The Tanzanian Coast in the First Millennium AD Studies in African Archaeology 7
The Tanzanian Coast in the First Millennium AD An archaeology of the iron-working, farming communities by FELIXCHAMI
With microscopic analyses by Anders Lindahl
Societas Archaeologica Upsaliensis. Uppsala 1994 Societas Archaeologica Upsaliensis. Uppsala 1994 Doctoral thesis at Uppsala University 1994
Printed with grants from Uppsala University and the Swedish Agency for Research Cooperation with Developing Countries (SAREC)
ABSTRACT Chami, F., 1994. The Tanzanian Coast in the First Millennium AD: an archaeology of the iron-working, farming communities (with microscopic analyses by Anders Lindahl). Studies in African Archaeology 7, Uppsala pp. 107, 27 figs, 19 maps, 3 plates. ISSN 0284-5040 ISBN 91-506-1037-6. This thesis deals with the cultural processes which took place on the coast of Bast Africa in the first millennium AD. The work focuses on the cultural origin of the Triangular Incised Ware (TIW) tradition, its temporal change and spread over the whole coast ofEastAfrica. The study, based on excavations at six sites, indicates that the central coast was the area of genesis in the 4th-5th centuries AD of this late, widespread tradition. TIW derived from the Early Iron-Working (EIW) communities that originated in the interlacustrine region of Bast Africa around the 5th century BC. The florescence of the tradition coincided with evidence of an increased population, im proved metal technology and the introduction of foreign trade. Consideration has also been given-to the spread of this tradition to the islands, the littoral and the deep hinterland. Similar cultural backgrounds and inter-regional trade could have facilitated the fast spread that took place in the 7th-9th centuries AD. The continuity of this tradition Was interrupted by the Is lamic-Swahili tradition at the beginning of the lOth century AD.
Key words: coast of Bast Africa, first millennium AD, early iron-working, fanning,Triangu lar Incised Ware, Sassanic-Islamic, trade. Felix Chami Faculty ofArts and Social Science, University ofDar-es-Salaam, P.0. Box 35050, Tanzania, and Department ofArchaeology, Uppsala University, Gustavianum, S-753 10 Uppsala:, Swe den.
Cov·er Design: Alicja Grenb6rger © Felix Chami English revised by Neil Tornkinson ISSN 0284-5040 ISBN 91-506-1037-6 Published by Societas Archaeologica Upsaliensis, Uppsala 1994 Editorial Committee: Bo Gräslund, Paul J. J. Sinclair, Christina Bendegard Distributed by the Department of Archaeology, Uppsala University, Gustavianum, S-753 10 Uppsala, Sweden Printed in Sweden by Almqvist & Wiksell, Uppsala 1994 CONTENTS
List of illustrations ...... 7 Acknowledgements...... 9 Preface ...... 11 1. Introduction ...... 13 I. I Conceptual definition ...... 13 1.2 Focus of the research ...... 14 1.3 Research background ...... 14 2. Problems and Statements ...... 17 3. Theory and Research Strategies ...... 19 3.1 Theory ...... 19 3.1.1 Traditional archaeology ...... 19 3.1.2 Modem "post-traditional" archaeology ...... 19 3.2 Research strategies ...... 20 3.2.1 Empirical techniques ...... 20 3.2.2 Hypothetico-deductive strategy ...... 22 3.2.3 The strategy of coherence ...... 24 4. lntellectual Background ...... 25 4.1 Anden! Graeco-Roman documents and reports by Arab, Chinese and Portuguese travellers ...... 25 4.1.1 Graeco-Roman documents ...... ,...... 25 4.1.2 Arabic and Chinese documen(s : ...... 26 4.1.3 Portuguese documents ...... 27 4.2 Chronicles ..... ·...... 27 4.3 Scholastic period ...... : . . 27 4.3.1 Early scholarship ...... 27 4.3.2 Later scholarship (1965-80) ...... 28 4.3.3 Recent scholarship ...... 30 5. Enviromnental Setting ...... 34 5.1 Introduction ...... '. . . 34 5.2 The extent of the East African coastal belt ...... 34 5.3 Physical environment...... 34 5.4 Clirnate ...... 35 5.5 Sea-level fiuctuation and marine terraces ...... 37 5.6 Soils ...... 39 5.7 Vegetation ...... 40 6. Techno-economic Subsystem ...... 43 6.1 _Production ...... 43 6.1.1 Metal-working ...... 43 6.1.2 Other crafts ...... 44 6.1.3 Subsistence ...... 45 6.2 Trade ...... 46 7. Field Work ...... 48 7.1 Mpiji...... 48 7 .2 Changwehela ...... 54 7.3 Kaole ...... 55 7.4 Masunguru ...... 58 6
7.5 Kiwangwa ...... 58 7.6 Misasa ...... 61 7.6.l Mapping and surveying ...... 63 7.6.2 Excavation and linds , ...... 63 8. Pottery Analysis: Decoration and Shape ...... 69 8.1 Introduction ...... 69 8.2 Analytical attributes ...... 69 8.2.2 Decoration ...... 70 8.2.3 Shape ...... 74 8.2.4 Placement of design elements ...... 80 8.3 Conclusion ...... 82 9. Pottery Analysis: Fabric ...... 83 9.1 Initial field and laboratory observations ...... 83 9.1.1 Paste ...... 83 9.1.2 The quality of the sherds ...... 83 9.2 Microscopic analyses (by Anders Lindahl) ...... 83 9.2.1 Introduction and formulated problems ...... 83 9.2.2 Material ...... 84 9.2.3 Method ...... 84 9.2.4 Results ...... 85 9.2.5 Conclusions ...... 88 10. Chronology ...... 90 10. l Stratigraphy ...... 90 10.2 Relative dating ...... : ...... 90 10.3 Radiocarbon dating ...... 91 10.4 Discnssion ...... 92 11. The Beginning, Spread and End of the TIW Tradition ...... , ...... 94 11.1 The beginning of the TIW tradition ...... 94 11.2 The spread of the TIW tradition ...... 98 11.3 The end of the TIW tradition ...... 99 11.4 Concluding remark:s ...... ·...... I 0 I References ...... 102 )
ACKNOWLEDGEMENTS
This work is a result of two phases of research in 1990 Individuals in the Uppsala Archaeology Department and 1991, a period of laboratory work in 1991-93 and have also contributed much to this work. Associate major library and supervisory chores in 1989-93. Such Prof. Paul Sinclair recruited me for the programme that a process normally involves many people and institu ultimately led into this thesis. He has supervised me in tions and it is difficult to mention them all. the field and in the writing of the thesis. He allowed me In the first spell of field work in 1990, I was assisted the free use of his books, and we spent much time by Messers. N. Muhanga (Antiquities Department), Ali together arguing about various positions presented in Kilindo (Archaeological Unit), Hamza Mahine this work. Prof. Bo Gräslund, as the head of Archaeol (Archaeological Unit), and Said Shomvi (University of ogy Department, contributed both psychologically and Dar-es-Salaam). In the second spell in 1991, I was academically. Ris caring attitude to his students gave assisted by Mr. J. Kimaro (Antiquities Department), me the encouragement to work hard. He participated in Miss Maro (Antiquities Department), Mr. Ndilana the field work while at Mpiji and supervised the writing (University of Dar-es-Salaarn), Mrs Tilya (Antiqnities of this book. Department), and Messers. Yunus (Tanzauia Harbours), All other members of staff and students in the A. Kweka (Museums), Ap. Kweka (Archaeological Archaeology Department worked very cooperatively Unit), C. Mwakyembe (Antiquities Department) and with me, creating a conducive atmosphere for the pro Seif (Antiquities Department). duction of this work. I thank Kjel and Helena Knutsson In all the villages where we camped, residents gave and other members of Gamla Posten/Composten for us all the support we required. The villages included their cooperative spirit. Thanks are also due to the East Mapinga, Masuguru, Kiwangwa and Misasa. While em African project team, Anna-Lena Forsberg, Elisa working at Kaole, the Bagamoyo Cultural College beth Green and Christina Bendegard, for all the hosted us for about two weeks. I thank the authorities assistance they offered. Frands Herschend helped me to there for allowing us the use of their facilities. calibrate the carbon-14 dates and Alicja Grenberger Paul Msemwa offered moral support in and outside drew rnost of the illustrations. the field work. He participated in the excavation of The University of Dar-es-Salaam as my employer Mpiji and helped in analysing the marine products, i.e. gave me a study leave that has led to this work. The shells. Dr. Tambila and Dr. Nyagava, then the respec Antiquities Department in the Ministry of Education tive heads of the History Department and Archaeology and Culture allowed me to conduct archaeological Unit of the University of Dar-es-Salaam, encouraged research and to export various artefacts for analysis me to join this study programme and helped me to get abroad. Uppsala University offered various facilities various clearances and university facilities, both in the and expertise. My thanks are due to the Swedish Cen field and in the laboratory. tral Board of National Antiquities and SAREC for I thank Prof. Bassey Andah, of the University of funding the research and providing the scholarship. I lbadan, for correcting and commenting on the first also thank Mr. Neil Tornkinson for language correc draft of this work. I also benefited from discussions at tions. different times with Prof. Peter Schmidt (my MA My farnily experienced difficulties throughout the supervisor), Prof. Graham Connah, Prof. John Sutton, period I was away in the field or in the schools. I am Mr Robert Soper, Prof. Pierre V erin, Prof. Henry more grateful than I can express here to my beloved Mutoro and Prof. Henry Wright. ones for their forbearance and support. PREFACE
This book is an archaeological work reviewing the elements, early alkaline products, for example, glazed extant knowledge of the ancient, East African, coastal pottery, glass and beads. communities and providing new data that contribute to The second is the Later TIW phase, also identified as such knowledge. Zanjian or proto-Islamic. This probably started at the In time, the work focuses on the first millennium AD. beginning of the 8th century and ended in the lOth-11 th This is important, because, apart from controversial, centuries AD. The diagnostic materials of the period, early documents and the early occupation of the urban apart from typical TIW pottery, are Islamic products, settlements dated to the 9th-10th centuries AD by the i.e. white, tin-glazed wares and early sgraffiato. use of relative dating, the pre-9th-century coast The TIW tradition seems to have had its origin in the remained terra incognita. As a corollary to this, the. earlier EIW tradition, contrary to the existing Arabic/ relevant scholarship was afflicted by speculations con Persian or Cushitic theory. Archaeologically observed ceming the nature and the origin of the people inhabit changes in their cultural norms can be perceived as a ing the first-millennium, coastal settlements, their way normal process of interna! adjustment for various rea of life and their cultural configuration. sons, for example, growing population, new contacts In space, the work generally focuses on the whole and trade opportunities. These would lead to re-organi coast of East Africa. The case studies are on the sites zation and the adoption of new ideologies. that fiourished on the littoral and in the hinterland of The TIW tradition seems to have been widely dis the central coast of Tanzania. In most of the literature, tributed all over the littoral and deep in the hinterland. both the, central and the southem coasts of East Africa This can be explained by the fäet that the people were have been portrayed as recipients of immigrants and farmers and metal-smelters with close, cultural links ideas from the Middle East and the northem Kenya between the littoral and the hinterland from the EIW coast. The immigrants attracted by trade and Islamic in period. The development of new ideologies, technolo religion were thought to have founded the early urban gies and· trade contacts easily affected such a !arge settlements later to be identified with the Swahili. region, owing to the common cultural and linguistic Empirical, positivistic and coherence strategies have heritage. been used to test various hypotheses pertaining to the It is also suggested that the change of trade agents origin, spread and cultural affiliations of the ancient from Persians to Islarnic Arabs might have destroyed coastal conununities. These encompass the understand many East African communities, especially those in the ing of the environment, the economic potentialities and hinterland. Many TIW settlements ended abrnptly, both the analysis of various archaeological finds, for exam on the littoral and in the hinterland. This was probably ple, Triangular Incised Ware (TIW) and carbon-14 due to a dramatic shift in trade alliances, new trade samples from sites excavated on the littoral and in the directions and probably a demand for new trade goods, hinterland. i.e. slaves. A new, littoral class of Islamic converts and Two phases have been identified for the first millen merchants would have engineered war, which is a nium coastal settlements. The first is the Early TIW necessary means to obtain slaves. Consequently, the phase, also identified as pre-8th-century or Azanian or Later TIW settlements fionrished mainly on the littoral pre-Islarnic. The beginning of this phase was probably and the islands, becoming prosperous, as may be seen in the .4th century hut it ll)ight have gradually develo from the construction of monumental architecture, and ped from the EIW communities of the early centuries apparently Islarnic. This was the beginning of the litto M. The diagnostic materials of the period include TIW ral-hinterland dichotomy on the coast of East Africa. pottery with many Early-Iron-Working (EIW) pottery :I: _) :J-'*- -~ "':f ';il !;'t c;j$ 1. INTRODUCTION i-;;, ~
~-;& -~ ~ -~i Tuere has been a great need amongst the schalars of the occupied by farmers in the period from the 4th century East African coast for information about the culture of BC in the inter!acustrine region to the 4th century AD in '~ ;§ the coast in the first millennium AD and even earlier. To Southern Africa (Phillipson 1985; Soper 1971a; Sin ;"i? cater for such a need, abundant literature has been pro clair 1987). The concept of EIW replaces that of the "1 duced since the beginning of this century. Early Iron Age (EIA), which has been thought to have The present contribution is based on recent research been imposed on the African situation, and that of ;i in a previously unexplored area. It is an expose of data Early Farming, which is now seen to comprise stone :4% from seven sites in the Kisarawe and Bagamoyo Dis using, food producers (Phillipson 1985, pp. 5, 113-47; 'Si tricts of Tanzania, linked to data from other first-mil Robertshaw 1990, p. 4; Shaw, Sinclair, Andah & leunium sites reported from other parts of the coast Okpoko 1993, p. 3). Later Iron-Working (LIW). Following the use of I;-;;t Sites with similar cultural materials have been found -~, on the coast of East Africa in the area from southern EIW, the period formerly defined as the Later !ron Age Si Somalia down to southern Mozambique. (LIA) is hereby changed to LIW. ':Cf ;J In this chapter, I intend to define a few concepts nsed Triangular Incised Ware (TIW). This stands for the -I in this thesis. I shall then discuss the focus and the pottery tradition that succeeds the EIW tradition along
-.(\~ coast extends to Morogoro in central Tanzania. In some tion" (Hall 1983, p. 52). '.g i!W of the literature (Horton 1984, p. 299; Datoo 1975, p. 3; Phase. This is a specific time segment, in which a ;1 Sutton 1990, p. 57; Nicholls 1971; Chittick 1975b), the minor change occurs in the assemblage or other key l~ coastal concept means the flat, 10-km strip along the material in the tradition (Hall 1983, p. 52) $ shore. This narrow strip is referred to in this work as Culture. This means groups or communities of peo ii~* the "littoral", while the land 100 m above sea leve! ple represented by an archaeological tradition. It :·i 0 beyond is the "hinterland". On the coast of Tanzania, encompasses the lifestyles and idiosyncrasies por ··11 the hinterland extehds beyond 200 km, reaching the trayed by such a tradition. According to Huffman -I* -i fringes of N guru and the U sambara Mountains (Map (1980, p. 124), 'f 'i 4). ''!• Style as apart of culture is learned and possessed within groups of -:f Early Iron-Working (EIW). This concept stands for z people, and the correlation between design style and specific groups '4 sites with the earliest evidence of iron-smelting in eas is well known (e.g. Boas 1927). Those portions of material culture which vary stylistically can be used to identify groups of people -;;;~ tern and southern Africa. On such sites, the presence of ;1 slag and tuyeres indicates smelting. These linds are even though the size of the group and their other characteristics are unknown. 'f associated with potsherds with bevelled and fluted rims i _f 1 and shoulders. It has been shown that such sites were The TIW tradition, therefore, stands for a first-millen- ~
i i j* 14
from the colonization framework to Africanization. Nlle R This is a shift of paradigm from which the coast of East Africa is seen as mainly African. There is much more combination of data from atchaeology, linguistics and ethnographic/oral tradition.
1.2 Focus of the research z <( The work focuses on the first-millennium sites found w along the littoral and the deep hinterland of the Tanza 0 0 nian coast now dated to the 4th-10th centuries AD. Pot tery with incisions or punctates bounded by triangles z <( characterises the sites of this period. The pottery has,
Cl therefore, been labelled ''Triangular Incised Wate" z (TIW), as defined in section LL Ruvuma Cape Delgado The present work re-assesses the extant theories R. •P1c;.omornlslaods explaining the nature of the first-millennium, coastal culture in relation to the coastal environment, techno 0 ~oza Angoche logy and economic activities. Pottery, which has been Zambez1 R. Quelimane used to identify and explain the origin and distribution _, MADAGASCA of different cultural groups in eastern and southern Sofala Africa (Phillipson 1976, p. 3; Soper 1971b, c; Huffrnan 1970), is subjected to further analysis. The extension of Cape Correntes 0 1000km this study to associated materials, i.e. metal objects, beads, imported cerarnics and glass, is aimed at widen ing our knowledge of the first-millennium-AD, local Map 1. The eastem coast of Africa (aft.er Wilson 1981). and inter-oceanic, exchange networks. To obtain a chronological perspective of the eatly coastal developments, atchaeological data pre-dating nium-AD, coastal culture present on the coast of eastern the TIW tradition will be discussed. This study will be Africa between the 4th and the lOth centuries AD. Ils supported by twenty C14 results obtained from seven cultural material, as indicated by the archaeological sites in the study ateas of Kisatawe and Bagamoyo. data, is discussed in Chapters 6 and 7. From the case-study atea, the work is aimed at spa Swahili. This denotes the coastal culture that devel tially extending our horizon. to cover the whole coast oped in the littoral settlements, probably from the end (Map 1) in the first millennium AD. The East African of the first millennium AD. The Swahili people are islands are involved in this work, because many of known to have been farmers, fishermen and traders them bave TIW sites with eatly, imported materials, i.e. who dwelled in the stone-built towns, some of whom Sassanid pottery. On account of the implication of its were Muslims (Allen 1980, 1982, 1993). trade with East Africa, the area atound the Gulf and the Early scholarship. This refers to the period c. 1900-- Arabian peninsula will also feature in the discussion. 1965, when the coastal, historical and cultural studies In general, the focus is on the first-millennium coast were made by using ancient and medieval travellers' of East Africa, the origin and distribution of its culture reports and the chronicles. The coast ofEastAfrica was and its relationship with the Indian Ocean trade. regarded as a colony of Arabs or Persians. Later scholarship. This refers arbitratily to the period 1965-1980, in which atchaeological techniques 1.3 Research background were introduced to support and clatify the reports and chronicles. Chronology was analysed by studying the The work has been generated within the context of the pottery imported from the Middle East and China. The "Urban Origins in Eastern Africa" project. One purpose coast was still seen as a colony of foreigners. of this project is to establish the chronology of the eatly Recent scholarship. The period from 1980 to the coastal settlements, while another deals with their present, in which there has been a shift of emphasis socib-economic organization and ethnic constitution of 15
"Periplus" document and Ptolemy's Geography (Cas son 1989). In dealing with EIW problems, I recognized that the period between the end of EIW (5th century AD) and the beginning of the LIW coastal urban settlements KENYA (9th-10th century AD) was unknown. According to Phillipson (1977, p. 155),
Despite intensive search for earlier sites, it has not yet prov ed possi ble to trace back the archaeological record of these east-coast trad ing societies beyond the ninth century AD. This date broadly coincides with the resumption of written references to the area after several centuries of silence following the Periplus.
TANZANIA INDIAN OCEAN Sutton (1990, p. 91) identified this gap as "discontinu ity tlnough the middle part of the first millennium AD". Recent studies in southem Africa have shown, how ever, that similar assumed periods of discontinuity 30~COMOROS 29' 'i27 could not withstand the evidence of new archaeological 28 (;26 linds (Huffman 1982; Sinclair 1987; Adamowicz 1991). According to Huffman (1982, p. 139), . . . there is at least one complete continuum from an EIA unit MOZAMBlQUE within Southem Africa to present-day Bantu speakers. I cannot point to a similar continuity in Bast Africa because local sequences have not been fully established. MADAGASCAR The opportunity to start examining this problem of discontinuity in East Africa came when the planned University of Dar-es-Salaam Field School series of sur veys along the coast continued in the area of our case study. Within the same area of Kisarawe, the research work in 1987 resulted in the discovery of two EIW and 35 two TIW sites (Fawcett & LaViolette 1990). The field
1-Tana, 2-Shanga, 3-Manda, 4-Ungwana, 5-Kwale, 6,7-East Usambara, 8-Amboni, work that followed the next year uncovered another 9-Mkokotoni, 10-Unguja Ukuu, 11-Mkadini, 12-Kilosa, 13-Dakawa, 14-Lugoba, TIW site in the centre of Dar-es-Salaam city (Karoma 15-Masuguru, 16-Kiwangwa, 17-Kaole, 18-Mpiji, 19-Dar-es-Salaam, 20-Limbo, 21-Misasa, 22,23-Kilwa, 24-Monapo trad. 25-Nampulatrad. 26-Dembeni, 27-Slma, 1989). And the 1989 Field School uncovered four more 28-Mro Dewa, 29-Mbeni, 30-Mbashile, 31-lrodo, 32-Hola Hala, 33-Bazaruto Island, 34-Chibuene, 35-!nhambane TIW sites in the Bagamoyo District, two along the shore and two about 55 km into the interior (Chami Map 2. East Africa: a section of first millenniuin AD EIW sites (rnarked with black triangles) and TIW sites (marked with black dots). 1990). Recent Field School surveys further inland from Bagamoyo to Morogoro towards central Tanzania found another three TIW sites in Lugoba, Kilosa and Dakawa (Halland, Kimathi and Reynold, pers. comm.), early urban settlements (Sinclair 1989). The early and my own observation of the materials). urban settlements known in East Africa
Mkadini (Chittick 1975a, pp. 151-3), Amboni and Five sites identified by the 1989 Dar-es-Salaam Field Usambara (Soper 1967a; Schmidt 1988), Kwale (Soper School researchers in Bagamoyo District in July-Sep< 1967b), Tana-Wenje (Phil!ipson 1979), Manda (Chit tember 1990 were proposed for a feasibility study, the tick 1984), Shanga (Horton 1984), Ungwana (Abungu results of which have been reported elsewhere (Chamj1 1989), Mijikenda and Takwa (Mutoro 1987) (Map 2). 1992a). Two sites on the littoral and two in the hinterc .. The Urban Origin project provided a good opportu land yielded positive results that, when combined wit!J,' nity to try to find out whether TIW sites could enable those from Misasa (Zakwati) in Kisarawe (Fawcett & us show the connection between the EIW and LIW LaViolette 1990), led to the fully fiedged research of' urban sites. July-October 1991 that resulted in the present work. 2. PROBLEMS AND STATEMENTS
The first problem concerns the fact that the TIW tradi Roman trade to East Africa in the first century AD. tion and its attendant culture ate said to have originated These pastoralists, according to him, founded the TIW on the northern Kenya coast. At present, two general settlements under study in this work. He has also stated models are proposed for this problem. On the one hand, that the EIW sites were founded on the coast after 500 the TIW sites, also known as the early Swahili (Horton AD (Horton & Mudida 1993, p. 672), a position com !987a, 1987b; Nurse 1983), are said to have been cre pletely untenable in the light of evidence from Sonth ated by the proto-Sabaki, Bantu speakers who moved Africa (Klapwijk 1974), Mozambique (Cruz e Silva to the Tana-Lamu region at the beginning of the second 1977; Sinclair 1987, 1991; Morais 1988), Tanzania half of the first millennium AD. After settling in the (Chami 1988b, 1992b) and Kenya (Soper 1967b). area, they learned from the passing traders and through M y aim in this work is, therefore, to re-evaluate the intermarriage with Arabs and Persians how to use and origin of the TIW tradition with the old and new data make boats and, as they started trading, they moved collected from the central coast of Tanzania. This geo southward to found Swahili settlements today identi graphical area is quite far away from Somalia and the fied by TIW pottery. Linguistic and oral traditions have Rift Valley areas associated with ancient Hamitic/Cus been used to support the hypothesis (Nurse 1983, pp. hitic occnpation (Phillipson 1985, p. 143: Robertshaw 136-7; Nurse & Spear 1985; Pouwels 1987). 1990). On the other hand, the central coast of Tanzania On the other hand, the northern coast of Kenya and has been shown to have been occupied by the EIW its extension northward to southern Somalia is thonght Kwale tradition (Soper 1971b; Chami 1988a, 1988b, to have been the land of Shungwaya. Cushitic pastora 1992b; Fawcett & LaViolette 1990). lists controlled and infiuenced the affairs of this region, The second problem is the widespread occurrence of including trade (Allen 1983, 1993 pp. 39-52). The TIW sites observed at a more or less similar period of spread of this culture by the Sabaki-speaking Swahili time, which has led some archaeologists to point to throughout the coast in the 9th-10th centuries led to the trade as the mode of dispersa!. TIW pottery is alleged beginning of the Swahili settlements (TIW sites) and to have been produced in the hinterland of the northem the Islamization of the whole coast (Horton 1984, coast and then traded all over the coast down to the l987b). A pastoral origin has been supported by refe Comoros and Mozambique (Horton 1984, p. 299; Sut rence to the seafarers' mention of the domestication of ton 1990, pp. 59-60). animals on the "Zanj" coast (see Masudi, Chinese This widespread trade is said to have been managed Bobali and Ibn Batuta-Freemans-Grenville 1975; Hor by the Islamized, Sabaki-speaking people, with whom ton 1984; Trimingham 1975), cattle and camel bones the Swahili people are linguistically affiliated (Hiunes from Shanga (Horton 1984, 1987b), and TIW pottery butch 1978; Nurse & Spear 1985); hence the supposed in connection with the "pastoral neolithic", ceramic tra linguistic verification of the northern origin of the TIW dition (Horton 1987b; Abungu 1989, pp. 147-9). tradition. The TIW sites, therefore, were thought to be The northern coast has also been thought to have had found only in the areas occupied by the Sabaki-speak a trade advantage in relation to the Red Sea and the ing people, including the littoral areas and the immedi northern Indian Ocean, because the monsoon winds ate hinterland between the Pangani River to the south and the oceanic currents are reliable at this latitude and and the Tana River to the north (Horton 1987, p. 315). permit rapid, seasonal navigation, both along the coast At the 1991 Zanzibar Urban Origin workshop, I sug of eastern Africa and across the Indian Ocean to wes gested a cultural unity between the littoral and the hin tern and southernAsia (Horton 1984, p. 266, 1987b). terland of Tanzania In this chapter, I discuss issues related to archaeologi diffusion of ideas from Bgypt and the Middle Bast to cal theory aud how it has affected research in eastem the rest of the continents dominated the archaeology of Africa. Orre cau then nnderstand why the archaeolo the time. InAfrica, the idea that every element of civili gists concemed with eastem African, coastal settle zation carne from the Middle Bast or the Far East has ments have reprodnced sinlilar results for decades come to be known as "Harnitic/Cushitic myth" (Sand (Chapter 4). The theoretical discussion will also ers 1969, p. 521; Amadi 1989, p. 80; Zwememaun inclnde au explauation of the particnlar strategies used 1983, pp. 15, 67). The Harnites, according to Zweme in this archaeological inquiry. mann (1983, p. 15), were divided into the Eastern Hamites, with whom Seligman included the ancient and recent Egyptians, the Nubians, the Galla, Somali, Danakil and most modern Ethiopians. The northem 3.1 Theory Hamites include the Berber, Tuareg, Tibu, Fulani, and the Guanche of the Canary Islands. 