Board of Trustees, Boston University

Early Swahili History Reconsidered Author(s): Thomas Spear Reviewed work(s): Source: The International Journal of African Historical Studies, Vol. 33, No. 2 (2000), pp. 257290 Published by: Boston University African Studies Center Stable URL: http://www.jstor.org/stable/220649 . Accessed: 22/08/2012 22:08
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Boston University African Studies Center and Board of Trustees, Boston University are collaborating with JSTOR to digitize, preserve and extend access to The International Journal of African Historical Studies.

http://www.jstor.org

ThelnternationalJournalofAfrican

MstoricalStudies,

Vol.33, No. 2 (2000)

257

EARLY SWAHILIHISTORYRECONSIDERED*
By Thomas Spear
The identityand historyof the Swahili-speaking peoples of the East Africancoast have long beencontentious, andtheycontinueto be so today.This is partlya function of the politics of historywithinhighly diverseand stratified societies,in which people frequentlyinvoke historicalclaims regardingforeign origins, social and religious status, or genealogy to substantiatecontemporary political, social, or economicpositions. But it is also due to claims made about complex societies by outsiders-travelers,government officials,archaeologists, linguists,anthropologists, and historians-that have long engendereddiverseclaims about their identityand roles in easternAfricanhistory. Fifteenyears ago, Derek Nurse and I analyzedthese conflictingclaims in the light of the archaeological, evidenceavailable to propose a linguistic,and historical new interpretative framework for viewingSwahilihistory, largelyin termsof internal forces.' Such a view,we felt, made bettersense of the available evidence,and we hoped it would encourageothers to conduct furtherresearchand modify it as needed.Thathascertainly and some 270 books, theses,and articleshave happened, This degree of scholarlyactivityis as welcome as it is subsequentlyappeared.2 unusualin Africanhistory,makingthis an opportune momentto review the literature andto reassessthe stateof coastalhistoriography The Bases of Swahili Culture The Swahilihavelong stresseddifferencesbetweenthemselvesand theirneighbors, theirputative descentfromPersian andArabimmigrants and theirown emphasizing "civilized" ways (uungwana). Swahili towns also struck outsiders as obvious foreigntransplants: The religionwas Islamand the fundamental bases of the culturecame from In language andthe materials of everydaylife local Africaninfluence abroad. was stronger. The standing architecture however reflects little African
*

Earliervcrsions of this paperwerc prcsentedat thc School of Oriental and African Studies, University of London, Oxford University, and thc Univcrsity of Wisconsin, wherc I received numerous helpful commcnts and suggestions. In addition, Jonathon Glassman, Richard Helnm, Mark Horton, Dcrck Nurse, Stcphen Corradini,and Jan Vansina kindly read and commcnted on earlierdrafts, and RichardHelm generously madc his unpublishedresearchon Kcnya available to mc. I Derek Nursc and Thomas Spear, TheSwvahili: Reconstructingthe History and Languageof an African Society, 800-1500 (Philadelphia, 1985). 2 See Thomas Spear, "Swahili History and Society to 1900: A Classified Bibliography," History in Africa 27(2000), 339-73. I was unable, however, to review The Swvahili(Oxford, 2001), a major new synthesis of Swahili archaeology, ethnography, and history, by John Middleton and Mark Horton,as it was not yet publishedwhen this article went to press.

