass.Plant-ash glasses generally contain >2.5% MgO, whereas mineral-soda glasses contain less, indeed usually much less MgO than this.

6.1. Mineral-soda glass

Variation in the quantities of K2O and Al2O3 allows the recog-nition of two major types of mineral-soda glass. The best known ofthese are glasses whose source of mineral soda was natron,obtained, perhaps exclusively, from the Wadi el-Natrun near Cairoin Egypt; this natron was used for the manufacture of Roman andByzantine glass. Natron glasses are characterized chemically byamounts of MgO and K2O that are both ca.<1.5% and often <1%, aswell as less than 4% Al2O3 (Brill, 2001a; Freestone, 2006; Freestoneet al., 2000; Henderson et al., 2004; Kato et al., 2009; Sayre andSmith, 1961). From about the early 9th century AD, however,Islamic glassmakers west of the Euphrates substituted plant-ash fornatron in their glass recipes (Henderson et al., 2004). Severalfactors, particularly political upheavals in Egypt that disrupted thesupply of natron (Shortland et al., 2006; Whitehouse, 2002), wereprobably responsible for this change. However, east of theEuphrates plant-ash glass was also manufactured during theSasanian period, mid-3rd to early 7th century AD (Brill, 1999, 2005;Freestone, 2006; Mirti et al., 2008, 2009; Smith, 1963).

The other type of mineral-soda glass, soda-alumina glass, wasrecognized from chemical analysis of Indian glasses (Brill, 1987). Itwas the most common glass type among several hundred glassbeads from archaeological sites in India and Southeast Asia datingbetween the 4th century BC and the 10th century AD (Dussubieux,2001). Dussubieux (2001:114e124; Dussubieux and Gratuze,2003:139, 141; Lankton and Dussubieux, 2006) termed this “m-Na-Al” glass, i.e. a mineral (“m”) soda (Na) glass with relatively highlevels of Al2O3. Characteristic levels of the major and minorelements of this mineral-soda glass were reported as: Al2O3 ¼ 10%;CaO ¼ 3%; MgO <1% and K2O >1.5% (Dussubieux and Gratuze,2003:139). Kock and Sode (1995, see also Brill, 2001b) described

present-day manufacture of glass of this type in the Firozabadregion froman evaporite deposit locally called reh, samples ofwhichhave been analyzed (Brill, 1999:II, 339). A similar salt called ooswasused until recently in glass manufacture in Gujarat (Brill, 2001b).

Recently, Dussubieux et al. (2008, 2009a) have identified fivesub-types of soda-alumina glass: m-Na-Al 1) a low uranium e highbarium (formerly known as lU-hBa) glass knownmostly from southIndia, Sri Lanka and Southeast Asia from the 5th century BC to the10th century AD; m-Na-Al 2) a high uranium e low barium(formerly known as hU-lBa) glass found in eastern and southernAfrica, Madagascar, and Chaul on the west coast of India datingfrom the 8th to the 19th century AD; m-Na-Al 3) a glass that is verysimilar to m-Na-Al 2 but containing more Cs (cesium) and knownonly from Southeast Asia from the 4th and 3rd centuries BC; m-Na-Al 4) a glass characterized by very low concentrations of lime andfound from the 14th century AD and later in Bangladesh, ona Chinese junk shipwrecked off the Sultanate of Brunei, and in threebeads discovered at the site of Muasya in Kenya; m-Na-Al 5) a glasswith relatively high lime and low potash concentrations, as well asvery little uranium in comparison with the other m-Na-Al glassesand known only from the 12th to 14th century AD site of Sardis inTurkey. Most, perhaps all, of our Southern African samples of soda-alumina glass best fit within sub-typem-Na-Al 2, but it is clear fromour data that this sub-type warrants division because the compo-sition of the Khami series beads is distinctly different from that ofthe earlier Southern African soda-alumina beads.

6.2. Plant-ash glass

Numerous beads made of plant-ash glass are also reported here.Starting in the 9th century, plant-ash glass became the mostcommon glass type found throughout the Mediterranean area.However, in the period towhich our southern African beads date, weknownot only ofMiddle Easternplant-ash glass but also of plant-ashglass of distinctly different composition found at sites in Sumatradated to the 9the15th centuries AD and known to have been tradedby Chinese merchants (Dussubieux and Kusimba, submitted forpublication; Dussubieux et al., 2009a; McKinnon and Brill, 1987).This Sumatran glass generally has more Al2O3 and less CaO than itsMiddle Eastern counterpart. Within Africa, vessel glass composi-tionally similar to the Sumatran glass has been found at Mtwapa inKenya (Dussubieux and Kusimba, in review), while beads that arebroadly similar chemically were discovered in Madagascar(Robertshaw et al., 2006). Relatively high Al2O3 plant-ash glasseswere also recovered at Qsar-es-Seghir in Morocco, though theypossess higher CaO levels than the Sumatran samples (Brill, 1999:II,175e6). Both low aluminae high lime (lAl-hCa) glass like that of theMiddle East and high alumina e low lime (hAl-lCa) glass similar tothat of Sumatra have been identified in the present study.

7. Chemistries of the southern African bead series

Each of the southern African bead series has distinctive glasschemistry. Beads of the following series were made from mineral-soda glasses: K2, K2 Garden Rollers (K2 GR), Indo-Pacific, andKhami. Plant-ash glasses were used to make Zhizo, Islamic,Mapungubwe Oblate, and Zimbabwe series beads. The mineral-soda glasses have more U3O8 but less MgO and CaO than theirplant-ash counterparts (Figs. 3e5; see also Table 2). Measurementsof other oxides also separate the bead series by chemistry. Forexample, mineral-soda glasses possess greater amounts of TiO2,commonly occurring as the mineral rutile in beach sands, andgreater concentrations of REE than do the plant-ash glasses.Calculation of the Ce and Eu anomalies (see Robertshaw et al.,2006) also shows that the mineral-soda glasses are characterized

Fig. 3. Plot of MgO* vs. K2O* in the reduced compositions of the southern Africanbeads. Beads with <10% Na2O* omitted.

Fig. 5. Box-plots of U concentrations in the Southern African beads. Note the loga-rithmic scale. Beads with <10% Na2O omitted.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1903

by a negative Eu anomaly while the plant-ash glasses have no ora positive Eu anomaly. A Ce anomaly, however, is not diagnostic ofany particular glass type (Fig. 6). Plant-ash glasses have more P2O5,an oxide whose concentrations also serve to separate the differentplant-ash bead series (Fig. 7; see also Fig. 8; Table 3). Zhizo beadsare distinctly different from the plant-ash beads of other series inpossessing much less Al2O3, as well as less CaO and more SiO2(Table 4; Fig. 4).

Fig. 4. Plot of CaO* vs. Al2O3* in the reduced compositions of the southern Africanbeads. Beads with <10% Na2O* omitted.

8. Provenance of the individual bead series

The chemistry of a glass reflects the geochemistry of the regionor regions where the ingredients used in glass manufacture wereacquired. Thus, glass manufactured in a particular region hasa distinctive and persistent chemical signature, assuming no majorchanges occur in the choice of raw materials. Once glass has beenmade, its chemical composition will not change significantly.However, raw materials, particularly coloring agents like cobalt,were sometimes widely traded prior to glass manufacture(Henderson, 1998). In addition glass beads were often made atcenters different from, and sometimes even continents away from,glassmanufacturing sites so in theoryglasses fromdifferent sourcescould have been used in making even a single bead. Glass may alsoundergo chemical weathering and surface hydration over time,particularly after burial, thereby altering its chemical compositionin predictable ways (Cox and Ford, 1993; Dussubieux et al., 2009b;Salviulo et al., 2004).