3.1.1 Traditional archaeology The Harnites, who were regarded as Caucasians in the Before the 1960s, archaeology was based on the classi late 19th century, have been associated with pastora fication aud description of artefacts, the results of lism aud their "negro" counterparts with agriculture. which were used to narrate evolutionary changes, Hence, "pastoralism and all its attributes becarne migrations of peoples and diffusion of ideas (Willey & endowed with an aura of superiority of culture" (Saud Sabloff 1980; Daniel 1950). This tradition begau ers 1969, pp. 529-30). Therefore all elements associ around the mid-19th century, during the period when ated with civilization in sub-Saharau Africa, Darwinian evolntionary ideas started to predominate in comprising metal-smelting, complex political institu academic circles. This coincided with the refining of tions, irrigation, the age-grade system and all "archaeo the Three Age System (Stone, Bronze and Iron Ages) logical findings of any magnitude", were attributed to by Thomsen and Worsaae in the mid-19th century. The the Harnites (Sanders 1969, p. 531). Three Age System was further classified into subsys Much study founded on the migratory aud diffusio tems, as in de Mortillet's division of the Stone Age into nist theories was infiuenced by the "Germau cultural the Chelleau, the Mousteriau, the Solutrean and the historical school of anthropology'' (Holl 1990, p. 300). Magdalenian stages (Daniel 1950; Gräslund 1987). The For details of this school, see Harris (1968, pp. 373- two perspectives were employed to show that humau 92), Zwememann (1983) and Shaw, Sinclair, Andah & cultures had evolved from earlier, primitive stages. Okpoko (1993, pp. 9-13). However, the archaeologists were not wholly evolu tiouary. According to Dauiel (1950, p. 45), in the works 3.1.2 Modern "post-traditional" archaeology of Thomsen and Worsaae, one found "the ideas of inva sion, diffusion and homotaxy which formed the frarne- . While African archaeology continues to suffer under work of the twentieth century". This combination of the plague of the "Harnitic/Cushitic myth" to the perspectives is probably better illustrated in Childe's present day (Holl 1990; Shaw 1990; Shaw, Sinclair, (1951) work, where, in spite of the explanation of the Andah & Okpoko 1993; see also Chapter 4 in this diffusion of ideas fröm the Nile Valley and Mesopota work), Buropean and American archaeology entered a mia, the Three Age System is correlated with au new era after 1960. anthropological, evolutionary scheme of savagery, bar Archaeologists adopted new paradigms aud episte barism and civilization. mological approaches (Clarke 1972; Binford 1968; On the thesis that "savages are ... totally without ini Willey & Sabloff 1980). Two competing, but also sym tiative, without the desire or the capacity for inveuting biotic strategies in the modem epistemology, compris a device, a myth, or an institutiou" and hence "all ing inductive and deductive inferences, were adopted major inventions were made but once by some chosen (Hempel 1965; Salmon 1982; Gibbon 1984). They people" (Childe 1951, pp. 12-3), migratious and the emanated from empirical and positivist (natura! sci- 20 ence) approaches to scientific inquiry. These strategies as a mere development of archaeological methodology,· became prominent in archaeology after the 1960s in the There is just an awareness among the archaeologists of rise of the "New Archaeology" (Binford 1968; Schiffer the various !imitations that encumber the various stra 1976; Trigger 1978; Willey & Sabloff 1980). The tegies employed for understanding past societies in th.; intention of their proponents was to break away from absence of concrete documentary or ethnographic evi~ . the non-explaining apathy of traditional archaeology. dence. However, instead of decrying one strategy in The new archaeologists (processual/behavioural) favour of another, I would advocate the use of different aimed at explaining past temporal and spatial proces strategies, depending on how well a particular strategy. ses. This was an attempt to understaud past cultural fits one's particular problem. Archaeologists using dif> behaviour in relation to environment, population, tech ferent strategies should be aware of the subjective nology, social organization and ideology. The quest for milieu in which they were brought up and try to contro!, laws and "middle range" theories was thought to be it as much as possible, so that it does not infiuence theif necessary, so as to help to develop snch a science; research results. Post-processual archaeologists should1 hence the adoption ofKnhn's (1970) paradigmatic lines be even more aware of this problem, since they havei of thought (Binford 1972; Clarke 1978). According to opted for an unsystematic method (poetic), which; Knhn (1970, p. 8), competing schools in academic makes them unaccountable for whatever they propaJ• studies are normally guided by a paradigm which he gate. As Renfrew (1989) has already pointed out, wei\ took to be "universally recognized scientific achieve may find ourselves writing fictitious stories, which is, ments that for a time provided mode! problems and worse than the academic work they have been trying to{. solutions toa community of practitioners". criticise. . To angment that endeavour to build up middle-range As far as pertains to this work, therefore, I have reali-\ theories, ethno-archaeology and experimental archae zed the possibility of utilizing both empirical and posi ology were initially utilized to provide models and tivist strategies. The choice of the approaches from the i hypotheses that could be tested either empirically or New Archaeology is based on the fäet that one can testi deductively (Binford 1972; Gould 1978; Schiffer the correspondence between the theory or hypotheses' 1976). The correspondence between the hypotheses or and the available data (Renfrew 1989). Various models' models developed with the archaeological data was developed in anthropology and archaeology will be· thought to help archaeologists to understand better the employed to help to explain the archaeological occur behavioural patterns and processes of past societies. rences on the coast of East Africa in the first millen-: In the subsequent developments, some scholars have niumAD. felt varied degrees of concern about the suitability of the methods of natura! science for archaeology. Some, though in favour of a scientific method, have refrained 3.2 Research strategies from extreme positivism (Trigger 1989, 1978; Renfrew 1989). They have thought that archaeology, as a study Three main strategies are concerned below. First, the of human behaviour, may require some general theory empirical techniques are those emanating from the . and consistent systematic methodology, but not neces inductive part of scientific research. This is a process of. sarily laws like those of the physical sciences. Others, building up information and ultimately knowledge while not decrying the scientific approach of the New from the observable data (Gibbon 1984, p. 74; 1989, Archaeologists, thought that the approach would have pp. 8-13). Secondly, an hypothetico-deductive app been better if structural and symbolic studies had been roach is used which is part of a deductive-nomological: diversified into a purely functioualistic approach in the mode! of science and is the procedure in which the New Archaeology (Rodder 1982). Ou the other hand, a deduced, "testable consequences from hypotheses" are group identifying the~selves as post-processual confronted with data (Gibbon 1984, p. 12; Salmon archaeologists have claimed that objectivity is unat 1982, p. 34). The third is the the strategy of coherence · tainable by the use of scientific method in archaeology. defined below. To them, scientific knowledge is subject to the present day ideologies and power politics. They have called, 3.2.1 Empirical techniques therefore, for the structural and symbolic interpretation of the archaeological remains, whereby the meaning of Surveying in archaeology is the basic technique of · the data yielded by archaeological sites could be inter locating sites and of deterrnining how and where the preted (Shanks & Tilley 1989). data should be collected (Flannery 1976a; Bower 1986; · In this work, I have regarded the above differences Plog 1976). For site location, two main methods-grid · 21 and transect sampling-have been proposed (Plog face-material distribution does not in all cases reftect 1976). The former is effective in a more open, arid the nature of the sub-surface disposition: ]andscape. For the wetter and wooded, forest land On orre hand, we have the optirnistic statement of Red.man and scape, transect survey is more appropriate. Watson (1970:280) that "surface and sub-surface artifact distribu tions are related so that a description of the first allows prediction of The problem of access can sometimes be sidestepped by surveying the second". On the other hand, we have the statement by my friend roads, especially if they are unpaved and deeply cut into the soil. the Real Mesoamerican Archaeologist (personal communicat:ion, Roads often sample such areas in a fairly systematic manner, and 1970) that "surface remains are just that~the junk you find on the sites are frequently well exposed in them (Bower 1986, p. 27). surface-and nothing more" (Flannery 1986, p. 51). While the transect-survey method has been employed Several techniques are available for site survey. They in some parts of Africa (Bower 1986; Sinclair 1987; include augering, drilling and magnetometry. In auger Hal! 1981; Odner 197la and b; Soper 1967a), the Tan ing, an anger is driven into the soil down to the archae zania littoral was first studied by this method in 1987. ologically sterile Jevel, bringing up samples of the The Uuiversity of Dar-es-Salaam Field School subsoil. This can be repeated systematically, following researchers 24 conducted there, and the results are discussed in Chap the first millennium BC. Another typical inconsistency ter 11 in relation to the problems in Chapter 2. was the idea that coastal pottery was produced in same locations to the north and traded southwards to the rest of the coast. This idea was propagated in spite of the i 3.2.3 The strategy of coherence fäet that it was known that EIW people reached the ' According to Renfrew (1989, p. 38), "the coherence coast by the beginning of the first millennium AD and ' approach emphasizes ... the extent to which the propo had mastered the technology of producing pottery. ' sition harmonizes with the existing frarnework of Moreover, the whole coast is known to have had clay knowledge". Same of the theories established for the for pottery-making, and the long-distance trade in local . coast of East Africa can be shown, by a mere examina pottery would have required every economic justifica- : tion of the literature, to be inconsistent with what is tian. known about the coastal sites. Same of these theories Ideas on the inconsistency of the theories will figure · were formulated to justify a particular, established directly in Chapter 11, where they will offer supple position, thus leaving behind much known information, mentary support to the new perspectives propounded which, if well taken account of, could have established there. a different opinion altogether. Typical of such theories In the next chapter, I highlight the sources, methodo is the assertion that the Sassanid pottery found on the logies and theories that have been responsible for our coastal settlements is Islamic. This has continued to be present knowledge of the early-first-millennium, the view of all schalars up to today, although the green/ coastal settlements. blue pottery is wel! known to have been produced from , > 4. INTELLECTUAL BACKGROUND There is abundant literature that offers accounts of the 4.1 Ancient Graeco-Roman documents and people, settlements and various economic and political reports by Arab, Chinese and Portuguese activities that took place in the first millennium AD on travellers the coast of East Africa. This can be divided into three main categories: (1) travellers' reports (ancient) and (2) This literature must be looked upon as non-academic. cbronicles, and (3) the scholarly reviews of both the It was written for different purposes, sometimes as a reports and the cbronicles and the subsequent academic guide to trade, for example, the Periplus, sometimes as research (scholastic). geographical information, for example, Ptolemy's The first part is the information about our coast dat Geography, sometimes as sailors' tales, for example, ing from the beginning of the 1st century AD, as nar the reports of Asian travellers, and sometimes as rated by travellers and other people from the reports to kings, for example, Portuguese accounts. As Mediterranean states and the Middle East who were I am going to discuss this literature in its different cate interested in travellers tales. It includes Graeco-Roman gories below, I would like to refer my readers to the documents of the early first millennium AD and Arabic more exhausti ve discussion of these documents in and Chinese documents of the late first millennium and Mathew (1963, 1975), Freeman-Grenville (1975), the early second millennium AD. From the sixteenth Trimingham (1975) and Horton (1984). century onwards, the Portuguese documents and cbron icles supplemented the ancient literature. The second 4.1.1 Graeco-Roman documents part, the cbronicles written as histories of coastal towns and their rulers, have been the most quoted evidence In this category, there are three documents: the anony for the Arabo-Persian colonization of the coast. mous Periplus oj the Erythrean Sea (AD 40-70), Ptole The third part is a process starting from the begin my's Geography (AD 140), and the account of Cosmas ning ofthe 19th century. European scholars and travel lndicopleustes (6th century). The Periplus, known as lers started documenting and studying the people and the "commercial handbook of the lst century AD" (Cas the history of East Africa. The process and the way in son 1989), is the earliest literature we know of that dis which the scholarly literature was accumulated is the cusses the coast of East Africa (Azania) in relation to subject of discussion in this chapter. the Middle East and Mediterranean regions (Map 3). It As Gräslund (1987, p.2) stated, it is only appropriate gives the routes, ports and goods involved in trade to judge scientific achievement by relating it to "the from the Red Sea to the rest of the Indian Ocean situation of its own time". I have therefore divided the including East Africa (Map 3). Ptolemy's Geography, scholastic literature into three categories to match the written a century later, offers sirnilar information about different periods of its production. Early scholarship is the routes and the locations of ports. It is, however, less arbitrarily the early part of this century up to 1965, informative and in some cases confuses the earlier pic when the early sources discussed above were solely ture drawn in the Periplus (Warmington 1963, p. 65; used to describe the past on the coast of East Africa. Mathew 1963, p. 96; Sutton 1990, p. 90). Because of The sources were used to prove or to support the then this, the Periplus has warranted several reviews, current migration and colonization theory. The latter including those of Schoff (1912), Huntingford (1980) scholarship is more related to academic research that and more recently Casson (1989). The third source, the ensued after the 1960s to assess the work of early account by Cosmas, is later and mentions little about scholarship in relation to the new information being the Azanian coast but concentrates more on the interior obtained from archaeological and linguistic researches. south of Axum. Recent scholarship has shifted slightly from the former From the first two early documents, we learn that paradigm to focus more on the African contribution to there was a market-place or emporium known as the early settlements on the coast of East Africa. Rhapta (after sewn boats) somewhere on the EastAfri can coast. Apart from the governing Homerites (not mentioned in Ptolemy), the land was occnpied by big- 26 30° 50° 70° 90° 0 o= 10° ) ' '' •···• Trade routes \ - -Principal caravan routes "', ' ' \\\\\Gold bearing region \ ' ---- Unknown lands \\ ' 0 1500km ' ,• ' '· ' ...... ----,,.-- ~,l i ~1 Map 3. EastAfrica and the Ancient World in the first century AD (afterFree man-Grenville 1988). bodied people who were either pirates or cultivators The early Arabic and Chinese travellers' documents (word indistinct; Casson 1989 and Horton 1990). They (c. 9th-12th centuries) are less helpful in identifying traded ivory and tortoise shell in exchange for meta! the locations of the places mentioned and in under objects. The two documents also mention other ports of standing the organization of the coastal people. These Azania to the north, Nikon and Serapion. The island of documents include those of Tuan Cheng-shih (AD 863), Menuthia is located "two runs" north of Rhapta. Buzurg Ibin Sahriyar of Rahorrnuz (mid lOth-century), In the account of Cosmas, the important information Al Masudi (late 9th century) and Ibin Hawqual (lOth is that the Axumites had pushed out the Romans and century) (Freeman-Grenville 1975) Arabs from the Red Sea, and the Persians were gaining Al Masudi is the most detailed writer, showing that the upper hand in the oceanic trade with Azania. The the coast was inhabited by black people, with their own Axumites are reported to have been using the land rulers ("fahue"). The Zanj coast had two areas known route to Bast Africa, trading beef for gold in a region as Sofala and Kumbalu Island, the !arter having a Mus south of the Bquator and close to the source of the Nile. lim community. There was a trade connection between Siraf/Oman and the Zanj coast. Sofala produced gold, which was traded together with ivory, tortoise shell and 4.1.2 Arabic and Chinese documents amber. Bananas and coconuts were cultivated. These have been referred to in some literature as Later Asiatic documents offer us a better picture of "medieval" documents. They were written either by the p!aces and activities on the coast. Docurnents of travellers who managed to reach the coast of Bast this period include those of Al-Idrisi (1100--66), Chao Africa (Zanj land) or by people based in the cities of Ju Kua (1226), Abu Al-Fida (1273-1331), Marco Polo the Middle Bast who were interested in travellers' (1295), Ibn Battfita (1331) andAbuAl-Mahasin (1441) tales. (Freeman-Grenville 1975). > 27 Place-names that we know today, for example, no other chronicle can be traced back to before the Unguja, Zanzibar, Mogadishu, Malindi, Mombasa, eighteenth century, and most are more recent. Kilwa and Sofala, are mentioned. Al-Idrisi and lbn Bat The importance of the chronicles relative to other tuta offer us better pictures of these places and their written sources is that they ascribe the origin of the activities (see Idrisi's map in Trimingham 1975, p. 138; early coastal towns to the Arabo-Persian immigrants. Freeman-Grenville 1972). The people cultivated fruits, Awareness of this is cardinal, because the subsequent, sorghum, sugar-cane, bananas, rice and camphor trees. 20th-century, scholastic literature was greatly influ Tuere was mining of iron at Malindi, Mombasa and enced by what is alluded to in the chronicles. Sofala for trade, and gold from Sofala. There was The Kilwa chronicle relates the immigration of the pearl-fishing and the cultivation of aromatic plants. The seven princes ofShiraz. About the lOth century AD, Al people also hunted for skins and ivory (see Freeman Hasan lbn Ali, Sultan of Shiraz, sailed from Persia with Grenville 1962, pp. 9-24). There is an indication from his six sons and some followers to the coast of East Battfita's report that, by the 14th century, the major set Africa. He founded settlements on the shore of the tlements along the littoral were generally Islamized mainland and the islands. His son Ali is alleged to have (Freeman-Grenville 1975, pp. 27-32). become the first ruler of Kilwa Island in 956 AD. The Pate and Lamu chronicles assert that early towns along the coast were founded by Syrians sent by the 4.1.3 Portuguese documents Umayyad Caliph al-Malik (AD 685-705) to settle there. Thirty-five coastal towns were founded. The Portuguese led the way in the European ambition The Book of Zanj (Kitab al Zunuj), a compilation of of having an alternative route to India and southern the chronicles of Shuugwaya, Mombasa and the Arabic Asia. This would help them to avoid restrictions and Treatise, is recent, being of the late 19th century. It nar blockades by the Muslims on the Red and Arabian rates how the coastal people were converted to Islam. Seas. This dream was fulfilled in 1498 by the expedi The work reiterates the earlier idea of the fouuding of tion of Vasco da Gama arouud the Cape of Good Hope, coastal towns by the Arabs. It is alleged that, in the era along the African coast and on to India. Several expedi of the Abbasid Caliph al-Mansur (754-775), an expedi tions like this generated much information in Europe tion was sent to crush the disloyal towns on the coast. about the people of the East African coast. The most Governors were sent by Harun al-Rashid to rule the relevant to our studies is the work by Joäo de Barros coast of East Africa (Chittick 1965, 1975, p. 32; entitled Decadas da Asia. This work repeats the chroni Mathew 1963, pp. 103-4). cles' information discussed below. The Portuguese document also shows that, for centuries, the coast had developed brisk trade-links with the Arab world. Major towns competing for this trade had developed, includ 4.3 Scholastic period ing Mogadishu, Mombasa and Kilwa. Kilwa is reported to have controlled the southern coast as far as 4.3.1 Early scholarship Sofala and the routes leading to the gold-producing We have recognized the first part of the 20th century, Zimbabwe. Zimbabwe and its state are mentioned for up to the mid-1960s, as the period of early scholarship the first time. The state was known to have controlled in the East African coastal studies. This period is sus the trade as far as the shore, where merchants linked to tained by the fäet that, before the mid-1960s, most Kilwa were operating. Cloth was a major item schalars had meagre archaeological and socio-linguis exchanged for gold (Freeman-Grenville 1962, pp. 120- tic data to guide or support their arguments. Their dis 4; see also Theal (1898) and Stranders (1968) for the courses were based on data from the sources discussed Portuguese documentation). above (Sections 3.1 and 3.2). A limited amouut of field data was available, i.e. observed ruins, coins collected on the surface and words in the Swahili language bor 4.2 Chronicles rowed or derived from the countries of the Middle East. For a better picture of this period, readers are The chronicles of the coastal towns date from the referred to the key authors of the time, including Bur second part of the second millennium AD. According to ton (1872), Stigand (1913), Pearce (1920), Ingrams Chittick (1975, p. 32) and Mathew (1963, pp. 102-4), (1931, 1962), Kenyon (1931), Robinson (1937, 1939), with the sole exception of the Kilwa chronicle, which Coupland (1938), Baker (1941), Gray (1951, 1952) and Was set down in the first half of the sixteenth century, Co!e (1963). > 28 Inthe late 1950s and early 1960s, some archaeologi thought that they were Bantu (Kenyon 1931, p. 263; ·. cal and linguistic works were introduced. Archaeologi Coupland 1938, p. 12), anda few argued that they rep- . cal work on the coast was aimed at the major resented a transient racial amalgam, in which there was monuments, then in ruins. Excavations were carried an element derived from a wave of emigration that out around the major house strnctures in Gedi and seems to have reached Madagascar from Indonesia • Kilwa to establish the chronology of the coastal civili (Mathew 1963, p. 95). The emporium of Rhapta was zations (Chittick 1958--62, 1965; Kirkman 1959, 1964; thought to have been in the area between the Pangani Freeman-Grenville 1958, 1960, 1962). The main inte and Ruliji Rivers (Pearce 1920; Ingrams 1931; Mathew rest was in verifying the chronicles' accounts of coastal 1963), the location around Dar-es-Salaam being the development. Data from such works started to be inte most favoured (Coupland 1938; Warrnington 1963; grated into the literature in the last few years of the Kenyon 1931). Menuthia Island was thought to be early scholarship period. Finds of the coins of Ptolemy either Pemba, Zanzibar or· Malia, with Zanzibar the · · VIII (116--80 BC), associated with a dagger, were most favoured (Pearce 1920; Warmington 1963; Coup reported around Dar-es-Salaam. Zanzibar coins of the land 1938; Ingrams 1931; Mathew 1963). Hellenistic, Parthian and Sassanian periods were used as evidence to supplement the ancient documeuts 4.3.2 Later scholarship (1965--80) (Freeman-Grenville 1960, 1963; Mathew 1963, Sutton 1966). These linds, however, were not made in any An important phenomenon marking the beginning of archaeological context (Freeman-Grenville 1960, pp. this period is the intensilication of archaeological and 32-4; Sutton 1966, p. 7). socio-linguistic research, both along the coast and in With regard to the linguistic literature, the first publi the hinterland. The literature of this period is blended cations about the origin and spread of the Bantu people with archaeological and linguistic data. started to appear in the late 1950s. The Bantu people, to The beginning of this period saw the occurrence of ; whom the Swahili were thought to be related, were two important events. First, the introduction of Azania, considered to have originated in the Cong~ameroon the journal of the British Institute (of History and highlands (Murdock 1959; Greenberg 1963; Guthrie Archaeology) in East Africa. This acted as a forum for 1962; Oliver 1966). Their rapid spread to the humid reporting varians archaeological, historical andlinguis- , region of tropical Africa was thought to have been tic linds, supplementing Tanganyika/Tanzania Notes . assisted by the adoption of Indonesian food plants, i.e. and Records and the Journal ofAfrican History. bananas, yams and taro, that allegedly had reached Secondly, the publication of archaeological reports West Africa from the Ethiopian highlands (Murdock of the lield work that had been going on along the coast ' 1959). Indonesians were thought to have come to eas of Kenya and Tanzania increased from the late 1950s tem Africa at the beginning of the first millennium AD, onwards. Kirkman was instrumental in publishing ):he a theory imagined to be supported by the presence in data from the Kenya coast, and his discussion of Gedi Madagascar of people with Far Eastem physical and (1964) in relation to other sites, including Ungwana ' linguistic features. The early literature on the Indone (1966), was soon referred to in the coastal literature. sian colonization has been discussed by Oliver (1966), Chittick was even more prolilic after his reports of the Verin (1975, 1986, pp. 26--52) and Jones (1971). Kilwa excavations (1966b), Unguja Ukuu (1966a) and A common feature of the early scholarship literature the discovery of the Lamu sites, especially on Manda is the theory that the coast of East Africa was continu (1967). Chittick then re-interpreted the linds of Pate ously under the umbrella of either the Arabs or the Per (1969) and later completed two volumes on the Kilwa sians or the Austronesians. The stone towns are said to excavations (1974). His linds and suggestions domi have been built by the colonizing immigrants. In the nated the literature of the coastal studies from the late literature of this period, the authors accept the authen 1960s to the early 1980s. This infiuence was further ticity of the early sources uncritically. strengthened by the publication of the Manda volume However, there is some debate on the identity of the (Chittick 1984). indigenous people and the location of places or towns In the same period, Chittick (1975b, 1977) published mentioned in the early documents. According to Cas other synthesizing works. He maintained the opinion son (1989, p. 136), many authors seem to have thought that the coast of East Africa was colonized by the that the indigenous people of the coast of Africa were Arabs. He supported this view with the study of the Harnitic/Cushitic people who had immigrated from the imported pottery and the architecture. At the conclu Horn of Africa (Cole 1954, 1963; Huntingford 1963, p. sion of his 1975b work, he wrote that the Arabic/Per 73; Ingrams 1962, p. 2; Sutton 1966, p. 7), A minority sian colouizers came to East Africa in small numbers > 29 from diverse regions, but sometimes there were "more These variants are based entirely on differences of pottery typology massive waves of irnrnigration from particular areas, and geographical distribution and should be viewed in this light, although they may reftect variations of culture in a broader sense. probably spread over years" (Chittick 1975b, p. 40). Augmenting the archaeological literature, there were The East African local variants were recognized at various other kinds of syuthesizing books or articles on Urewe around the lake zones, Lelesu in central Tauza the coastal settlements or their early inhabitauts. Sutton nia aud K wale on the coast (Soper 197 lc). From the (1966, 1973) discussed the trade on the coast, extend studies, it was hypothesized !hat the EIW people had ing his focus to Zimbabwe; Verin (1966) wrote on the spread from the interlacustrine region, where they were Indonesiau question on Madagascar, Ricks (1970) on established by the mid first millennium BC, to the rest the relationship between the Persian Gulf and Bast of eastern aud southern Africa along three migratory Africa, Allen (1974) and Pouwels (1974) on the routes, the western, central aud eastern (Huffrnan 1970; peopling ofthe coastal settlements, Trimingham (1975) Phillipson 1976, 1977; Soper, 1971b). and Chittick (1977) on the early documents in relation What is crucial for East Africau coastal archaeology to the new finds aud Allen (1977) and Datoo (1970) on is the shared view !hat the EIW people from the interla the settlement pattern. Morgan's (1973) geography of custrine region occupied the coast of East Africa in the East Africa also dealt with the coast, discussing its early centuries AD, producing pottery known as Kwale environment, economic activities, people and history. (Soper 1967b, !971b, c). On reaching the coast in the Phillipson's (1977) chapter on the coast introduced a early part of the first millennium AD, the producers of new viewpoint by indicating !hat the alleged 9th-cen Kwale pottery are alleged to have migrated to southeru tury pottery had elements of the EIW tradition. This Africa to occupy the southern part of Malawi (Robin meaut that the early urbau settlements had au Africau son 1971) and southern Mozarnbique (Cruz e Silva foundation, in opposition to the prevailing Arabo-Per 1977) and then to Trausvaal in South Africa (Klapwijk sian theory. His clairns were further strengthened by his 1974). 1979 report on the Taua Valley TlW sites. He again The archaeological work in the hinterland was argued for a connection between the TIW aud the EIW closely related with the linguistic work !hat had been traditions. From his Taua site, the TIW tradition came focused chiefly on the Bautu speakers (Phillipson to be labelled "Tana" (Horton 1984). 1977; Ehret & Posnansky 1982). As regarded the coast In the hinterlaud, research took place in the period and the eastern highlauds of Tanzauia aud Kenya, lin 1965-80. Much of this was directed to the iron-work guists aud historiaus were involved in a debate on how iug sites, later to be known as the "Bantu Project" Bantu speakers had originated from the rnythical laud (Soper 197la). This endeavour had in fact started in the of Shungwaya. Mcintosh (1968) accepted the myth as early l 960s, when some archaeologists had already true for all north-eastern Bautu-speakers, including the attempted to explain the spread of EIW cornmunities in Chaga, Taita aud Kikuyu. Others (Saberwal 1967; easteru aud southern Africa (Posnansky 1961; Fagau Fadimau 1973) accepted the myth but saw it as apply 1965). However, serious attempts to campare the pot ing only to fewer groups, including the Mijikenda, tery from different EIW sites, in order to establish the Pokomo, Taita aud Segeju. Hienesbutch (1976) clari direction of migration aud the group affiliation started fied the problem by showing !hat the groups that shared to appear after 1966. Posnansky (1967, 1968) conti the myth were represented by Sabaki-Seuta-speakers, nued to be instrumental in the interlacustrine region, cornprising the Kenya coastal groups of Pokomo aud and later Schmidt (1978). Lelesu in central Tanzania Mijikenda aud the Swahili of the coastal, stone towns. was reported by Sutton (1968). Soper (1967b, c, 1977) In terms of the development of knowledge, the aud Odner (197la, b) operated in the irnrnediate hinter researchers of the later scholarship period, compared laud and the eastern highlauds of Tanzania and Kenya. with their predecessors, gradually became aware of the Meauwhile, similar studies were being carried out in possible, early occupation of the coast by Africaus, for central aud southern Africa (Robinson 1973; Phillipson example, the EIW (Bantu) people, though they were 1968a, 1974; Fagau 1965; 1970, 1976; Klapwijk 1974; not quite convinced. Their awareness was owing to the Cruz e Silva 1977), the result being the establishment finding of au EIW site at Kwale aud in South Pare of an iron-working tradition, according to Soper dated to the 3rd century AD (Chittick 1975b, p. 18). (1971b, p. 6), known as the "Southem African Early Chittick (1977, p. 189) later expressed his scepticisrn Iron Age Industrial Complex" or just the "Early Iron again when he argued !hat the coast had no sites pre Age". Local variauts within the complex were identi dating the 9th-century period aud none that had never fied (Soper 197lb, c; Huffrnau 1970). According to had trade goods from the Islarnic world, specifying Soper (1971b, p. 8): such sites in the hinterland. He argued !hat such hinter- 30 land sites had no cultural relationship with the coastal se, Africans had been producing steel iron, a developc; si tes. ment not witnessed in the other continents (fora rebut-J Apart from the above awareness, some scholars con tal, see Rehder 1986). tinued to believe that a "Cushitic-speaking population Given the mid-first-millennium BC dates reported by' like the stone bowl users of the Kenya Rift Valley" Schmidt, questions started arising on whether archaeo-; (Oliver 1978, pp. 373-4) had occupied the coast in the logists and linguists could agree on the question of the . first millennium AD. origin ofthe EIW communities. Gramly (1978, p. 108}, On the other hand, the coastal researchers still failed for instance, highlighted the fact that the linguistically . to relate the littoral and the hinterland or the urban set assurned origin of the Bantu in the Niger-Benue region'. tlements and the EIW tradition. This problem was (primary nuclear area) and the subsequent spread from: neglected, despite the hint already given by Phillipson the secondary area south of the Congo forest contra- . (cf. above) that the TIW tradition at the beginning of dicted the early archaeological dates from the Lakei the occupations of the stone towns was closely linked Victoria region. with the EIW tradition. The assurnption continued to be Graruly (1978, pp. 109-112) went further, refusing' that the coast had no relationship with the agticultural the migration approach and suggesting that "Baittu was; hinterland (Chittick 1977, p. 189). The question of cul spoken for millennia in many of the same regions·· tural change was neglected both in the hinterland and where it is found today and that the Negro and other; in the littoral. African populations have remained in lands which they ' As the period of later scholarship came to an end, occupied before the advent of food-production, cera-; however, some scholars seemed to have discovered mics, and metals". that the problem causing all the vacillation above was the Hamitic/Cushitic and migration paradigm of the 4.3.3 Recent scholarship traditional archaeology (Chapter 3). Allen (1977) heral ded a new era of theoretical framework by decrying the Research and theory after 1980 were revolutionary, in theory that Arabs and other Asians had founded the the sense asserted by Kuhn (1970, p. 4). Strategically, coastal settlements. He argued (p. 361) that such theory archaeological excavations have been supplemented by was based exclusively on documentary sources, "most ethno-archaeology and economic, spatial and ceramic of them written by casual visitors from outside Africa, analysis models, both on the littoral and in the hinter- . supplemented by archaeological evidence from Kilwa, land; the nature of iron technology has been addressed ; Gedi and a few other si tes". for the first time, and there has been a paradigmatic ; In the hinterland, iron-working research was also shift towards the recognition of Africans' involvement under review. According to Garlake (1978), researchers in the early coastal civilization spreading from Somalia ' in that field had failed to demonstrate diachronic pro to Madagascar. cesses because the paradigm in operation was that of For the littoral stretch, this period can reasonably