258

THOMAS SPEAR

influenceand its forms are entirely alien to those of the hinterland-its originswereoutsideEastAfricaandit is presentfully fledged in the earliest known buildings....The culturewas provincial-initiative was always from 3 abroad. Swahilibuildingsand styles, like theirMuslimreligion,seemed to emergeabruptly in the eleventhand twelfthcenturies, and with no known local antecedents, archaein Persiaor the Arabianpeninsula. ologistsconcludedthatthey musthaveoriginated Similarly, linguistspointedto the fact thatthe Swahililanguagewas first writtenin Arabicscript and containedlarge numbersof Arabicwords, while ethnographers noted that ustaarabu-Arabness-denoted the lifestyle of a sophisticated, urbane elite, manyof whomboreArabnamesand,some felt, appearance. Finally,historians of Swahili themselvesclaimingforeign origins. pointedto the historicaltraditions Theevidenceall seemed to point to the conclusionthatSwahilisocieties developed fromArabroots,putdownby merchants and clericswho began tradingand settling AD. alongthe EastAfricancoastduringthe latterfirstmillennium By the 1980s, though,these views were beginningto change. Archaeologists andMarkHortonstartedto show thatthe Muslim stone towns like Neville Chittick did not suddenlyemergeon a previouslybarren coast,but were precededfrom the of farmers, ninthcenturyby slowly expandinglocal communities fishers,and traders who capitalizedon their location to trade with visiting merchants.As trade expandedfrom the twelfthcenturyon, they began to build in coral and to adopt in the fourteenth and fifteenthcenturies. Islam,reachingtheir fullest development andstylesdevelopedfrom within,largelyfree Throughout, local buildingtechniques of foreigninfluence.4 were exploringthe Africanroots of the SwahililanAt the sametime,linguists guage. The Swahili dialects,now spoken from Somalia to Mozambique,are so closely relatedthey probablydeveloped from a single ancestrallanguage.5That language,Swahili,is a Bantu language,closely relatedto other languagesspoken more distantlyto alongthe Kenyacoast-the Sabakilanguages-which are related languages currently spoken in northeasternTanzania-the Northeast Coast NortheastCoast speakers their historicaldevelopment, languages.Reconstructing Tanzania coastby early in the first millenwere probably presentalongthe northern into Kenya.By the middle niumAD, fromwhencesome beganto spreadnorthwards of the first millennium AD, these Sabakispeakersextendedalong the Kenya coast, dialectsandlanguages, and by the intoseparate wheretheytoo beganto differentiate
3 P. S. Garlakc, The Early Islamic Architectureof the East Africani Coast (Nairobi, 1966),

2. 4 Janies dcVcrc Allen, "Swahili Culturc and thc Naturc of East Coast Scttlencent," Jouirtialof Africani Historical Stutdies 14 (1981), 306-34; H. Ncvillc Chittick, lintertiationial TradinigCeniter onzthe East Africani Coast (Nairohi, 1974); idem, Manda:. Kilwa: AtnIslanmic Excavations at atnIslanidPort onzthe Coast of KenYa (London, 1984); Mark Horton, Shaniga, Report (Nairobi, 1980). 1980: AnilItzerinm 5 For discussion of a possiblc altcrnativc niodel, seC TomnGuldemann, "Ist Swahili cinc monogenctischc Einheit? Bctrachtungenaus der Sicht periphercr Varictaten untcr besondercr Arbeitspapiere 30 (1992), 35-62. Bcrucksichtigungder Vcrbalmorphologic,"Afrikaniiscche