8.1. Zhizo series

Zhizo beads, identified both morphologically and chemically,occur not only in southern Africa but also in west Africa, at Igbo-Ukwu (Nigeria), Gao (Mali), and Kissi (Burkina Faso), and rarely ineast Africa at Tumbe (Pemba Island, Tanzania) and Shanga (Kenya)(Wood and Robertshaw, 2009).

Zhizo beads were manufactured from lAl-hCa plant-ash soda-lime-silica glass, indicative of a Middle Eastern origin, rather thanfrom a hAl-lCa glass typical of South Asia. Zhizo bead compositionsaremost similar to thehigh (byMiddle Eastern standards) Al2O3 glasssamples from Nishapur in Iran and differ markedly from samplesfrom further west in theMiddle East in that they possess more Al2O3and less CaO, aswell as generally higher combined totals ofMgO andK2O (Table 5). Zhizo beads are also similar chemically to earlierSasanian glasses from east of the Euphrates River (Mirti et al., 2008,2009). The high levels of Al2O3 in the Zhizo beads, compared withglasses fromwest of the Euphrates, is indicative of the use of sand asthe silica source. Similarly the greater combined totals of MgO andK2O that characterize glasses from east of the Euphrates may beexplained either by the use of different plant species in ash prepa-ration or by the fact that the high MgO levels reflect the MgO-richalluvium of the Euphrates and Tigris valleys (Freestone, 2006).

Table 2Descriptive statistics (%) for the oxides of the reduced compositions of the beadseries. (Beads with <10% Na2O* were omitted from the calculations.)

Bead series N Minimum Maximum Mean Std. Dev.

Zhizo Na2O* 16 10.17 16.38 13.15 1.75MgO* 16 2.50 7.17 4.31 1.37Al2O3* 16 2.29 4.09 3.26 0.58SiO2* 16 65.56 75.94 69.62 2.88K2O* 16 1.69 4.49 3.23 0.70CaO* 16 3.39 9.13 5.50 1.52Fe2O3* 16 .42 2.79 0.94 0.56

K2 Na2O* 29 10.91 22.37 16.22 2.80MgO* 29 .16 .83 0.43 0.18Al2O3* 29 6.40 17.66 11.85 3.09SiO2* 29 57.89 73.58 64.51 4.01K2O* 29 1.98 5.11 3.34 0.73CaO* 29 1.53 3.37 2.34 0.49Fe2O3* 29 .55 3.46 1.30 0.67

K2 GR Na2O* 11 10.26 19.93 14.36 3.00MgO* 11 .26 .47 0.37 0.07Al2O3* 11 9.53 20.69 16.63 3.40SiO2* 11 55.09 67.54 61.05 3.74K2O* 11 1.82 8.31 3.39 1.86CaO* 11 1.84 4.10 2.85 0.86Fe2O3* 11 .84 1.72 1.35 0.32

Indo-Pacific Na2O* 38 11.26 20.23 14.75 2.29MgO* 38 .24 1.78 0.59 0.32Al2O3* 38 4.21 21.09 13.00 3.93SiO2* 38 55.26 71.96 63.08 4.73K2O* 38 1.90 5.50 3.46 0.97CaO* 38 1.73 4.84 2.85 0.82Fe2O3* 38 .64 5.84 2.27 1.33

Islamic Na2O* 3 10.55 18.20 13.71 4.00MgO* 3 4.45 5.19 4.83 0.37Al2O3* 3 3.79 7.57 6.05 2.00SiO2* 3 58.87 65.74 63.21 3.77K2O* 3 3.69 4.02 3.91 0.19CaO* 3 5.33 8.40 6.63 1.59Fe2O3* 3 1.12 2.65 1.66 0.85

Map Oblate Na2O* 57 10.38 19.51 13.47 1.83MgO* 57 2.68 10.16 5.80 1.87Al2O3* 57 5.01 11.93 7.67 1.48SiO2* 57 51.34 70.06 61.88 4.96K2O* 57 2.02 5.31 3.47 0.63CaO* 57 3.35 12.41 6.66 1.78Fe2O3* 57 .52 2.54 1.04 0.33

Zimbabwe Na2O* 90 10.31 20.89 15.81 2.27MgO* 90 2.97 6.02 4.33 0.69Al2O3* 90 2.98 9.77 6.71 0.82SiO2* 90 51.21 67.26 60.98 2.95K2O* 90 2.21 5.13 3.74 0.63CaO* 90 4.27 11.30 6.94 1.34Fe2O3* 90 .64 3.26 1.48 0.50

Khami Na2O* 81 10.42 31.76 18.66 3.98MgO* 81 .47 2.73 1.21 0.55Al2O3* 81 5.39 16.22 9.81 2.01SiO2* 81 48.65 74.95 61.40 4.78K2O* 81 1.45 9.20 2.82 1.13CaO* 81 1.71 6.61 3.39 0.92Fe2O3* 81 .76 7.30 2.70 1.37

Fig. 6. Plot of the Ce vs. Eu anomalies in the Lanthanide series of elements. Beads with<10% Na2O* omitted.

Fig. 7. Plot of TiO2 vs. P2O5 in the compositions of the southern African beads. Beadswith <10% Na2O omitted. PR215, with 3.74% P2O5 is not shown.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121904

Lead-isotope analysis supports the hypothesis of a MiddleEastern provenance east of the Euphrates for the glass used tomakeZhizo beads (Fenn et al., 2009). However, two caveats are neces-sary: 1) none of the analyses of Middle Eastern glass was under-taken on beads, but rather on vessel glass, cullet, and other items;2) the beads themselves were possibly manufactured in one ormore places, perhaps far distant from the workshop(s) where theraw glass was made.

8.2. K2, K2 garden roller, and Indo-Pacific series

The K2, Garden Roller (K2 GR), and Indo-Pacific beads weremade from soda-alumina glasses, whose reduced compositions are

indistinguishable apart from the greater quantity of Al2O3 in the K2Garden Rollers and of Fe2O3 in the Indo-Pacific beads. Chemicallythis glass is that identified by Davison (1972) as the M1 ChemicalGroup. The enriched Al2O3 in the K2 GR beads may derive perhapsfrom the addition of small amounts of alumina-rich local sand intothe glass batch or more likely from the molds in which the beadswere fired. While there is about a 5% difference between the K2 andK2 GR beads in the mean concentrations of Al2O3 (Table 2), there is

Fig. 8. Plot of V2O5 vs. Cr2O3 in the compositions of the southern African beads. Beadswith <10% Na2O omitted.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1905

considerable overlap in the ranges of Al2O3 in the two types. Amoremodest enrichment of about 2% Al2O3 in the making of the K2 GRbeads is in line with the observation that the higher Al2O3 levels insome of the Type 2 glass samples from al-Raqqa (Syria) “.mayhave interacted with a ceramic (such as the wall of a glass-meltingcrucible) to the extent that aluminia migrated into the glass”(Henderson, 2000:84).

Indo-Pacific series beads possess more Fe2O3 than the K2 and K2GR beads because iron was used as a coloring agent in green glassand as a reducing agent, encouraging the precipitation of copper asmetal or cuprite, in brownish-red glass. Indo-Pacific beads also havesignificantly more TiO2 and Cr2O3 than do K2 beads (but not K2 GR

Table 3Comparison by bead series of the means of minor and trace oxides not commonly used as(p < 0.05). For each statistically significant pair, the letter key of the smaller category appepossess a larger mean concentration of P2O5 than do the Indo-Pacific beads (A) and thesample size was 1. (Beads with <10% Na2O were omitted from the calculations.)