be traditional archaeology. Reviewing Phillipson's (1977) argued to have been inaugurated by an issue of Paid work, Garlake (1978, p. 459) put forward an idea that euma (1982), which was published in honour of James marked the tendencies in the post-1980 scholarship: Kirkman. In this publication, Allen continued to depart from the earlier period of scholarship by suggesting Archaeologists interested in Iron Age Africa seem to have reached that the people identifying themselves as Shiraz an impasse. They are not able to come to tenns with causes of (alleged to have founded towns) wereAfricans, who, in change in the societies they study . ... As a ''traditional" archaeolo gist, Phillipson sees bis work as providing "the lower storey", ev en their bid to assert their status among the coastal people, ''the foundations for the prehistorians' edifice" on which the "new" were posited as Persians. Allen argued that archaeolo archaeology can develop its own concems. But the foundations of a gical finds and linguistic and oral tradition do not suit science do not rest on data but on paradigms. It is the theoretical un the Persian origin of the coastal people and their settle derpinning of Phillipson's edifice-one we at present all inhabit that most needs renovation. · ments. Shepherd's article extended our attention to the southeru coast of East Africa, where the early Swahili In the last few years of this period, there was also an traders are known to have gone as far as southem attempt to understand the smelting technology used in Mozambique. She showed how the traders crossed the Africa. This was a positive shift from the regular pot Mozambique Channel to Madagascar and the Comoros tery studies. The experimental method advocated in the to avoid the Mozambican current. Sinclair reported on New Archaeology was applied for the first time to the the Chibuene (TIW) settlement in southeru Mozam study of iron-smelting (Schmidt 1980, 1978; Merwe bique, which has similar cultural materials to those of 1980). It was found that, from the mid first millennium Kilwa. For the first time, the people responsible for 31 'f(W were shown to have reached the southem part of Swahili successors, who later spread it south to the rest eastem Africa. of the coast: ··The Jingnists Nurse (1983) and Nurse and Spear Definite evidence that same of the inhabitants of Shanga were at :(1985) contributed also to the change of paradigm and least pastoral in origin comes from the identification of a few camel methodology. By the use of comparative and glatta- bones from the site (Horton 1984. p. 232). . chronological Jinguistics, they suggested that the Swa The Swahili people, in collaboration with alleged, hili tradition occupying the coastal towns down to the incoming, foreign traders who settled and intermarried Comoros had originated from the Lamu Archipelago with the Africans, continued to be marine-oriented. region. They showed that the culture had its origin in From this northem position, a new coastal culture was the Bantu speakers, who, on settling in the Larnu area bom and spread to the rest of the southem coast along in the latter part of the first millennium, saw the pass the shore. "Geographical fäctors, such as fävourable ing Middle Bast traders, leamed to make boats and to winds and currents and vegetation, put the northem trade from them, and, as they started to intermarry with Swahili coast in a particularly good position with these people, a Swahili group was formed, who then regard to the trade of the western Indian Ocean" (Hor rooved southward to found new settlements; hence the ton 1984, p. 266). origin of the early coastal towns (Nurse 1983, p. 315). On the second problem, Horton (1984, p. 299) In other publications, the southem extension of the deduced that, since the founders of the TIW tradition TIW tradition was further supported by Wright (1984), were traders and not farmers, they would have little to who published the archaeological report on the early do with the hinterland. Because of this, he thought, the occupation of Dembeni and other sites in the Comoros. finding of TIW pottery in the hinterland, i.e. in the In the same year (1984), coastal scholarship saw the U sarobara hills, was a function of trade from the litto production of two other works from northem Kenya by ral, as this pottery was made by non-farming people. Chittick on Manda and another by Horton on Shanga, Horton does not discuss, however, how the descendants giving us the northemmost perspective of the early, of the Cushitic/Pastoral people caroe to adopt the Bantu coastal, TIW culture. In the work by Chittick, we still language. see the survival of the old paradigm, He regards Manda In his follow-up articles, Horton (1987a) went a fur as an Arab colony but is for the first tiroe aware of the ther step to argue that the early Swahili people were not !arge quantity of local wares on the site that formed just the recipients of trade goods from the Arabian !ra ''the greatest part by volume even in the early periods" ders but were also seafarers sailing all the way to the (1984, p. 217). Chittick offers a flimsy explanation of Red Sea and to the Persian Gulf to convey goods. This how local pottery predominated in a settlement of marks the period of great prosperity for the coast, when immigrants. He suggests the existence, in the vicinity items like gold, precious stones and ivory from Bast of the stone-built towns, of African settlements that Africa were reaching as far as Burope. would have produced the pottery and traded it to the In another work, Horton (1987b) made a step back town-dwellers. wards to balance his theory of Pastoral/Cushitic origin Horton, just like Allen (1977, 1982), takes advantage by arguing that the early African founders of the early of this problem to suggest that the founders of the TIW Jittoral settlements were Bantu Sabaki speakers. This settlements, i.e. Manda, Shanga and Kilwa, were Afri was probably owing to the fäet that Jinguists had found cans rather than Arabian immigrants. By the use of that the language of the early Swahili had been Bantu. L. substantive economic medels, Horton grapples with Capitalising on this, Horton (1987b, p. 315) jumped two major issues, one being the origin of those early inta another difficulty by arguing that the spread of Africans and the other conceming the fäet that the TIW TIW tradition southwards would have ended at the pottery similar to what had been found in association River Pangani in the hinterland, except for the littoral, with Sassanid pottery in the earliest levels of the Larou where the Swahili carried it further south to Mozaro Archipelago had been found elsewhere, i.e. in Usaro bique and the Comoros. This is the supposed distribu bara, Unguja Ukuu, Kilwa, Chibuene and the Comoros. tion of the Sabaki-speaking groups. He, however, As regards the first problem, Horton found himself argued that the TIW pottery tradition had nothing to do being carried inta the well-entrenched, traditional para with the BIW tradition but was a pastoralist tradition digm by arguing that the early Africans who founded (1987b, p. 315). the littoral settlements were Pastoral/Cushitic in origin. In his 1990 article, Horton reiterated his Harnitic/ After being attracted to the coast for temporary or sea Cushitic paradigm by denying the existence of the BIW f sonal trade fairs, they settled down and passed on their sites along the coast in the early centuries AD (Horton tradition, including trading and pottery-making, to their 1990). 32 The early-first-century 33 approach in southem Africa. In proposing this, I am tion of the finds. In the next chapter, I discuss the envi quite aware of the ambiguities embedded in archaeo ronmental factors that rnight have played some role in Jogical data, and in Chapter 3 I have suggested the need the past cultural adaptation and change. for curbing our subjective tendencies in the interpreta- ,. i f f 1 ' 5. ENVIRONMENTAL SETTING 5.1 Introduction coast, in order to create a better understanding of the resources and obstacles that would have conditioned It has already been mentioned that the aim of this work ancient settlements. My aim is to set the cultural procc is to deal mainly with cultural chronology, change and esses in their environmental background. spread in the first millennium AD on the East African coast. Part of this aim entails the recognition that the processes of change and spread took place in particular, 5 .2 The extent of the East African coastal belt ) environmental contexts. It is therefore necessary to describe these environmental contexts, which could The coast of East Africa is part of the eastern-African\ have affected or conditioned the processes of change coastal belt, which extends from southem Somalia (1 o; under study. N.) to the mouth of the Limpopo River (25° S.) in The interplay between enviromnent and culture has southern Mozambique (Map 4). According to White) become one of the central studies in anthropology, (1983), it is a strip about 3000 km long and between 50. archaeology and geography (Steward 1956; Ellen and 200 km wide, except where it penetrates furthe :· 1982; Carlstein 1980; Steward & Murphy 1977; Witt inland along broad river valleys like that of the fogel 1957; Carneiro 1970). Although environmenta Ruvuma. It is very narrow on the northem coast o( lism had a strong infiuence in America between 1940 Mozambique and wider on the central coast of Tanza-i and 1960, according to Willey and Sabloff (1980, pp. nia, where itreaches the hills ofNguru and Uluguru. · 149-50): The coastal strip includes the major islands of East: Africa. Zanzibar has an area of 1660 km2 and extends. Such concerns in envirorunental archaeology in America had a much earlier inception in European prehistory. and American 85 km from north to south and 39 km from east to west.: 2 archaeologists undoubtedly were infiuenced by these pioneer old Pemba is 980 km in area. The total area of the coastal' world studies. For instance, in Scandinavia. an interest in environ strip is 336,000 km2 (White 1983; Morgan 1973). mental reconstruction can be traced well back into the nineteenth century and, in Great Britain, Sir Cyril Fox, 0. G. S. Crawford and I shall confine the following discussion to the coast· Grahame Clark had long been working along these lines. ofTanzania and Kenya (former British EastAfrica). As, far as the environment of the southern coast is eon-: The major infiuences of 20th-century environmenta cerned, I may refer to Verin (1986), Sinclair (1987),; lism, however, came from the German Friedrich Rat Morais (1988) and White (1983). zel' s "anthropological school", with the predorninant message that differences in natural habitat "were often regarded as sufficient to explain cultural diversity" 5.3 Physical environment (Ellen 1982, p. 2). Most ofthe researches conducted by the German school were concentrated in Africa The environment of the East African coast is mostly' (Zwernemanu 1983). The incliuation of these scholars that of the land below 200 m above sea-leveL Only a; to diffusionism led to its downfall in the early 20th cen few scattered hills and higher plateaux occur. They • tury, leaving the American cultural sphere and ecologi include the Shimba Hills (c. 400 m), Mrima Hill in cal studies to predorninate (Zwernemann 1983; Harris Kenya, the Pugu Hills, the Rondo (Mwera) Plateau in 1968). In recent years, some early, mechanistic, eco Tanzania, and the Maconde Plateau (986 m) in northem logical studies have been discredited for their environ Mozambique. On the outskirts of the coastal zone are . mental determinism. They have now been sup the eastern Usambara Monntains (1500 m) and the plemented by econornic and system theories (Harris Ulugurn and Ngnru Mountains (1000 m) (Map 4). 1968; Clarke 1978; Ellen 1982; Gibbon 1984). Seven major rivers enter the Indian Ocean, but none My intention in this chapter is not to submerge this of them is navigable. Given the leve! of development ' work in the environmental aspects of either ecological, reached by the early-second-millennium culture, the' geographical or spatial studies. These deserve their absence of navigable rivers has been used to explain · own research. This work just draws attention to the why there was minimal extension of the early trading • natural features, climate, soils and vegetation of the connnunities to the deep hinterland (Datoo 1975, p. 6; > 35 SOMALIA sonal, depending on the rainfall. However, some rivers, for example, the Rufiji, could be controlled for irriga tion agriculture (Morgan 1973, pp. 192-94). The coast of East Africa also does not possess any KENYA gulfs or big lakes. This again has offset the chances of extending the prosperity of the littoral to the hinterland communities through water cornmunication. North of Bagamoyo, the coast runs north-east to south-west, l coinciding with the Ruvu-Mombasa fault (Åse 1987; Temple 1970, 1971). This trend is offset by the Dar-es Salaam Peninsula, the projection of land extending from the Warni River in the north to the Rufiji River in the south. According to Temple (1970, p. 21), This unit forms a roughly rectilinear block, offsetting south-east TANZANIA ward the major trend of the rnainland coast south of the Wami delta. Dar-E s-Salaam From Ras Dege the general north-east alignment is taken up again as far as the Rufiji delta. This region, defined in the east by the coast, in the west by the Rufiji depression, is a rather homogeneous area of of moderate relief. I', In relation to the settlements, however, the projecting peninsula is assumed to have been the point where ancient ships, driven by trade winds via Zanzibar, made their landfalls (Datoo 1970; KITwan 1986; Casson 1989). While coral reefs created shelter for vessels in early times, a number of small, tidal inlets or rias also gave rise to good harbours. According to Datoo (1975, p. 5), these became good locations for local ports, espe cially on the offshore islands dotting the coast from the Lamu Archipelago in the north to Kilwa in the south. Zl However, these coral reefs are today dangerous to large, modem vessels. Taken together, these factors 500km poor rivers, absence of inland lakes, absence of gulfs, and a more or less straight littoral-meant that most of the iniportant settlements fiourished on a north-south line along the shore (Map 1). 1 - Arabuko-Sokoke hills Coral reefs and offshore islands also provided rich 2 - Shimba hills waters for fishing. Abundant marine life, i.e. shellfish, 3 - W .Usambara M ts low-water fish and crabs, is to be found, and there is 4 - E.Usambara Mts I much evidence from the archaeological sites that indi 5 - Nguru Mts l. cates the ancient exploitation of these resources (Chit 6- Uluguru Mts I tick 1974; Horton & Mudida 1993; Charni 1992a; 1' / Marginal coastal zone Msemwa 1992). The presence of the coral reefs along 1. the shore has also afforded cement (mortar) andrag for Map 4. The coastal belt of Bast Africa showing by the similar kind of vege 1 tation: Zanzibar-Inhambane (ZI) Regional Mosaic (White 1983, p. 185). house construction since the latter part of the first mil lennium AD. The Swalllli towns and tombs, as opposed to the hinterland settlements, were made of such mate Morgan 1973, pp. 192-94). The rivers concemed are rials, refiecting adaptation to the natura! environrnent. the Gn!ana and Tana to the north, the Pangani, Ruvu and Rufiji in the centre, and the Ruvurua to the south. Owing to the unreliability of the rainfall and thus the 5.4 Climate drastic fiuctuation of water reginies, no documented, ,,. .irrigation agriculture has been practiced along the riv The whole coast of East Africa has a mean temperature ers of the coast. Cultivation in the valleys has been sea- of c. 26°C but this diminishes steadily southward. The 36 range is quite minimal. Warrner temperatures, coupled1( with warrner waters, have provided favourable condi< '1 tions for the growth of marine life, such as the polyps·,~. responsible for coral reefs. The importance of these1} reefs for the·coast of East Africa from early limes hasi :1 been discussed above. ·} According to White (1983, p. 185; Map 5), the rain~.L fall is mostly between 20 in. and 50 in. (508 mm and'~ 1270 mm) per year. It increases closerto the equator ow& the southern and northern coasts of Kenya and Tanza-:§ nia respectively, including the islands of Pemba and:?; Zanzibar. The East U sambara Mountains, Zanzibar and ·i Pemba receive the highest annual rainfall (Amani, 76.6 ~ in./1946 mm; Wete, 77.3 in./1946 mm). The rainfall is adequate for agriculture, as is shown in section 5.6. ,. Among the most crucial, climatic factors relative to :; the developments on the coast of East Africa are the : 1 monsoon winds. These winds bring precipitation and i! stormy weather to the coast, and these effects differ't from those of the north-eastern and south-western · · winds (Map 5). Altemating with the position of the sun, !'. 0 the north-east monsoon blows between November and ,:lll. () Febrnary, a period of fair weather bringing precipita- 1E rn tion to the coast south of the equator; the south-west].~ )> monsoon occnrs from May to August, when the wmds :t; <: are very strong and bring precipitation north of the ,• Equator. The reversal of the wind direction causes two '·i annual rainfall seasons on the northern and southern ,, coasts of Tanzania and Kenya respectively, plus the ·;, islands of Pemba and Zanzibar. This makes the coastal il. strip wet enough to support forests and perennial crops, 't i.e. bananas, coconuts and fruits. · ~ Another respect in which the winds are important ·1 relates to commerce. Since the EIW period, the mon- ;j soon winds have also been used to propel boats (or il ships) bringing !raders to and from East Africa, hence -~ the expression "trade winds". The knowledge of how .~ the winds operated became an absolute necessity for 1j the trade links between East Africa and the northern 'I states of the Red Sea, the Persian Gulf, India and other :l parts of Asia. The same was true for the links between .'j the East African local settlements ranging from Soma- ;~ lia to southem Mozambique. Different Arabian !raders '] MOZAMBIOUE who happened to have knowledge of this climatic fac-. Il 100 Mi\es tor in trade are known to have kept it secret from the ,; Mediterranean !raders for centuries, in order to main tain their trade monopoly (Casson 1989, p. 11). Discussions of how the !raders exploited the winds have been carried on by Datoo (1975), Horton (1984) . and Casson (1989), among others. According to Casson · (1989, p. 283), :··. ·· · : 30-50 mches /////// //////// Map 5. East African coastal rainfall pro what made both routes possible were the monsoons, the winds of ;::;::;::;::;::;:;: over 50 inches bability (abstracted from Russell 1977).' the Arab i an sea and western Indian Ocean that blow from the north- 37 Mtoni terrace, the area with i 1st millennium AD Hill site (Surveyed and mappedJ. s HEIGHT (m)r-----__,_ 10 s 8 B.ea.ch terrace, with the 13th century i 6 ruins. Surveyed. 1 4 i Fig. 1. A cross section of the terraces at 2 i Kaole, Bagamoyo. a = lOth century AD sea level, b = 15th century AD sea level, c =present high tide level. 440 400 360 320 280 240 200 160 120 80 40 DISTANCE {ml east ,- in fair weather, the return trip was dangerous, owing to there has been in the geophysical sciences an active strong winds accompanied by storms. Because of this, discussion on the existence and development of the the dhows set their northward course either with the marine lerraces in Bast Africa (Alexander 1968, 1969; build-up of the south-west monsoon in April or with Cooke 1974; Temple 1970, 1971; Kaaya, Boenigk the tail-end of the monsoon in September (Datoo 1975, 1986). p. 4). The marine terraces on the coast of Bast Africa are The monsoons also detennined the trade movements better observed in the area between Dar-es-Salaarn (to all over the coast. The winds reached the southem coast the south) and Malindi in Kenya (to the north). Several of Tanzania in January over its greatest areal extent. lerraces have been observed and narned after the loca From there, the winds are no longer steady, and move tion where they were first identified (Alexander 1968, ment southward depended on the strong, Mozarnbique, p. 138). They are Sakura, Tanga and Mtoni. Beach ler ocean current. To retum from the Mozarnbican ports, races are also observed in some places. Sirnilar lerraces may differ in height in different >r !raders went across the southwarcl-flowing Mozarn e bican currents to the west coast of Madagascar, where places. According lo Alexander (Åse 1978, p. 220), the they entered northward-flowing currents that carded Sakura terrace ranges from 30 up to 115 m above sea " !hem to the Comoro Islands. There, the south-west level, the Tanga terrace from 18 to 30 m above sealevel monsoons are strong enough to drive dhows to the and the Mtoni terrace from 3 to 11 m. coast of Tanzania. A foreigner interested in trading as In the area around Bagarnoyo, Cilek (Åse 1987, p. ~r; 282) found !hat "the cliff of the Mtoni terrace is about 8 Il far south as Sofala had to be prepared to spend about a l year in Bast Africa. The trade to !hese southem parts of to 12 m high and typically developed between Baga es East Africa was, therefore, easily monopolized by the moyo and Kaole, where it forms the shore line". Kilwa traders, who had easy access to the gold-produc Between the Mtoni terrace and the present beach, Chit ing hinterland of Zinlbabwe, rnaking it possible for the tick (1962), Cooke (1974) and Åse (1987) bave identi northem !raders to get goods without delay. Hence, the fied another recent terrace, a beach terrace about 2 to 3 prosperity ofthe Kilwa town (Datoo 1975). m high (Fig. I). ls Controversial dates based on coral and shells have I) been presented for the formation of the lerraces (Åse 1987, p. 286; Kaaya, Boenigk 1986, p. 314). Sakura 'Il 5.5 Sea-level fluctuation and marine terraces and Tanga are dated to different periods of the Pleis It is also inlportant, in discussing the TIW sites of the tocene before 24,000 BP and Mtoni to the late postgla of first millennium AD, to bring the reader's attention to cial age between 1500 and 6000 BP. The beach terrace h- 38 Yellow-red sandy clay loams Dark red friable clays Dark red ta red loamy sands Red- Brown- laterite soils NE Dark red sandy clay loams Dark red to reddish brown YH ca\careous sandy clay loams FR Dark red sandy clay \oams Brown to dark brown FO calcareous clay \oams 1127:7~~~ Strong brown to pale yellow WE/LF loarmy sands with laterite horizon QC/QF Coastal sands Brown to yellow-red sandy NE clay loams with laterite horizon 'TANG~ Brown to yellow-red sandy LF clay loams with laterite horizon ••~Pemba VP GP/JE/GH 0 VP Saline and alkali c\ays 0 rn Coral rag )> I-V ~ BC BC Brown friable clays cifMafia Coastal and marina deposits,fine LF liliili!i1"'ili!iLlil grained marl, calcareous sand clay MOZAMBIQUE Map 6. Bast African coastal so ils (abstracted from Russell 1977. NE, BC etc according ta FAO/UNESCO soil map 1977). 0 50 100 MILES 39 growing days • 1-74 ~ 75- 89 Il 90-119 -120.:.149 150-179 ~ 180-209 tiiliil 011210-239 on Map 7. Agro-ecological zones of the central coast of East Africa. 1e clay (Fig. 1) is dated to "late medieval times (c. 500 BP)'', effect of that in relation to the settlements' realignment which is said to correspond wen with the Kaole ruin or the dislocation of resources is yet to be studied. site, originally founded on the beach terrace between the 13th and 16th centuries AD (Åse 1987, p. 286). The implication of these dates is that the archaeolo 5.6 Soils gists should not expect to find pre-4th-century-AD, iron working and Stone Age sites on the Mtoni and beach The description of the soils below has been abstracted terraces. By then, these terraces were under water. As from Milne (1935), Russen (1977) and the FAO indicated elsewhere (Chami 1992a, p. 32), such sites UNESCO description of general African soils (1977, could be found some distance beyond the present-day 1978). The FAO-UNESCO (1988) soil types QC, BC, · littoral. NE, etc. have been compared with Russen (1977) to As it will be discussed in Chapter 8, the sites of produce Map 6. Mpiji and Kaole Hill are located ou the Mtoni terrace, The soils of the coast are derived from three paren! and the Changwehela and Kaole ruins on the beach ter rocks of sedimentary nature. The inland margin of this race. The dates of the Changwehela and Kaole ruin rock, fonning the coastal foot-plateau above 300 m, sites (Table 21 and Fig. 27) indicate that the last, major, consists of continental sediments of Karoo age, largely 7.NE. sea-level change !hat created the beach terrace of sandstone and shales. The ocean side consists occurred about the end of the first millennium. The entirely of marine deposits, which have been laid down > 40 Table 1. Agricultural suitability of the soils on the coast of East Africa. Abbreviations: CAS = cassava, WH = wheat, SOR = sorghum, :MIL= millet, BEA= bean, SOY = soybean, SCA = sugar-cane, MAI = maize, WPT = wbite potatoes, SPT = sweet potatoes, RIC = rice. In the matrix, S = suitable, VS = very suitable. SO!LS CAS WH SOR MIL BEA SOY SCA MAJ WPT SPT RIC F VS s vs VS vs vs vs vs VS vs vs QC VS VS vs VS vs vs VS VS VS OF NE s s s s s s s s s s s VP VS vs vs vs VS vs vs VS GP/GHIJE s s {~ I ;t) B-C s s s s s s s s s s s AO/AF vs vs vs vs VS VS vs vs vs vs vs ~-§; YH FR vs vs vs vs vs VS vs vs vs vs VS FO vs vs vs vs vs vs vs vs vs vs vs WE vs vs vs vs VS vs vs vs vs vs vs successively against the edge of the older mass. In the coast of East Africa has been shown to have the Kenya, an outcrop of Jurassic sedirnentary rock occurs potentiality to cultivate the various crops illustrated in; between the two above. The plain soils below 200 m Table l. above sea-level are yellowish-brown, sandy clays, acid Given the number of growing days, ranging from in reaction and poor in organic matter. They are perme 150 to 209 in the agro-ecological zones 5 and 6 (Map · able to a depth varying from 30 to 120 cm below the 7, FAO 1980, p. 109), many crops can be cultivated on i surface. The soils are calcarious, owing to the Ill.arine, the coast3! strip of East Africa (FAO 1980, p. 109), as i. sedimentary, paren! rock (Map 6: QC and QF). shown in Table l. These include sorghum, millet, ! The river plains (valleys) and the lagoons have rich beans, sugar cane and rice, which have a longer history deposits of clay and loams. These include the Tana and in Africa than other plants of recent introduction, i.e. Webi-Shebelle valleys to the north, some parts of cassava and maize. We have already seen in Chapter 6 Malindi, Pangani on the northern coast of Tanzania, the that mos! of these, including bananas and coconuts, had Ruvu and Rufiji valleys and the Ruvuma vaHey to the already been domesticated on the soils of East Africa at south (Map 6: GP/JE/GH, VP). the end of the first millennium AD. The area above 3000 m on the mountains forming Given !hese agricultural potentialities, supple the fringes of the ·coastal belt, U sambara, Uluguru and mented by fishing and hunting (Chapter 6), and assum-, Nguru, has, according to Milne (1935, p. 90) ing that the conditions in the period under study were i the same as today (Hamilton 1982), the coast of East ' ... red-brown, orange-brown or yellow-brown, pervious, structure Africa could have sustained !arge agricultural commu- i less leached soils, whose bright subsoil colours persist practically to nities. the soil surface in spite of a high content of organic matter in the foot. The soils are acid throughout the profile, and their clay-frac tion shows them to be fairly far advanced in laterisation (silicafses quioxide ratio 1.5 to 1.0 or less). They are derived from rocks of the "ancient crystalline complex", mainly of acid types-gneiss, granu 5.7 Vegetation lite, pegmatite; and the dissection of the Usambara mountain block is sucb that they are mostly found on step slopes or narro.w ridges. White (1983) provides the most detailed account of the vegetation of the coast of East Africa. What should be Except where these soils have been disturbed by shift noted here is that the vegetation of the study area is ing cultivation, in areas such as eastern Usambara, such defined as the Zanzibar-Inhambane flora! mosaic, soils support rain-forest.resources, i.e. wood and other placed by White (1983, p. 59) under mapping unit 16, biomass. These kinds of soil are part of a sub-section of which he identifies as the East African coastal mosaic. the great soil group that has, as yet, received no better Under this mapping unit, the Zanzibar-Inhambane name than "the laterised red loams" (Milne 1935, p. 90; mosaic falls under units 16a and 16b (Map 8). Unit 16a i Map 6: BC, FR, NE, AO/FO). comprises vegetation which has been "so extensively ' The agricultural potentialities of the coast of East modified by man that it is impossible to map the diffe Africa can now be assessed. A study combining the rent physiognomic types separately". Unit 16b maps length of the growing period and the soil types has the patches of forest. been conducted (Map 7 and Table l).According to this, Typical remnants of such wooded land are seen in 41 Afro-alpine Heath and Moorland 2 ~Wi!\!1'1""'Jil Forests at low and medium altitudes Brachystegia-other species WoodJands 3 of Tanzania ("Miombo") W07771 Savanna-like communities derived 4 from forest (includes forest remnants) 16a 5 Woodland-Bushland intermediate 6 Dry Bushland and Thicket 54b 7 Desert Scrub 8 Mangrove ~ 16• <: ANGA~ D •• • Pemba ):> <: ~nz;bar 0 BAGA MOYO () rn ):> <: ; ' MOZAMBIQUE Map 8. Bast African coastal vegetation (abstracted from Russell 1977. For 16a, 16b, 54b se Chapter 5). 0 50 100 Miles 42 the wettest part of the coastal fringes, the U sambara, What is also clear today is that the wooded/forested Uluguru and Nguru Mountains (Map 8:2). Intensive lowland of the East African coast is infested with tsetse agricultural activities have transfonned the ancient for f!ies (Bourn 1978; Hamilton 1982, pp. 19, 22) and !hat'. est on the moister, coastal lowland into savanna-like serious agriculture has to go hand in hand with clearing . · communities (Map 8: 4) vegetation and the active protection of crops against According to White (1983, p. 189), this secondary wild animals, i.e. primates and swine. From the forests, grassland and wooded grassland are typical and exten people still obtain fruits, honey, hunted animals and sive on both the islands and the rnainland. They are a birds. These could have been more abundant in ancient mosaic of agricultural crops, grassy fallows and secon limes. dary thickets, according to White (1983, p. 189), domi Conseqnently, although some cattle, goat and sheep. nated by "lantana", and orchards of "cocos", bones have been found on the TIW sites, the people "anacardium" and "mangifera", which, -when the can now occupying the coast of East Africa rarely keep cat" opy is not too dense, often have a carpet of grass. In tle, goats or sheep, owing to the tsetse f!ies, the lack of places, trees from the original forest have been left pasture, due to the encroachment of the forest, and the standing. The pahns, mangoes and cashew nuts are absence of tradition. Only in areas largely cleared of locally conspicuous. wood, for example, Kaole, is this practised, giving the Swamp forests (Map 8:8) occur on the lagoons, estu people a mixed economy. It has, however, been shown aries, creeks and silted parts of the shore. Mangrove in southern Africa that areas infested with tsetse f!ies forests are typical of these places and have been today maintained domesticated animals in different exploited by coastal communities from ancient limes periods in the past (Maggs & Whitelaw 1991, p. 16). for their hard wood for local and export purposes. They This means that the wooded/forested areas now also host a variety of marine food resources, for exam infested by the flies might have expanded or contracted ple, shellfish, crabs and fish (Msernwa 1990). in the past (Sinclair 1987, p. 47), owing to different The remaining major part of the coast is woodland conditions, including clearing and burning (Morais bushland intennediate (Map 8:5) in northern Tanzania 1988, pp. 30--33). The idea, therefore, that the 11W and Kenya. South of the Pangani River is the tsetse sites were found and occupied by pastoralists (Horton, infected, miombo woodland (Map 8:3). 