EARLY SWAHILI HISTORY RECONSIDERED

259

ninth century, Swahili had emerged, along with Comorian, liwana, Pokomo, and Mijikenda, as a distinct language.6 Swahili was thus clearly a Bantu language, both structurally and lexically, but how could one account for the large amount of Arabic vocabulary it contained and its use of Arabic script in the 19th century? All languages borrow words from other languages, and Swahili speakers have borrowed extensively from Arabic. But such borrowing was largely limited to certain cultural fields where Arab influence was strongest (such as jurisprudence, trade, religion, non-indigenous flora, and maritime affairs), and it occurred relatively late in the development of Swahili, when Arab influence along the coast was increasing. Arabic's influence on Swahili was thus relatively late and restricted,and it had little impact on the development of the basic elements of early Swahili. The new archaeological and linguistic interpretations of early Swahili history appeared to fit together remarkablywell and enabled us to hypothesize that Swahilispeaking societies first emerged around the Lamu Archipelago in the early ninth century, from whence Swahili speakers spread along the coast to establish a number of maritime settlements by the eleventh century, when increasing trade and Arab influence led them to expand their settlements more rapidly, build in stone, and adopt Islam. There were some problems with this interpretation,however, and it now needs rethinking in the light of new data. The first concerned the historical traditions of Swahili themselves. Most Swahili have their own traditions concerning their origins and subsequent development. Many tell of early immigrants from Shiraz who established the earliest settlements and ruling dynasties on the coast. It is unlikely that these traditions were literally true, however, for there is little archaeological, cultural, or linguistic evidence of Persian influence on Swahili, and the role of Shiraz in Indian Ocean trade had already declined before Shirazi began to appear in Swahili traditions. Not only that, but many of the traditions state that the Shirazi came not from Shiraz, but from Shungwaya, the legendary Swahili homeland along the Kenyan-Somali coast. Those who claimed to be Shirazi, thus, were probably early Swahili who migrated south to establish themselves as the first families, or waungwvana, of the major towns of the coast. Given the common practice of adopting the names of prestigious places as family names, or nisbas, they must have taken the name Shiraz as their nisba, just as many were to later take names from Oman in the nineteenth century.7

6 Nurse and Spear, The Swvahili, 32-67; Thomas J. Hinnebusch, "Swahili: Genetic supplement6 (1976), 95-108; Thomas Affiliations and Evidence," Sticdiesin African Lingutistics, J. Hinnebusch, Derek Nurse, and M. J. Mould, Stuidies in the Classification of Eastern Bantiu (Hamburg, 1981); Derek Nurse, "A TentativeClassification of the Primary Dialects of L.,anguages Philippson, "The Bantu Languages Swahili," SUGIA 4 (1982), 165-205; Derek Nurse and G6rard of East Africa: A Lexico-Statistical Survey" in EdgarC. Polomc and Clifford P. Hill, eds., in Tanzania (Oxford, 1980), 26-67; idemIl, "Historical Implications of the Language Langtuage Map of East Africa" in L. Bouquiaux, ed., L'Erpansion Bantou(e (Paris, 1980), II, 685-714. 7 Nurse and Spear, The Swvahili, 68-79.

260

THOMAS SPEAR

The Shirazi traditions also relate how the Shirazi paid tribute to the local inhabitants, married their daughters, and had sons by them, who then became the legitimate rulers of the town through their dual inheritance.The Shirazi traditions are thus elegant models of the historical development of Swahili societies. To be a Shirazi was to be a real Swahili who settled along the coast, paid tribute to and intermarriedwith the local inhabitants, and subsequently became the ruling class. Subsequent traditions expanded on this history to show how the towns then developed into complex, stratified societies with the expansion of trade from the twelfth to the fifteenth century.8 A second problem, not so easily resolved, concerned a chronological gap in the archaeological, linguistic, and documentawyevidence between the second and ninth centuries. Linguists feel that the first Northeast Coast (NEC) speakers reached the Tanzanian coast no later than the second-third century AD, contemporary with the earliest evidence for Kwale ware pottery and associated iron working. The distribution of Kwale ware and NEC also coincides, making the association of NEC speakers with Kwale ware and the early iron age likely. The earliest attested date for the next series of settlements-those of the earliest Swahili sites associated with Tana ware (also known as Wenje ware, Kitchen ware, Triangular Incised ware, or Maore ware)-was not until the eighth century, a period linguists associate with the emergence of Swahili as a distinct language, however, leaving a gap of six centuries between initial coastal settlements associated with NEC and Kwale ware and the subsequent emergence of Swahili and Tana traditions in northern Kenya. Similarly, no documents exist between the earliest references to coastal towns in the Periplus and Ptolemy's Geogrcaphyin the first and second centuries AD and those of Arab travelers visiting the coast from the tenth. While NEC was slowly evolving into Sabaki and Swahili, then, there was no archaeological data for the development from Kwale to Tana ware, if indeed they were related, or documentary evidence for other events along the coast. Further,we associated the subsequent migrations of Pokomo and Mijikenda to their current locations with their own traditions of having been forced to migrate from Shungwaya by Oromo expanding down the coast in the early sixteenth century. There was thus an even larger gap of thirteen centuries between the earliest known Bantu settlement behind the Kenya coast and the subsequent settlement of the area by its current Mijikenda inhabitants.Was the hinterland vacant all the intervening time? Had Sabaki speakers really all gone north prior to returning south from the 9th century, or had they simply gradually expanded along the Kenya and Somali coasts, during which time their dialects slowly differentiated into the languages presently spoken in the area?9In short, while the Shungwaya traditions made sense linguistically in terms of the probable historical development of the Sabaki languages and influences on them, there was a large gap in the archaeological and documentary data that needed to be accounted for.