Bead series

Indo-Pacific K2 K2 GR K

n 38 29 11 8

Alkali Mineral Mineral Mineral M

(A) (C) (D) (E

P2O5

TiO2 C F G H F G H F G H CV2O5 F G H F G H F G H ACr2O3 C CZnORb2O F G H E F G H A E F G H FSrOY2O3 F G H F G H F G H FZrO2 E F G H E F G H F G H FMoO2 F H F H F H FCs2O F G H F G H F G H FBaO G G G GHfO2 E F E F G H E F G H FThO2 F G H FU3O8 F G H F G H A

beads), while the K2 GR beads have significantlymore Rb2O than dothe Indo-Pacific beads. Rb2O is significantly correlatedwith Al2O3 inthe K2 GR beads (Pearson's correlation (pc) ¼ 0.676; p < 0.01),suggesting that the high Al2O3 concentration in the K2 GR beadsexplains this difference from the Indo-Pacific beads' chemistry.

As noted above, Dussubieux et al. (2009a) identified five sub-types of soda-alumina glass. The beads of these three series allappear to fall broadly withinm-Na-Al 2, characterized by high U andlow Ba concentrations. Table 6 compares the U, Ba, Sr, and Zrconcentrations in these southern African bead series, plus the Khamiseries (discussed below), with soda-alumina glass beads fromvarious sites (Dussubieux et al., 2008): Mtwapa, an urban site on theKenyan coast dating approximately to the 10the18th centuries(Kusimba, 1999) that contains many Indo-Pacific and Khami beads(Wood, pers. obs.); Mahilaka, a trading settlement of similar age inMadagascar,whose tested soda-alumina beadswith>10%Na2O (andhence uncorroded) comprise 10 Indo-Pacific and 2 K2 beads(Robertshaw et al., 2006); and the 9the19th century levels of Chaul,a port on the west coast of India (Gogte, 2003). Three of the fourelements have very similar average concentrations at all the sites.Only Zr shows considerable variation between sites. The high meanlevels of Zr in the southern African beads may indicate that thesource of the sand used in glass manufacture was close to an oldmountain belt.

There is an exception to the rule that all the K2, K2 GR, and Indo-Pacific beads consist of soda-alumina glass. This is the white glass,occurring as irregular swirls in the blue-green glass of PR065, a K2GR bead. Chemically this looks to be a typical Middle Eastern plant-ash glass. Perhaps it resulted from curated Zhizo beads, which oftencorrode to a whitish color, being poorly mixed with K2 bead glass.Apart from this exception, the glass used to manufacture the K2, K2GR and Indo-Pacific beads found in southern Africa, as well as eastAfrica and Madagascar, is clearly of South Asian origin. We cannotyet be more specific about its provenance, though the west coast ofIndia seems a plausible candidate (Dussubieux et al., 2008, 2009a).

8.3. Islamic series

Only two southern African beads of the Islamic series wereanalyzed, PR068 and PR177. These multicolored beads have varied

coloring or opacifying agents showing those with statistically significant differencesars under the category with the larger mean. For example, Mapungubwe Oblates (F)difference is statistically significant. The Islamic series is not included because the

hami Map Oblate Zhizo Zimbabwe

1 56 16 89

ineral Plant-ash Plant-ash Plant-ash

) (F) (G) (H)

A C D E A C D E A C D E FD F G H G GC D F G H

A C D E F HE

G HA C D E A C D E A C D E

G H

H F HG H

G A C E F GG HHC D F G H

Table 4Comparison by bead series of themeans of the oxides of the reduced compositions showing those with significant differences (p< 0.05). For each significant pair, the letter keyof the smaller category appears under the category with the larger mean. (Beads with <10% Na2O* were omitted from the calculations.)

Bead series

Indo-Pacific Islamic K2 K2 GR Khami Map Oblate Zhizo Zimbabwe

n 38 3 29 11 81 57 16 90

Alkali Mineral Plant-ash Mineral Mineral Mineral Plant-ash Plant-ash Plant-ash

(A) (B) (C) (D) (E) (F) (G) (H)

Na2O* F G A C D F G H F GMgO* A C D E A C A C D E G H A C D E A C D EAl2O3* B E F G H B E F G H A B C E F G H F G H G GSiO2* E H A C D E F HK2O* E E ECaO* A C D E C A C D E G A C D E A C D E GFe2O* C F G H C D F G H

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121906

chemistries, partly as a result of coloring agents. A third bead of thisseries, PR178, came from al-Fustat (Cairo). The reduced composi-tions of the various colored glasses of these three beads (Table 7) donot all fit comfortably within the known chemistries of MiddleEastern glass of the Islamic period, though the white and yellowglassesof the al-Fustatbeadand the redglass of PR068 look likegoodcandidates for Middle Eastern plant-ash glass. The other analysesexhibit much higher Al2O3 levels than is normal for Middle Easternglass and not all of these data can be explained as a byproduct ofcorrosion. Thereare also someunusuallyelevatedvalues recorded asa result of the use of coloring agents: the black glass of PR068possesses 2.29%MnO2; thewhiteglass of PR177has13.25%SnO2 andthat of PR1786.43%, though there is far less SnO2 (0.59%) in thewhiteglass of PR068; the yellow glasses of PR177 and PR178 respectivelypossess 21.72% and 20.99% PbO, with correlated quantities of SnO2indicative of the use of lead stannate as the yellow colorant. Thevaried chemistries of the glasses comprising these beads suggestthey were made with a mixture of recycled and imported glasses.These two rare, decorated Islamic series beads were found on

Table 5Reduced compositions of Zhizo andMiddle Eastern plant-ash glasses. There are two sub-ty1995); this table reports only the high alumina (�2.5% Al2O3*) glasses. Several plant-ash gdistinguished here. None of the al-Raqqa plant-ash glass types per se is a good match forNa2O* were omitted from the calculations. Comparative data from Brill (1999), vol 2; Fre

Zhizo Nishapur Siraf

Southern Africa Iran Iran

n 16 14 9

Mean Std. Dev. Mean Std. Dev. Mean

Na2O* 13.15 1.75 17.53 1.88 13.11MgO* 4.31 1.37 3.79 0.85 3.07Al2O3* 3.26 0.58 3.29 0.38 1.42SiO2* 69.62 2.88 64.24 3.00 71.03K2O* 3.23 0.70 3.18 0.97 2.97CaO* 5.50 1.52 6.86 1.03 7.10Fe2O3* 0.94 0.56 1.12 0.35 1.29

Fustat Serce Limani

Egypt Turkey

n 20 95

Mean Std. Dev. Mean Std. Dev

Na2O* 11.78 4.44 13.00 0.94MgO* 5.09 1.41 2.70 0.43Al2O3* 1.34 0.57 1.90 0.37SiO2* 70.50 2.64 n/a n/aK2O* 2.38 0.32 2.66 0.35CaO* 8.23 3.29 9.45 1.24Fe2O3* 0.68 0.34 n/a n/a

Mapungubwe Hill, where the king and other royal functionariesresided. PR177came fromthe so-calledGoldGrave areawhile PR068was found at bedrock near grave No. 14, which contained consid-erable quantities of gold ornaments and many other glass beads(Wood, 2005:57). As only twoof threedecoratedbeads found todatein southern Africa in this period (the third e from the same part ofthe site e has disappeared), and being so unlike all other importedbeads, one might assume they were royal treasures.

8.4. Mapungubwe Oblate and Zimbabwe series

The similar Mapungubwe Oblate and Zimbabwe bead series canbe distinguished from each other by the higher concentrations ofP2O5, BaO, and Na2O, but lowerMgO in the Zimbabwe series (Figs. 9and 10). Both of these bead series are manufactured from hAl-lCaplant-ash glass, which Davison (1972) recognized as the Mapun-gubwe Chemical Group. However, variation in glass compositionalso exists within the two bead series: among the MapungubweOblates, PR019 possesses considerable PbO, while PR028 and PR081

pes of glass evident at Nishapur separable on the basis of their alumina content (Brill,lass types are represented also at al-Raqqa (Henderson et al., 2004), but these are notthe Zhizo beads (see Robertshaw et al., 2009 for discussion). Zhizo beads with <10%estone (2002), Henderson et al. (2004), Mirti et al. (2008, 2009).