1990) will certainly require re-evaluation (Chapters 6 What should be noted in relation to this work is that and 11). the geographical zone that has been identified with the It should be noted that, while it is possible to show distribution öf TIW sites is generally forested or how the occupants of the littoral adapted to several of wooded. Research aimed at re-drawing the picture of l:he factors discussed above, the same task would the ancient coastal vegetation would probably shed require more research as regards the hinterland occu light on the nature of the ancient settlements, the forest/ pants. We know of too few settlements of this early wood resources then available and the problems cre period in the hinterland to develop a clear, adaptive ated by such heavy vegetation. What can be noted in pattem. On the few sites that we know of, however, this archaeological research is that the coastal, their subsistence and crafts depended very much on the : savanna-like vegetation is a product of man's slash good agricultural soils, the lacustrine resources availa- .! and-burn agriculture (above) and commerce (Elton ble around the settlements, the availability of enough i 1874). As regards the latter, it has been reported that rainfall and clean water for daily use, and the existence the "msandarusi-gum-producing tree" was a dominant of raw materials for metal-smelting, i.e. ferrous, late- · forest tree all over the coast, thickly lining the shore in ritic rocks and red soils. These aspects will be dis- . ancient limes. The depletion of the ancient forest can cussed in the next chapter, in which I shall discuss the be shown today by the fäet that the tree can be found economic activities of the ancient coastal communities, only far towards the hinterland on the slopes of the which refiect their efforts to harness the environment Nguru and Uluguru Mountains over 200 km away from for subsistence and commercial purposes. the shore (Elton 1874, p. 230). > 6. TECHNO-ECONOMIC SUBSYSTEM The importance of the techno-economic system has exploitation of the natura! envirorrment (especially been discussed in detail in Chapter 2. Gibbon ( 1984, p. forests ), and significant economic, political and possi 180) has defined the techno-economic system as that bly religions influences. part of a social system that mediates between human Much less is known about the farmers occupying the beings and their natura! surroundings: coast of East Africa in the early part of the first millen nium AD. However, their knowledge of iron-smelting More specifically, at the community level the technoeconomic sys tem is a buffer between the human population, the rest of a comrnu gave them a technological advantage in adapting to the nity's sociocuJtural subsystems, and the natural setting. marine and terrestrial environment of the coast. A study conducted at the Limbo site on the Kisarawe Gibbon has also identified the main elements of coast of Tanzania has shown that intensive smelting techno-economic systems as took place there at the beginning of our era (Charni modes of subsistence and production, tools and machines, proce 1988a, 1988b, 1992). A trench measuring 2 x 1 m, dures for allocating goods and services, rules for regulating eco excavated toa depth of 3 m, yielded about 113.96 kg of nonric activities, and the knowledge that makes all of these slag (Figs 2a and b ). Together with the material col activities possible (p. 180). lected from the other three trenches, over 150 kg of This chapter is therefore divided into two sections: pro slag were collected. Such intensive production has duction, which deals with various technologies and been attributed to an effort to cope with a demand for crafts and with subsistence activities; and trade, which the tools required for clearing coastal woodland, allti- · is important in the allocation and distribution of goods. cultrue, warfare, fishing, boat-building and probably In the discussion, mast information is obtained from the archaeological data. An attempt to link the first-mil lennium, techno-economic activities with both ethno graphic and documentary data has been made by lforton (1984, pp. 246-306). 6.1 Production .6.1.1 Metal-working 0 The EIW communities of East Africa knew how to \eve[scm amounlkgs Smelt iron and make tools and weapons from probably Fig. 2a. Limbo: amount of slag from trench 3a and b = 40 cm, then 10 cm. lhe middle of the first millennium BC onwards. On the Intervals. Total= 113.9 kg. assurrtpti.on that these early communities were also llllticultural, the adoption of iron technology offered them great possibilities of expanding by exploiting vir gin land, clearing of woodland and conquering enemies 20 · (Phillipson 1985, p. 148). 15 1 . The earliest, iron-using si tes in East Africa have been 10 · f9und in the interlacustrine region. Studies of iron 5 ,smelting there have shown that intensive and sophisti . cated technology was known that led to the production wvutsrqponmlkj ihgfedc a-b of steel (Schmidt 1980, p. 340; 1978). Such technology levelscm In the interlacustrine region resulted, according to Fig. 2b. Limbo: amount of slag from trench 3. From the earliest occupation Schmidt, in dense concentrations of population, over- (w) to the latest (d). > ( ,- 44 trade. The few meta! pieces fouud on the site included clair 1982, p. 162). Similar copper objects recovered :ii elongated nai!s and awls. from the mainland sites of Tanzania are reported in W Other EIW sites on the coast have shown signs of Chapter 7. • ilC4 smelting, thongh not as much as Limbo. At the Kwale The Kilwa and Manda excavations have also shown ,;j site, an iron arrowhead, some slag anda piece of tuyere that other metals, for example, gold and lead, were·~ /'. were recovered (Soper 1967). In Mozambique, Ada exploited and worked at the end of the first millennium;Jj mowicz (1992) has reported smelting at the Nampula AD. These metals could have been exploited from ear-1 j sites dated to the second century AD (Sinclair, Morais, !ier times. Continued research on the coastal sites may :~ Adamowicz & Duarte 1993, p. 421). In Zitondo, "the offer us a better picture of !hese other metals. Lead has;~ earliest known occurrence of a major smelting site in been found to have been worked at the TIW site of:~ Southern Africa", dated in the lst-3th centuries AD, Mpiji, which is dated to the 7th century (Chapter 7). :j.-., "105 samples of slag have been recovered from the occupation levels in 22 of the excavated trenches" ; ;1 (Sinclair 1988, p. 7; Morais 1988). 6.1.2 Other crafts : J In the TIW tradition, the smelting of iron continued Apart from metalwork, the coastal people developed :~ unabated. Chittick's excavations at Kilwa (1974, pp. other crafts which were probably important for their il 438-59) and Manda (1984, pp. 203-12) and Horton's adaptation to the coastal enviromnent. Pottery-making '~ at Shanga (1984, pp. 260-61) yielded some slag and for both domestic use and the storage of water and!°I many iron objects. They comprised knives, arrow foodstuffs appears to have been a dominant craft from';) heads, hooks, wire wound round fibre cores, rings and the early farming period. All iron-working sites found'l bangles. Fumaces .have been found in Manda (Chittick on the coastal strip have yielded pottery (Soper 1967a;:j 1984, p. 211) and Shanga (Horton 1984, p. 260). In Sinclair 1987; Morais 1988; Chami 1992a and b; Ada-'J Chibuene in southem Mozambique, "twenty-one lumps mowicz 1991). Chittick (1974, p. 317), for instance,i~ of ferruginous material, two hook fragments, one nail recovered over one million sherds from Kilwa alone,] i :0 and one spike were found" in layers pre-dating 1000 AD and Horton (1984, p. 257) recovered one hundred thou-;~ (Sinclair 1982, p. 162). All TIW sites excavated on the sand from Shanga. l;I mainland and reported in Chapter 7 yielded some slag, To facilitate fishing and communication among thel'll pieces of tuyere and iron objects. At the 9th-10th-cen coastal communities, the construction and use of boatsr~ tury sites on the Comoros, Dembeni and Sima, might have started from early in the first millennium:t AD. Big fish identified on the littoral sites by their bone~~ ... slag and iron fragments were common. ... It is clear not only :.,j:\ that iron tools were used on these sites, but also that iron was remains required boats !hat would take fishermen to theil locally smelted and forged (Wright 1984, p. 45). deep waters (Chittick 1974, p. 235; Horton & Mudida,)I 1993_; cf. Chapter 7 in this work). Finds ~f EIW potter~-,fl Apart from iron-smelting, there are strong indications on s1tes such as Unguja Ukuu and Kilwa Kis1wam.:1~ !hat other kinds of metals were worked on the coast of (Sinclair, Morais, Adamowicz & Duarte 1993, p. 427Jjl East Africa. The earliest evidence of copper-smelting and Bazaruto in Mozambique (Sinclair 1987) indicate,:I on the coast comes probably from the reported copper !hat the EIW farmers had managed to sail across the: (I crucibles at Kisiju (Harding 1960). While Harding channels to the islands. This is also testified by thefl reported !hat the associated materials could be dated Periplus repor! !hat the people of Rhapta sailed in sewniil between the 4th and the 15th centuries AD (p. 136), boats (Casson 198_9). In_ the LIW period, from _the mid-i··.·!·1'..::.•:... · Chittick's crucibles at Kilwa were later !han the 9th dle of the first rmllennmm AD, the boat-making crafö: century AD. However, the early adoption of copper must have been improved so !hat the people of the., smelting technology along the coast is confirmed by coastal communities could sail long distances for trade~.. ·.1.. the finding of many copper objects at almost all the This is supported by the proliferation of TIW settle-:1 TIW sites. The copper objects found at Kilwa and ments on the littoral and on the islands of Zanzibar;.. :.l.1'• Manda (Chittick 1974, pp. 438-59; 1984, pp. 203-12) (Chittick 1966a, Horton & Clark 1985), Pemba (La'r comprised rings, beads, chains, bangles, fish-hooks, Violette,pers. comm.) and the Comoros (Wright 1984).il curved copper wire, vessels, meta! sheeting, nails and Apart from improving boat technology, the coastal:; kohl sticks. In Ungwana, Abungu (1989, p. 174) communities had also probably mastered the knowl-::· reported copper chains, rings, and rods/kohl sticks. In edge of trade winds which was necessary for sailing~ Chibuene, the copper objects included a fragment of long distances. '): copper sheeting, a spike, a bracelet fragment of wire Bead-making seem to have been another major pre- 5 twisted round a fibre core and two copper beads (Sin- occupation. This is indicated by the finding of many', 45 "bead-grinders" (Chittick 1974, pp. 414-15, 1984, pp. This evidence is also coupled with the widespread 155-56; Abungu 1989, pp. 72-3; Chami 1992b; see occurrence of grindstones, indicating widespread culti aJso Chapter 7). The !arge number of "bead-grinders" vation in the first millennium AD (Maggs & Whitelaw found on TIW sites indicates that prodnction might 1991, p. 17). Documentary evidence from the Periplus have been geared to Jocal needs and trade (Chittick AD ( 40--70) also indicates that the inhabitants of Rhapta 1984, pp. 28-9). Considerable quantities of finished Jocated in the vicinity of present-day Dar-es-Salaam and unfinished beads have been found at Kilwa, Manda were farmers (Casson 1989, p. 253). and other sites (Chapter 7). Other beads were made In the succeeding TIW period, more archaeological from bone, ivory, clay, and precious stones (Chapter 7). evidence for farming is available, giving further evi Rock crystals were also worked, as !hese were found at dence that cultivation was probably widespread over a the Jower Jevels of Shanga and Manda as finished ]arge region of sub-Saharan Africa. Coconut seeds and beads, wasters and lumps. "East Africa is well known rnillet and rice have been found in Ngazidja in the as a source of rock crystals to Egypt" (Horton 1984, p. Comoros (Wright 1984, p. 261), and Horton (1984, p. 255). 261) has reported a coconut scraper. Coconut shells Cloth-making, probably from cotton, is indicated by have also been found at Misasa, Mpiji, and Kaole the finding of spindle whorls at Kilwa (Chittick 1984, (Chapter 7). Sorghum has been reported from Kilwa p. 236), Manda (Chittick 1984, pp. 156-57) and (Chittick 1974, p. 236) and Nampula (Sinclair, Morais, Shanga (Horton 1984, p. 252) and Ungwana (1989). Adamowizc & Duarte 1993). In the TIW period, it seems, people Jearned the initial Arabic documents report the cultivation of crops technique of obtaining lime. Burnt coral stones and such as sorghum, millet, bananas, rice and sugar cane coral and shell floors have been found in the TIW occu at the beginning of the second millennium AD (Horton pational layers at Kilwa (Chittick 1974, p. 22), Mwali 1984, pp. 229-30; Freeman-Grenville 1972). Mdjini on the Comoros (Chanudet 1991), Unguja Domestic animals rnight also have supplemented the Ukuu (Juma, pers. comm.) and Kaole (Chapter 7). The food crops in some settlements. Goat and sheep banes use of lime and coral stones for serious construction have been reported from Chibuene (Sinclair 1982), . was introduced at the end of the first millennium AD, Manda (Chittick 1984), Shanga (Horton 1984), the · and by the 13th century hundreds of settlements dotting Comoros (Wright 1984, p. 53), and probably Mpiji the Jittoral had houses and tombs built in this material (Chapter 7). Cattle and camel bones have been recove (Chittick 1974, 1984; Horton 1984, 1987b; Freeman red from Shanga (Horton 1984, pp. 231-32). Other ani Grenville 1962). Mudbrick houses have been reported mals included the cat, chickens and domestic fowls from the northern site of Manda, dating from the begin (Table 4). This implies that the ancient people of the of the TIW occupations (Chitttick 1984, p. 217). coast practised rnixed farming, a tradition still followed houses in this period, however, were of mud and on the coast today, where the land is open enough to wattle (Allen 1980; Wright 1984; Chitick 1984). The allow some domestication of ]arge animals. Mixed exploitation of wood, especially mangrove poles, for farming, however, does not mean pastoralism . .house construction went hand in hand with building For the littoral and lacustrine sites, fishing and shell technology, as may be seen in the mins on the coast. fishing were definitely another major source of food. All such sites have been reported to have yielded !arge arnounts of fish-bones, shells, turtle and crabs (Chittick 1984, 1974; Horton 1984; Wright 1984; Sinclair 1982; 6.1.3 Subsistence Horton & Mudida 1993). Banes of dugong, crocodile, ··The primary subsistence activity of the EIW communi turtle and whale, together with shells and crabs, have . ties was probably farming (Phillipson 1985, pp. 147- been recovered in !arge amounts at Mpiji (Table 4). Morais 1988, pp. 137-43; Sinclair 1987, pp. 143- Hunting for wild animals is indicated by the finding of 44). Little archaeological data, however, exists to offer iron arrowheads at the Mpiji site (Fig. 11) and the · a better picture of the early farrning activities. Only in bones of wild anirnals, comprising the dik-dik, duiker, : Silver Leaves, a si te identified with Kwale pottery, bush buck and bovid, and various kinds of birds at ha.ve seed impressions in pottery been reported as bul Shanga (Horton 1984, p. 233). Adding to the Shanga fush millet (Klawijk 1974). list, the Mpiji site yielded a lot of bones, providing evi dence of a wide range of hunted animals, including the ~ultivated millet and possibly sorghum and Citrillus species were dik-dik, bovids, dugong, ovicaprids and pig (Table 4). r~covered from the 6th- and Sth-century sites of Magogo and Ndon dondwane in the Thukela Basin, and cultivated sorghum was identi Hence, both rnarine and terrestial resources were fied as SK17 in the Kruger Park (Maggs & Whitelaw 1991, p. 17). equally exploited. ) 46 6.2 Trade began in Rhodesia in about AD 600", while Sutto1' (1973) suggested a date in the 9th-19th centuries AD; In the literature dealing with the development of the Recent archaeological work has, however, indicated a East African coastal settlements, the question of trade date in the 8th century for the beginning of gold-work~ has figured as the most important. Some literature has ing in the Bulawayan system (Sinclair 1987, p. 148; been devoted entirely to trade (Sutton 1973; Ricks Duarte 1993, p. 42). Itis agreed, however, that the gold' 1970; Horton 1987a; Shepherd 1982). Over emphasis trade became important after the end of the first millen~ ·· on trade has conditioned scholars to attribute the early nium AD, especially the 12th-l 4th centuries AD (Datoo coastal development to the Middle East traders and 1975, p. 17; Sutton 1973; Sinclair 1987, p. 149). immigrants (Chapter 4 ). While this one-sided approach Although little is known about the trade in gum ara" is avoided in the present work, I recognize the eco bic in Africa before the 18th century (Elton 1874; nomic importance of trade in the development of the Webb 1985), many grains of fossilized (glassy) guq} early coastal culture. arabic have been found on the Limbo EIW site (Chami From the first century AD, traders were known to 1988a) and on the Misasa, Mpiji and Kiwangwa TIW have exploited the monsoon winds to visit East Africa. sites ( Chapter 7). From Elton' s report ( see Chapter 5); According to the Periplus (Casson 1989, p. 61), the we learn that the forest of "msandarusi"-gum-produc"' principal imports into the ports of East Africa were ing trees that once covered the coast had been depletedi "spears from Muza of local workmanship; axes; after centuries of gum-copal exploitation. No doubt the; knives; small awls; and numerous types of glass association of these products at the above sites implie~ stone". To certain places, wine and grain were brought such trade in ancient times. As mentioned above, in considerable quantities. East African ports exported except for Limbo, these sites were connected with the. a great amount of ivory but inferior to that from Adulis; interoceanic trade. rhinoceros horn; best-quality tortoise shell after the For the first millennium AD, very little is known: Indian; a little nautilus shell. about the trade going on between the East African comJ What confuses coastal researchers today is the fäet munities; on the one hand, between the littoral commu' that at the beginning of the first millennium AD, the nities and, on the other, between the littoral and th~ coastal people were producing iron (Soper l 967b; Cruz hinterland communities. But it is obvious that the TIW e Silva 1977; Charni 1992b). Why should the same settlements on the mainland and the islands had trade · people have imported what they were producing? It links by which the goods sent to and received from the may have been that the Azanians exported iron ore and interoceanic trade were collected and distributed. It i~ imported finished goods or rather supplemented their likely that there were major ports, i.e. Kilwa, Manda; products with better-quality ones. Unguja Ukuu, Kaole and Chibuene, which would hav~ After the middle of the first millennium AD, many played the role of gateway communities (Datoo 1975,1 TIW settlements fiourished all over the coast. The pp. 28-41; Hodge 1982; Horton 1984, p. 306). In suc . excavated sites of this period have yielded a variety of centres, elites could exist. The wealth resulting frotu imported goods, comprising glazed ware from the Per such groups of people is indicated archaeologically by sian Gulf and China, glass vessels and glass and pre the variety of expensive and prestigious, local and före cious-stone beads (Chittick 1974, 1984; Horton 1984; eign goods, for example, meta! objects, glazed wares,, Wright 1984; Sinclair 1984; cf. Chapter 7). glass objects and monumental constructions. Feedet Horton (1984, pp. 266--82; 1987b) has gone further ports would then develop, acting as intermediaries. than the archaeological data to exarnine the Arabic and between the centres and the hinterland communities. Chinese documents on the trade goods (Freeman-Gren It is likely, therefore, that local traders spread goods:. ville 1962). In the latter part of the first millennium AD, like glazed ware, glass and beads to the rest of the coast the major imports that would not have survived archae (Abungu & Mutoro 1993; cf. Chapter 7 below). ThiS ologically comprised cloth, rice, soap, wheat, indigo, could well have involved foodstuffs, fish and salt (Hor-1 butter and oils. The items exported comprised slaves, ton 1984) and the products of local crafts, i.e. shell ambergris, ivory, crystals, skins and tortoise shell (Hor beads, iron and copper products, graphite and to some ton 1984, pp. 262-71; Masao & Muttoro 1988, pp. 613; degree probably pottery. Datoo 1975, pp. 17). The question of how far the first-milleunium trade Gold was probably exploited from this period but is went to the hinterland remains unresolved. However, it first mentioned by Al-Masudi in the lOth century (Free can now be shown that it reached more than 200 km. man-Grenville 1962, pp. 15-7). According to Datoo into the hinterland. The finding of TIW sites over 300 (1975, p. 17), Summers estimates that "gold-mining km inland from the Tanzanian coast (Chapters 1 and 7) > 47 . with cultural affiliations with the littoral sites indicates first-millennium-AD sites of Botswana and South Africa that trade links between the littoral and the hinterland (Maggs & Whitelaw 1991, p. 15; Horton 1987a, p. 86). probably went fnrther than was previously thought The significance of the emergence of trading elites, (Chapters 2 and 4 ). In southern Africa, glass beads and gateway and feeder communities, and far-hinterland, marine shells traded from the eastern coast reached the peripheral communities is discussed in Chapter 11. 1 1 i r" s s ;t s r ll e . .e it n 0 1) 7. FIELD WORK As I said in Chapter 2, there are three issues related to in the hinterland of Bagarnoyo (see Maps 10 and 11)/ the TIW tradition which this work addresses: first, the Misasa had been identified and test-excavated by the idea that the tradition first emerged on the northem University of Dar-es-Salaam field researchers in 197 . Kenya coast; secondly, that the TIW pottery was spread (La Violette, Fawcett & Schmidt 1989; Fawcett &, ' through trade; and thirdly, that the TIW tradition was a LaViolette 1990). The rest were identified by the Unid cultnre oriented to the ocean rather than to the hinter versity of Dar-es-Salaam field researchers in 1989, bu~ land. The identification of several TIW sites both in the not excavated (Chami 1990). · hinterland and on the littoral of the central coast of The field research reported here was organized in;, Tanzania offered a chance to evaluate these issues. two phases. Phase One, which was between June an&. The central coast of Tanzania and its hinterland lie July 1990, consisted of reconnaissance. It was aimed td between the Wami River to the north and the Rufiji resurvey the sites identified by the 1989 Field School iti River to the south. On this coast are located the histori the Bagarnoyo District and to conduct test excavations, cal towns of Bagarnoyo, Dar-es-Salaarn, Kisiju and, to By this, I would be able to assess the potentialities o the south, Kilwa. The offshore islands comprise Zanzi the sites for further comprehensive research. Phas bar, Mafia and Chole and K wale near Kisiju. The hin Two, July to November 1991, was earmarked for mor terland extends for about 300 km into central Tanzania, comprehensive work on the sites found to yield mor~ covering the regions of Pwani, Morogoro and Dodoma. information on the problems mentioned in Chapter ~ These inland regions are linked to the littoral by the above. The sites preferred for this phase comprisecj Rufiji, Ruvu and Wami Rivers (Map 9). Misasa, Mpiji and Kiwangwa. ., Taking the central coast of Tanzania as a case study, Below, the discussion starts with the sites on th~ two factors make it a perfect area for investigating the Bagarnoyo littoral, moving into the Bagarnoyo hinter'• above problems: land, and then to Misasa in the Kisiju hinterland. ! First, the area is remote (about 500 km) from the northem-Kenyan pastoral, Neolithic (Cushitic) infiu ence. Similarly, the case-study area is away from the 7.1 Mpiji northem coastal strip alleged to have been favonred by the trade winds that made commercial transactions pos The site of Mpiji (6° 32'S., 39° 08'E.) is 27 km south of sible between the region and the other ports on the Bagarnoyo town, 6 km to the east of Mapinga villag~ periphery of the Indian Ocean (Horton 1984, p. 266; on the Dar-es-Salaarn-Bagarnoyo road, and 1 ~ also Abungu 1989, p. 31). north-west of the mouth of the Mpiji River (Map 12)\ Secondly, assuming that the EIW people were immi The site is located where the marine terraces (cf. Chap', grants from western Tanzania, the case-study area is ter 5 above) are quite outstanding. It lies where the, located in the settled zone, where the early, iron-using Mtoni terrace slopes gently inta the old terrace (Fig. 3)1 communities would have settled and established them The 3-km-wide, Mtoni terrace is of sandy-loarn clay, selves first before spreading northward and southward covered mainly by patches of secondary forest as yef to the rest of the eastem-African coast (Soper 197lb; uucleared. The forest plays host to a number of wil4 Phillipson 1985, pp. 173--4; Chami 1992b). EIW sites animals, inc!uding swine and primates. The beach ter! have been discovered here, providing evidence for race is 2 km wide and colonised by salt-tolerant plants\ EIW occupation from the beginning of the Christian including mangroves (where the sea water stil} time (Map 2). reaches ). It is a perfect habitat for crabs and shellfish) In this chapter, I report on the field work conducted There is a clear indication that the site was a port. Tu~ on the six sites in the case-study area. The sites com fact that today the shore is 2 km away implies that the prise Misasa on the Kisiju hinterland, Mpiji and sea-level has dropped or that there was siltation aftef Changwehela on the littoral north of Dar-es-Salaarn, the 7th-century AD date of the site. These events could Kaole near Bagarnoyo, and Kiwangwa and Masuguru have caused the ground-water leve! to go down, as iS 49 TANGA ~ KW PEMBW Pangarn 0 () "')> " Kilwa Map 9. The central littoral and Masoko hinterland of Tanzania. r l Ma~)O. The hin terland of Kisiju with Misasa and Limbo si tes. > 50 ,, LUGOBA }J I j •' ",, ")~" #~ '\\ -cf(l \~ 7 / \ CHAUNZE 11 \\ " To Dar es Sa/~am EXCAVATED SITES IN BAGAMOYO DISTRICT OC------'<10km e- S!TES Map 11. Excavated sites in the Bagamoyo district. 510 ------S1S 1 • I N DIA N 0 C EAN t bound surface road -~ loose su rface road ...... _...... trunk road ~ river • woodland • scrub ~ palms -50--- contours (V.I. 50m) 9270 0 2 3km • sos S10 S15 Map 12. The 1ittoral of Mpiji and Changwehela. 51 220 olive gray sand-loam 200 17\'/I with few artefacts 180 light olive brown sand clay 1so very dark gray sand loam with - cultural material concentration 140 m olive calcarious sandy clay 120 - ~ pale olive sand 100 IIIlllJ] 80 light gray sand beach 60 40 20 0 -20 -40 -60 2 1 4 s a 1 a 9 w n • m 14 w • rr • 19 20 21 22 u 24 Fig. 3. Mpiji. Cross-sect:ion of the Mtoni and beach te~aces. indicated by the dry marsh on the site (Map 13). (blades, arrowheads, projectile points, wire wound W11etl1er they had an effect on the settlements will be round fibre cores, rings, hooks ), copper objects ( copper discussed in Chapter 11. beads, copper chain, copper bangle), irnported pottery In Phase One, the site was found to be undisturbed. It (Sassanid pottery, Chinese stoneware ), beads ( of glass, \Vas mapped by using the theodolite (Map 13). Stakes shell and clay) (see Figs 10-14, Plates 1-2), gum :Were placed at intervals of 5 m on the notth-south and copal, bones and shells (Tables 4 and 5a, b ). east-west diagonals. These enabled us to establish grid Trenches 2 and 3 were outside the area of material on which a hand anger, 15 cm long and 5 cm concentration (Map 13). Few artefacts were recovered (;;;.•nu~, was used to establish the area of material concen from these trenches, with their homogeneous, brow . Samples of soils from the augering holes were nish, sandy-loam-clay deposit. Trench 2 was dug down iUso tested by applying phosphate. The result showed to 90 cm, hut only leve! 3 (20-30 cm) yielded artefacts. • .tl:ie concentration of organic matter to have coincided In trench 3, only the first leve! yielded artefacts. ·.. With the area of cultural-material concentration (Map Six fragments of charcoal were recovered for C14 ,;14). dating: orre was from leve! 4 (30-40 cm), two from Three trenches were then excavated by using trow level 5 (40-50 cm), two from leve! 6 (50-60 cm) and .)ls. The first trench measured l x 2 m. It was placed at one from level 8 (70-80 cm). It was not possible to lhe centre of the cultural-material concentration identi- determine whether these fragments of charcoal were ;; fied by augering. The first three levels (0-30 cm) were from a tree trunk or branches. of olive-grey, compact sandy-loam-clay. Few artefacts Fragments weighing over 3 g were selected for C14 were recovered from these levels. The next three (30- dating. One fragment was chosen from leve! 6 and 6D cm) were of very dark grey, compact sandy-loam another from level 8. The results are discussed in Chap clay with high concentrations of cultural materials. The ter 10 (Table 21 and Fig. 27, Lab. Nos. Ua-2087 and l<(st two levels (60-80 cm) were of brownish, calcari Ua-2088). ous sandy-loam-clay with few cultural materials (Fig. The Mpiji site was one of those selected for Phase 3). The materials collected comprised TIW pottery, Two research, because it had yielded a great variety of bead-grinders, tuyere fragments, slag, iron objects artefacts (above) and the dates obtained from the above > 52 ~ contours with 0,5m interval ;:::~.:::-. road raised sand beach 'f dry marsh - possible ancient lake e baobab tree >C DP datum point 1 D excavated trenches. excavated trench 0 5 10 15 20m 4 5 Map 13. Mpiji site plan. > 53 ' . 60W 40 20 if '. 20 4oE ~e s-"' @ 111 st>-~ll" \>-~Il s'll t>-\11"s • ; .. .. Rll'l'- lf\I>-~" 0 ; 20 ~zo e ,; ; ; , ; 0 /. 0 , / / / I / ___ rnangrova.swamp limil / ' ; @ 0- -0 .. , / ,,/ • very llttle blue colour • blue colour around the soil " lil blue colour with tendency to radlate ,__zo 11!1 @ 20- 0 yellow colour after solution an • blue colour with beam radiations datum poinl • • " • ~surface material concentration @ ~40 .. • " 40- • • 0 9 10 Z?m f-605 !il 60- 60 ~o zp 0. 20 40 Map 14. Spot test results fo:r Mpiji. charcoal samples were of the 7th century, earlier than only (Terebralia palustria). The soil was compacl, the conventional, 9th-century date of the early coastal brownish loam. settlements (Chapter 2). It was mentioned above that the material concentra The previonsly surveyed area was extended both to tion on the site is found on a slope of the Mloni lerrace the north and to the south-east. More extensive excava eroded down lo the beach terrace. To investigate the :• tion was conducted in the area identified previously as deposits on the slope, a trench (cross-section), 0.5 m · having rich deposits of cultural material. A 2 x 2 m wide and 13.5 m long, joining the upper and lower ter · extension of the previous trench I (Map 13) was exca races via the excavated trenches, was dug (Fig. 3) ·'[: .. :vated to the south, and later a 2 x 2 m trench was exca down to the sterile leve!. It was found thal the area on !.Nated to the west towards the upper terrace. The the slope had been used as a dump by the upper-terrace .. ·accurrence of cultural material and the nature of the occupants. The soil of the beach terrace was olive-grey '· soil followed the pattem previously described above sand in the upper 20-cm layer and then whitish beach f{Phase One ). 1bis lime, the change was in the occur sand. rence of cultural material past the 80-cm leve! of the The linds listed (Table 2) include those found in previous excavations to I 00 cm, owing to the sloping Phase One. TIW pottery forrned the bulle of the mate .. gtound. rial, with a total of 5538 sherds, 937 (16.9%) being Away from this area of material concentration, decorated (Fig. 9). Two hundred and förty bead-grin another 2 x 1 m trench was excavated lo the south-easl ders, mostly rnade of potsherds, were also recovered of the sile (trench 4, Map 13). It was placed in an area (Fig. 10). Iron-work was indicated by 16 fragments of identified by surface material. tuyere and 20 of slag, 89 iron objecls, three iron arrow · Apart from the pottery and glass collected from the heads, and six hooks (Fig. 11). Copper-working was surface, the lower levels down to 50 cm yielded shells indicaled by 26 beads, a chain and a bangle (Fig. 12). > 54 Table 2. Amounts ofremains collected from Mpiji in 1990 and 1991. The top layers of same of the trenches had no remaini (1991, trenches I & 2). Abbreviations: Tr =trench, Lv = level; (a) = decorated sherds, (b) = undecorated sherds, (c) = bead grinders, (d) = tuyere, (e) =slag, (f) = iron fragments, (g)~ = iron arrow, (h) = iron book, (i) = knife, (k) = copper chain, (1) = copper bangle, (m) = lead, (n) = imported sherd, (o) = glass fragments, (p) = glass bead, (tj) = shell bead, (r) = other beads (bane, clay), (s) = gum copal, (t) =bane, (u) = shell; +=present. "c. sect" at the bottom of the table = cross-section. ' Tr Lv a b c d e f g h k m n 0 p q r s u 1990 1 0 25 19 7 1 12 2 8 17 2 + 20 3 21 83 8 2 + 35 4 24 47 7 1 2 + 40 5 30 101 19 2 1 4 + 155 6 30 144 8 3 2 2 7 2 + 115 7 14 36 2 4 2 + 90 8 7 32 2 2 2 + 65 2 3 10 35 1 6 6 + 31 3 1 1 1 4 1991 1 1 16 9 2 17 8 210 35 3 85 202 21 11 1 9 270 68 4 104 346 30 3 4 24 1 1 3 34 26 2 297 90 5 89 320 30 7 4 1 1 1 7 74 382 280 6 46 109 23 3 2 1 4 4 7 36 800 215 7 84 229 14 1 9 2 3 6 4 6 40 623 192 8 74 147 39 4 3 1 4 5 2 22 340 102 9 69 194 1 3 3 3 2 4 6 187 30 10 6 78 1 1 14 2 2 2 13 1 2 3 30 224 3 9 1 5 1 2 4 74 335 16 9 1 3 22 23 5 5' 19 55 1 1 2 3 2 6 3 13 1 2 8 7 3 6 30 8 2 5 3 0 15 1 2 17 2 4 1 8 3 4 32 4 33 5 5 47 6 4 34 I 7 13 71 63 4 1-5 7 25 2 20 c.sect. 30 73 3 9 Astonishingly, 11 pieces of lead (Fig. 12), not found in banes are very decayed. The countable pieces of bone the previous excavations, were recovered. They were amounted to more than 3271, and the shells to 1600'.' bent to create a hollow cavity, perhaps for a string. The analysis of the bones (Table 4) was made by An irnportant element of the finds is the imported Kimengich, of the Kenya Museum, and that of the material. About 27 pieces of Sassanian Islamic sherds, shells (Table 5a, b) by Msemwa, of the Tanzania most of them in the form of bead-grinders and highly Museum. The subsistence implications of these finds brittle, one sherd of white-glazed, Chinese porcelain have already been discussed in Chapter 6. and 152 pieces of early glass were recovered (Fig. 14 and Plate 1). Most of the glass fragments were of greenish and yellowish tints, with the surface decaying 7 .2 Changwehela into a golden patina. Four glass and 215 shell beads were also recovered (Fig. 13 and Plate 2). Two more The site of Changwehela (6° 32'S., 38° 34'E.) has beer\ grains of gum arabic were recovered. described before in relation to the ethnographic dat~ The occupants of the site consumed a lot of fish and collected in the area about its history (Chami 1990). It shellfish, with fewer terrestrial animals. Most of the is 23 km south of Bagamoyo, 3 km north of Mpiji, and ) 55 X X the surface covered with potsherds scatter - salt water level (beginning of mangrove and salt swampl ~'-V vegetation on the flat top af the ridge scale~ 1:40 Fig. 4a. A section of the slope at the excavated area at Changwehela. 