8 Ibid., 80-98. 9 As subsequentlyarguedby MartinWalsh, "MijikcndaOrigins:A Review of the Evidence," TransafricanJo4 rna! of History 21 (1992), 1-18.

EARLY SWAHILI HISTORY RECONSIDERED

261

The Archaeology of Early Swahili Societies A recent explosion of archaeological researchhas begun to fill out the historical landscape. This research has been of two types:individual site reports,which stress the internal of individual and area surveys,which show development communities, the relationship of Swahilicommunities with one another andwith others. MarkHorton's comprehensive site report,Shanga, is a modernclassic, carefully reconstructing the development of one of the oldesttownsof the coast from its beginningsin the 8th centuryto its demise in the 15th. Horton details the town's pre-Muslimorigins, something that had only been implicit in Chittick's earlier reportson KilwaandManda,thuscharting for the firsttime the evolutionof a small coastalfishing and iron workingcommunityinto a large and prosperousMuslim town.10Horton'sstudyis a modelof carefulstratigraphy trading and precisedating, based on radiocarbonand importedpottery,"'though his historical inferences and he sometimescomes to occasionallymove beyond the bounds of his data,12 13 differentconclusionsin different publications. Accordingto Horton,Shangawas first occupiedca. AD 760-780 by a small groupof local fishermenand craftsmen.14 They lived in circular thatched or rectanhutswithina small enclosed area,fished in the adjacent gulartimber-and-daub bay, workediron,groundshell beads,and produceda characteristic early type of Tana tradition at the time.They also tradedwith potterytypicalof the coastandhinterland merchantsfrom the Persian Gulf, as evidenced by the presence of Persian waresdatedto the late 8th centuryand by the constructionof a Sassanian-Islamic wooden mosque and the first Muslim burialsby 850. In return,Shanga exported tortoiseshell,ambergris, rockcrystal,slaves,gumcopal,andiron. ivory,timber, Ironproduction andtradeincreased, andthe villagegrew in size and scale until whenpeoplebeganto use carved the 10thcentury, porites coral to build a new town wall, mosque, and houses outside the centralenclosure. The influence of Islam
10 Mark Horton,Shanga: TheArchaeology of a Muslim TradingCommunity on the Coast of East Africa (London, 1996). Cf. Chittick, Kilwa; Chittick, Manda; Mark Horton, "Asiatic Colonisation of the East African Coast: The Manda Evidence," Journal of the Royal Asiatic Society 1 (1986), 201-13.

11 Horton, Shanga, 14-16, 274-3 10, but cf. idem, "EarlyMuslim TradingSettlements on the East African Coast: New Evidence from Shanga,"AntiquariesJournal 67 (1987), 295-96. 12 E.g., the extent of putative Southern Custhitic influence on early Swahili speech and economy (Horton, Shanga, 410-11) or the reputed similarities between Mijikenda kayas and Swahili central squares[Horton, Shanga, 8, 85; idem, "Swahili Architecture,Space and Social Structure" in Michael Parker-Pearson and Colin Richards,eds., Architecture and Order (London, 1994), 153-54]. See below for furtherdiscussion of these points. 13 While some of these differences,such as the nature of Southern Cushitic influence on early Swahili (cf. Horton, "Early Settlement" and Horton, Shanga, 410-11), clearly represent developments in his own thinking, others, such as changes in overall dating and periodization(cf. Horton, "Muslim TradingSettlements,"303-08, and Horton,Shanga, 394-406), pass unnoted and unexplained. 14 I draw here largely on Horton, Shanga, 87-148, 394-427. Note differencesin dating and periodization in Horton, "MuslimTradingSettlements,"303-12.