Al-Raqqa Tyre

Syria Lebanon

102 10

Std. Dev. Mean Std. Dev. Mean Std. Dev.

2.46 14.27 1.42 13.64 1.170.39 4.25 1.19 3.34 0.610.60 2.03 1.06 1.74 0.293.26 68.82 2.56 67.74 3.260.64 2.77 0.60 2.73 0.100.93 6.99 1.77 9.83 2.170.97 0.87 0.59 0.97 1.03

Sasanian Type 1 Sasanian Type 2

Iraq Iraq

48 19

. Mean Std. Dev. Mean Std. Dev.

16.97 1.47 18.61 1.354.27 0.45 7.36 1.342.46 0.54 1.64 0.45

64.50 1.81 62.29 2.443.46 0.61 3.15 0.627.22 1.08 6.34 1.171.12 0.31 0.61 0.17

Fig. 9. Plot of BaO vs. P2O5 in the compositions of the Mapungubwe Oblate andZimbabwe series beads. Beads with <10% Na2O omitted.

Table 6Comparison of the average concentrations (ppm) with relative standard deviationsof various elements of the K2, GR, Indo-Pacific, and Khami bead series with beadsfrom sites in Kenya, Madagascar, and India. (Southern African and Malagasy beadswith <10% Na2O were omitted from the calculations.)

K2 &K2GRseries

Indo-Pacificseries

Khamiseries

Mtwapa(Kenya)

Mahilaka(Madagascar)

Chaul(India)

n 42 38 81 43 12 27U 68 � 68 84 � 59 189 � 100 109 � 80 98 � 79 105 � 41Ba 468 � 125 443 � 128 442 � 155 357 � 124 499 � 98 340 � 79Sr 196 � 68 218 � 84 228 � 99 213 � 79 193 � 52 222 � 49Zr 411 � 198 455 � 480 253 � 168 157 � 68 322 � 138 148 � 61

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1907

contain >10% CaO; among the Zimbabwe series, most beads fromThulamela possess low levels of Na2O, normally associated withcorrosion, of which they exhibit no visible signs, while PR926 fromHlamba Mlonga is unusual both for its comparatively low Al2O3level and for its method of manufacture, having beenwound ratherthan drawn. Corrosion is a feature of black Mapungubwe Oblateswhich decay to a yellowish and then whitish color. PR790ePR794represent a series of black beads that exhibit increasing visiblecorrosion, demonstrated chemically by their declining Na2O levels(see also Prinsloo and Colomban, 2008:87e8).

This hAl-lCa plant-ash glass has also been found at several othersites (Table 8), all of which are approximately contemporary withsouthern African sites containing Mapungubwe Oblate andZimbabwe series beads. Indeed, Mahilaka beads of this glass mostlybelong to the Zimbabwe series. Despite broad similarities, there issome variation between the assemblages suggesting that it isunlikely that all these glasses have the same provenance; forexample, the average concentration of Zr (111 � 10 ppm) in theMtwapa assemblage (Dussubieux and Kusimba in review) is inter-mediate between those of the Mapungubwe Oblate (88 � 19 ppm)and Zimbabwe (147 � 27 ppm) bead series. Looking further afield,the lower Al2O3 and MgO levels, as well as the high CaO:Al2O3 ratio,of theQsar-es-Seghir glass (Table 8) indicate that this is a poormatchfor the other African hAl-lCa plant-ash glasses, both chemically andgeographically, as all the other sites are located in the Indian Oceanregion. The glass from the port of Kota Cina on Sumatra, for which anIndian origin has been suggested (McKinnon and Brill, 1987), alsoshows compositional differences from the Mapungubwe Oblate andZimbabwe series beads. However, on present evidence, the highAl2O3 concentrations in these two bead series favors a hypothesis ofSouth Asian or perhaps Southeast Asian, rather thanMiddle Eastern,origins, though the current state of knowledge pertaining to plant-ash glass production in the second millennium AD in South andSoutheast Asia is woefully inadequate.

8.5. Khami series

After the plant-ash glasses of the Mapungubwe Oblate andZimbabwe series, the Khami bead series represents a return to hU-

Table 7Reduced compositions (%) of the Islamic series beads from Mapungubwe and al-Fustat.

ANID PR068 (black) PR068 (red) PR068 (white) PR177 (black) PR177

Site Mapungubwe Mapungubwe Mapungubwe Mapungubwe MapuColor Black Red White Black WhiteNa2O* 4.96 10.55 7.09 18.20 2.20MgO* 1.47 5.19 1.90 4.86 1.36Al2O3* 10.41 3.79 12.24 6.79 11.63SiO2* 68.59 65.74 67.12 58.87 71.66K2O* 3.67 3.69 2.38 4.01 4.97CaO* 8.45 8.40 7.58 6.15 6.23Fe2O3* 2.45 2.65 1.70 1.12 1.95

lBa soda-alumina glass. The Khami series glass is distinguishedfrom its earlier soda-alumina counterparts (the K2, K2 GR, andIndo-Pacific series) by its higher Na2O, MgO, and CaO and lowerAl2O3 concentrations (Tables 2 and 4). It can also be distinguishedon the basis of several trace elements (Table 3; see also Wood et al.,2009). Indeed, its high average U3O8 concentration differentiatesthe Khami bead series from the soda-alumina glasses of not onlythe earlier southern African bead series but also from the soda-alumina glasses of Madagascar, as well as the samples fromMtwapa (Kenya) and Chaul (India) used to define the m-Na-Al 2glass type (Table 6).

There is also variation within the Khami series glasses. MgOconcentrations are higher in most of the beads from Botswana(Bosutswe, Tora Nju, Toutswe) than in beads from Zimbabwe andSouth Africa (Fig. 11). Reaching a maximum of 2.65% MgO (PR116 atBosutswe), one might assume that some of these beads were madefrom plant-ash glass, particularly when it is borne in mind thatplant-ash glass beads of the Mapungubwe Oblate series also occurat Bosutswe. However, these are neither cases of misidentificationof bead series nor of glass type because all the Khami series beads inBotswana, including those with relatively high concentrations ofMgO, possess plenty of U3O8, in marked contrast to the plant-ashglasses of both the Mapungubwe Oblate and Zimbabwe series. We

(white) PR177 (yellow) PR178 (black) PR178 (white) PR178 (yellow)

ngubwe Mapungubwe Fustat Fustat FustatYellow Black White Yellow12.38 12.99 12.88 12.134.45 2.68 3.10 2.807.57 8.35 2.50 2.1365.01 62.63 70.57 70.484.02 4.20 2.70 2.685.33 6.23 7.66 9.061.22 2.92 0.59 0.72

Fig. 11. Box-plots of MgO* concentrations by country in the Khami series beads. Beadswith <10% Na2O* omitted.

Fig. 10. Plot of MgO* vs. Na2O* in the reduced compositions of the MapungubweOblate and Zimbabwe series beads. Beads with <10% Na2O* omitted.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121908

have no explanation for the higher MgO levels in Khami series glassfrom Botswana.

The provenance of the glass from which the Khami bead serieswasmanufactured is most probably located in South Asia. However,it is doubtful that all the soda-alumina glass comes from the sameregion within the South Asian subcontinent because of the chem-ical variation between the bead series. In particular the Khamiseries has a significantly (p< 0.05) higher Eu anomaly, a result of Eufractionation from igneous processes, than the K2 and Indo-Pacificseries beads (Fig. 12). The Khami series Eu anomaly correspondsbroadly with that of intermediate igneous rocks such as diorite andandesite, whereas many of the K2 series beads and some of theIndo-Pacific series lie closer to the reference sample for granite (seeRobertshaw et al., 2006:105 for further discussion).