6 km north-east of Mapinga (on the Dar-Bagamoyo Chittick 1970; Mtnri 1974; Sutton 1990). It is sitnated road, Map 12). The site is on a ridge bordered on the 4 km south of Bagamoyo, the present-day Kaole vil north and east by salt-pans; to the west by forest on the lage being on the beach terrace. The first-millennium Mtoni terrace, extending all the way to Mpiji; and to site is on the edge of the Mtoni terrace, from which the south by mangrove swamps. The site itself is cov artefacts are being eroded to the lower beach terrace ered with thick, thorny brush. (Fig. I and Map 16). This site is covered by patches of Only one trench was dug in the area, where nume thick brush, extending southwards from the village and rous potsherds were being eroded from a wall founda the ruins. The road to the ruins cuts through the site tion cut across the northern edge of the beach-terrace (Map 16). I have called the first-millennium-AD site ., ., ,u,.,v. This disturbance occurred probably :~~; dark gray brown sandy clay ~ very dark gray brown sandy clay 7.3 Kaole ·::..:.·· ·-·:...· yel!owish gray with white sand mottling Kaole (6° 29'S., 38° 57'E.) is probably well known for darkish gray sand with yellowish mott1ings its famous, 13th-14th-centnry-AD tombs, which have ye!low sand with no material n ~ttracted many people visiting Dar-es-Salaam and brown sand 1 11 bog ~ B~gamoyo. The site has been reported since the late [t 1950s (Chittick 1958-62; 1962) and has featured in Fig. 4b. The eastern profile of trench 1 at Changwehela. (Other wall profiles d i:iany other publications (Freeman-Grenville 1962; are uniform). 56 INDIAN OGEAN Open sea Q""°'oB a , ", ~ ~q '> RUIN SITE Oa"' "'fl \~ "'aoo t::i "g;j p~ 5J a UP-HtLL SITE B (t;\ Harbour l KAOLE SUB-ISLAND ~ Proposed Harbour 0 100m Protected sea Map 15. Kaole sub-island with hill and ruin sites. 40 E --- lill DID • 0 ------0 - "' -- - -- I t 13 ----=--- ~ trench 1 JIIID trench 2 !il trench 3 § trench 4 ~ road ta the Kaole Ruins I P contour (V.I. 1m) 408 I' A datum point O 4 8m -~-~ 0 20 40 E Map 16. Kaole Hill site. > 57 the cultural-material concentration was mapped. The sub-surface concentration of material was identified by augering over the open area. Systematic augering would have required much work in clearing the brush. Trenches 1, 2, and 4 were excavated in the area of cultural-material concentration. They were dug with trowels. The stratigraphy of the upper units above was similar, with a top layer of dark-brown, sandy loam (to c. 30 cm) overlying first dark-to-gray and then basi cally reddish-yellow, loam deposits (Figs Sa, b). Artefacts of different periods were collected from the levels. The lower levets (60-90 cm) yielded first-mil lennium, TIW-tradition materials. The 30-60-cm leve! /· /. / /:/;/.·/:( /,'>/. / .._ .. · (.-/.: . . ; . : ' •.. · : . -;,:·y·/-./ /../ / /:/../"/ /.°/.. /./ /./.( _; /.·(·./.:( /.:-< had artefacts typical of the Kaole ruin phase, dated to ;:;-:;./:;".>'._(. /:"/.·/ ~--: the 13th-1Sth centuries (Mturi 1974). The surface le ;-.·:~y////"~~-/~/// ~///·~~~ vets had more recent material. scale~ 1:20 Trench 3 was dug lower down the hill, where a stone Fig. 5a. The eastem profile of trench 1 at the Kaole Hill site. (Other profiles wall was observed. The first two levels consisted of are uniform). very dark-grey loam. This was followed at 20 cm by the top part of the structure first thought to be a wall. It comprised a horizontally continuous layer of coral rag, mixed with shells, as well as other artefacts packed together like a pavement. This continued down to 45 ,,,,,,,,,,,,,,,,,,,, ,,,,,,,,,,,,,,,,,,,,,,,, cm below the surface. Below this layer, a reddish-yel ,,,,,,,,,,,,,,,,,,,,,,,, low layer with coral and shell pieces was encountered. ,,,,, ,,,,, The artefacts collected from trench 3 were those of the Kaole ruin phase (Fig. Se). Only one sherd of the TIW tradition was found in association with the coral and shell fragments at the lower leve!. Some previous exca vations in the ruin area had also indicated !hat the earli est leve! there had few sherds of the TIW tradition (Chami 1990, 1992a). This indicates !hat the ruin scale~ 1:2 O. period was a direct continuation of the TIW tradition. Fig. 5b. The western profile of trench 4 at the Kaole Hill site. (Other profiles The following material was collected from the site: are uniform). S45 potsherds of the first and second millennia AD (Fig. 9 and Table 6); llS "bead-grinders" (Fig. 10), some daub and lime concretions, 29 fragments of imported very dark gray sand-loam cerantics (early Islantic, sgraffiato and Chinese ), beads very dark gray s.and-loam of glass and copper, corroded iron, slag and fragments dark brown sand-loam of tnyere, remaills of coconnt, and fish and mammal dark gray sand-loam bones. dark gray sand-loam with cultural material concentration Several charcoal samples were recovered from dif whitish coral concentration ferent layers. Three of !hem were selected for Cl4 reddish brown sand-loam with first ·millennium AD material analysis. One was from trench 1, leve! 9, one from re~dish yellow sand-loam .with no material trench 3, leve! 6 and one from trench 4, leve! 7 (Table ////////// Fig. 5c. The western profile of trench 3 at the Kaole Hill .:. . : (.· .· .. site. (Other profiles are uniform). scale~ 1:40 > 58 MASUGURU SITE MAP 0 1 Un~ extavated <11d the Unit · Number. ::: Road /;$' Conlour IV.l.O·Sm_) A Datum point 2nd Datum point 3rd Datum point scale"' 1:2 O S_ca\e: very dark grayish brown clay 0 •. .. reddish brown/yellowish brown clay Pig. 6. The eastern profile of trench 2 at Masuguru. m, except for the first one, were dug down to 50 cm t& the beginning of sterile deposits. At the depth of 30 cm, the stratigraphy changed from dark-grey, loose, cla~ deposits, to sticky, reddish-yellow, clay deposits (Pig! 6). Except for unit 1, which showed disturbance at tht\ lowest leve!, the rest of the trenches were· undisturbed below 20 cm, providing reliable artefact and charcoa) samples. The materials collected were mainly of the' TIW tradition, with only small amounts in the upperl mos! levets. They included potsherds (458), live bead.c grinders, 10 lumps of daub, 50 fragments of slag, lQ fragments of tuyeres, 10 iron fragments and two coppet sheets. Two copper beads and two recent glass beads were also collected. Other linds comprised one marin~ Map 17. Masuguru si te map. shell (Terebralia palustria), and bones of birds and bovid. Charcoal samples were collected from the undisturbed layers. Two were selected for analysis. 21 and Pig. 27, Lab. Nos. Ua-2092, Ua-2094 and Ua- from trench 2, leve! 4, and from trench 4, leve! 4. The: 2093 respectively). The results are discussed in Chap results are discussed in Chapter 10 (Table 21 and Pig, ter 10. 27, Lab. Nos. Ua-2095 and Ua-2096). A survey away from the site located some second". millennium sites half a kilometre south-west of the före 7.4 Masuguru est in an area, in which maize was currently being cul'' · tivated. No work was done on these sites. Masuguru (6° 22'S., 38° 28'E.) is 55 km to the west of Bagamoyo (Map 11). It is the site of a present-day vil lage situated on a hill bordered by a man-made lake to 7.5 Kiwangwa the west, forest to the south and west, and open agricul tural land to the north (Map 17). The soils on the site The working trip in the 1990 season ended at Kiwan-, are leached, reddish-brown clays, bnt the soils of the gwa (6° 23'S., 38° 34'E.), 44 km to the west of Baga-c forested and agricultural area are dark, compact clays moyo (Map 11). This site is about 2 km west of the: and the !atter are used for growing maize and millet. present small town of Kiwangwa on the road to Masu-• Systematic survey of the site was hampered by the guru (Map 11). The few people living around the site houses. However, by augering along the streets and the are cattle-keepers of both the Wakwere and Masai eth-•': backyards of the houses, we were able to identify areas nic groups. with material concentrations, in which all four trenches The site is surrounded by patches of bushland and·· were excavated. The trenches, which all measured lx 1 !arge valleys with streams running to the south, west • > 59 ' ' 40E ' 60 , , , , , , / , :::;::.::,,.-·J -:;. /i ,," " ,,,, " t ' ' Dtrench 1 § trench 2 " ,,,, ,, lIIIJ trench 3 ,,,, I ,, w 1r:v contour (V .I. 0-5m) \\ A datum point ,, ",, 0 ,,,, ",, ",, "\\ " O 4 B 12 m 20 o~',:,,, ",,,, 1 20W 0 20 ~\ Map 18. Kiwangwa site. and north; these valleys are used for grazing. The farm the excavation was to go through the disturbed layer ing land is to the east and north-east. and to recover material from the undisturbed one. Unfortunately, the major part of the site has been All the material collected from this site was of the complete!y destroyed by quarrying, as it is located on first millennium AD. Finds included potsherds (452), lateritic soil with many quartz and ferrous concretions. one bead-grinder, one fragment of glass, 95 pieces of The quarrying activities have removed the top layer of clayish soil over a !arge area, in order to reach the sub swface grave!. A !arge number of mounds has been created, some of which may be seen in Map 18, and ,,,,,,,,,,,,,,,, ,,,,,,,,,,,,,,,,,,,,,,,, on!y one small area of 10 x 5 m remained untouched. ,,,,,,,,,,,,,,,,,,,,,,,,, Fortunately, this small area contained a concentration ,,,,,,,,,,,,,,,,,,,,,,,,, ,,,,,,,,,,,,,,,,,,,,,,,,, of cultural material providing the data presented in this ,,,,,,,,,,,,,,,,,,,,,,,, ·. work. ,,,,,,,,,,,,,,,,,,,,,,,, ,,,,,,,,,,,,,,,,,,,,,,,, The first trench was dug in the undisturbed area. The ,,,,,,,,,,,,,,,,,,,,,,,, stratigraphy varied from very dark-grey, compact, ' ,,,,,,,'\. · sandy c!ay to brown, compact, sandy clay at 40 cm ·; (Fig. 7). Other trenches were excavated in the disturbed scale~ 1:40 areas. The second trench was located in a piled-up • mound with the purpose of recovering some of the very dark gray clay quantities of pottery being eroded from it. A third trench was placed in the displaced, sub-surface soil very dark brown clay piled up on an undisturbed underlayer. The purpose of Fig. 7. The western profile of trench 1 at Kiwangwa. 60 '", \ 4, 0 4,5 ,, 3,5 " ' \ \ ',, '\ \:i \\ \ ' '2.._ contour (interval 0,5) '\ \ \ \ \ \ \ 0 \ \ \ I ~ path: 156m to the main road I \ \ \ \ I \ \ \ \ \ D excavated trenches 1 to 6 I ® \ \ \ \ \ \ \ 0 \ e test pits with high material concentration: \ \ ~ cSI.\ I \ 0 test pits with little material concentration'. I I I 0 testpits with no cu!tural material ~ .: " \ I I 40 \ 1c/DP3 .\ I \ 0 XDP1 datum points 1 to 3 \ 1,, I so I \ \ I I ., 't marginal areas of the lake \ \ \ \ \ \ \ \ \ \ o 10 20 30 40 som \ ® \ \ 60 \ \ 0 \ \ • \ \ \ \ I \ e\ \ ' I I I I I I ', \ I \ \ \ I \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ ' ' ', \ ', Map 19.. Misasa: elevation, excavation and test pits. slag and 11 fragments of tuyere. Charcoal samples Kiwangwa was another site earmarked for the Phase were collected from the lower levels of units 1 and 3, Two excavations. The site, which is 44 km from th~ lev els 5 and 7. The detailed results are given in Chapter shore, was found to have interesting local pottery with 10 (Table 21 and Fig. 27, Lab. Nos. Ua-2097 and Ua- convincing, EIW pottery elements. It also had early 2098). glass and copper beads found on the early littoral sites. The last day of the research was used to survey the The site may help to throw light on the link betwee{i vicinity of the site. About 1 km to the north-east, a the TIW and the EIW traditions, and the littoral an~ major geological feature, consisting of a steep escarp hinterland relations. The undisturbed area previously ment greater than 100 m in height, extends for several identified as having a material concentration was excar k:iloinetres from the south-west to the north-east. The vated further. · floor forms a gently sloping valley, trending north-east, A 2 x 4 m extension of the previous trench 1 was du& with a seasonal stream and forest. A survey on the to the west. Digging continued down to the 70-cnj lower and upper parts of the escarpment located a late leve!, where a sterile layer was encountered. A secon4 Stone Age site, probably with some pottery. No work 2 x 1 m trench was placed eastward of trench 3 and dug was done on this site. down to 70 cm below the surface. A tllird trench was all 10 20 30 40 50 60 70 80 tl,l~@ very dark gray sand with post 14th century finds black sand with concentration af TIW tradition remains mll!le gray sand with few remains Fig. 8. Misasa: trench 2 eastem profile. (Other walls are uni- ., - 10cm waJI remains f----1 form). > 61 ': Table 3. Amounts of remains collected from Misasa. Not all tbe upper Jayers had remains (trenches I, 2 & 5) . ._. Abbreviations: 2x = the extension of trench 2; Tr = trench, Lv = level; (a) = decorated sherds, (b) = undecorated sherds, {c) = bead grinder, (d) = tuyere, (e) · :::: -slag, (f) = iron fragments, (g) = iron hook, (h) = iron bangle, (i) = iron ring, G) = iron key-like object, (k) = copper beads, (1) = copper bangle, (m) = irnported ceramic, (n) =glass fragments, (o) =glass beads, (p) = otherbeads (bane, clay), (q) = shell, (r) =bane. Tr Lv a b c d e f g h k m n 0 p q r 0 1 4 23 60 8 9 1 5 82 267 11 1 1 2 6 97 344 4 2 2 2 7 43 103 10 2 8 2 3 2 3 6 22 2 4 20 96 2 5 218 704 6 15 22 5 2 5 6 51 125 1 3 1 7 5 22 1 1 8 2 12 2x 1 15 2 5 23 1 1 3 7 21 2 4 7 47 1 2 5 164 887 9 14 14 4 2 4 4 6 43 155 3 2 7 4 12 8 4 21 3 1 7 33 2 6 27 3 8 491 4 51 183 1 6 1 5 148 485 2 3 1 6 2 3 2 5 6 35 147 1 2 2 2 2 7 3 17 1 1 4 1 2 4 2 3 20 2 3 12 4 1 5 3 10 2 3 6 14 86 4 4 2 7 22 65 2 2 3 2 8 3 25 1 'd 2 3 9 2 3 23 95 3 3 ,, " 4 103 361 3 6 1 4 ';::11 5 4 9 0e.,stward 2 x 2 m extension of trench 1 and was exca from trench 2, leve! 6, are presented in Chapter 10 vated down to 30 cm (Map 18). (Table 21, Lab. Nos. Ua-2599 .and Ua-2598 respec _; The artefacts include those found in Phase Oue tively). ·· (meutioned above ). Local TIW pottery was the pre- dominant material. Of the 3377 sherds, 571 (16.9%) Were decorated. Four bead-grinders, 27 fragments of 7.6 Misasa tµyere, 198 fragments of slag, 19 fragments ofiron, six 9opper beads, six Sassanid sherds and three glass frag The Misasa site (7° 28'S., 39° 12'W.) is in the district of ments were recovered. The finds on this site clearly Kisarawe, about 85 km south of Dar-es-Salaam. It is demonstrate closer littoral-hinterland linl<:s in the mid accessible by the Dar-es-Salaam-Kilwa road, branch d!e of the first millennium AD than were previously ing off to the east at Nyarnironda. From Nyamironda, C:?nceived. The site has also yielded obvious transi the road continues through Lukanga to the village of ti9na1 pottery, shedding light on the link between the Misasa for about 6 km. A !arge lake, Zakwati, is found TIW and the EIW tradition (Fig. 16). close to this village (Map 10). The road from Misasa to Two more charcoal samples were co!lected for ana the coastal plain touches the southern horder of the lysis. One sample from trench 1, level 4, and another lake. Kisiju on the Indian Ocean littoral is situated 62 a b e k q p 0 I 2 cm. '=='--' 63 terland and the littoral. Secondly, the site is located in ' an area knowu to have EIW sites, a factor thought to be crucial for investigating the possible relations between .) the EIW and the TIW traditions. Thirdly, it was thought worth while to assess the extraordinarily early, absolute dates for the TIW tradition mentioned above. Fourthly, the site was reported as !arge and rich. 7 .6.1 Mapping and surveying The mapping of the site was complicated by thick ve getation, comprising both forest trees and abandoned, cashew-nut trees. It was difficult and time-consuming to survey the whole eastern side of the lake, as it first had to be cleared. However, moving northward along a narrow path used by the inhabitants, we managed to clear 300-m-long corridors al 30 m intervals, running perpendicular to the margin of the lake across the plain. Fig. 10. Grooved objects (potsherds, coral or sand stone) known as 'bead Auger and hoe pits were dug every 10 m along the cor grinders' found in the TIW sites. ridors dowu to 70 cm. In the lower leve!, cultural mate rial was found. A 1-km stretch of the plain was systematically surveyed in this way, and a further 1-km stretch northward was surveyed haphazardly. It was about 15 km to the north-east across the River Luhute found that the southernmost part of the plain had been (Rufute). The Zakwati lake, which is today an irnpor occupied, an area of about 4 hectares (Map 19). tant source of fish, is 4 km long and 0.5 km at its widest The site was then mapped and, following a grid with part. The TIW site was located at the south-east limit, 5-m intervals, augering was conducted to determine where the lake flows out ~· ·~~ ~ fA1 ~ vg r a b c d e f . . ' . . • n ng h \' Fig. 11. Assorted iron objects from TIW occupatid-. nal layers. a & d = points from Miasa, b & c = key-like objecJS from Misasa, g & h = rings from Misasa, i & j = '!f row heads from Mpiji, k = knife from Mpiji. --~ k tion and probably post-TIW-tradition, house construc bralia palustria), two pieces of coconut shell tion (Table 3). pieces of gum arabic were also recovered. Heavil A total of 5788 TIW potsherds was found in trenches burnt lumps of red soil and lateritic stones indica 1-5. One thousand, two hundred aud nineteen (21 %) of that iron ore was srnelted from red, lateritic soils. Simtc these were decorated (Fig. 9). Other artefacts inc!uded !ar ore has been thought to have been utilized at thy 28 bead-grinders (Fig. 10), 32 fragments of tuyeres, 83 Limbo, early-iron-using site (Chami 1988b), and at fragments of slag, two iron rings, four key-like objects, Kilwa (Chittick 1974, p. 236). one hook and many miscellaneous iron objects, rang ing from wire wound on a fibre core to iron knives (Fig. 11). Seventeen copper beads (Fig. 12), two copper Fig. 13. Assorted beads from TIW occupational layers. a = joined copper beads (?) from Mpiji, b = a tubelike copper bead (?) frotri bangles and two copper rings were also found. Other Masuguru, c = three conical caper beads from Misasa, Mpiji and Kiwangwa; artefacts similar to those found on littoral sites (cf. d & e = drawn bluish beads with gass bubbles from Misasa and Mpiji, f -=:' reddish simple wound bead from Misasa, g = green drawn bead without gasS Chittick 1974) were eight Sassanid sherds, seven glass bubbles from Misasa, h = drawn bead read outside and white inside front and stone beads and 45 pieces of yellowish-green and Misasa, i = carnelian sherd from Misasa, j & k = shell beads from Mpiji, 1 :::= bone bead from Mpiji, m = fish bone bead from Mpiji, n = clay bead fro~ colourless glass (Figs 13-14, Plates 1, 2). A few, heav Mpiji, o = round clay object from Misasa. ily weathered, bird and fish bones, a rnarine shell (Tere- (For glass bead identification see V an der Sleen 1956). '"t I 65 c- - 0• a b c ' !' I d I I I ·i::- ''·' .:;;,; i. - - i' . I ll Fig. 12. Copper/bronze and lead objects from TIW occupational layers. a = copper bangle from Mpiji, b :::: bronze (?) ring from Misasa., c = copper ring from Misasa, d :::: cop I per chain from Mpiji, e & f = lead objects from I ! 0 Mpiji. I e f I I e I 1-11/ 111. . . . a 0 b ~ "c d @ e I! • 0 l l I ~ I I I @ f g 0 h I - I . t?\ . . --fI '...::.# . I 0 ~ k m n O' 0 0 > 66 ~.•.·...• ' Plate 1. The first two sherds are green/blue decorated shoulders (?).The t.hU1i ·'. fragment is a green/blue base sherd. The frrst object in the second row is 4 greenish flask and to the right are early alkaline green/yellowish and b1ue glass sherds. , .<. I Plate 2. V arious types of beads found in th~ , TIW sites. In the first row, the first two are drawn green beads, the third is a red wound bead, the fifth and sixth are drawn green and read wittj· white inside beads, and the last is cameliarl,: In the second raw, the second and the thir'4 are made of bone, then there are two of cop~ per, and the last is made af clay. Below arifi shell beads. > 67 b a e d Pig. 14. Sassanid objects and glass ware from Early TIW occupational layers. sherd (Plate 1), d = a bottom of a flask (Plate 1), e = rim/neck af a greenish a & b = Sassanid decorated shoulder ? sherds (Plate 1 ), c = a Sassanid base glass bottle. 68 Table 4. Bones collected from Mpiji (include Crustacea). Table 5. Shells from Mpiji and Changwehela. Species No Species No Mammals A.MPUI Dik-dik 10 1. Telebralia palustris L. 2141 Neotragus moschatus 23 2. Melongenidae (chank shells) 4139 Bovid 18 3. Pleuroploca trapezium L. 126 Cercopithecus sp. 5 4. Chicoreus ramosus L. 35 Carnivora ( cat) 1 5. Anad.ara antiquata L. 49 Dugong 5 6. Olive shells 101 Cephalophus sp. 2 7. Land snail (Achatina africana) 9 Ovicaprid 12 8. Cypreacis rufa 11 Suid 1 9. Cowry shells 40 Whale 1 10. Conus 1 Reptiles 11. Lambis lambis 4 Crocodile 8 12. Nerita sp. 16 Turtle 365 13. Moon shell 13 Crustacea 14. Donax sp. 2 Crab 42 15. Macona sp 4 Birds 16. Pinctada sp. 8 Chicken . 1 17. Stromhus sp. 4 Others 13 Fish B. CHANGWEHELA Lethrinidae (changu) 121 1. Telebralia palustris 74 Rock cod (tewa) 2 2. Melongenidae (chank shells) 12 Sparidas (sea bean) 4 3. Oysters (rocky) 2 Parrot fish (scarus sp.) 3 ludet 126 Other fish (unidentified) 764 .:C Several fragments of charcoal were recovered from highest material concentration and hence indicated tli~ each trench. Four sarnples weighing over 3 g were period of major occupation on the site. The results of selected for c14 analysis. One was from trench I. leve! the above charcoal analyses are presented in Chapt~k 80 representing the earliest occupational layer. Two 10 (Table 21 and Fig. 27, Lab. Nos. Ua-2593, Ua-2594, were from trench 2, levels 7 and 5 and one from trench Ua-2595 and Ua-2597). · 3, level 5. Leve! 5 in the first four trenches has the t: i, I 8. POTTERY ANALYSIS: DECORATION AND SHAPE 8.1 Introduction include a line of punctates (Fig. 9:k, p & t), comb stamping (Fig. 9:n, s & u), bands of cross-hatching Three important aspects must be noted before we (Fig. 9:e), horizontal and oblique incisions (Fig. 9:d, i embark on tbis analysis. Firstly, in Chapters 6 and 7, it & m; Plate 3; Fig. 16a and Table 18, design element a). has been shown that pottery is by far the most frequent Bevels, fiutes and thickened rims appear frequently in artefact on the TIW sites. Its type of decoration is by the TIW tradition (Fig. 9:e, d, h & m, Fig. 16b; Plate 3). now the only outstanding identifier of the expansive Typical decorative elements in the TIW tradition, i.e. tradition which I have called TIW in tbis work. The triangles (Fig. 9:a, o ), and double lines of punctates/ question of whether pottery can be used to identify a stamps (Fig. 9:v) have also been found in the EIW tra tradition or a cultnral or ethnic group has been dis dition in small amounts (Fig. 15a:c, e, 1 & r respec cussed in Chapter 2. What is essential in this chapter is tively). Tpe recognition of such elements shared the fäet that from the 4th to the lOth century AD, the between the two traditions has led to the analysis of the whole coast of East Africa and its hinterland were TIW pottery as a means of discerning the cultural affin occupied by communities who were producing similar ities along the coast of Bast Africa both diachronically kinds of pottery. Whether these communities belonged and synchronically. to a similar ethnic group or not, the important aspect is that they had some cultural similarity or economic con tact. Orre of the principal aims of tbis work has been to 8.2. Analytical attributes study the decoration, forms and paste of this pottery, in order to obtain information ou its users' origins, spread The analytical attributes of archaeological artefacts are and degree of cultural and economic affiliations. different, depending on the kind of object analysed, Secondly, much work has a!ready been reported on whether stones or pottery. The attributes range over the iron-working sites pre-dating the TIW sites on the "the material artefacts made, size, shape and surface coast (Soper 1967a; Chami 1988a, 1988b, 1992b; Sin treatrnent" (Doran & Hodson 1975, p. 99). In pottery clair 1987; Morais 1988; Sinclair, Morais, Adamowicz analysis, many aspects of the above attributes have & Duarte 1993; Cruz e Silva 1977). The pottery found been suggested, ranging over thickness, surface condi at the coastal EIW si tes (Kwale ware ), dated to the tion, weight, colour, temper, dimensions, shape, deco early centuries AD, has been analysed and compared ration, surface finish, etc. (Shepard 1963; Hulthen, with its counterparts from the hinterland sites (Soper 1974). Different archaeologists have, however, used 197lc; Sinclair 1987; Morais 1988; Chami 1988b). different attributes in their studies, depending on the Thirdly, the main decorative motifs for K wale ware nature of their problems. In relation to cultural affinity, are said to include bands of comb-stamping (Soper for instance, decorative techniques have been found to 1967a, pp. 5-13, Fig. 2b, c; cf. Fig. 15a:h in this work); be "less informative" (Collett & Robertshaw 1983, p. oblique incisions (Fig. 15a:a, g); herring-bone pattems 109). (Fig. 15a:d); horizontal incisions (Fig. 15b:j, q); cross Consequently, in giving a better picture of the hatching (Soper 1967a, pp. 9, Fig. 6b); and punctates/ ancient Nubian pottery and its culture, Nordström stamps (Fig. 15b:l, r). Punctates are also used to delimit (1972) dealt with the attributes hierarchically, employ other bands (i.e. Figs 15a:f and 15b:j, n). Bevels and ing as many as he could identify. They included surface fiutes on the rims or shoulders appear on most pots of properties, technical properties, classification of the the Kwale tradition. Otherwise, the rims are tbickened. fabrics, vessel shapes and decoration. In an attempt to Bevels are more confined to the rims (Fig. 15a:d, f, g & clarify the EIW pottery of East Africa, Soper (197lb) h), while there are fiutes to both rim and body (Figs used an eclectic approach, by selecting 50 attributes 15a:a and 15b:i, j, p & r). (traits) ranging from "vessel shapes, rim morphology, Some of the above EIW decorative elements appear bases, finish and decorative techniques and motifs" (p. in different frequencies on the TIW pottery. They 16). In trying to classify the pottery of southern Africa, 70 b a e 0 2CM Fig. 15a. Typical EIW pottery from the site of Limbo. -~ 8.2.2 Decoration Huffman (1976, 1980) concentrated more on the struc -:} ture of motif and decoration placement. In differentiat Different terminologies have been used in differerit ing the Pastoral Neolithic pottery at the inland sites of works for decorative structures appearing on pottery. Bast Africa, Collett and Robertshaw (1983) have dealt These comprise motifs (Huffman 1976, 1980; Collett more with the decorative motifs and shape. & Robertshaw 1983; Soper 1971b); decorative patterns For this work, I have hierarchically selected four (Nordströrn 1972, pp. 76-7); structure of desig:µc attributes: decoration, shape, and motif placement (dis (Hulthen 1974, p. 25); and decorative format (Sinclair· cussed in this chapter) and fabric (discussed in Chapter 1987, p. 164). All !hese terms, however, seem to con 9). I have chosen !hese attributes for analysis below, note one idea, expressed by Nordström (1972, pp. 76"- because I am convinced that they can shed more light 7) as patterns which are either impressions or incisions on the questions pertaining to the origin and spread of that can be either single or in bands or both in combinat the TIW tradition. tion. 71 0 2cm. I ,- ,- m " ~ Fig. 15b.Typical EIW pottery from the site of Limbo. Hulthen (1974, p. 25) has used the term "element" to levels, in order to find ont when a particular design ele mean those independent, diagnostic patterns in a com ment was introduced into the site and its distribution bination motif, for example, a horizontal, zigzag line over the region. and a vertica!, rectilinear band filled with a rhombic, A total of 42 independent design elements (Fig. 17) thecker pattern. was identified. These comprised oblique incisions, '' Recognising that many motifs in the TIW assem horizontal incisions and incised or starnped triang]es, blage are such combinations of elements (Plate 3), the regular and irregular comb-stamps, lines of punctates, recording of vatious motifs went hand in hand with the lip decoration, bevels and flutes. Elements such as isolation of elements from the combinations (Fig. 17). thickened rims, bumishing and appliques were also These elements were then subjected to proportional included in the list. åna!ysis. One aim was to study the frequency of diffe The numbers and percentages of decorated sherds for rent design elements at different sites and in different the five TIW sites are shown in Table 6 and Fig. 18. > 72 a a ' ' b b I c Fig. 16b. TIW (transitional JX>Uery with clear EIW elements i.e. bevels antJ :flutes. - a. Kiwangwa R!N, b. Kiwangwa R/N/SH, c. Kiwangwa R/N. Abbreviations: N = Neck, R =Rim, SH = Sboulder 0 2 ' Fig. 16a. 1TW (transitional pottery with clear EIW elements Le. bevels and flutes. tered and compared (Table 7). Fora single line of punc; a. Misasa N/SH, b. Misasa R/N, c. Kiwangwa BO, d. Kiwangwa graphited tate the results in Table 8 were found · bowl fragment. It should be noted that this design element witj Abbreviations: BO = Bowl, N = Neck, R =Rim, SH :::: Shoulder mostly functional and was used to bound decorativ~ motifs in a band (see Plate 3, 3rd row, last sherd). It hat! the same function in the EIW pottery (Fig. 15), oq The Misasa site was found to have the greatest percen which it also has the highest decorative frequencj( tage of decorated sherds, followed by Masuguru, and (Charni 1988b). the Kaole site had the smallest percentage of decora If we put aside pnnctates as a design element o~ tion. It should be noted !hat Misasa has early dates in account of their functional purpose, triangles would be the 4th to the 6th century and Kaole in the 9th century the most prominent design element, as shown in Table AD (Chapter 10). This could mean that the earlier the 9. In fäet, triangles would still have the highest fre" site, the greater is the percentage of decoration. The quency at Kaole, even when a line of punctates is giveii high percentage of decoration for Masugnru is proba equal weight. Masuguru being in the furthest part of bly another indication that the hinterland sites had the hinterland (55 km from the shore) studied, thi~ greater arnounts of decoration. result indicates that the triangular design may hav~ The frequencies of 42 design elements were regis- been more popular at the hinterland sites. The same 73 Flate 3. TIW potsherds with motifs of elements combination. Observe, for Bevels in sherd two, row three, combined with wider cross hatching; in sherd example, how following elements are combined differently in different five, row three, with EIW band of oblique/vertical incision bound by puncta sherds. tes. Double stabs (horizontal) in sherd one, row one, combined with applique and Narrow cross hatching in sherd two, row two, combined with roulette; in zigzag double incision; in sherd two, row one, cornbined with rim thickening sherd two, row four, with oblique lines and sherd three, row four, with triang and triple (or more) zigzag incision. les. observation was recorded for pottery from Morogoro A few other motifs are quite common, including a sites about 200 km from the shore (Kimathi, personal band of oblique incisions (continuous or non-contiuu communication). On the other hand, the highest fre ous) and double-zigzag incisions (Table 11). Ob!ique quency of triangles observed (at Kaole) indicates that and double-zigzag incisions read high, next to trian in the late dates of TIW tradition triangles became the gles, at Kaole, indicating that, as the tradition faded most dominant marker of the tradition after shedding away, the three design elements became dominant. The many of the EIW pottery elements (see below). frequency of the other elements can be seen in Table 7. Other outstanding design elements are lip incisions In the analysis, the design elements were also or stabs (which tend to be placed on the lip bevels), divided into two categories: those thought to be more bevels, flutes and thickened rims. These are shown in affiliated