262

THOMAS SPEAR

had builta new Friday mosque continuedto expand,and by 1050, the community for the first time. all the adultmen in the community largeenoughto accommodate of Tanaware, They also began to producetextiles,developedmore refinedvariants and importedgreaterquantitiesof white glazed, sgraffiato,and Chinese pottery. Muslimtrading community. Shangawas becominga prosperous when the Friday mosque was The town was destroyed shortly thereafter, burnedand the site leveled,but it began to recoverby 1075. A new largermosque and importscontinued.By 1250-1325, was built, new stone houses constructed, Arabia(as indicated by a shift from Persian Shangahadshiftedits tradeto southern textile and iron pottery),largely abandoned sgraffiatoto Arabianblack-on-yellow economy. People now built in coral production, andbecomea developedmercantile Shangareachedits peakbetween1325 and 1375, as people rag,usingmudmortar. pottery builtnew housesandtwo new mosquesof coralrag and lime,and imported the town went into slow decline, and by became more plentiful. Thereafter, as 1400-1425, it was in ruins,its Fridaymosqueburned,and its houses abandoned, left for Siyu andPate. its inhabitants natureof the town's development throughout. Hortonstressesthe evolutionary new materials and techBuildingplansandstylesevolved in situ as people adapted their niquesto olderforms.New stone houseswerebuilton older mud foundations; untillatein theirdevelopment; the overall to be pitchedandthatched roofscontinued even as individual townplanwas retained mosques,tombs,and houses were slowly to Mecca;and the first small timbermosques builtby foreignmerchants reoriented were repeatedlyredesigned,rebuilt in coral, and enlarged to accommodatean expandinglocal Muslim population.In the meantime,older style thatchedroof, the poorerpartsof town.15 timber-and-daub throughout buildingspredominated elsewhere along the coast expand Horton's Archaeological excavations in northernSomaliaconfirm picture.'6 Surveys west and east of Cape Guardafui Romanaccountsof earliercoastaltrade,from the 1st centuryBCto the 5th century AD, long before the adventof Islam.'7 Further south, Mogadishu, Gezira, and based on the presenceof SassanianBarawa maydate from the 8th or 9th century, Islamicpottery,while Merka,Munghia,Kismayu,Bur Gao, and the BajuniIslands 18 dateonly fromthe 11thcenturyor later.

15 Horton, Shanga, 26-62, 83-85, 170-229, 235-42. For an attempt to reassert the alien natureof Swahili architecture,see Linda W. Donley-Reid, "A StructuringStructure:The Swahili House" in Susan Kent, ed., Domestic Architecture and the Use of Space (Cambridge, 1990), 114-26, and Horton's response, "SwahiliArchitecture." 16 Capably summarizedby Chapurukha M. Kusimbain The Rise and Fall of Swahili States (WalnutCreek, Calif., 1999), 90-140, and Middleton and Horton, The Swahili. 17 H. Neville Chittick, "An Archaeological Reconnaissancein the Horn: The British Somali Expedition, 1975," Azania 11 (1976), 117-33; Matthew C. Smith and Henry T. Wright, 'The Ceramics from Ras Hafun in Somalia: Notes on a Classical Maritime Site," Azania 23 (1988), 115-41. 18 H. Neville Chittick, "An Archaeological Reconnaissance of the Southern Somali Coast," Azania 4 (1969), 115-30; idem, "MediaevalMogadishu," Paideuma, 28 (1982), 45-62; Hilary