9. Coloring agents

The most common colorants intentionally added to glass duringmanufacture are transition metal compounds (Bamford, 1977).Different types of glass, produced under different manufacturingconditions, may produce a range of colors from the same transition

Table 8Reduced compositions (means and standard deviations) of the Mapungubwe Oblate andregions. (Southern African beads with <10% Na2O were omitted from the calculations.)

Southern Africa Madagascar Kenya

Map Oblate Zimbabwe Robertshaw et al., 2006 Mtwapa

Dussubie

n 57 90 15 30Na2O* 13.5 � 1.8 15.8 � 2.3 13.7 � 1.8 18.8 � 1MgO* 5.8 � 1.9 4.3 � 0.7 4.1 � 0.5 5.2 � 0.4Al2O3* 7.7 � 1.5 6.7 � 0.8 5.2 � 1.2 5.7 � 0.6SiO2* 61.9 � 5.0 61.0 � 3.0 67.7 � 2.9 59.3 � 1K2O* 3.5 � 0.6 3.7 � 0.6 3.3 � 0.7 2.7 � 0.3CaO* 6.7 � 1.8 6.9 � 1.3 4.5 � 0.7 5.1 � 0.5Fe2O3* 1.0 � 0.3 1.5 � 0.5 1.4 � 0.7 1.7 � 0.3

metal impurity or colorant. Other factors may also influence glasscolor; for example, adding iron to glass in a reducing furnace willproduce a blue-green glass, but the same formula in an oxygen-richfurnace will produce a pale yellow-green to red glass.

Iron, lead, tin, copper and cobalt were clearly used as coloringagents in the southern African beads. A compound of lead and tin,lead stannate (PbSnO3), was the preferred yellow coloring agent,particularly for the plant-ash glasses, while iron, dissolved as Fe3þ,was sometimes used to make yellow in the mineral-soda glasses.Iron acts as a reducing agent in brownish-red glasses, pulling thecopper into the reduced state so that it precipitates as metal orcuprite (Nassau, 2001). Among the southern African beads mineral-soda glasses tend to contain more iron relative to copper thanplant-ash glasses. Copper was the major colorant for blue-greenand green glasses. The quantity of copper is correlated with that oftin in the composition of the blue-green and green beads of theZimbabwe series (Pc ¼ 0.382; p ¼ 0.018), suggesting the use ofbronze, while copper concentrations correlate with those of zinc inbeads of these colors in the Khami series (Pc ¼ 0.437; p ¼ 0.048),hinting at brass, though the actual concentrations of zinc aregenerally very low.

Coloring agents with potential for assisting in indentifyingprovenience are manganese and cobalt. MnO was used as a decolor-ant in the IslamicMiddle East, where it commonly comprises 0.5e1%of the glass (Brill, 2001a:29), a figure that corresponds to its averageconcentration (0.81% � 0.6) in the Zhizo bead series. The other bead

Zimbabwe bead series compared with glasses of similar compositions from other

Sumatra Morocco

Kota Cina Qsar-es-Seghir

ux and Kusimba, submitted for publication Brill 1999 Brill 1999

11 18.3 18.5 � 1.0 19.4 � 1.3

5.6 � 1.0 3.2 � 0.84.7 � 0.7 4.3 � 0.6

.9 63.5 � 2.2 60.4 � 2.23.0 � 0.4 3.3 � 0.63.6 � 0.3 6.2 � 0.91.2 � 0.2 2.0 � 1.0

Fig. 12. Plot of the Ce vs. Eu anomalies in soda-alumina glasses compared withgeological reference samples. Beads with <10% Na2O* omitted.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1909

series of plant-ash glasses have lower MnO concentrations(MapungubweOblate: 0.38%� 0.55; Zimbabwe: 0.28%� 0.35), whilethe average MnO concentrations of the bead series made withmineral-soda glasses are all considerably less than 0.1%.

Cobalt was frequently used to color glass a deep blue, butsources are relatively few and they may be identified by their traceelements. There are no analyzed glass beads from the K2, K2 GR,and Indo-Pacific series that show the use of cobalt as a coloringagent; all contain less than 100 ppm CoO. CoO is present inamounts in excess of 100 ppm and thus presumably deliberatelyadded in beads of the other series, though there is only a singleexample from the Islamic series (Table 9).

Assuming that the source of the glass of the Zhizo series lies eastof the Euphrates in theMiddle East (as discussed earlier), the sourceof the cobalt in the Zhizo series beads is likely to be one or more ofthe known ores found in Iran, one of which is erythrite, an arsenicalcobalt found at Qamsar (Freestone and Stapleton, 1998:123), andanother is a zincian cobalt found at Anorak near Tabriz (Henderson,

Table 9Pearson correlation coefficients and their significance between cobalt oxide (CoO) and vawith >100 ppm CoO and >10% Na2O were included in the analysis.

Bead series MnO Fe2O3 CoO

Zhizo CoO Pearson Correlation .252 .412 1Sig. (2-tailed) .547 .310n 8 8 8

Map oblate CoO Pearson Correlation �.408 �.388 1Sig. (2-tailed) .422 .447n 6 6 6

Zimbabwe CoO Pearson Correlation �.288 �.459 1Sig. (2-tailed) .391 .156n 11 11 11

Khami CoO Pearson Correlation �.290 �.414 1Sig. (2-tailed) .228 .078n 19 19 19

*Correlation is significant at the 0.05 level (2-tailed).**Correlation is significant at the 0.01 level (2-tailed).

1998:118). Other cobalt minerals, includingMn-rich asbolane, wereexploited at Qamsar, as documented in the 1302 treatise by Ab�u'lQ�asim (Allan, 1973). Cobalt pigments found in 9th century blueglazes on Islamic Samarra faiences derive from this Qamsar source(Kleinmann, 1991). While cobalt of the Zhizo series beads is posi-tively correlated only with InO (Table 9), the ratios of cobalt to otheroxides indicate that there are probably at least two cobalt mineralsrepresented in the Zhizo beads, one a zincian cobalt and the othera Mn-rich cobalt, though there seems to be no arsenical cobalt.Thus, the chemical results do not contradict a hypothesis of anIranian origin for the cobalt of the Zhizo beads. If the glass, but notthe actual beads, was made in Iran, then it is also possible thatIranian cobalt may have been traded along with raw glass to theregion where the beads were manufactured.

The cobalt in the Mapungubwe Oblate, Zimbabwe and Khamiseries, all of which are probably of south Asian origin, appears to beof different mineral origin from that of the Zhizo series. The CoO inall three series is highly positively correlated with As2O3. There arealso several statistically significant negative correlations, notablybetween CoO and InO in the Zimbabwe series, a marked contrastwith the Zhizo series (Table 9). All the Mapungubwe Oblate andZimbabwe cobalt-rich beads have a CoO:As2O3 ratio of approxi-mately 1:1. Previously we used this ratio to identify the cobaltmineral used in Zimbabwe series beads found in Madagascar ascobaltite, for which there is a source in Rajasthan, India(Robertshaw et al., 2006:102); however, arsenic is volatile and maybe lost in the preparation of the pigment through heating(Dussubieux et al., 2008:816; Zucchiatti et al., 2006). Therefore,while the cobalt mineral used was certainly arsenic-rich, we cannotwith confidence identify it as cobaltite.

For the Khami series there is more variation between the ratiosof the quantities of CoO to those of the other oxides than is found inthe Mapungubwe Oblate and Zimbabwe series. Although arsenic-rich cobalt dominates the Khami series, other cobalt minerals mayalso be present; for example, many of the cobalt-colored beads ofthe Khami series are relatively rich in Ni (see also Wood et al.,2009), which is very different from almost all the MapungubweOblate and Zimbabwe cobalt-rich beads.