to the EIW tradition and those typical of the · Table 10. It shonld be noted that the last three elements TIW tradition (Table 7). above are closely associated with the EIW tradition The results showed that Misasa and Kiwangwa have (Pigs 15 and 16). Their more frequent appearance at the more affiliations to the EIW tradition than the other sites found to have early occupatioual dates, for exam three sites (Table 12 and Pig. 19). Indeed, several typi ple, Misasa, and their reduction at the later sites, for cal EIW potsherds (Pig. 16 and Plate 3) have been · example, Kaole, indicates a chronological relationship recovered from the two sites. Kaole shows little con between Early TIW sites and the EIW tradition. nection with the EIW tradition, an observation justified > 74 MOTIF SIGN MODEL DESCRIPTION m branch-!ike incision a l/////////!t oblique incision n archical incision b ~ zigzag double incision 0 tlutes c &A 6. incised triangles I triangles filled with punctates p narrow cross hatching d bevels q singla zigzaz incision horizontal incision • wider cross hatching \/ /////// /\ ,,,,,,, applique s laddar incision ...... g duble stabs :-::.::: .. · ...... irragular stamps h lOOC>C 11111 • • • •• lip decoration u 111111 bold incislon ...... punctates V \/ singla oblique inclsion incised pendant triangles w ;·.If)-::: irregular incision k comb-stamps X 11 Il 11 11 double square incislon burnish y wedge nicks Fig. l 7a. TIW design elements. ' ' ' ' ' z ~ ~ shell impression by its post-8th-century dates (Chapter 10). The indica tion is therefore that the earlier the site, the more EIW Pig. l Th. TIW design elements. design elements it has. The same pictnre was probably obtained again when the analysis was repeated at those sites with thicker, cnltnral-material levels, i.e. Misasa ing Shepard (1963), developed one that recognizes two and Mpiji. The analysis leve! by leve! indicated that the categories of restricted and unrestricted vessels, eithf?t earlier the leve!, the greater is the number of EIW pot jars (pots) or bowls. The contours of various parts öf tery elements that it has (Figs 20 and 21). the vessels, for example, shoulder and neck, are impor; tant in determining in which category a pottery vess~l should be included. This particular mode! has be~h 8.2.3 Shape adopted by many schalars (Hulthfo 1977, p. 2Q; Several models are used for determining the shapes of Holthoer 1977, pp. 43-55; Sinclair 1985, p. 8, an!! pottery vessels. Nordström (1972, pp. 68-74), follow- 1987, pp. 164-7; Morais 1988, pp. 63-4). · In eastern and southern Africa, students of EIW po~ tery have used other types of models obtained froi)i Table 6. Percentages of decorated sherds. pottery reconstruction. Soper (1971c) gives a mode! Si te Total No.of deco- Percentages made up of necked pots, narrow-mouthed, globulåf sherds rated sherds (%) vessels, rough, open bowls, carinated forms, and triar\ Misasa 5 788 1 219 21.0 gular handles or spouts. Sirnilar models can be seen iU Mpiji 5 538 937 16.9 Kiwangwa 3 377 571 16.9 Collett and Robertshaw (1983, p. 109) and Huffma)i, Masuguru 458 96 20.0 (1980). Kaole 545 48 8.6 The analytical results obtained from the use of 14e 75 above two types of models, however, do not differ. In aa )· .. .. ~ neck stabs using the former mode!, according to Sinclair (1987, p. 165): bb \)(()(/ irregular obllque incision The categories are useful in that, even when working with severely 1 fragmented assemblages, they provide a similar range of shapes to cc ::I11 horizontal/vertical incision those mentioned by other worke_rs in the literature, i.e. Phillipson (1977), Huffman (1974b) etc., while avoiding some of the problems resulting from working with reconstructed vessel counts. dd triangle and horizontal incision I have adopted the former mode!, in order to be able to identify shapes from the fragmentary assemblages col thickened rims •• lected from the live sites. X The concept of restriction means t:hat "the maximum ff L\ triangles in horiz. incision diameter (major point) lies on the body, dividing the latter into two parts, the lower and the upper body. The gg triple or more zigzaz incislon rim diameter is always smaller than the maximum diameter... " (Nordström 1972, p. 71) (Fig. 22:a, b). hh vertical double stabs Unrestriction means that "no part of the body is wider than the orifice, and the maximum diameter (major point) is measured at the rim" (Nordström 1972, p. 71; .. oblique double stabs ii Fig. 22:d). In the restricted-vessel category, the TIW assem jj ·. oblique singla stabs blage has necked vessels with an inflected contour and vessels with a composite contour (jars) (Nordström kk roulette 1972, p. 71). These are recognized in Sinclair (1985, p. 8) as independent, restricted and dependent, restricted, respectively (after Shepard 1963). Restricted bowls Il lÅ ffÅ jqined incised quadrangles with a simple contour have also been identified. These types of vessels are illustrated in Fig 24i:a, b & c. mm 111111111111111 vertical incision In the unrestricted-vessel category, the TIW assem .... -· ...... blage has been found to have vessels with a simple nn EIA horizontal incision contour (open bowls), as seen in Fig. 24ii:d) (see Nord ström 1972, p. 71; Sinclair 1985, p. 8). 00 EIA cross hatching From the shape models (Fig. 22), therefore, seven strnctural-shape categ'?ries of potsherds were recog nized (Sinclair 1985, p. 8; 1987, p. 165; Table 13). EIA oblique incision pp These categories are as follows: Category 1 could be Fig. 17c. TIW design elements. derived from any vessel, and category 3 could be Fig. 18. Intersite comparison of the occurence of decorated sherds. Mlsasa Mpijl Klwangwa Masuguru Kaole si tes > 76 Table 7. TIW design elements: EIW and typicalTIW elements separated and compared. Misasa Mpiji Kiwangwa Masuguru Kaole freq % freq % freq % freq % freq % EIW-elements d 94 7.6 72 5.8 35 7.6 5 2.6 4 7.7 e 72 5.8 46 3.7 5 1.0 5 2.6 251 20.5 224 18.0 103 22.5 58 30.2 8 15.4 k 6 0.4 2 0.1 3 0.6 2 1.0 m 2 0.1 2 0.1 0 25 2.0 3 0.2 4 0.8 p 2 0.1 2 0.1 1 0.2 dd 17 1.4 4 0.3 2 0.4 ee 80 6.5 132 10.6 34 7.4 5 2.6 nn 2 0.1 00 6 0.4 pp 2 0.1 3 0.6 Total 551 44.5 495 39.4 190 41.3 75 39 12 23.1 TIW-elements a 76 6.2 123 9.9 52 11.4 23 11.9 7 13.5 b 60 4.9 42 3.3 55 12.0 14 7.2 9 17.3 c 157 12.8 173 13.9 61 13.3 51 26.5 12 23.0 f 1 3 0.2 5 1.0 g 46 3.6 29 2.2 5 1.0 h 124 JO.I 158 12.7 21 4.6 2 1.0 j I 0.2 1 39 3.1 99 7.9 23 5.0 6 3.1 1 1.9 Il 3 0.2 3 0.2 q 45 3.6 18 1.4 14 3.0 9 4.6 s 18 1.4 3 0.2 12 2.6 4 2.0 1.9 t JO 0.7 2 0.4 3 1.5 u JO 0.7 2 0.1 6 1.2 V 13 1.0 18 1.4 5 1.0 4 2.0 2 3.8 w 3 0.2 5 0.3 1 0.2 I 0.5 X 3 0.2 2 0.1 y 0.2 z 8 0.6 1 1.9 aa 3 0.2 I 6 11.5 bb 6 0.4 cc 5 0.3 ff 5 0.3 gg 26 2.1 43 3.4 8 1.7 bh 5 0.3 6 0.4 9.0 ii 5 0.3 5 0.3 4 0.8 jj I 0.0 kk 2 0.1 11 2 0.1 mm 4 0.3 3 0.2 Total 672 53.1 745 58.8 266 55.4 117 60.8 40 76.9 derived from either independent, restricted j ars or category 6 from restricted bowls and T from nnrJ" restricted bowls. Category 2 is derived from restricted stricted bowls (Fig. 22). jars, either dependent or independent. However, cate- For the purpose of comparing the frequencies of thp gory 4 can only be derived from independent, restricted four types of vessel, therefore, only the last four cat~ vessels, category 5 from dependent, restricted vessels, gories are important. But, for comparing jars and ''\ Table 8. Decorative frequency of a line of puuctates. Table9. Decorative frequency of triangles. Site Freguency (%) Site Frequency (o/o) Misasa 20.5 Misasa 12.8 Mpiji 18.0 Mpiji 13.9 Kiwangwa 22.5 Kiwangwa 13.3 Masuguru 30.2 Masuguru 26.5 Kaole 15.4 Kaole 26.9 ) 77 Fig. 19. Intersite EIW & TIW elements compared. Mlsasa Mpljl Klwangwa Masuguru Kaole ~tes 50 • TIWelem's - EIW elerr(s Pig. 20. Misasa EIW and TIW compared. 8 7 6 5 4 levels cl l 100 fra::iuency "!.. I 75 .50 25 0 Fig. 21. Mpiji EIW and TIW compared. 10 9 8 7 6 5 4 levels Table 10. Decorative frequency of lipincisions, bevels, flutes and thickened rims. Table 11. Decorative frequency of oblique incisions and double- and single-zigzag incisions. Site Frequency (%) Oblique Double Single Lip inc./st. Bevels Flutes Thick. rims Si te inc. zigzag inc. zigzag inc. l'vlisasa !OJ 7.6 2.0 65 Misasa 6.2 4.9 3.6 Mpiji 12.7 5.8 0.2 10.6 Mpiji 9.9 3.3 1.4 Kiwangwa 4.6 7.6 0.8 7.4 Kiwangv.ra 11.4 12.0 3.9 Masuguru LO 2.6 2.6 Masuguru 11.9 7.2 4.6 Kaole 3.8 Kaole 13.5 17.3 78 Rim orifice plain Rim Neck Neck inflection point SHoulder major point -4"---"m"'a'°'xi,.,m'-"""m'--d'"i"am....,,et"'e'--r _ major point_.__~m~•x~i~m~"~m_d~ia~m~•~'~•'~ BOciy BOciy BAse BAse a b Rim SHoulder major point Ri~m~~m~a=x~im="m~d=i•=m~•='~"'~ orifice plain major point,_._,m,,,a,,xe:im"""'m"--'d"ia=m~e"te,,_r_~ B0dy BOciy Fig. 22. Pottery vessel shape models. ;: a = independent restricted vessel, b. depen(fo' BAse BAse restricted vessel, c = restricted bowl, d = Q;. bowl. c d ft.. eq.::;::LI""'l'o:_:;'-.%:______-, 100r Fig. 23. Intersite pottery shape cornparison. sltes Table 12. Inter-site EIW and TIW comparison of decorated Table 13. Categories nsed to determine pottery forms. elements. Abbreviations: R=rim, N=neck, SH=shoulder, BO=body and BA=l/ Site EIW a:ffiliates Typical TIW Category 1: R; BO/BA; BO; BA. % % Category 2: RJN; N. Category 3: SH/BO/BA; SHIBO; SH. -'-: Misasa 41.4 59.5 Category 4: R/N/SHIBOIBA; R/N/SHIBO; R/N/SH; N/SHIBOIBA; NISJlf Mpiji 37.9 62.0 BO;N/SH. . Kiwangwa 39.8 56.5 Category 5: R/NIBOIBA; R/NIBO; NIBO/BA; N/BO. Masuguru 36.6 63.4 Category 6: R/SHIBO/BA; R/SHIBO; R/SH. Kaole 19.2 80.8 Category 7: R/BO/BA; R/BO. 79 ------ b ~~----- 34cm ---~~ ------~------______I~ d , fig. 24i. A reconstruction of vessel forms of TIW sites. a = EIW jar from Misasa, b, c & d = jars from Mpiji, Kiwangwa and Masuguru respectively. bowls, category 2 can be included. Category 1 is less Fig. 23 were obtained. The tendency was for the earlier informative, since all the types of vessels have rims, sites, i.e. Misasa, to have a greater frequency of pots bodies and bases. As regards the TIW pottery, all the and the later ones, i.e. Kaole, to have a smaller fre bases found are round. In the followiug comparative quency. iµtalysis, although data from all the categories are Iisted A reconstruction of the vessel forms (Fig. 24i & ii) (Table 14), only categories 2, 4, 5, 6 and 7 have been was also made to determine the morphology of the ves :\.\sed to compare the frequeucies of pots and bow Is for sels. Measurements of rim diameters and heights of (\\ffereut sites (Table 15). TIW vessels were analysed. It became clear that ves- When categories 2, 4, 5, 6 and 7 in Table 15 were ilsed to find the vessel-type frequency, the results in Table 15. Frequences and percentages of pols and bowls Table 14. Number of potsherds from different vessel ca from different sites. tegories. Site Pors Bowls Site Cat. 1 Cat.2 Cat 3 Cat.4 Cat. 5 Cat. 6 Cat. 7 Freq. % Freq. % Mlsasa 177 710 282 103 10 33 49 Misasa 823 90.9 82 9.0 Mpiji 11& 624 120 74 2 39 91 Mpiji 700 84.3 130 15.6 Ki,wangwa 65 358 84 51 2 35 54 Kiwangwa 411 82.2 89 17.8 Masuguru 64 134 23 5 2 3 22 Masuguru 141 84.9 25 15.0 Kaole 126 65 6 8 8 13 Kaole 73 77.6 21 22.3 •'i 80 -=:--=-.:::-.=:Jib;;cl!!------+------~=====::::1 ! I I 1i 11s,s cm J \ I I \ I I ' , I / / ' ...... ____ ,!.. ______,' a b .:.1· Fig. 24ii. A reconstruction of vessel forms of TIW si tes. a = terminal TIW bowls from Masunguru and Ka~le respectively, d = 12th-13th century ve-.· jar with elongated and flared neck from Kaole, b & c = open and restricted fromKaole. sels with EIW pottery elements (Fig. 24i:a) tend to be 8.2.4 Placement oj design elements more globular, the widest part of the vessel (major point) being larger !han !hat of the rim, and the height After getting some knowledge of the TIW pottety of the vessels being greater !han the major-point dia decoration and forms, it was thought to be worth majt, meters. However, for typical TIW vessels from Mpiji, ing an analysis of where various design elements we~e Kiwangwa, and Masuguru (Fig. 24i:b, c & d), it was placed on the vessels. The main pmpose of this ex(\r" very difticult on some vessels to deterrnine whether the cise was again to show how far the design mentalitie . major point was at the rim or on the body. The vessels of different communities in the TIW tradition wef acquired a bag-like shape (Chittick 1974, p. 320), the similar or dissimilar. neck curvature being so slight !hat it was difficult to Huffman (1970, 1980) and Collett and Robertsha\\i observe. (1983) have analysed the placement of decoration fro' The Kaole vessels represented a fading out of the reconstructed sherds. The use of the vessel categori~s TIW tradition. Necked vessels tend to be transformed discussed above (Fig. 22 and Table 13) has, however, into wide-open jars with the major point at the rim, and been found more effective for determining the placi;-: the rims being obliquely everted (Fig. 24i:a; see also ment of decoration from even fragmentary shertls Chittick 1974, p. 339, type 2 vessels). At the beginning (Morais 1988, p. 119). After using a similar mode! ~o of the Swahili tradition, probably in the 12th-13th cen deterrnine the shapes of the vessels, the same mode! is turies, a squat pot was introduced, which is more like a used to determine placement, as shown below. frying-pan !han a pot (Fig. 24ii:d). Design elements from Misasa, Mpiji and Kiwangwa 81 Table 16. Misasa, frequency of placement of design Table 17. Mpiji, frequency of placement of design elements. elements. Abbreviations: SH =shoulder, Ind. rest. =independent restricted·, Dep. rest. Abbreviations: SH = shoulder, Ind. rest.-::::. independent restricted, Dep. rest ::: dependent restricted, R. bowl = restricted bowl, 0. bowl = open bowl. = dependent restricted, R. bowl = restricted bowl., 0. bowl = open bowl Design lnd. Dep. R. 0. Design Ind. Dep. R. 0. elem. Rim Neck SH rest. rest. bowl bowl elem. Rim Neck SH rest. rest. bowl bowl a 76 a 123 b 55 1 3 b 34 8 c 78 41 22 c 149 7 17 d 94 d 72 e 72 e 46 f 1 f 3 g 46 g 20 6 h 124 h 158 2 246 2 221 2 j i k 6 k 2 1 21 18 1 10 19 60 ID 2 m 2 n 2 1 n 3 0 20 5 0 3 p 2 p 2 q 45 q 8 42 4 r 20 16 2 35 2 s 2 8 2 t u JO u 2 V 13 V 18 w 2 w 2 3 X 3 X 2 y y z z 3 5 aa 3 aa 1 bb 6 bb cc 5 cc I dd 2 15 dd 4 ee 80 ee 133 ff 5 ff gg 18 4 4 gg 30 12 hh 4 hh 6 :': ii ii 5 jj jj kk 2 kk u 2 il mm 4 mm 3 .iin 2 nn 00 00 6 /PP pp 2 jvere subjected to this analysis. These sites, having so only slightly were recorded in the category which i'eceived much more field attention, also yielded much they covered most, i.e. neck or shoulder. fuore pottery. The design elements were recorded It was quite interesting also to note that bowls of according to their placement on every sherd (Tables 16, both categories were virtually undecorated, except for 17 & 18). For design elements a, b, c, etc., see Pig. 17, burnishing. On the other hand, burnishing was app!ied and for the shape categories, see Pig. 22 and Table 13. specifically to bowls, except for a few cases at Mpiji, The general results indicate that, apart from the rim where pots were burnished. And, except for a few cases specific elements, i.e. lip incision or stamps, bevels and at Misasa, comb-starnping was applied on restricted rim thickening, the general tendency of the artisans of bowls. the TIW tradition was to decorate on the neck and It is apparent that the decorative difference between slioulder. In fäet there was a good nurnber of design pots and bowls indicates a functional difference. It is elements that fell between the neck and the shoulder. likely that bowls, being small, would have had rela When this was clearly seen, they were recorded in cate tively small areas for incisions and stamps and, being gory 4-independent, restricted pots. Those which did used for table purposes ( eating and drinking), would 1! 82 Table 18. Kiwangwa, frequency of placement of design look more beautiful when burnished and were possib\y' elements. easier to wash. With the knowledge derived from beau Abbreviations: SH :::: shoulder, lnd. rest. = independent restricted, Dep. rest. tiful, imported plates and bowls, burnishing wouJ :::: dependent restricted, R. bowl:::: restricted bowl, 0. bowl:::: open bowl. probably have been used to compensate for the glazfug Design Ind. Dep. R. 0. technique. At the same tiine, island settlements, {e. elem. Rim Neck SH rest. rest. bowl bowl Kilwa and the Comoros, were red-slipping their bowl a 52 (Chittick 1974; Wright 1984). b 43 9 1 c 26 35 d 35 e 5 f 5 8.3 Conclusion g 5 h 21 The extreme differences observed between the Misa~ 4 97 2 i 3 and the Kaole sites in both decoration and vessel fonh k 3 indicate a chronological difference, the former offeriii 1 14 7 m a liuk with the earlier EIW tradition in the 4th and th n 5th centuries AD, when pols were more globular $ 0 3 decorated (Chami 1988b), and the latter indicating th p q 7 7 fading away of the TIW tradition, when the vessdl r 2 were less decorated and less globular (Fig. 24; Chitti s 10 2 t 2 1974, p. 320). Ceramic finds from the remaining si\~ u 6 of Mpiji, Kiwangwa and Masuguru seem to be cor1'e~ V 5 w lated with each other, indicating the period when tji X TIW tradition was at its peak in the 7th century. How y z ever, they seem to have much in common with the ei\r. aa lier site of Misasa in terms of the EIW pottery elemen' bb and iinported materials. The chronological aspects al cc dd 2 discussed in detail in Chapter 11. · ee 34 ff gg 5 3 hh ii 4 jj kk 11 mm '' i i nn !i1 00 pp 3 > 9. POTTERY ANALYSIS: FABRIC The purpose of analysing the pottery fabric was to grave!, including lateritic concretions and felspar. obtain more information on the type of clay and temper Some of the tempering material was so coarse that, used, in order to assess the problem of the pottery trade fälling out of the sherds, it leaves holes which can discussed in Chapter 2. Some elements of decoration sometimes be confused with decorative stamps. and pottery quality have been used below to supple On the other hand, in some parts of Masuguru, the ment the fabric studies in understanding the problem. soils have deposits of coarse buff quartz in association · The most advanced methods of investigating pottery with chunks of black and dark-grey, shining, graphitic paste are petrographic microscopy and thermal analy particles. Large pieces of graphitic stones were also s\s. These have been discussed in detail in Hulthen recovered from the archaeological contexts, indicating 0977, pp. 20-1) and Lindahl (1986, pp. 28-31). In a common use of the mineral and hence a possible eastem Africa, pottery from Mozarnbique and Zimba source of the burnishing graphite. The paste of the pot bwe have been subjected to such analyses (Hulthen tery from this site has all the above components, 1988). Similar methods have been applied to pottery including particles of graphite. from the five sites discussed in this work, together with !)lat from the Limbo EIW site reported elsewhere 9.1.2 The quality ofthe sherds (Charni 1988b, 1992b). The results are reported by Lin- 4ah1 in section 9.2. The quality and strength of the sherds were also exarni ned. Some sherds from Mpiji and Misasa were pro duced from fine clay and were so well fired that their 9.1 Initial field and laboratory observations hardness is half that of porcelain. They arnount to about 10% of all sherds in Misasa and 8% in Mpiji. This type Same few observations were made initially in the field of highly fired and hardened pottery was not found in and later in the laboratory, before submitting a few Kiwangwa, Masuguru, and Kaole, indicating that it samples for petrographic studies. A magnifyiug glass was a local technological innovation in some areas of )Vas used for laboratory observation. This method was the coast. The fäet that the decoration and forms of the pcecessitated by the fact that, as we moved from one site pottery are of the TIW tradition argues against the idea \O another, the pottery fabric seemed to portray diffe that they were imported. rent featnres and qualities. From the preliminary analysis, it was concluded that, in most cases, the TIW pottery at different sites was produced locally. The paste of the pottery from the four 9.1.1 Paste sites showed local particulation, which strongly sup Misasa, Mpiji, and Kaole sherds were found to have ports the idea of local manufacture. To check on this been tempered with very fine sand. In the pottery from hypothesis, several sherds from different sites were Kaole and Mpiji, the sand is of a glittering type, which submitted for petrological analysis. In the following rriay be beach sand. A few sherds from Kaole and Mpiji section, Dr. Anders Lindabl presents the result. Were also decorated with shells (design element z in Fig. 17; cf. Fig. 25), a confirmation that they were ptobably produced close to the shore. The two interior sites of Kiwangwa and Masuguru 9 .2 Microscopic analyses (by Anders offered an even better understanding of the source of Lindahl) the paste material. Kiwangwa, for instance, is sitnated 9 .2.1 Introduction and formulated problems man area rich in coarse buff quartz, most of it attached to lumps of red, lateritic grave! (iron concentration). The laboratory analyses of the sherd material have Whitish stones associated in the grave! have been iden been carried out at the Laboratory for Cerarnic tified by Dr. Anders Lindabl as felspar. The paste of the Research, Departrnent of Quarternary Geology, Uni Kiwangwa pottery has all the rock components of versity of Lund, Sweden. The aim of these laboratory 84 (2) Is it possible to make suggestions about which vessels originated at the same site? (3) Is it possible to determine which vessels are cån• a temporary? 9.2.2 Material The material comprises 26 sherds from the six site (Table 19). Five sherds from Misasa (sample Nos 1,,2, 3 and 5) are dated to the 7th centnry AD, sample N o, 4' is dated to the 4th centnry AD, 3 sherds from Limbo' (sample Nos 6-7) are dated to the 5th century ADi sherds from Kaole (sample Nos 9-12) are dated to th 9th centnry AD), 5 sherds from Masuguru (sample !'fri b 18-22) are dated to the 7th century AD and 4 sheril from Mpiji (sample Nos 23-26) are dated to the it,_, centnry AD. 9.2.3 Method ,~:·((;A In order to provide answers to the above questions;t å number of unambiguous studies and calculations haye, :~;~;_ been made. Bach individual stndy reflects a speci(ic' part of the cerarnic process and only the comb" · analysis results will give a more complete picture 9 the ancient craft and its products. In this investigation, the analysis is concentrated o c petrographic microscopy. Petrographic micoscropy'i carried out on thin sections of pottery, i.e. a piece o! · sherd which has been ground to a uniform thicknesso 0 1 2cm 30 µm. The thin section is analysed under a polarizi(i microscope in magnifications ranging from 25x tQ Fig. 25. Shell decorated sherds from Mpijj.. 1OOOx. This analysis makes it possible to identify 85 Table 19. Description of sherds analysed petrologically. . Abbreviations: Tr = trench; Th = thickness; Wt = weight; Dec = decoration (for decoration abbreviation see Fig. 17). Tr Th. Wt sample Site (lev) (mm) (gm) Dec. Sherd type l. Misasa 1 (6) 15 25.7 obl. inc. neck 2. Misasa 1 (5) 12.7 31.3 plain rim 3. Misasa 2 (5) 12.5 18.5 hor. hatch. neck 4. Misasa 1 (7) 12.0 24.3 er. hatch neck 5. Misasa 2 (6) 11.0 24.3 obL hatch neck 6. Limbo 4 (12) 9.0 13.7 bevelled neck 7. Limbo 4 (13) 10.0 16.2 bevelled neck 8. Limbo 3 (13) 9.0 22.3 hor. inc. shoulder 9. Kaole 1 (8) 9.0 46.7 triangle neck 10. Kaole 1 (8) 9.0 26.3 plain rim/body ll. Kaole 1 (8) 9.0 11.6 er. hatch neck 12. Kaole 1 (8) 12.0 48.0 grooves body/rim 13. Kiwangwa l (5) 11.0 16.9 triangle neck 14. Kiwangwa l (5) 9.0 13.0 hor. hatch neck/shoulder 15. Kiwangwa l (5) 11.0 51.2 triangle neck 16. Kiwangwa l (5) 13.0 19.8 plain rim/body 17. Kiwangwa 1 (5) 10.0 11.2 plain rim/body 18. Masuguru 3 (4) 12.0 14.2 triangle shoulder 19. Masuguru 3 (4) 13.0 44.8 obl. zigz shoulder punctates 20. Masuguru 4 (4) 11.0 37.4 triangle neck 21. Masuguru 4 (4) 12.0 12.0 plain rim/body 22. Masuguru 3 (4) 11.0 12.4 plain rim/body 23. Mpiji l (6) 9.0 19.0 dobl. zigz neck . 24. Mpiji 1 (6) 9.0 41.2 triangle neck 25. Mpiji 1 (7) 9.0 5l.1 triangle/rim neck cross hatch. 26. Mpiji 1 (6) 10.0 81.7 plain rim/body parts of the thin section were taken, in order to mini quartz; very few grains of amphiboles and pyroxenes . mize erroneous groupings due to local variations (Lin have been observed. The ware has also been tempered . dahl 1986, p. 29). with chamotte. Apart from two !arge grains, the ore material consists of very small grains sparsely scattered in the ware. ·. 9.2.4 Results Sample 3 (Fig. 26b ). MIPHO group Il, a medium-to ·· All the clays are ferruginous and free of calcium car coarse, silty clay with a minor amount of very fine bonates. No organic matter or diatoms were observed sand. The max. grain size is 1.9 mm and the mean in any of the samples. max.grain size is calculated as 1.2 mm. The amount of Sample 1 (Fig. 26a). MIPHO group 1, a fine clay coarse fractions has been estimated as 10%. The mine with an admixture-most likely a deliberately added ral content consists mainly of quartz, a small amount of . temper----0f coarse silt and sand. The max. grain size is feldspar anda few grains of mica. Furthermore, there is 1.6 mm and the mean max. grain size is calculated as a rich representation of ferrihydrite in the ware. The .1.3 mm. The amount of coarse fractions (grains larger grains of ore material are few and very small . . !han 0.1 mm) has been estimated as 13%. The mineral Sample 4 (Pig. 26c ). MIPHO group I, a fine, dense , content is basically quartz, with very few grains of clay with very sparse contents of sil! and sand. The amphiboles and pyroxenes. In addition to the sand tem- max. grain size is 1.1 mm and the mean max. grain size per, the ware has also been tempered with chamotte. is calculated as 1.0 mm. The amount of coarse fractions Some grains show traits of being sintered. Few and has been estimated as 7%. The mineral content consists small grains of ore material occur. mainly of quartz and very few amphiboles and Sample 2 (Fig. 26a). MIPHO group I, a fine dense pyroxenes. The ware has been tempered with chamotte, clay with an admixture-most likely a deliberately of which some grains are sintered. Very few and small added temper----0f coarse silt and sand. The max. grain grains of ore material occur. size is 2.0 mm and the mean max. grain size is calcu Sample 5 (Fig. 26a). MIPHO group I, a fine clay with lated as 1.3 mm. The amount of coarse fractions has an admixture-most likely a deliberately added tem been estimated as 17%. The mineral content is mainly per-of coarse silt and sand. The max. grain size is 1.3 86 .. 0 • a ? 'i i~' r:: sorted medium si Ity, D feldspar 6 quartz D fine clay [2J sandy clay [2J I weathered feldspar - mica LJ unsorted sandy clay sorted silty clay [2J unsorted somewhat unsorted coarse si Ity, • amph ibole/pyroxen e .\ ore silty, sandy clay LJ sandy clay unsorted silty, sandy sorted coarse silty, c/' ferri hydrite 0 chamotte ~ clay [2J very fine sandy clay sorted medium to coarse [1J unldentified mineral ' clay pellets LJ silty clay The hatched pattern indicates the amount af microscopically observed silit and sand grains in the ware. Few lines means few grains and several lines means a large number af visible grains. Fig. 26 a-k. The paste of the EJW and TIW potsherds: section from petrological analysis. mm and the mean max. grain size is calculated as LO 1.3 mm. The amount of coarse fractions has been estj mm. The amount of coarse fractions has been estimated mated as 14%. The mineral content is basically quarl.;l as 13%. The mineral content consists mainly of quartz and a srnaller proportion of feldspar, as well as a fe~ and very few amphiboles and pyroxenes are observed. small grains of amphiboles and pyroxenes. A small The ware has been tempered with chamotte, of which number of concentrations of fenihydrite and clay pel some grains are sintered. Very few and small grains of lets are scattered in the ware. The ware comprises a rel: ore material occur. atively ]arge amount of ore material in grain sizes frof\l Sample 6 (Fig. 26d). MIPHO group III, an unsorted, 0.03 to 0.3 mm. c somewhat silty and sandy clay. The max. grain size is Sample 7 (Fig. 26d). MIPHO group III, an unsorted., 2.1 mm and the mean max. grain size is calculated as somewhat silty and sandy clay. The max. grain size is l 87 J. 7 mm and the mean max. grain size is calculated as samples 6-8 and the grains are also coarser. The grain I J.2 mm. The amount of coarse fractions has been esti size variation is between 0.06 and 0.3 mm. The ware l mated as 14%. The mineral content is basically quartz also includes small concentrations of ferrihydrite. and a smaller proportion of feldspar, as well as a few Sample 12 (Fig. 26b). MIPHO group Il, a sorted, small grains of amphiboles and pyroxenes. A small medium-to-coarse-silty clay with a minor amount of number of concentrations of ferrihydrite and clay pel very fine sand. The max. grain size is 1.4 mm and the lets are scattered in the ware. The ware comprises a mean max. grain size is calculated as 1.0 mm. The relatively !arge amount of ore material in grain sizes amount of coarse fractions has been estimated as 9%. from 0.03 to 0.3 mm. The clay is partly sintered. The mineral content consists mainly of quartz, a small Sample 8 (Fig. 26d). MIPHO group III, an unsorted, amount of feldspar and a few grains of amphiboles, somewhat silty and sandy clay. The max. grain size is pyroxenes and mica. Furtherrnore, there is an abundant 1.9 mm and the mean max. grain size is calculated as representation of ferrihydrite-rich grains (less than 0.45 1.4 mm. The amount of coarse fractions has been esti mm) in the ware. The grains of ore material are few and mated as 18%. The mineral content is basically quartz very small. and a smaller proportion of felspar, as well as a few Sample 13 (Fig. 26g). An unsorted clay with the small grains of amphiboles and pyroxenes. A small majority of grains within the coarse-silt fractions. The number of concentrations of ferrihydrite and clay pel max. grain size is 1. 7 mm and the mean max. grain size lets are scattered in the ware. The ware comprises a is calculated as 1.2 mm. The amount of coarse fractions relatively !arge amount of ore material in grain sizes has been estimated as 17%. The mineral content con from 0.03 to 0.3 mm. The clay is partly sintered. sists mainly of quartz and a small amount of feldspar Sample 9 (Fig. 26e). MIPHO group IV, an unsorted, with a few weathered grains. A few small grains of silty and sandy clay. The content of silt is larger and the amphiboles and pyroxenes, and a very few grains of grains are also coarser than those in samples 6-8. The ore material, occur. max. grain size is 1. 8 mm and the mean max. grain size Sample 14 (Fig. 26h). MIPHO group V, a sorted, is calculated as 1.1 mm. The amount of coarse fractions medium-silty to very fine clay with a sparse content of has been estimated as 16%. The mineral content con coarser sand. The amount of sill is larger than in sam sists mainly of quartz, as well as a small amount of ples 6-8. The max. grain size is 2.3 mm and the mean feldspar and a few grains of amphiboles and max. grain size is calculated as 1.9 mm. The amount of pyroxenes. Some grains of feldspar show traces of coarse fractions has been estimated as 17%. The mine . weathering. The amount of ore material is somewhat ral content consists of quartz and a small amount of greater than in samples 6-8 and the grains are coarser, feldspar. Grains of amphiboles, pyroxenes and mica are . with a grain-size variation between 0.06 and 0.3 mm. more frequent than in samples 1-13. The clay also con The ware also includes a small amount of ferrihydrite. tains ferrihydrite and clay pellets of varying sizes. Few Sample 10 (Fig. 26f). A sorted, coarse-silty and very grains of ore material have been observed. Part of the ' fine, sandy clay, in which the majority of the grain sizes clay, as well as a few clay pellets, have begun to sinter fall within the range 0.03--0.2 mm. The max. grain size towards the outside of the vessel wall. js 0.9 mm and the mean max. grain size is calculated as Sample 15 (Fig. 26h). MIPHO group V, a sorted, 0.6 mm. The amount of coarse fractions has been esti medium-silty clay with a sparse content of coarser j mated as 6%. The mineral composition consists mainly sand. The amount of silt is larger than in samples 6-8. 1öf quartz and a small amount of feldspar. The grains of The max. grain size is 2.1 mm and the mean max. grain I ; amphiboles and pyroxenes are more frequent than in size is calculated as 1.4 mm. The amount of coarse samples 1-9. The ware is also relatively rich in ore fractions has been estimated as 14%. The mineral con 1' I' ·material and in small concentrations of ferrihydrite. tent consists of quartz and a small amount of feldspar. ' I Sample 11 (Fig. 26e ). MIPHO group IV, an unsorted, Grains of amphiboles, pyroxenes and mica are more silty and sandy clay. The content of silt is larger and the frequent than in samples 1-13. The clay also contains I grains are also coarser than in samples 6-8. The max. concentrations of ferrihydrite and clay pellets of vary grain size is 1.3 mm and the mean max. grain is calcu ing sizes. A few grains of ore material occur. Part of the lated as 0.9 mm. The amount of coarse fractions has clay, as well as a few clay pellets, have begun to sinter been estimated as 11 %. The mineral content consists towards the outside of the vessel wall. lllainly of quartz, as well as a small amount of feldspar Sample 16 (Fig. 26h). MIPHO group V, a sorted, and a few grains of amphiboles and pyroxenes; some medium-silty to very fine clay with practically no con grains of feldspar show traces of weathering. The tent of coarser saud. The amount of silt is larger than in amount of ore material is somewhat greater than in samples 6-8. The max. grain size is 0.8 mm and the i!l ~------~~~-"-~___j;; ) 88 Table 20. Fabric grain-size measnrements. Sample 21 (Fig. 26i). MIPHO group VI, an unsortect., sandy clay. The max. grain size is 2.8 rnrn and the me • Max. grain Mean grain Greater than Sample size size 0.1 mm(%) max. grain size is calculated as 2.2 mm. The arnount 0 coarse fractions has been estimated as 21 %. The min~. I. 1.63 1.31 13 2. 