EARLY SWAHILI HISTORY RECONSIDERED

263

EarlyRomantradeis also attestedin UngujaUkuu on the southeastcoast of Zanzibar from the 5th century,followed by Persiantradein the 8 h-9th. The site in the 9th century,however, and like the comparable terminates periodat Shanga,it containsplentyof local Tanawaresimilarto thatfound elsewherebut no mosques or stone buildings of any sort. Subsequent sites on Zanzibarand PembaMkokotoni, Kizimkazi, Ras Mkumbuu, Mkia wa Ngombe, Mtambwe Mkuu, Zanzibartown, and Tumbatu-all date from the I Ith and 12th centuriesand are Muslim sites lackingearly local pottery,but with Sassanian,sgraffiato,and early Qingbaiglazedstoneware imports.'9 Within the LamuArchipelago and on the adjacent mainland, Manda,Pate,and with Shanga,based on the common presenceof Ungwanawere all contemporary and all becameMuslimcommercial earlyTanawareand Sassanian-Islamic pottery, centers,producingiron and copper work,timber,ivory,beads, textiles,Tana ware, andexport.20 andlime for localconsumption Sites as early as these also occur at the southernend of the coast at Kilwa, Chibuene,and the Comoro Islands.The early finds at Kilwa, ca. 800-1150, are remarkably similarto those at Shanga and Manda,includingearly Tana tradition timber-and-daub pottery(termed"KitchenWare"by Chittick), rectangular houses, andweaving,and occasionalimported extensivefishing,beadgrinding, ironworking, concludesthatthe early settlement Sassanianandsgraffiatopottery.While Chittick theclose parallelswith Shangaand Mandaand the absence was not autochthonous, of Muslim influenceindicatethatthis too was an early pre-MuslimSwahili town, This was followedby a at the sametimeas those in the north.2' initiallyestablished in 1150-1300, with the firstcoraland lime buildings, periodof increasing prosperity the first importedglass more imported pottery(sgraffiatoand Chinese porcelain), of spindlewhorls,and the first mosques beadsand locally struckcoins, a profusion

Costa Sanseverino, "ArchaeologicalRemains on the Southern Somali Coast," Azania 18 (1983), 151-64. 19 Abdulrahman Worlds: Pottery from the M. Juma, "The Swahili and the Mediterranean Late Roman Period from Zanzibar,"Antiquity70 (1996), 148-54; Mark Horton and CatherineM. Clark, "ZanziharArchaeological Survey," Azania 20 (1985), 167-7 1; idem, The Zanzibar Archaeological Survey, 1984-85 (Zanzibar, 1985); Mark Horton, H. W. Brown and W. A. Oddy, "The Mtambwe Hoard,"Azania 21 (1986), 115-23. We await furtherreportson Horton's recent work on Zanzibar. 20 Chittick, Manda; Horton, "Asiatic Colonisation"; H. Neville Chittick, "Discoveries in the Lamu Archipelago,"Azania 2 (1967), 37-67; idem, "A New Look at the History of Pate," Journal of African History 10 (1969), 375-91; Thomas H. Wilson and Athman Lali Omar, "Excavationsat Pate on the East African Coast" in G. Pwiti and R. Soper, eds., Aspects of African Archaeology (Harare, 1996), 543-54; idem, "Archaeological Investigations at Pate," Azania 32 (1997), 31-76; George Abungu, "Pate: A Swahili Town Revisited," Kenya Past and Present, 28 (1996), 50-60; James S. Kirkman,(In,gwanaon the Tana (The Hague, 1966); George Abungu, "Communitieson the River T'ana,Kenya: An Archaeological Study of Relations bctween the Delta and the River Basin, 700-1890 AD" (Ph.D. thesis, Cambridge, 1990). 21 H. Neville Chittick, "Kilwa: A Preliminary Report," Azania 1 (1966), 5-25; idem, Kilwa, 1, 17-24, 235-45. For a recent overview of the archaeologyand history of Kilwa, see J. E. G. Sutton, "Kilwa: A History of the Ancient Swahili Town," Azania 33 (1998), 113-69.