The interpretation that the makers of the beads of the differentseries employed different cobalt sources is borne out by a clusteranalysis of the beads using the quantities of the oxides listed inTable 9 as the variables (Fig. 13).

10. Discussion and conclusions

Each of the southern African bead series defined on morpholog-ical traits has a characteristic chemistry, though some series share

rious oxides that may be represented in the mineral source of the cobalt. Only beads

NiO ZnO CuO As2O3 InO PbO

�.335 .344 .286 .683 .738* .302.417 .405 .493 .062 .037 .468

8 8 8 8 8 8�.856* �.561 �.853* .998** �.597 �.543.030 .247 .031 .000 .210 .265

6 6 6 6 6 6�.331 �.650* �.753** .899** �.720* �.688*.320 .030 .008 .000 .013 .019

11 11 11 11 11 11.424 �.306 �.546* .604** �.307 �.302.071 .202 .016 .006 .200 .209

19 19 19 19 19 19

Fig. 13. Hierarchical average-linkage cluster analysis, using squared Euclidean distance with the values standardized to z-scores by variable, of cobalt-rich beads using the oxideslisted in Table 9. Only beads with >100 ppm CoO and >10% Na2O were included in the analysis.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121910

the sameor very closely similar chemistries. Beads of the Zhizo serieswere made from a plant-ash glass probably manufactured in Iran.Zhizo beads of the same basic chemical composition have also beendiscovered in West Africa and in very small numbers on the EastAfrican coast. They date from approximately the 8th to themid-10thcentury. The appearance of the first beads of the K2 series in aboutAD 980 and those of the Indo-Pacific series shortly thereafter marka radical shift in the chemistry and probable origin of the glassimported into southern Africa. These two series, plus the closelyrelated K2 GR beads, were made from hU-lBa soda-alumina (m-Na-Al 2) glass, almost certainlymanufactured at workshops somewherein South Asia. This shift may be part of a broader reorientation oftrading networks in the Indian Ocean world with South Asianproducts and ports benefitting from increased trade with sub-Saharan Africa at the expense perhaps of ports on the Persian Gulf.Moreover, these beads are far more numerous in southern Africansites than their Zhizo predecessors, presumably as a result of an

intensification of trade spurred in turn by demand for prestigeobjects by the emerging leaders of the first polities in southernAfrica. Another change occurs in the mid-13th century when vastnumbers of MapungubweOblates flood into southern Africa and theK2 and Indo-Pacific beads disappear from circulation. These newbeads and the Zimbabwe series beads that succeed them in the 14thcentury may well derive from the same broad South Asian region astheir K2 and Indo-Pacific predecessors, but they were made froma plant-ash glass rather than the soda-alumina glass of the K2 andIndo-Pacific series. It is also during the occupation of Mapungubwein the last half of the 13th century thatwefind the rare Islamic seriesbeads that are probably of Middle Eastern origin, perhaps from al-Fustat. The final shift in the glass bead sequence prior to the arrival ofthe Portuguese on the southern Africa coast is the appearance of theKhami series in the early 15th century. The Khami series representsa return to the soda-alumina glass of earlier centuries, being closelyrelated both chemically and morphologically to the Indo-Pacific

Table 10Tentative correlations between the date of changes in the southern African beadseries and historical events in the Indian Ocean world.

Approx. date Transition Historical events

950/1000 AD Zhizo/K2 & IP Collapse of Abbasid Empire;Fatimids enter Egypt and foundCairo (Fustat)as center of an empire

1200/1240 K2/Mapungubwe Decline of Chola Empire (S. India)and Srivijaya kingdom (SE Asia);arrival of Muslims in NW India

1300 Mapungubwe/Zimbabwe Delhi Sultanate's conquest ofGujarat

1400 Zimbabwe/Khami Sacking of Delhi by Tamerlane(1398); Vijayanagara Empire inSouth India; Chinese visits to Africa

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e1912 1911

series. The Khami series glass is probably of SouthAsian origin. Thesebeads are slightly larger than those of the Indo-Pacific series andincludenewcolors, includinghues rich in cobalt, an additive thatwasnot employed in earlier soda-alumina glasses.

Changes through time in bead types and glass sources requireexplanation. This is a difficult task forwhich oneway forwardwouldappear to involve correlation of the dates for the changes observedin the glass with major events in the history of the Indian Oceanworld, though correlation is not the same as causation (Table 10). Amorenuanced approach to the problem involves the realization thatchanges in the pattern of the circulation of trade goods in the IndianOcean world that involved the import of glass beads to southernAfrica may have been triggered by events affecting one or more ofthe three sets of players in this trade: themanufacturers of the glassand of the beads (not necessarily the same people), the merchantswho transported the beads and transshipped them at variousentrepôts, and finally the consumers in the African interior.

Bouts of political instability, such as some of the events noted inTable 10, would have disrupted the production of glass and beads, aswell as trade, since, for example, the decline of the Srivijaya king-dom's control of the surrounding seas must have encouraged piracy(see Francis, 2002). Similarly, competition between the city-states ofthe eastern African coast would have affected the volume of tradepassing through different cities, which in turn had to maintainstrong ties to the peoples of the African interior in order to acquiresufficient products, notably gold and ivory, for export. The demandfor imported beads, as well as cloth, in southern Africawas probablysufficiently great that when one source dried up, merchants wouldhave scrambled to find new sources, thereby promoting economic,as well as technological and institutional, changes in societies, suchas those of northwest India, that attempted to satisfy the demand.Similarly, when the Portuguese attempted to take over the trade tothe eastern African coast, they learned that Africans' taste for Indiancloth and beads was so strong that they were compelled to fetchthem from India (Pearson, 1998:134). Finally, it may be noted thatChina seems never to have been an important source of glass beadsfor the African market despite Zheng He's trip or trips to thecontinent in the early 15th century with an armada of about 250ships and 27,000 men (Dreyer, 2007). Chinese beads were allwound, i.e. individually made, and thus probably far more expen-sive to produce than the drawn beads of South Asia; very few havebeen found on African sites (Wood, pers. obs.).

Irrespective of whether the historical events mentioned herewere really those that most influenced the changes in the beadsshipped to southern Africa, it is clear that southern Africa wasinvolved in Indian Ocean history. The archaeology of southernAfrica during the Iron Age can no longer be examined in isolationfrom the rest of the Indian Ocean. Southern Africa was not simplybuffeted by thewinds of Indian Ocean history (see also Pwiti, 2005).

Just as the collapse of the Abbasid Empire and the rise of the CholaEmpire may have brought new beads and traders to southernAfrica, so events in southern Africa, such as the shift in the seat ofpower from Mapungubwe to Great Zimbabwe and with it perhapschanges in the levels of gold extraction, may have reverberatedaround the Indian Ocean; indeed, perhaps even beyond the IndianOcean world, for it has been suggested that the enormous demandfor gold in Europe in the 13th and 14th centuries was met in partwith gold from southern Africa (Sutton, 1997:237e8).

Acknowledgements

The beads used in the analyses reported here were kindlyprovided for study by Jim Denbow; Carolyn Thorp; Lyn Wadley; EdWilmsen; Jan Boeyens; the Van Riet Lowe collection, University oftheWitwatersrand; University of Pretoria and the National CulturalHistory Museum, Pretoria. The researchwas funded by the NationalScience Foundation (BCS-0209681). Laure Dussubieux kindlyallowed us to use some of her unpublished data in our analyses andprovided copies of her unpublished papers. Thanks to JenniferBerdan for help with the figures.

Appendix. Supplementary material

Supplementary material associated with this article can befound in the online version, at doi:10.1016/j.jas.2010.02.016.