1.96 1.28 17 ral composition consists of quartz and partly weathered 3. 1.86 1.19 10 feldspar, as well as a few grains of amphiboles aiid 4. 1.07 1.01 7 5. 1.28 1.04 13 pyroxenes. To be noted are the grains of a not yet iden,• 6. 2.08 1.26 14 tified mineral. There are also a few grains of clay pel 7. 1.70 1.21 14 8. 1.89 1.42 18 lets. The ore material is sparse and basically within ti\e 9. 1.84 I.Il 16 coarse silt/fine sand fraction. · 10. 0.94 0.62 6 Sample 23-26 (Fig. 26k). MIPHO group VII, 11. 1.31 9.92 11 .a 12. 1.36 1.02 9 sorted, silty clay with a small content of coarse, si)ty 13. 1.73 1.21 17 sand. The max. grain size is 2.0 rnrn and the mean mai. 14. 2.34 1.85 17 15. 2.07 1.41 14 grain size is calculated as 1.1 rnrn. The amount df 16. 0.84 0.76 5 coarse fractions has been estimated as 12%. The min~ 17. 2.64 1.71 21 18. 3.53 1.54 20 ral composition consists largely of quartz and feldspaj, 19. 297 1.47 18 as well as a few grains of amphiboles and pyroxenes. 20. 2.71 1.71 18 21. 2.79 2.20 21 The clay is comparatively rich in ore material, in which 22. 1.81 1.35 17 a grain size between 0.08 and 0.25 rnrn predominates~ 23. 2.31 1.12 12 A snmmary of the grain-size analysis is provided 24. 1.54 1.17 14 lll 25. 1.50 1.06 12 Table 20. · 26. 1.31 1.24 13 9 .2.5 Conclusions <, The investigated material from Misasa, Limbo, Kaol~; mean max. grain size is calculated as 0.8 rnrn. The Kiwangwa, Masuguru and Mpiji consists of very di):~ amount of coarse fractions has been estirnated as 5%. tinctive ware-groups, which to a large extent are rela.t~ The mineral content consists of quartz and a smaller to the different sites. proportion of feldspar. Grains of amphiboles, The Misasa site has three different types of war~; pyroxenes and rnica are more frequent than in samples The predominant type is made of a very fin.e clay, J 1-13. The clay also contains ferrihydrite and clay pel which a temper of quartz and chamotte has been add lets of varying sizes. Few grains of ore material. (sherds Nos 1, 2 and 5; Fig. 26a). The vessels manufag Sample 17 (Fig. 26i). MIPHO group VI, an unsorted, tured in this fashion are dated to the 7th century Ap sandy clay. The max. grain size is 2.6 rnrn and the mean However, one of the test sherds (No. 4), which is da ·· rnax. grain size is calculated as l.7 rnrn. The amount of to the 4th century, is made out of the same fine clay ~ coarse fractions has been estirnated as 21 %. The mine the three others but it is only tempered with chamotj ral composition consists of quartz and partly weathered (Fig. 26c). This may imply that chamotte temJlr feldspar, as well as a few grains of amphiboles and belongs to an older tradition of manufacturing tecll' pyroxenes. To be noted are the grains of a not yet iden niques that lived on even when the tradition of vesser tified mineral. There are also a few grains of clay pel production had changed into the use of crushed qu lets. The ore material is sparse and basically within the as the dominant type of temper. -~ coarse silt/fine sand fraction. The fifth sample (No. 3) consists of a much coars.e Sample 18-20, 22 (Fig. 26j). MIPHO group VII, a clay, whichhas no resemblance to the others (Fig. 26Q-). coarse, unsorted, silty and sandy clay. The max. grain The larger amount of silt, the presence of mica and th size ranges from 2. 7 to 3 .5 rnrn. The mean max. grain concentrations of ferrihydrite make this ware differe. size is calculated as 1.5 mm. The arnount of coarse and clearly points to a different source for the clay. Tll fractions has been estirnated as 18%. Large, composite ware has its closest parallel in one sherd from Kao) grains of quartz and feldspar occur, as well as of sin (No. 12). ' tered quartz and feldspar grains. The feldspar is greatly The samples from the Limbo site (Nos 6-8) are weathered. Amphiboles, pyroxenes and mica occur very similar to one another. The clay has a relative\ more frequently, compared with all the other samples. fine matrix, which has a natura! temper of coarse s(l The clay is comparatively rich in ore material, most of and sand. They all display the same composition of or whi9h is within a grain-size range of 0.1-0.3 material, ferrihydrites and clay pellets that makes the- ) 89 unique, compared with sherds from the other sites (Fig. cerned. Four of the live sherds (Nos 18, 19, 20 & 22) 26d). have an almost identical ware (Fig. 26j). The coarse, The Kaole site hasa more varied composition of pot unsorted, silty and sandy clay contains several !arge tery ware. Two sherds (Nos 9 & 11 ), however, have a composite grains of quartz and feldspar, as well as a similarity in clay type and mineral content that strongly !arge amount of heavily weathered feldspar, rnica, indicates that they were manufactured in the same tra amphiboles, pyroxenes and ore material. There is very dition and of clay from the same source (Fig. 26e ). The little doubt that all these vessels have been formed of vessel represented by sample No. 10 has been made of clay from the same deposit. a sorted, very coarse, ferruginous clay that is unique The ware of sherd No. 21 has an almost identical among the investigated sherds (Fig. 26f). As mentioned composition to that of sherd No. 17 from Kiwangwa above, sample No. 12 hasa similar clay composition to (Fig. 26i). sample No. 3 from Misasa (Fig. 26b). There are slight All the vessels from the Mpiji site (sample Nos 23- differences in the mineral compositions of these two 26) have a raw-material composition that is practically samples. However, this difference is not greater than identical (Fig. 26k). The sorted clay is rich in grains of the normal variation within the same clay deposit. quartz and feldspar, as well as grains of ore material. It The material from the Kiwangwa site shows the is therefore most likely that they were all made of clay same pattern as Kaole: a nucleus of vessels with the from the same source. same ware composition (sample Nos 14, 15 & 16) (Fig. As regards answering the questions put forward in 26h), one unique sherd (sample No. 13) (Fig. 26g), and the introduction, this pilot investigation strongly indi one sherd (No. 17) which can be related to a material cates that the majority of the vessels within a site show from another site (No. 12 from Masuguru) (Fig. 26i). a sirnilar or almost identical composition of the clay The majority of the vessels have been manufactured and that there are great differences among the sites. from a medium-silty to fine sandy clay containing The results may therefore, with the greatest certainty, quartz and some feldspar, mica, arnphiboles and be interpreted as meaning that the pottery at each indi pyroxenes. The clay also includes a fair amount of clay vidual site was made predominantly from raw material pellets and concentrations of ferrihydrite. The ware of collected from only one clay deposit. In the cases of sample No. 16 almost completely lacks grains of the Limbo and Mpiji, the results do not display any exter sandy fraction. A possible explanation is that the clay nal infiuence on the pottery. At the other sites, sherds of was dug at a different level of the clay deposit, com other origins have been observed only in a few cases. pared with the clay used to make the other two vessels. Misasa and Masuguru, for instance, have only one The ware of sample No. 13 has a very uniform grain sherd each that can be related to an external origin. size, with the coarse silt to fine sandy fractions, which Kaole and Kiwangwa display a more complex inven largely consist of quartz and a small amount of feld tory. The majority of the vessels are of local origin. spar. This type of ware has not been identified among One sherd at each site is unique among the investigated the sherds from the other sites. vessel sherds. The ware of sample No. 17 has a very fine matrix, It is difficult to determine which vessels are contem mixed with an unsorted, coarse fraction of sand. In the porary, on the basis of these laboratory analyses. The · mineral composition, heavily deformed grains of a very difference between sample No. 4 and samples Nos 1, 2 strong, brown colour (PPL) have been observed. It has and 5, however, suggests a change of the handicraft at has not yet been possible to ascertain what type of one site over a period of 300 years. mineral this is. In short, the locally produced pottery-at the village The Masuguru site has a more uniform inventory of or town level-to a very !arge extent outnumbers the vessels as far as the composition of the clay is eon- imported wares. > 10. CHRONOLOGY To be able to review properly the theories of the origin down for a long time before it was bnrnt. This woUJd and spread of the early and later iron-working cornmu also offer an age 400 years older than the norniai nities along the coast of East Africa, the chronological (Wright 1984, pp. 54-5). ,. . aspects must be discussed thoroughly. These comprise The use of 18 radiocarbon dates in this work h~s the stratigraphical ordering of the material, the strati been checked by taking into account the fäet that they graphical association of imported and local artefacts were collected from different sites, including those i for relative dating, and the absolute dating by carbon- the hinterland, which have nothing to do with mariue 14 samples from different cultural layers. water. Small chunks of charcoal collected from diffe: Relative dating has been the main dating technique rent levels and units of different sites make it unlikeiy used with certainty on the coast of East Africa. The that all would have come from the trunk of one tree. •\ Chinese and Islamic ceramics fouud in archaeological contexts on the East African coastal sites have gene rally been used as the objects of such dating technique (Chittick 1974, p. 319; Horton 1987b, p. 294). This 10.1 Stratigraphy approach has been said to agree with the occasional It has been shown in Chapter 7 that, apart from u\ coin linds, the radiocarbon dates, the excavation of Kaole site, the other TIW sites under study belongedJo kilns and the investigation of occupation in China and one major period of occupation. For instance, altho11gh the Near East. On this criterion, occupation in the the sites of Misasa, Mpiji and Masuguru bad indi~'a; Lamu Archipelago has been dated from the eighth cen tions of post-TIW occupations in the vicinity, the la\(e tury onwards, "probably from aboutA.D. 750" (Horton is not a direct, cultural continuation of the former; 1987b, p. 294). have, however, noted that, prior to the TIW occupati~ The certainty attributed to this method without chal in Misasa, there existed an EIW tradition in the vii'i, lenge may, however, have led many archaeologists nity, of which the cultural traits might have been c.~' astray. Many radiocarbon dates were thrown away, ried forward in the pottery of the TIW tradition. It tja· becanse they did not match the dates obtained from the been shown in the previous chapter that some EIW p?t relative dating (Chami 1992a). It is now even appear tery sherds have been identified in the context of Eai;ly ing that the justification for using imported ceramics to TIW occupations in Misasa, Mpiji and Kiwang~a; date settlements on the coast of East Africa was not This has led to the conclusion that the TIW sites w~re well founded. It has been shown that "neither is there chronologically later than the EIW tradition. much dating evidence from excavated Islamic and Chi The study in Kaole has also shown that a less-de~o nese kiln-sites" (Tampoe 1989, p. 69). rated and ultimately a plain-ware tradition marked tj,~ The use of the relative-dating method in this work final phase of the TIW tradition. Therefore, a straµ• has been carefully checked with both the existing dates graphic sequence of artefacts from the five sites studlf of the local sequences and the new radiocarbon results. in this work should begin with the EIW tradition at tli It was mentioned above that one of the reasons for bottom, followed by the TIW tradition and theu ' neglecting radiocarbon dates was that most of them did plain-ware tradition. not fit the agreed relative dating in the 9th-10th centu ries AD. However, another reason was based on the idea that dates obtained from materials associated with the waters of the Indian Ocean, i.e. shells and charcoal 10.2 Relative dating from mangrove wood, would be 440 years older than the real age. The reason for this is attributed to the Pottery with EIW elements has been found in the lowef "natura! variations in atmospheric carbon-14" and the layers at Misasa, Kiwangwa and Masugum. As w~s "carbon-14 content of Indian Ocean water" (Wright also observed (see Figs 20 and 21), these elements. 1984, pp. 54-5). It was also thought that the charcoal decreased in the upper layers. Since the EIW tradition· samples could be from the inner rings of a tree laid on the coast has been dated between the lst century BC 91 Table 21. Cl4-results from six sites (calibration according to Stuiver & Becker 1986). Lab. no. Si te Trench Lev el Radiocarbon Age BP One sigma (caL) Ua-2593 Misasa l 8 1975+/-50 BC 89 (25, 43. 46) AD 68 Ua-2594 Misasa 2 7 1725+/-45 AD 239 (261. 279, 293, 297, 335) 385 Ua-2595 Misasa 2 5 1295+/-55 AD 659 (685) 775 Ua-2597 Misasa 3 5 1485+/-50 AD 536 (575) 639 Ua-2592 Mpiji 4 8 1340+/-60 AD 642 (665) 765 Ua-2087 Mpiji l 6 1420+/-60 AD 576 (640) 658 Ua-2088 Mpiji 1 6 1390+/-80 AD 582 (644) 678 Ua-2598 Kiwangwa 2 6 1405+/-55 AD 599 (642) 662 Ua-2599 Kiwangwa 1 4 1275+/-45 AD 671 (691, 700, 710, 749, 765)797 Ua-2097 Kiwangwa 1 5 1440+/-60 AD 556 (608, 627, 638) 650 Ua-2098 Kiwangwa 3 7 1250+/-60 AD 676 (735, 772) 863 Ua-2095 Masuguru 2 4 1290+/-60 AD 660 (686, 754, 757) 797 Ua-2096 Masuguru 4 4 1430+/-60 AD 562 (639) 654 Ua-2092 Kaole l 9 1270+1-60 AD 667 (692, 699, 712, 748, 767) 801 Ua-2093 Kaole 4 7 1130+/-60 AD 781 (895, 922, 939) 984 Ua-2094 Kaole 3 6 720+1-60 AD 1258 (1279) 1285 Ua-2089 Changwehela 8 870+1-50 AD 1042 (1163, 1174, 1188) 1221 Ua-2090 Changwehela 9 650+/-60 AD 1279 (1296, 1375) 1391 and the 5th century AD (Soper 1967b, Chami 1988a, b; such Islamic ware) or later than AD 1000 (the beginning . Sinclair 1991), this would indicate that the TIW tradi oflustre and sgraffiato) (Chittick 1974, p. 319) . tion bad its formative period around the 4th-6th centu -; ries AD. Also in the main layers of the TIW occupation, at 10.3 Radiocarbon dating 60-40 cm, 70-40 cm and 50-40 cm at Misasa, Mpiji and Kiwangwa respectively, TIW pottery is associated As good samples of charcoal were recovered from with imported artefacts. These include the Sassanid many of the levets excavated (Chapter 8), a number of ware, glass and beads. The production of alkaline charcoal samples were subrnitted for dating. A total of glazed products started in Roman and Parthian limes, 18 charcoal samples were analysed from the six sites from the 3rd century BC to the beginning of the 3rd cen excavated in both Phases One and Two. In Table 21, tury AD. The use of such products continued throughout the first 15 samples were collected from the levets of I the Sassanid period (c. 226--642) and in the early the TIW tradition, and hence they were the most appro 'i Islamic period to about AD 750 (Lane 1947, pp. 8-9; priate for the discussion in this section. The last three Whitehouse 1972, p. 70; Pope 1939). The absence of are from the post-TIW tradition (and hence worthy of a I any typical Islamic goods associated with the TIW discussion elsewhere). il materials at the three sites indicates a date for these Accordiug to Table 21 and Fig. 27, therefore, the !, j! sites before the 8th century, the period before Islarnic TIW tradition can now be dated to between the 4th and 11 '-1I I · art came into existence (Lane 1947; Tampoe 1989). the !Oth century AD. The Misasa site is more informa : The association of the EIW elements and the Sassanid tive for the early period. The charcoal sample recove pottery would therefore mean a date between the 4th red from the early, faint, occupational layer (80 cm) has century and the end of the 7th century AD. offered the ear!iest date in the !st century AD (Ua- Close study of the Kaole chronology, however, 2593). The charcoal from this Jayer was associated offered further chronological data. The lowest layer of with a few fragments of iron. As no pottery was reco the site at the Kaole hill is represented by the TIW tra vered from this ear!iest layer, it cannot be regarded as dition. Unlike the other sites where Sassanid ware was the beginning of the TIW tradition. It should rather be · recovered, the Kaole site was found to have early regarded as that of the past EIW tradition. A sirnilar Islamic, white-glazed ware. In the succeeding, non date was ohtained for the EIW site at Limbo in the TIW layers, lustre and sgraffiato ware were found. It is vicinity. A charcoal sample from a subsequent layer (65 concluded that the TIW occupation at the Kaole site cm) in the same trench in Misasa, well associated with cou]d not be dated before AD 750 (the beginning of a good numher of TIW sherds, gave a date in the 4th \--c:J J,,i';f •------~~~~~~~---'"" 92 EarlyTlW that the formative period for the TIW tradition was fu~' . rel prob cum percentage 100 4th-6th centuries AD, and that it flourished along fue' V 90 littoral and in the hinterland of the central coast of T.\n: 1,1\ 7 80 zania in the 7th century AD. .. \I 70 /\) Turniug to the Kaole site, the radiocarbon dates''.in 60 I the 8th and 9th centuries derived from the TIW layer 50 (Ua-2092 & Ua-2093) indicate !hat the last phase of fue 40 I TIW tradition coincided with the spread of Islam to. 30 ~ 20 East Africa. This agrees well with the finding of eaity ~ )/\ \ 10 Islamic goods in the same layer. . ..· Jff1 ~A"" I I 0 "'- 0 At this juncture, therefore, it is sufficient to concltrr!e -400 -200 0 200 400 600 800 1000 1200 Calender YeaJS from the relative and absolute datings that the TIW tta dition manifests two phases: an Early TIW ph~se LateTIW before the 8th century and a Later TlW phase after the ref prob cum percerttage 100 . 7th century and probably up to the llth century AD: I 90 termed the first phase Azanian (cf. Chapter 4 in tlii I 80 volume; Casson 1990; Sutton 1990, pp. 89-91) or pte ~V XI\ 70 Islamic and the second phase Zanjian ( cf. Chapter 4_: __ r 60 I this volume; Trimingham, 1995) or Islamic. - 50 The Early TIW phase is characterized by the pottify 40 and the radiocarbon dates from Misasa, Mpiji, Kiwaj,c I 30 I 20 gwa and Masuguru. In terms of pottery (Chapter 8), tµ I 10 assemblages from these sites have many EIW pott~'. I \ 0 0 elements and diverse elements of TIW decoration. Sils -400 -200 0 200 400 600 800 1000 1200 Calender years sanid pottery marks the pre-Islamic trade. t The Later TIW phase is represented by the potti Fig. 27. Probability distribution and cwnulative probabilities from C14 tests and the dates from Kaole. The pottery has few E '., from a = Misasa, Mpiji, Kiwangwa and Masuguru. The hatched area is the contribution made by the test Ua-2593); b = Kaole. (Calibration according to elements, with triangles and oblique- and double-zlg Stuiver & Becker 1986). zag incisions predominating. There is a diminis "' interest in decoration, indicated by many sherds bei\l undecorated and having a smaller number of decorati~ elements. This period is marked by the coming '1o century AD (Ua-2594). This the Manda site. The radiocarbon dates in the 6th and AD 630 reported by Chittick (1966a, pp. 9-10) from 7th centnries obtained from the lowest layer at Manda Period 1 indicates that the TIW tradition there was (Horton 1987, p. 295) would, if accepted, place the founded in the Early TIW phase. TIW tradition there in the early phase. However, these The same problem applies to the Comoros sites, dates were refused, because they do not fit the 8th- where TIW si tes ha ve been dated to the 9th and 10 cen 11 th-century, imported-ceramics mode! (Chittick 1967, turies AD (Wright 1984). Existing earlier dates have p. 49; Horton 1987a, p. 296). Manda has therefore been been neglected, owing to assumptions about the Indian dated unjustifiably to the Later TIW phase. Ocean sea-water effect (Wright 1984, pp. 54--5). Abungu (1989, p. 212) dated the TlW site of Ung In Mozambique, the Chibuene site dated earlier to wana to the lOth century by using the above, relative the 9th century, following the above, relative-dating dating mode!. His imported pottery wares for the 910- mode! (Sinclair 1982), has now been dated to the 5th- 1150 period are green/blue Sassanian Islamic, Islamic 7th centuries in recognition of the problems inherent in white-glazed, and incised sgraffiato. The alkaline the imported-ceramics mode!. According to Sinclair glazed, green/blue pottery found in the early phase of (1991, p. 190), the Islamic period in the Middle East came to be the lower occupation at Chibuene, which has been dated to the later termed "Sassanian Islamic" owing to "its origin in the first mfllennium, contemporary with Kilwa lb on the basis of Sassanian times" (Tampoe 1989, p. 31) and "ultimately ceramics and a series of carbon dates, is further dated from mid to from the Parthian blue-glazed wares" (Chittick 1966a, late first millennium. p. 162). However, contrary to Abungu's relative dating, This new position puts Chibuene in the Early TIW it is unlikely that its production continued beyond AD phase. Similarly, Adamowicz has dated the Monapo 815 (Tampoe 1989, pp. 78-9). TIW layer to AD 600-900 (Sinclair, Morais, Adamow Unguja Ukuu was origiually dated by a gold coin of icz & Duarte 1993, p. 425). Caliph Harun al-Rashid, who ruled in AD 798-9 (Chit In Madagascar, the Irodo site is dated to the 8th and tick 1966a, p. 163). However, the coin did not come 9th centuries but has no TIW pottery or Sassanian pot from the archaeological context but was found by tery (Venn 1986; Sinclair 1991, p. 195). treasure hunters. The absolute date helped to strengthen To summarize, the result of the analysis of the asso the idea that the TIW sites had their origin in the Later ciation of cultural material and radiocarbon dates has TIW phase after the 7th century. Recent excavation, given important, new, chronological information about however, has dated the early occupation at Unguja the first-millennium-AD settlement along the coast of Ukuu to the 4th century AD (Juma 1993). This agrees East Africa. very well with the dates from Misasa, which firmly put The Early TIW phase dates from the 4th to the end of the origin of the TIW tradition in the early phase. the 7th century, a period associated with Sassanid trade Although the Kilwa Periods la and lb show ele goods (alkaline-glazed ware and glass). The Later TIW . ments of the Early TIW phase observed in Misasa, phase started at the beginning of the 8th century and Mpiji and Kiwangwa, Chittick (1974, p. 319) has nsed survived to the 1Oth century. This is an era related to imported ceramics to date the occupations to the 9th the introduction of Islamic products. The implication of • and lOth century AD, and hence to the Later TIW phase. this new recognition will be discussed in the next chap • However, it should be noted that a radiocarbon date of ter. I \ l ! i. 11. THE BEGINNING, SPREAD AND END OF THE TIW TRADITION My case so far has been that, in the East African coastal around Misasa has other EIW sites, which, althoriglt areas, a major cultural tradition developed, based on its reported, await proper study (Fawcett & LaViolette• own socio-economic structures and natural resources. 1990). EIW sites have also been found in other areas. The tradition has been identified in this work by its tri with TIW sites. They comprise Kilwa and Unguja angular, incised ware, as the TlW tradition. The socio Ukuu (Sinclair, Morais, Adarnowicz & Duarte 1993ip, economic strength of this tradition enticed trade from 427; pers. observation), Dar-es-Salaarn, and Kilosaiti abroad, beginning perhaps in the early part of the first the far hinterland (Msemwa 1992; Chami 1991). On millennium AD. the northern Mozarnbican coast, EIW sites have b \ :.:. .. in all sections of this chapter, as we perceive the TIW the proponents of the 8th-9th-century date hay tradition developing from the EIW tradition. The sec changed their minds. A date in the first century AD ;i . tion on the end of the tradition discusses the factors now thought to be appropriate for the beginning of tjl · affecting the cultural process on the coast of East TIW tradition on the central and southern coast. It ii!is Africa in the first millennium AD and hence gives a bet also been thought that the tradition would have pfe ter understanding of the subsequent cnltural and eco ceded the early iron-using settlements. According·f nomic relationship between the hinterland and the Horton (1990, pp. 96-7), littoral. The early first-century Unpublished reports also indicate that the charcoal It has also been shown (Chapter 6) that the shore sarnples rnight have been contaminated by chernicals between Bagamoyo and Kisiju is aligned to the sea in or possible recovery problems. It is no doubt still too such a way that any ship driven by strong, north-east early to detennine the date of the beginning of the TIW erly trade winds would naturally land in this area, if not tradition. Moreover, the Cushitic aspect is incompre blocked by Zanzibar Island in the same area, 15 km hensible, since no single Cushitic or Pastoral Neolithic from the mainland. This is also the area where strong, site has been found on the coast of East Africa. soutb-easterly trade winds start driving the northbound The lowest leve! with TIW-tradition pottery at Mis ships to the northem Indian Ocean and the Red Sea asa has in this work been dated to 335 ±45 AD. This is (Datoo 1975). also the period of the decline of the EIW tradition at Secondly, the area is climatologically wet enough for Limbo and soon on the whole coast. The TIW tradition agriculture and tropical forests, especially behind Kis matured at Misasa at the end of the 6th century and at iju in the Kisarawe zone (Map 10). This particular fac the other sites reported above in the 7th century, coin tor, coupled with tbe inland lakes, for example, ciding with the introduction of imported glass, Sassa Zakwati and Manzi, was one of tbe major reasons why nid ware and glass beads. By this period, the EIW it was decided to survey the district of Kisarawe for tradition on the central coast had completely been early sites in 1987 (Chami 1988a). transfonned into the TIW tradition. Except for the Thirdly, it has been suggested linguistically tbat this Chibuene and Monapo TIW sites reported to have had central coast was possibly where the coastal languages mid-lst-millennium dates (Sinclair, Morais, Adarno in East Africa originated. "From presently available wicz & Duarte 1993), otherTIW sites on the coasthave Iingnistic studies, Sabaki Bantu first appeared some been reported to have had post-7th-century AD dates where around the Ruvu River in northem Tanzania and associated with Islamic ware (Horton 1987a; Chittick spread northwards, possibly as far as the Juba River" 1984, 1974; Wright 1984; Abungu 1989). (Pouwels 1987, p. 21). Two phases for the TIW tradition (Chapter 10) which Fourthly, historically, it has been strongly suggested recognize the new chronology have already been sug in the works interpreting the Periplus document that gested: the Early TIW phase is the period from the end the most southem emporium visited by the Graeco of the early iron-using tradition (in the 4th and 5th cen Romans was around Dar-es-Salaam via Zanzibar turies AD) to the maturity of the TIW tradition (in the (Datoo 1970; Kirwan 1989; Casson 1989). The penin 6th and 7th centuries AD). The Later TIW phase is the sula which ships could not avoid landing at when sail period from the beginning of the 8th century to the end ing southward from Menuthia (Zanzibar) is tbought to of the first millennium AD. The early phase would pro have been Dar-es-Salaam. The peninsula extends bably help us to understand the pre-Islamic, Azanian southward to tbe bay of Kisiju and tbe estuary of the coast, which is reported in the Graeco-Roman docu River Luhute (Rufute). According to tbe Periplus, the ments, and the later phase would probably explain the emporium was some distance from a bay towards the Islamic period when the Arabs recognized the coast as interior along a river. The emporium's name (Rhapta) the land of Zanj. was also the name of tbe river and the bay. The mean Where did the TIW tradition originate? From the data ing of Rhapta was "sewn boats" (Casson 1989, p. 61). discussed above, the TIW tradition probably first began The River Luhute (Rufute), which enters Kisiju bay, on the central coast of Tanzania or further to the south. has a meaning very close to tbe Periplus' Rhapta. The most probable area-to which I would direct much Rhapta is certainly a corruption of an ancient river . more research-is between the Wami River in the north name very sirnilar to the present-day Rufute . . and the Rnfiji River in the south (Map 8). Several fac- Fifthly, the area between Dar-es-Salaam and Kisiju tors favour this central coast. towards the hinterland requires much more research, First, the area receives three major rivers bringing for tbe following archaeological reasons: waters from the deep hinterland of present-day Tanza First, Harding (1960) reported a first-millennium-AD . nia. Although not navigable, these rivers have provided site in Kisiju. He suggested a date in tbe 4th-6th centu a natura] connection to the heart of the region assumed ries from tbe use of imported artefacts: to be the origin of the early iron-using communities in Western Tanzania and the interlacustrine region. The The fragment of glass ... is very similar to glass found in Anglo river valleys, deltas, estuaries and muddy, shallow sea Saxon graves. The glass is thin, clear, and full of nllnute, drawn Waters would also have functioned as the basis for dif out, air spaces. It is cobalt blue in colour and has a slightly raised ribbon-like pattem in white opaque glass. It was recovered from the l ferent subsistence activities (Chapter 6), exploitation of swamp in association with the Frankish bead ... and with the earli J mangrove wood on the estuaries, and fishing. est Indian beads. 1 1.i______...... ,. __ ....,_~~--~~-----! i", 96 :i f] - Prof. Gräslund, of the Department of Archaeology at coast (Soper 1967a, Chami 1988a), more than 201:IW !i Uppsala University, managed to meet Mr D. M. Wilson sites are now known (Map 2), most of them havfug:• (quoted in Harding 1960, p. 138), who confirmed that sizes ranging from two to more than fifteen hecta'tes I he had seen the Frankish bead. The Kisiju site was also (Chapter 7; Chittick 1966, 1984; Horton 1984; Hortoti reported to have copper crucibles and many local pot & Clark 1985; Wright 1984, 1993). Thls isa clear indi-' sherds. cation that there was a population growth. Secondly, 12 km towards the hinterland is situated The third and fourth aspects would probably. be the centre of the lst-7th-century-AD sites, comprising among the factors that could help to explain how cul those of the E!W and the Early TIW phase discussed in tural change took place among the early-first-millen-' this work (Chami 1988a, 1988b, 1992b; Fawcett & nium inhabitants of the coast Similar factors >ire' LaViolette 1990). They include the Limbo EIW and the known to have triggered homeostatic readjustmentjin the formation of Mesoamerica and the Aegean, leadin o Misasa TlW sites. ,b How 98 The earliest sites studied in northem Madagascar. i.e. sanid infiuence started to be produced (Tampoe 1989,, Irodo (9th century AD) and Mahilaka (12th century AD), p. 94). have shown connections with the coast of East Africa, instead of Austronesia (V erin 1975, p. 184, 1986). Verin (1975, p. 191) thought !hat the solution to the Austronesian problem could be solved by the finding of 11.2 The spread of the TIW tradition "sites on the African coast anterior to the ninth cen tury", whose material would show Austronesian links. The distribution of TIW sites and the nature of th~ir All the early TIW sites dealt with in tbis work are· ante spread to many parts of the coast of East Africa h~ve : rior to the 9th century. No find, however, can be attri also been the concems of recent research (Horton · buted to the Austronesians. 1987a; Fawcett and LaViolette 1990). It has been sdg- ' It has also been shown in Chapter 3 !hat many crops gested !hat the distribution of the TIW tradition sifes · and cultural items found in Africa have been attributed favoured the areas occupied by Sabaki-speakljig toAustronesians (Jones 1971; Verin 1975, pp. 175-7). groups, i.e. the littoral and the islands of eastern Afci~a These crops include coconuts, rice, bananas, yams, and and the 100-km hinterland of the northem coast north• sugar cane and the cultural items include xylophones, of Pangani and south of the Tana River (Horton 198ia, dug-out boats, outrigger canoes, board games, and p. 315). This hurried conclusion was founded on \)ie metal crafts. The speculation !hat !hese items came assumption discussed above, that the northem Kenya from eastern Asia awaits concrete evidence. Some of coast saw the genesis of the tradition. The TIW pottetY• !hese items could also have been exported from Africa. was !hus thought to have been produced in that regipu Whatever forces contributed to the cultural transfor and spread througb trade to the rest of the East Africiln mation on the first-millennium coast of East Africa, a coast. fully-fledged TIW tradition was widely distributed over the littoral and in the immediate hinterland by the The Tana tradition pottery may have been traded up the T produced locally within the sites (Chapter 9). Although This work has presented strong evidence !hat the TIW some trade iu local pottery caunot be ruled out, I sug tradition was founded by farmers occupying the littoral gest !hat it 101 First, the Islamic crusade used war as a meaus of in Chapter 3. The weaknesses in previous works have conquest aud to replace indigenous ideologies by the been discussed in the same chapter. Muslim faith. After having conquered Persia, no doubt The work has shown that the coast of East Africa had they would have replaced the Sassauians in their role as settlements of the TIW tradition in the first millennium traders in East Africa, at the same time imposing their AD. They were distributed over the islauds, the littoral ideology on the population by force. This could easily and the deep hinterlaud of the Tauzauian coast. Their be performed by local converts, who a!ready controlled inhabitauts developed various meta! crafts aud traded the littoral settlements. with states in Arabia, the Persiau Gulf and the Mediter Secondly, the rapid expansion of the Muslim empire ranian. in the Middle East created a need for labour aud con Two phases have been identified. The Early TIW struction material, for example maugrove poles, things phase evolved from the EIW tradition around the 4th which are well known to have been obtained from East century AD aud tenninated around the end of the 7th Africa (Hitti 1956; Ricks 1970). Hodges and White century. This has been found to have had a strong foun house (1983, p. 151) observe that: dation on the central coast of Tauzania and to have been related to the Azauiau period of the Graeco In southem Iraq, the annies of Zanj (African) slaves, who worked in the fields and sugar plantations, rebelled in 868. The slave revolt Roman aud Sassanid trade to East Africa. Iasted fourteen years, Abungu, G. 1989. Communities on the River Tana, Kenya: an oamerican Village, Flannery, K. (ed.), 345-65. New York: Ac.a archaeological study of relations between the Delta and River derrric Press. Basin 700-1800 AD. Cambridge: Unpublished Ph.D. Thesis. Burton, R. F. 1872. Zanzibar, City, Island and Coast, 2 vols. LO$-·. Abungu, G. H. & H. W. Mutoro 1993. Coast-Interior settlements don: Tinsley Brothers. and social relations in the Kenya coastal hinterland. In The Carlstein, T. 1980. 'lime, Resources, Society and Ecology. Lunq.: ;· Archaeology oj Africa: jood, metats amt towns, Shaw, T., P. J. J. InfotiyckAB. Sinclair, B. Andah & A. Okpoko (eds), 694-704. London: Carneiro, R. L. 1970. A theory of the origin of the state. Science_ Routledge. 169, 733-8. •. Adams, W. Y. 1979. On the argument from ceramics to history: a Casson, L. 1989. Periplus Maris Erythreai. Princeton: Universf.fy challenge based on evidence from medieval Nubia. Current Press. · Anthropology 20(4), 727-44. Chami, F. A. 1988a. A coastal EIA site in Kisarawe district Nya:J;.e Adamowicz, L. 1990. Newly discovered Stone Age cultures in Akuma 30, 34-5. i· Nampula Province, northern Mozambique. In Urban Origins in - 1988b. The EIA Site in Kisarawe. Unpublished MA th'18./s; Eastem Africa: proceedings oj the 1990 workshop Harare and Brown University. ~ Great Zimbabwe, Sinclair, P. J. J. & G. Pwiti (eds), 137-97. - 1990. Changwehela: a submerged slave town of Bagamoyo: :in Stockholm: The Swedish Central Board of National Antiquities. Urban Origins in Eastem Aftica: proceedings oj the 1990 work Alexander, C. S. 1968. The marine terraces of the north-east coast slwp Harare and Great Zimbabwe, Sinclair P. J. J. & G. pvefi .- ofTanganyika. ZRitschriftfar Geomorphologie 7, 133-54. (eds), 124-8. Stockholm: The Swedish Central Board ofNatio::· - 1969. Beach ridges in north-eastem Tanzania. The. Geographical Antiquities. Review 59, 104-22. - l 992a. Current archaeological research in the Bagamoyo Q~..:' Allen, J. de V. 1974. Swahili culture reconsidered. Azania 9, 105- trict, Tanzania. In Urban Origins in Eastern Africa: proceedi ·· 38. oj the 1991 Workshop in Zanzibar, Sinclair, P. J. J. & A. J. - 1980. Settlement patterns on the East African coast, c. AD 800- (eds), 16-34. Stockholm: The Swedish Central Board ofNatio 1900. In Proceedings oj the 8th Panafrican Congress oj Prehis Antiquities. tory and Quaternary Studies, Leaky, R. E. & B. A. Ogot (eds), - 1992b. Limbo: early iron-working in south-eastern Tanzama· 361-3. Nairobi: InternatiOnal Louis Leakey Memorial Institute Azania 27, 45-52. .. for African Prehistory. Chanudet, P. C. 1988. Contribution a l'etude du peuplement de .f - 1982. The Sbiraz problem in East African coastal history. Paid de Moheli. Paris: Unpublished Ph.D. thesis. euma 28, 9-27. Childe, V. G. 1951. Social Evolution. London: Watts & Co. '' - 1983. Shungwaya, the Mijikenda and the traditions. Interna Chittick, N. 1958-62. Annual Repor! of the Department of Antiq\ii tional Journal ojAfrican Historical Studies 16, 455-85. ties. Dar-es-Salaam. VI - 1983. Swahili Origins. London: James Curry Ltd. - 1962. Recent discoveries in Tanganyika. Actes du congF1 Amadi, I. R. 1989. The northem factors in the history of sub-Saha Panafricain de pr6histoire et de 1'6tude du quatemaire. Tervm:~'. ran Africa: the Hamitic hypothesis revised. Transafrican Journal Musee Royal de l'Afrique Centrale 3, 215-23. ·~ ojHistory 18, 80-9. - 1965. The 'Shirazi' Colonization of East Africa. Journal ofA', Ammerman, A. J. 1981. Surveys and archaeological research. can History 6, 275-94. ~ Annual Review ojAnthropology I 0, 63-88. - 1966a. Unguja Ukuu: the earliest imported pottery Baker, E. C. 1941. Notes on the Shiraz of East Africa. Tanganyika Abbasid dinar. Azania l, 161-3. Notes and Records 11, 1-11. 1966b. Kilwa: apreliminaryreport.Azania I, 1-36. .. Binford, L. R. 1968. Archaeological perspectives. In New Perspec - 1967. Discoveries in the Lamu archipelago. Azania 2, 37-68. --~ tives in Archaeology, Binford, S. & L. R. Binford (eds), 5-32. - 1969.An new look at the history of Pate. Journal ojAfrican f{ts Chicago: Aldine. tory 10 (3), 375-91. ,. - 1972. An Archaeological Perspective. New Seminar Press. - 1970. Relics of the past in the region of Dar-es-Salaam. Tanzati/ Boum, D. 1978. Cattle, rainfall and tsetse fiies in Africa. Journal of Notes and Records 71, 65-8. Arid Environment 1, 49-61. - 1974. Kilwa: an Islamic trading city on the East African coast4_ Bower, J. 1986. A survey of surveys: aspects of smface archaeol vols. The Finds. Nairobi: British Institute in Eastem Africa. ogy in sub-Saharan Africa. African Archaeological Review 4, 21- - 1975a. An early salt-working site on the Tanzanian coast. Aza~i 40. 10, 151-3. - 1988. Evolution of Stone Age food-producing cultures in Bast - 1975b. The peopling ofthe EastAfrican coast. In EastAfrica Africa. In Prehistoric Cultures and Environments in the Late the Orient: cultural syntheses in pre-colonial times, Chittick, ,N·: Quaternary oj Africa, Bower, J. & D. Lubell (eds), 91-114. & R. Rotberg (eds). New York: Africana Publishing Company,l Oxford: British Archaeological Report. - 1977. The Bast Coast, Madagascar and the Indian Ocean. In Tb ,· Brunfiel, E. 1976. Regional growth in the eastem valley of Mexico: Cambridge History ojAfrica Vol 3, R. Oliver & J. D. Page (eds),. a test of the 'Population Pressure' hypothesis. In The Early Mes- Cambridge: Cambridge University Press. -,, · 103 - 1984. Manda: excavations at an island porton the Kenya coast. Fawcett, W. & A. LaViolette 1990. Iron Age settlement around Nairobi: British Institute in Eastem Africa. Mkiu, south-eastem Tanzania. Azania 25, 19-25. Clarke, D. L. 1972. Models and paradigms in contemporary archae Flannery, K. V. 1976a. Sampling by intensive surface collection. In ology. In Models in Archaeology, Clarke, D. L. (ed.), 1-60. Lon The Early Mesoamerican village, Flannery, K. V. (ed.), 51-62. don: Methuen. London: Academic Press. - 1978. Analytical Archaeology, 2nd edition. New York: ColUmbia Flannery, K. V. 1976b. The trouble with regional sampling. In The University Press. Early Mesoamerican Vzllage, Flannery, K. V. (ed.), 159-60. Lon Cole, S. 1953. The Stone of Bast Africa. In History af East Africa, don: Academic Press. Oliver, R. & G. Mathew (eds). Oxford: Clarendon Press. Freeman-Grenville, G. S. P. 1958. Some recent archaeological work - 1954. The Prehistory ojEast Africa. Montreal: Penguin Books. on the Tanganyika coast. Man 58 (155), 106-12. Collett, D. P. 1985. The spread of early iron-producing communi - 1960. East African coin finds and their historical significance. ties in eastern and southern Africa. Unpublished Ph.D. thesis, Journal afAfrican History I, 11-37 Cambridge University. - 1962. The Medieval History oj the Coast oj Tanganyika. London: - 1993. Metaphors and representations associated with precolonial Oxford University Press. iron-smelting in eastem and southern Africa. In The Archaeology ~ 1975. The East African Coast: select documents from the first to oj Africa: jood, metats and towns, Shaw, T., P. J. J. Sinclair, B. the earlier nineteeth century. Oxford: Clarendon Press. Andah &A. Okpoko (eds), 500-11. London: Routledge. - 1988. The Swahili Coast: 2nd to J9th centuries. London: Vario Collett, D. P. & P. T. Robertshaw 1983. Pottery traditions of early rum Reprints. pastoral communities in Kenya. Azania 18, 107-25. Garlake, P. S. 1978. Review of the later prehlstory of eastern and Cooke, H. J. 1974. The coastal geomorphology ofTanga, Tanzania. southern Africa. Journal ojAfrican History 20, 457. Geographical Review 64, 515-35. Gibbon, G. 1984. Anthropological Archaeology. New York: Colum Coupland, R. 1938. East Africa and its Invaders. Oxford: Claren bia University Press. don Press. - 1989. Explanation inArchaeology. New York: Basil Blackwell. Cruz e Silva, T. 1977. First indication of early IronAge in southem Gould. R. A. (ed). 1978. Exploration.>1 in Etlmo-archaeology. Albu Mozambique: Matola IV 1/68. In Proceedings ojthe 8th Panajri~ querque: University ofNew Mexico Press. can Congress oj Prehistory and Quatemary Studies, R. E. Gramly, R. M. 1978. Expansion of Bantu-speakers versus develop Leakey & B. Ogot (eds), 349. Nairobi: The International Louis ment of Bantu Ianguage in situ: an archaeologist's perspective. Leakey Memorial Institute for African Prehistory. South African Archaeological Bulletin 33, 107-12 Daniel, G. E. 1978. One Hundred and Fifty Years af Archaeology. Gray, J. 1951-1952. A bistory of Kilwa. Tanganyika Notes and London: Duckworth. Records 31, 1-37. Datoo, B. A. 1970. Rhapta: the location and importance of East - 1963. Zanzibar and the Coastal Bell, 1840-84. InHistory ofEast Africa's first port. Azania 5, 65-77. Africa, Oliver, R. & G. Mathew (eds). Oxford: Clarendon Press. - 1975. Port Development in East Ajrica. Dar-es-Salaam: Bast Greenberg, J. H. 1963. Studies in African Linguistic Classification. African Llterature Bureau. New Haven: Compass Press. Doran, J. E. & F. R. Hodson 1975. Mathematics and Computers in Guthrie, M. 1962. Some developments in the pre-history of the Archaeology. Edinburgh: Edinburgh University Press. Bantu languages. Journal ojAfrican History 3(2), 273-82. Drennan, R. D. 1976. Religion and social evolution in formative Gräslund, B. 1987. The Birth oj Prehistoric Chronology. Cam Mesomerica. In The Early Mesoamerican Vrllage, Flannery, K. bridge: Cambridge University Press. (ed.), 345-68. New York: Academic Press. Hall, M. 1981. Settlement Patterns in the !ron Age oj Zululand. Duarte, R. -T. 1993. Northem Mozambique in the Swahili world. Cambridge: British Archaeological Repor!. Studies in African Archaeology 4. Maputo: Eduardo Mondlane - 1983. Tribes, traditions and numbers: the American mode! in University: Stockholm: The Swedish Central Board of National southem African Iron Age ceramic studies. The South African Antiquities; Uppsala: Societas Archaeologica Upsaliensis. Archaeological Bulletin 38, 51-7. Ehret, C. & M. Posnansky 1982. The Archaeological and Linguistic Hamilton, A. C. 1982. Environmental History oj East Africa. Lon Reconstruction of Ajrican History. Berkeley: University of Cali don: Academic Press. fornia Press. Harding, J. R. 1960. On Some Crucibles and Associated Finds from Ellen, R. 1982. Environment, Subsistence and System. Cambridge: the Coast ofTanganyika. Man 60, 136-9. Cambridge University Press. Harris, M. 1968. The Rise oj Anthropological Theory. New York: Elton, F. 1874. On the copal districts south of Dar-es-Salaam, East Harper and Row Publishers. African coast. The Journal ojthe Royal Geographical Society 44, Hempel, C. G. 1965. Aspects of Scientific Explanation. New York: 230-31. The Free Press. • Fadiman, J. 1973. Early History of the Meru of Mt. Kenya. Journal Hinnebusch, T. J. 1976. The Shungwaya Hypothesis: a linguistic · afAfrican History 14, 9-27. reappraisal. In East African Culture History, Gallagher, T. (ed.). Fagan, B. M. 1965. SouthernAfrica during the IronAge. New York: New York: Syracuse University. Frederic Praeger. Hirth, K. G. 1978. Interregional trade and the formation of prehis FAO. 1978. Report on the Agro-ecological Zones Project, Vol. I. toric gateway communities. AmericanAntiquity 43 (1), 35-45. Rome. Hitti, P. K. 1956. History ojthe Arabs. London: Macmillan _FAO-UNESCO 1977. Soil Map oj the World. Volurne VI: Africa. Rodder, I. 1982. Theoretical Archaeology: a reactionary view. In Paris: UNESCO. Symbolic and Structural Archaeology, Rodder, I. (ed.). Cam FAO-UNESCO. 1988. Soil Map ofthe World. Wageuingen: ISRJC. bridge: Cambridge University Press. ~AO-UNFPA. 1980. Land Resources for Populations of the Future. Hodges, R. & D. Whitehouse 1983. Mohammed, Charlemagne and Rome. the Origins oj Europe. New York: Comell University Press. 104 Roll, A. 1990. West Africa archaeology. In A History of African - 1989. Pre-Islarnic Pottery in the Centre of Dar-es-Salaam. Kurfi Archaeology, Robertshaw, P. (ed.). London: James Currey. buka, 1(1), 45-52. Holthoer, R. 1977. New Kingdom Pharaonic sites: the pottery. Kenyon, K. M. 1931. Sketch of the exploration and set1tlernenfof Copenhagen: Munksgaard. the Bast Coast of Africa. In Zimbabwe Culture: ruins and Horton, M. 1984. The Early Settlement of the Northern Swahili tions, Caton-Thompson, G. (ed.), 260...74. Oxford: C!Etren1don Coast. Unpublished Ph.D. thesis, Cambridge. Press. - 1987a. The Swahili Corridor. Scientific American 255, 86-93. Kiriarna, H. 0. 1990. The Iron-Using Communities of Southein - 1987b. Early Muslim trading settlementS on the Bast African Kenya and North-eastern Tanzania c. 0-15th AD. Cambridle, coast: new evidence from Shanga. Antiquaries Journal 61, 290- Unpubl. Ph.D. Thesis. 322. Kirkman, F. S. A. 1959. Excavation at Ras Mkumbuu on the Is1arict - 1990. The Periplus and EastAfrica. Azania 25, 95-9. of Pemba. Tanganyika Notes and Records 53, 161-78. Horton, M. & C. M. Clark 1985. Zanzibar archaeological survey - 1964. Men and Monuments on the East Afn·can Coast. London: 1984-5. Azania 20, 167-71. Lutterworth Press. - Horton, M. & N. Mudida 1993. Exploitation of marine resources: - 1966. Ungwana on the Tana. The Hague: Mouton. evidence for the origin of the Swahili communities of East Kirwan, L. P. 1986. Rhapta, metropolis of Azania. Azania 21, 9'{'- Africa. In The Archaeology of Africa: food, metals and towns. 114. Shaw, T., P. J. J. Sinclair, B. Andah & A. Okpoko (eds), 673-93. Klapwijk, M. A. 1974. A preliminary report on pottery from tlje London: Routledge. North-eastern Transvaal, South Africa. South African Archae9- Huffman, T. N. 1970. The Early !ron Age and the spread of the logical Bulletin 29, 19-23. Bantu. South African Archaeological Bulletin 25, 3-21. Kroeber, A. 1944. Configurations of Culture Growth. Berkeli)\: - 1976. Gogomere pottery from the Tunnel site, Gogomere. South University of California Press. . AfricanArchaeological Bulletin 31, 31-53. Kuhn, T. S. 1970. The Structure oj Scientific Revolutions. Chicaib: : - 1980. Ceramics, classification and ceramic entitieS. African Stu The Univers_ity of Chicago Press. dies 39(2), 123-74. Lamm, C. J. 1935. Glass from Iran in the National Museum, StoC:~- - - 1982. Archaeology and Ethnohistory of the African Iron Age. halm. London: Kegan Paul. Annual Review ofAnthropology 11, 133-50. Lane, A. 1947. Early Islamic Pottery: Mesopotamia, Egypt a~d - 1984. Leopard's kopje and the nature of the Iron Age in Bantu Persia. London: Faber and Faber. Africa. Zimbabwea 1. Rarare: National Museums and Monu Lane, A. & R. E. Serjeant 1948. Pottery and glass fragments from ments Administration of Zimbabwe. the Aden littoral, with historical notes. Journal oj the Royal A:Si- Hulth6n, B. 1974. On the Documentation oj Pottery. Lund: Acta atic Society, 108-33. .. · Archaeölogica Lundensia. LaViolette, A. 1989. Preliminary Report: Archaeological Investii-3. - 1977. On Ceramic Technology during the Scanian Neolithic and tions at Pujini, Pemba June-July 1989. Unpublished repor! of Bronze Age. Stockholm: Akademilitteratur. Zanzibar Antiquities Department. · - 1988. Report on a pilot investigation of ceramics from Mozarn LaViolette, A., W. Fawcett, & P. Sclunidt 1989. The Coast and tlie bique and Zimbabwe. In Analyses oj Slag, /ron, Ceramics and Hinterland: University of Dar-es-Salaam Archaeological Fie~d Animal Banes from Excavations in Mozambique, Sinclair, P., M. Schools, 1987-88. Nyame Akwna 32, 38-46. .· •. Törnblom, C. Bohm, B. Sigvallius & B. Hulten (eds), 35-53. Lindahl, A. 1986. Information through Sherds. Lund: Studies'in' Maputo: Eduardo Mondlane University; Stockholm: The Swed Medieval Archaeology 3. Lund Ph.D. thesis. ish Central Board of National Antiquities Maggs, T. & G. Whltelaw 1991. A review of recentArchaeologic~: Huntingford, G. W. B. 1963. The Peopling of the lnterior of East Research on Food Producing Communities in Southem Afri~. Africa by its modem inhabitants. In History oj East Africa, Journal ojAjrican History 32, 3-24. Oliver, R. & G. Mathew (eds). Oxford. Masao, F. 1979. The Later Stone Age and the Rock Paintings-1f:' - 1980. The Periplus oj the Erythreaan Sea. London: The Hakluyt Central Tanzania. Wiesbaden: Frantz Steiner Verlag GMBH. '( - Society. Masao, F. & H. Mutoro 1988. The East African coast and the - Ingrarns, W. H. 1931. Zanzibar: its History and its People. London: Comoro Islands. fu General History oj Ajrica, Frasi, M. L. & I:~ Wotherby. Hrbak (eds), 586--016. London: Heinernan/UNESCO. lngrams, K. 1962. A History of East Africa. London: Longmans. Mathew, G. 1963. The EastA:frican Coast Until the Coming oft_!le" Jones, A. M. 1971. Africa and Indonesia. Leiden: E. J. Brill. Portuguese. In History oj East Ajrica, Oliver, R. & G. Math~W Juma, A. 1993. Unguja Ukuu: chronology of the site. Paper pre (eds), 94-168. Oxford: Clarendon Press. sented at the World Archaeological Inter-conference 1993 in - 1975. The Dating and the Significance of the Periplus of tb Mombaså. Erythrean Sea. In East Africa and the Orient, Chittick, N. & R: Juma, A. & A. Löfgren 1992. A site guide to Unguja Ukuu and the Rotberg (eds), 147--63. New York: Africana Publishing co. 1 fieldwork done in 1990. In Urban Origins in EastemAfrica: pro Mattem, E. 1992. Kilwa 1989. Excavation Report. In Urban Ofi ceedings oj 1991 Workshop in z.anzibar, Sinclair, P. & A. Juma gins in Eastern Africa: proceedings oj the 1991 Workshop in Zah~: (eds), I Q...14. Stockholm: The Swedish Centra!Board of National zibar, Sinclair, P. & A. Juma (eds), 36-57. Stockholm: The Antiquities. Swedish Central Board of National Antiquities. · , Kaaya, C. Z. & ·w. Boenigk 1986. Sedimentological and geomor Mclntosh, B. G. 1968. The Eastern Bantu Peoples. In Zamani,_:a, phological studies on the Pleistocene Tanga terrace north of Dar survey of East African History, Ogot, B. A. & J. A. Kieran, 19~.-: es-Salaam. Zeitschriftfii-r Geomorphologie 20, 304-15. 215. Nairobi. Karoma, J. 1982. New Evidence of Palaeohistoric Human Activi Merwe, N. J. van der. 1980. Production of high-carbon steel in 1'1e_ ties along the Kilwa Coast, Southem Tanzania. University of African Iron Age: the direct steel process. In Proceedings of tfle Dar-es-Salaam. Unpublished paper. 8th Panafrican Congress oj Prehistory and Qurternary Studies, 105 Leakey, R. E. & B. A. Ogot (eds), 331-4. Nairobi: International in the American southwest. Cambridge: Cambridge University Louis Leakey Memoria1 lnstitute for African Prehistory. Press. Milne, G. 1935. Some suggested units of classification and map Pope, A. U. 1939. A Survey oj PersianArt 2, 1466-1470. Oxford. ping, particularly for EastAfrican Soils. Soil Research 4 (3), 13- Posnansky, M. 1961. Dimple-based Pottery from Uganda. Man 61, 198. article 168. Morais, J. 1988. The Early Farming Communities of Southem - 1967. The Iron Age in East Africa. In Background to Evolution in Mozambique. Studies in African Archaeology 3. Maputo: Edu Africa, Bishop, W. W. & J. D. Clark (eds), 629-49. Chicago. ardo Mondlane University & Stockholm: The Swedish Central - 1968. Bantu Genesis: archaeological reflections. Journal ojAfri Board ofNational Antiquities. can History 9(1), 1-11. Morgan, W. T. W. 1973. East Africa. London: Longman. Pouwels, R. L. 1974. Tenth-centmy Settlement on the EastAfrican Msemwa, P. 1990. The Determinants of Secondary Refuse Disposal Coast. Azania 9, 65-74. Patterns: an ethnoID-chaeological study. In Urban Origins in East - 1987. Horn and Crescent: cultural change and traditional Islam em Africa: proceedings ofthe 1990 Workshop Harare and Great on the East African coast, 800-900. Cambridge: Cambridge Uni Zimbabwe, Sinclair P. & G. Pwiti (eds), 116-20. Stockholm: The versity Press. Swedish Central Board of National Antiquities. Radimilahy, C. 1992. Travaux de reconnaissance l'arch6ologique a - 1992. The nature of the Dar-es-Salaam archaeological record. In Mahilaka en 1989-1990. In Urban Origins in Eastem Africa: Urban Origins in Eastern Africa: proceedings of the 1991 Work proceedings of the 1991 Workshop in Zanzibar, Sinclair, P. J. J. & shop in Zanzibar, Sinclair, P. & A. Juma (eds), 58-67. Stock A. Juma (eds), 222-9. Stockholm: The Swedish Central Board of holm: The Swedish Central Board ofNationalAnt:i-quities. National Antiquities. Mturi,A. 1974. A guide to the Ruins oj Kaole. Dar-es-Salaam: Min Rehder, J. E. 1986. U se of prehistoric air in primitive furnaces: istry ofNational Culture and Youth, Antiquities Division. comment on views of Avery and Schmidt. Journal oj Field Murdock, G. P. 1959.Africa, its Peoples and Their Cultu,-e Hlstory. Archaeology 13, 351-3. New York: McGraw Hill. Renfrew, C. 1969. Trade and culture process in Europe. Current Mutoro, H. 1987. An archaeological study of the Mijikenda Kaya Anthropology l 0, 151-69. settlement. Unpubl. PhD thesis, University of California, Los - 1989. Comments on archaeology into the 1990s. Norwegian Angeles. Archeological Review 22, 33-41. Nicholls, C. S. 1971. The Swahili Coast. London: George Allen & Rice, P. M. 1987. Pottery Analysis. Chicago: Chicago University UnwinLtd. Press. Nordström, H. Å. 1972. Neolithic and A-Group sites: The Scandi Ricks, T. M. 1970. Persian Gulf Seafaring aod East Africa: ninth navian Joint Expedition to Sudanese Nubia, Il!. Stockholm: twelfth centuries. African Historical Studies III(2), 339-57. Scandinavian University Books. Robertshaw, P. 1990. Early Pastoralists oj South-westem Kenya. Nurse, D. 1983. A Linguistic Reconstruction of Swabili Origins. Nairobi: British Institute in Eastern Africa. Azania 18, 127-50. Robinsons, A. E. 1937. The Shirazi Colonization of Bast Africa Nurse, D. & T. Spear 1985. The Swahili. Philadelphia: University of Tanganyika Notes and Records 3, 40-81. Pennsylvania Press. - 1939. Shirazi Colonization of Bast Africa: Vumba. Tanganyika Oliver, R. 1966. The problem of the Bantu expansion. Journal oj Notes andRecords 7, 92-112. African History 7, 361-76. Robinson, K. R. 1973. The pottery sequence from Malawi briefiy - 1978. The Emergeilce of Bantu Africa. Cambridge History oj compared with that already established south of the Zambezi. Africa, 373-74. Cambridge. Amoldia6(l8), 1-11. Odner, K. 1971a. Usangi Hospital and other archaeological sites in Rodney, W. 1972. How Europe Underdeveloped Africa. Dar-es north Pare Mountains, north-eastem Tanzania. Azania 6, 89-130. Salaam: Tanzania Publishing House. - 1971b. A preliminary report of an archaeological survey on the Russell, E. W. 1977. The Natura! Resources oj EastAfrica. Nairobi: slopes of Kilimanjaro. Azania 6, 131-50. Bast African Literature Bureau. Pearce, F. B. 1920. 'Zanzibar: The Island Metropolis of Eastem Saberwal, S. C. 1967. Historical notes on the Embu of Central Africa. London: Fisher Unwin. Kenya. Journal ofAjrican History 8, 29-38. Phillipson, D. W. 1974. Iron Age history and archaeology in Zam Sackett, J. R. 1977. Style, Function, and Artefact Variability in bia. Journal ojAfrican History 16, 1-25. Paleolithic Assemblages. In The Explanation oj Culture Change, - 1975. The chronology of the Iron Age in BantuAfrica. Journal oj Renfrew, C. (ed.). Pittsburgh: University of Pittsburgh Press. African History 16, 321-42. Salmon, M. H. 1982. Philosophy and Archaeology. New York: Aca - 1976. The Barly Iron Age in eastern and southern Africa: critical demic Press. re-appraisal. Azania 11, 1-23. Sanders, E. R. 1969. The Hamitic Hypothesis: its origin and func - 1977. The Later Prehistory oj Eastem and Southem Africa. Lon tions intime perspective. Journal ofAjrican History, 10(4), 521- don: Heinemann. 32. 1979. Some Iron Age Sites in the Lower Tatla Valley. Azania 14, Sanders, W. T. & B. J. Price 1968. Mesoamerica. New York: Ran 155-60. don House. 1985. African Archaeology. Cambridge: Cambridge University Sanderson, S. K. 1990. Social Evolution. Cornwall: Basil Black Press. well. Plog, S. 1976. Relative efficiencies of sampling techniques for Schiffer, M. B. 1976. Behavioral Archaeology. London: Academic archaeological surveys. In The Early Mesoamerican Village, Press. Flannery, K. T. (ed.). London: Academic Press. Schiffer, M. B. ed. 1982. Advances in Archaeological Method and - 1980. Stylistic Variation in Prehistoric Ceramics: design analysis Theory. Vol. 5. London: Academic Press. 106 Schmidt, P. 1978. Historical Archaeology: a structural approach in Steward, J. & R. Murphy 1977. Evolution and Ecology. London:-' an African culture. Westport: Greenwood Press. University of Illinois Press. - 1980. Steel production in prehistoric Africa: insights from eth Stigand, C. H. 1913. The Land ofZinj. London. noarchaeology in West Lake, Tanzania. In Proceedings of the Sth Strandes, J. 1968. The Portuguese Period in Ea.st Africa. Dar-es Panafrican Congress of Prehistory arul Quaternary Studies, Salaam: East African Literature Bureau. Leakey, R. & B. A. Ogot (eds), 335-40. Nairobi: International Stuiver, M. & B. Becker 1986. High-precision decadal calibration, Louis Leakey Memorial Institute for African Prehistory. of the radiocarbon time scale, AD 1950--2500 BC. Radiocarbo~1 - 1983. Further evidence for an advance prehistoric iron technol 28, 863-910. ogy in Tanzania. Journal oj FieldArchaeology 10, 421-34. Sutton, J. 1966. The East African Coa.st. Nairobi: General Printers.' - 1988. Eastem expressions of the 'Mwitu' tradition. Nyame 1968. Archaeological sites in Usandawe. Azania 3, 167-74. Akurna 30, 36-7. - 1973. Early Yrade in East Africa. Nairobi: Afropress. Schmidt, P. & S. T. Childs 1985. Innovations and industry during - 1990. A Thousand Years of East Africa. Nairobi: British Institute; the early lron Age in east Africa: the KM2 and KM3 sites of in Eastem Africa. ri northern Tanzania. AfricanArchaeological Review 3, 53-94. Tampoe, M. 1989. Marine Yrade between China and the West:; 'l Schoff, W. H. 1912. The Periplus of the Erythrean Sea. New York. Oxford: British Archaeological Report. :-. i Shanks, M. & C. Tilley 1989. Archaeology into the 1990s. Norwe gianArcheological Review 22, 1-11. Temple, P. H. 1970. Geomorphology. Tanzania Notes and Recordi.,i Shaw, T. 1990. Bones in Africa: Presidential Address 1989. Pro 71, 20-54. ceedings oj the Prehistoric Society 56, 1-10. - 1971. Discussion of C. S. Alexander's paper on marine terraceJ:) Shennan, S. 1988. Quantifying Archaeology. Edinburgh: Edinburgh of the northeast coast of Tanganyika. Zeitschrift far Geomoph.o!1 University Press. ogie 15, 236-40. Shepard, A. 0. 1963. Cerarnics for the Archaeologist. Carnegie Theal, G. 1898. Records af South Eastem Africa (1893-1903, 9; Institution ofWashington, Publication 609. Washington D.C. vols ). Cape Town: Government of Cape Colony. Shepherd, G. 1982. The making of the Swahili: a view from the Trigger, B. G. 1978. Time and Traditions. New York: Columbia southem end of the East African coast. Paideuma 28, 129-147. University Press. _ Sinclair, P. J. J. 1982. Chibuene: an early tracling site in southem - 1989. Cornments on Archaeology into the 1990s. Norwegiaf! Mozambique. Paideuma 28, 149-64. Archeological Review 22, 28-31. - 1985. Ethnoarchaeological surveys ofthe Save river valley, south Trirningham, J. S. 1964. lslam in East Africa. London: Oxford Unfl central Mozambique: Working papers in Cultural Anthropology. versity Press. --:~ Uppsala: Departrnent of Cultural Anthropology. - 1975. The Arab Geographers and the East African coast. In Ea/t - 1987. Space, Time, Social Formation: a territorial approach to Africa ond the Orient, Chittick, N. & R. Rotberg (eds), 115-4~. the archaeology and anthropology oj Zimbabwe and Mozam New York: Africana Publishing Co. -"~ bique, c. 0-1700 AD. Aun 9. Uppsala: Societas Archaeologica Toussaint, A. 1966. History oj the Indian Or;ean. Londö#; Upsaliensis. Rouledge. - 1989. Urban Origins in Eastem Africa: a regional cooperation Van der Sleen, W. G. N. 1956. Ancient glass beads with special ref programme in archaeolögy and related disciplines. Kumbuka 1, erence to the beads of Bast and Central Africa and the Indiiifl 1:59-63. Ocean. Man 27, 203-17. - 1991. Archaeology in eastem Africa: an overview of current Verin, P. 1966. L'Archeologie a Madagascar (with English surr\- chronological issues. Journal ojAfrican History 32, 173-219. mary). Azania I, 119-38. : Sinclair, P. J. J., Kokonya, M., Meneses, M. & J. Rakatoarisoa - 1975. Austronesian contributions to the culture of Madagascaj:: 1992. The impact of information technology on the archaeology some archaeological problems. In Ea.st Africa and the Orient, :@. of southem and eastem Africa-the first decades. In Archaeology Chittick & R. Rotberg (eds), New York: Africana Publishing cd; • and the Injonnation Age: a global perspective, Reilly, P & S. - 1986. The History oj Civilization in North Madagascar. Rotte_t- Rahtz (eds), 30-41. London: Routledge. dam: A. A. Balkema. ' Sinclair, P. J. J., Morais, J., Adamowicz, L. & R. Duarte 1993. A Wannington, E. H. 1963. Africa in ancient and medieval times._·tn perspective on archaeological research in Mozambique. In The The Cambridge History af the British Empire, Walker, E. (ed), Archaeology ojAfrica: food, metals and towns, Shaw, T., P. J. J. vol. 8. Cambridge. Sinclair, B. Andah & A. Okpoko (eds), 409-31. London: Webb, J. L. 1985. The trade in gum arabic: prelude to French coil Routledge. quest in Senegal. Journal ojAfrican History 25, 129-31. Sinclair, P. J. J., Shaw, T. & B. Andah 1993. Introduction. In The White, F. 1983. The Vegetation ofAfrica. Paris: UNESCO. Archaeology oj Africa: jood, metals and towns, Shaw, T., P. J. J. Whitehouse, D. 1972. Excavations at Siraf: Fifth interim repoft. Sinclair, B. Andah & A. Okpoko (eds), 1-31. London: Iran 10, 63-88. ; Routledge. Willey, G. R. & J. Sabloff 1980. A History of American Archaeål- Soper, R. 1967a. Iron Age in north-eastem Tanzania. Azania 2, 19- ogy. San Francisco: W. H. Freeman and Company. <, 37. Wilson, T. H. 1982. Spatial analysis and settlement pattems on ~e - 1967b. Kwale: an early Iron Age site in south-eastem Kenya. EastAfrican Coast. Paideuma 28, 201-19. Azania 2, 1-17. Wittfogel, K. 1957. Oriental Despotism. New Haven: Yale Univ·IJ-.'. - !97Ja; Bantu Project. Azania 5, 1-4. sity Press. - 197lb. Ageneral review ofEarly IronAge of the southern half of Wright, H. T. 1984. Early seafarers of the Comoro Islands: t$-e Africa. Azonia 5, 5-37. Dembeni Phase of the IXth-Xth centuries AD. Azania 19, 13-6Q'. - 1971c. Early Iron Africa pottery types from East Africa: compar - 1993. Trade and Politics on the Eastem littoral of Africa, AD ative analysis. Azania 5, 39-52. 80G-1300. In The Archaeology ofAfrica: food, metals and towf;<; 107 Shaw, T., P. J. J. Sinclair, B. Andah & A. Okpoko (eds), 657-71. - 1981. Studies of shores and shore displacement on the southern London: Routledge. coast of Kenya-especially in Kilifi District. Geografiska Annaler Zwernemann, J. J. 1983. Culture HistoryandAfricanAnthropology. 63A, 303-10. Uppsala: Department of Cultnral Anthropology. - 1987. Sea-level changes an the East Coast of Africa