References

Allan, J.W., 1973. Ab�u'l Q�asim's treatise on ceramics. Iran 11, 111e120.Bamford, C.R., 1977. Colour Generation and Control in Glass. Elsevier Publishing,

New York.Barkoudah, Y., Henderson, J., 2006. Plant ashes from Syria and the manufacture of

ancient glass: ethnographic and scientific aspects. J. Glass Stud. 48, 297e321.Brill, R.H., 1987. Chemical analyses of some early Indian glasses. In: Bhardwaj, H.C.

(Ed.), Archaeometry of Glass. Proceedings of the Archaeometry Session of theXIV International Congress on Glass 1986 New Delhi India. Indian CeramicSociety, Calcutta, pp. 1e25.

Brill, R.H., 1995. Chemical analyses of some glass fragments from Nishapur in theCorning Museum of Glass. In: Kröger, J. (Ed.), Nishapur: Glass of the EarlyIslamic Period. The Metropolitan Museum of Art, New York, pp. 211e233.

Brill, R.H., 1999. Chemical Analyses of Early Glasses. The Corning Museum of Glass,Corning, NY.

Brill, R.H., 2001a. Some thoughts on the chemistry and technology of Islamic glass.In: Carboni, S., Whitehouse, D. (Eds.), Glass of the Sultans. Yale University Press,New Haven and London, pp. 25e45.

Brill, R.H., 2001b. The glassmakers of Firozabad and the glassmakers of Kapadwanj:two pilot video projects. In: Annales du 15e Congrès de l'Association Inter-nationale pour l'Histoire du Verre. The Corning Museum of Glass, Corning, NY,pp. 267e268.

Brill, R.H., 2005. Chemical analyses of some Sasanian glasses from Iraq. In:Whitehouse, D. (Ed.), Sasanian and Post-Sasanian Glass in the Corning Museumof Glass, Appendix 2. The Corning Museum of Glass, Corning, New York, pp.65e88.

Cox, G.A., Ford, B.A., 1993. The long-term corrosion of glass by groundwater.J. Mater. Sci. 28, 5637e5647.

Davison, C.C., 1972. Glass Beads in African Archaeology: Results of Neutron Acti-vation Analysis, Supplemented by Results of X-ray Fluorescence Analysis. PhD.thesis. Lawrence Berkeley laboratory (LBL-1240), Berkeley, CA.

Davison, C.C., 1973. Chemical resemblance of garden roller and M1 glass beads. Afr.Stud. 32, 247e257.

Davison, C.C., Clark, J.D., 1974. Trade wind beads: an interim report of chemicalstudies. Azania 9, 75e86.

Davison, C.C., Clark, J.D., 1976. Transvaal heirloom beads and Rhodesian archaeo-logical sites. Afr. Stud. 35, 123e137.

Dreyer, E.L., 2007. Zheng He: China and the Oceans in the Early Ming Dynasty,1405e1433. Pearson Education, New York.

Dussubieux, L., 2001. L'apport de l'ablation couplée à la caractérisation des verres:application à l'étude du verre archéologique de l'Océan Indien. Doctoral thesis.Université d'Orleans, France.

Dussubieux, L., Gratuze, B., 2003. Non-destructive characterization of glass beads: anapplication to the study of glass trade between India and Southeast Asia. In:Karlström, A., Källén, A. (Eds.), Fishbones andGlittering Emblems: SoutheastAsianArchaeology 2002. Museum of Far Eastern Antiquities, Stockholm, pp. 135e148.

Dussubieux, L., Gratuze, B., Blet-Lemarquand, M., 2009a. Mineral soda aluminaglass: occurrence and meaning. J. Arch. Sci 37, 1646e1655.

P. Robertshaw et al. / Journal of Archaeological Science 37 (2010) 1898e19121912

Dussubieux L., Kusimba, C.M. Glass vessels in sub-Saharan Africa: compositionalstudy of some samples from Kenya. In: Stevenson, C., Ioannis, L. (Eds.),Proceedings of the International Specialized Workshop e The Dating andProvenance of Obsidian and Ancient Manufactured Glasses, Delphi, Greece,21e24 Feb, 2008, submitted for publication.

Dussubieux, L., Kusimba, C.M., Gogte, V., Kusimba, S.B., Gratuze, B., Oka, R., 2008.The trading of ancient glass beads: new analytical data from South Asian andEast African soda-alumina glass beads. Archaeometry 50, 797e821.

Dussubieux, L., Perret, D., Surachman, E., 2009a. Compositional analysis of ancientglass fragments from North Sumatra, Indonesia. In: Perret, D., Surachman, S.(Eds.), Histoire de Barus III: Regards sur une Place Marchande de l'Océan Indien(XIIe e mi-XVIIes). Association Archipel/EFEO, Paris, pp. 384e417.

Dussubieux, L., Robertshaw, P., Glascock, M.D., 2009b. LA-ICP-MS analysis ofAfrican glass bead composition: laboratory inter-comparison with anemphasis on the impact of corrosion on data interpretation. Int. J. MassSpectrom. 284, 152e161.

Fenn, T., Robertshaw, P., Wood, M., Chesley, J., Ruiz, J., 2009. Isotopic analyses of late1st millennium AD glass beads from Igbo-Ukwu, Nigeria. In: Paper Presented atthe 74th Annual Meeting of the Society for American Archaeology, Atlanta, GA.

Francis, P., 2002. Asia's Maritime Bead Trade: 300 B.C. to the Present. University ofHawaíi Press, Honolulu.

Freestone, I.C., 2002. Composition and affinities of glass from the furnaces on theIsland Site, Tyre. J. Glass Stud. 44, 67e77.

Freestone, I.C., 2006. Glass production in late antiquity and the early Islamic period:a geochemical perspective. Special Publication 257. In: Maggetti, M., Messiga, B.(Eds.), Geomaterials in Cultural Heritage. Geological Society of London, London,pp. 201e216.

Freestone, I.C., Gorin-Rosen, Y., Hughes, M., 2000. Primary glass from Israel and theproduction of glass in late Antiquity and the early Islamic period. In: Nenna,M.-D.(Ed.), La Route de Verre: Ateliers Primaries et Secondaires du Second Millénaireav. J.-C. au Moyen Âge. Maison de l'Orient Méditerranéen, Lyon, pp. 65e85.

Freestone, I.C., Stapleton, C.P., 1998. Composition and technology of Islamic enam-elled glass of the thirteenth and fourteenth Centuries. In: Ward, R. (Ed.), Gildedand Enamelled Glass from the Middle East. British Museum Press, London, pp.122e128.

Glascock, M.D., 1999. An inter-laboratory comparison of elemental compositions fortwo obsidian sources. IAOS Bull. 23, 13e25.

Gogte, V., 2003. The archaeology of maritime trade at Chaul, western coast of India.Man. Environ. 28, 67e74.

Gratuze, B., Blet-Lemarquand, M., Barrandon, J.-N., 2001. Mass spectrometry withlaser sampling: a new tool to characterize archaeological materials. J. Radioanal.Nucl. Chem. 247, 645e656.

Hall, M., 1990. Farmers, Kings and Traders: The People of Southern Africa 200e1860.University of Chicago Press, Chicago.

Henderson, J., 1998. Blue and other coloured translucent glass decorated withenamels: possible evidence for trade in cobalt-blue colourants. In: Ward, R.(Ed.), Gilded and Enamelled Glass from the Middle East. British Museum Press,London, pp. 116e121.

Henderson, J., 2000. The Science and Archaeology of Materials. Routledge, Londonand New York.

Henderson, J., McLoughlin, S.D., McPhail, D.S., 2004. Radical changes in Islamic glasstechnology: evidence for conservatism and experimentation with new glassrecipes from early and middle Islamic Raqqa, Syria. Archaeometry 46, 439e468.

Huffman, T.N., 2000. Mapungubwe and the origins of the Zimbabwe culture.Goodwin Series 8. In: Leslie, M., Maggs, T. (Eds.), African Naissance: The Lim-popo Valley 1000 Years Ago. South African Archaeological Society, Cape Town,pp. 14e29.

Huffman, T.N., 2007. Handbook to the Iron Age: The Archaeology of Pre-ColonialFarming Societies in Southern Africa. University of KwaZulu-Natal Press,Scottsville, South Africa.

Kato, N., Nakai, I., Shindo, Y., 2009. Change in chemical composition of early Islamicglass excavated in Raya, Sinai Peninsula, Egypt: on-site analyses usinga portable X-ray fluorescence spectrometer. J. Archaeol. Sci. 36, 1698e1707.

Killick, D.J., 2009. Agency, dependency, and long-distance trade: east Africa and theIslamic world, ca. 700e1500 CE. In: Falconer, S.E., Redman, C.L. (Eds.), Politiesand Power: Archaeological Perspectives on the Landscapes of Early States.University of Arizona Press, Tucson, pp. 179e207.

Kleinmann, B., 1991. Cobalt-pigments in the early Islamic blue glazes and thereconstruction of the way of their manufacture. Archaeometry 90, 327e336.

Kock, J., Sode, T., 1995. Glass, Glassbeads and Glassmakers in Northern India. THOTPrint, Vanlose, Denmark.

Kusimba, C.M., 1999. Material symbols among the pre-colonial Swahili of the EastAfrican coast. In: Robb, J. (Ed.), Material Symbols: Culture and Economy inPrehistory. Center for Archaeological Investigations, Southern Illinois Univer-sity, Carbondale, pp. 318e341.

Lankton, J.W., Dussubieux, L., 2006. Early glass in the Asian maritime trade:a review and an interpretation of compositional analyses. J. Glass Stud. 48,121e144.

McKinnon, E.E., Brill, R.H., 1987. Chemical analyses of some glasses from Sumatra.Section 2. In: Bhardwaj, H.C. (Ed.), Archaeometry of Glass. Proceedings of theArchaeometry Session of the XIV International Congress on Glass 1986 NewDelhi India. India Ceramic Society, Calcutta, pp. 1e14.

Mirti, P., Pace, M., Negro Ponzi, M.M., Aceto, M., 2008. ICP-MS analysis of glassfragments of Parthian and Sasanian epoch from Seleucia and Veh Arda�s�ır(Central Iraq). Archaeometry 50, 591e605.

Mirti, P., Pace, M., Malandrino, M., Negro Ponzi, M., 2009. Sasanian glass from VehArda�s�ır: new evidences by ICP-MS analysis. J. Archaeol. Sci. 36, 1061e1069.

Nassau, K., 2001. The Physics and Chemistry of Color: The Fifteen Causes of Color,second ed. John Wiley & Sons, New York.

Pearson, M.N., 1998. Port Cities and Intruders: The Swahili Coast, India, and Portugalin the Early Modern Era. Johns Hopkins University Press, Baltimore, MD.

Prinsloo, L.C., Colomban, P., 2008. A Raman spectroscopic study of the Mapungubweoblates: glass trade beads excavated at an Iron Age archaeological site in SouthAfrica. J. Raman Spectrosc. 39, 79e90.

Pwiti, G., 2005. Southern Africa and the east African coast. In: Stahl, A.B. (Ed.),African Archaeology: a Critical Introduction. Blackwell, Oxford, pp. 378e391.

Robertshaw, P., Glascock, M.D., Wood, M., Popelka-Filcoff, R.S., 2003. Chemicalanalysis of ancient African glass beads: a very preliminary report. J. Afr. Arch. 1,139e146.

Robertshaw, P., Magnavita, S., Wood, M., Melchiorre, E., Popelka-Filcoff, R.S.,Glascock, M.D., 2009. Glass beads from Kissi (Burkina Faso): chemical analysisand archaeological interpretation. In: Magnavita, S., Koté, L., Breunig, P, Idé, O.A.(Eds.), Crossroads: Cultural and Technological Developments in 1st MillenniumBC/AD West Africa. Journal of African Archaeology Monograph Series 2. AfricaMagna Verlag, Frankfurt a. M, pp. 105e118.

Robertshaw, P., Rasoarifetra, B., Wood, M., Melchiorre, E., Popelka-Filcoff, R.S.,Glascock, M.D., 2006. Chemical analysis of glass beads from Madagascar. J. Afr.Arch. 4, 91e109.

Saitowitz, S.J., 1996. Glass beads as indicators of contact and trade in SouthernAfrica ca. AD 900 e AD 1250. Unpublished PhD Thesis. University of Cape Town,South Africa.

Saitowitz, S.J., Reid, D.L., van der Merwe, N.J., 1996. Glass bead trade from IslamicEgypt to South Africa, c. AD 900e1250. S. Afr. J. Sci. 92, 101e104.

Salviulo, G., Silvestri, A., Molin, G., Bertoncello, R., 2004. An archaeometric study ofthe bulk and surface weathering characteristics of Early Medieval (5the7thcentury) glass from the Po valley, northern Italy. J. Archaeol. Sci. 31, 295e306.

Sayre, E.V., Smith, R.W., 1961. Compositional categories of ancient glass. Science 133,1824e1826.

Shortland, A., Schachner, L., Freestone, I., Tite, M., 2006. Natron as a flux in the earlyvitreous materials industry: sources, beginnings and reasons for decline. J.Archaeol. Sci. 33, 521e530.

Smith, R.W., 1963. Archaeological evaluation of analyses of ancient glass. In:Matson, F.R., Rindone, G. (Eds.), Advances in Glass Technology Part 2. PlenumPress, New York, pp. 283e290.

Speakman, R.J., Neff, H., 2005. The application of laser ablation ICP-MS to the studyof archaeological materials-an introduction. In: Speakman, R.J., Neff, H. (Eds.),Laser Ablation ICP-MS in Archaeological Research. University of New MexicoPress, Albuquerque, pp. 1e14.

Sutton, J.E.G., 1997. The African lords of the intercontinental gold trade before theblack Death: al-Hasan bin Sulaiman of Kilwa and Mansa Musa of Mali. Antiq. J.77, 221e242.

Taylor, S.R., McLennan, S.M., 1985. The Continental Crust: Its Composition andEvolution. Blackwell Scientific, Oxford.

Tite, M.S., Shortland, A., Maniatis, Y., Kavoussanaki, D., Harris, S.A., 2006. Thecomposition of the soda-rich and mixed alkali plant ashes used in theproduction of glass. J. Archaeol. Sci. 33, 1284e1292.

Turekian, K.K., Wedepohl, K.H., 1961. Distribution of the elements in some majorunits of the Earth's crust. Geol. Soc. Amer. Bull. 72, 175e192.

Whitehouse, D., 2002. The transition from natron to plant ash in the levant. J. GlassStud. 44, 193e196.

Wood, M., 2000. Making Connections: Relationships Between International Tradeand Glass Beads from the Shashe-Limpopo Area. South African ArchaeologicalSociety, Goodwin Series 8, pp. 78e90.

Wood, M., 2005. Glass beads and pre-European trade in the Shashe-Limpoporegion. Unpublished M.A. Thesis. University of the Witwatersrand, South Africa.

Wood, M., Dussubieux, L., Wadley, L., 2009. A cache of w5000 glass beads from theSibudu Cave Iron Age occupation. S. Afr. Humanities 21, 239e261.

Zucchiatti, A., Bouquillon, A., Katona, I., D'Alessandro, A., 2006. The ‘Della RobbiaBlue’: a case study for the use of cobalt pigments in ceramics during the ItalianRenaissance. Archaeometry 48, 131e152.