264

THOMAS SPEAR

and Muslim graves. Kilwa had become "a substantialand prosperous town." growthbetween1300 and 1500, with further Finally,therewas a periodof marked increasesin imported pottery(now mostly Chineseceladon)and extensivebuilding changesin local pottery, anda decline in stone,dramatic new buildingstyles,marked in weavingandironwork, just like Shanga. was Chibuene,more than 1,500 km. south of Kilwain southernMozambique, also first occupiedin the 8th or 9th century,as indicatedby a range of imported materials, local potteryidenticalto that found in Manda,Shanga, Kilwa, and the Unlike the others,however, Comoros,and extensivebead and iron manufacturing. did not prosper, andit was abandoned ca. 1000 and not reoccuthe earlysettlement pied untilca. 1450.22 Extensivediscoveriesin the Comoro Islands confinn the simultaneousdevelComoopmentof townsalongthe lengthof the coast. While not Swahili-speaking, 300 riansspeaka closely relatedlanguageto Swahiliand theirislandcommunities, with those of the coast km. off the southernTanzanian coast,developedin parallel the Dembeni phase (9"'-10th itself. The earliestperiodof Comoriandevelopment, of a single main villageon each of century), was characterized by the establishment the four islands. People lived in rectangular pole-and-mud houses; fished extenrice,millet,coconut,legumes, sively withhook,line,and net;raisedgoats;cultivated and probablybananasand taro;producediron, madeplainand red-slippedpottery often identicalto Tana ware at Manda,Kilwa,and Chibuene;and importedsome The Dembenisites were,then,classic early Swahilisites, Sassanian-Islamic pottery. seeds for the first time,we andwith the firmdiscoveryby flotationof domesticated subsistedon bothfishingandfarming.23 can now say thatcoastalcommunities centuriesand a Tradeand towns in the Comoros expandedin the I "'-13"' withthe MiddleEast of smallerrural number villagesgrewup.Tradewas conducted andMadagascar opaquewhite (as established by the presenceof Sassanian-Islamic, glaze, and sgraffiatopotteryfrom the PersianGulf and chorite-schistvessels from Madagascar). People began to raise chickens and they continuedto make local potterysimilarto thatat Mandaand Kilwa.The first stone mosques were builtearly in the period, possibly over earliertimberones, and later rebuiltand enlarged.
22 Paul J. Sinclair, "Chibuene:An Early T'rading Site in Southern Mozambique,"Paideimtalc acd Social Fornation. A Territorial Approaclh to the 28 (1982), 149-64; idem, Space, Titmze c. 0-1700 (Uppsala, 1987), Anthropology,of Zimbabwe and Mozamiibiquce, Archaeology acmd 86-90. Sinclair has subsequentlysuggested pushing the date of Chibuene back to the mid-Ist millenniumbased on a single fragmentof green glazed ware, dated at Hafun from the Ist c. BC-5th c AD, but this is problematicgiven the prevalenceof local ceramics similar to those at Kilwa and dates: Paul J. Sinclair, "Archaeologyin Eastern Africa: An Overview of associated radiocarbon CurrentChronological Issues," Jouirnalof African Histoty 32 (1991), 190. 23 Henry T. Wright, "EarlySeafarersof the Comoro Islands: The Dembeni Phase of the IXth-Xth Centuries AD," Azania 19 (1984), 13-59; idem, "EarlyIslam, Oceanic Trade,and Town Development on Nzwani: The Comorian Archipelago in the XIth-XVth Centuries AD," Azacnia 27 (1992), 81-128; Claude Allibert, Mayotte (Paris, 1984); Claude Allibert, Alain Argant, and JacquelineArgant,"Le site de Bagamoyo (Mayotte,Archipeldes Comoros),"EtuldesOcEan indien 2 (1983), 5-40; idem, "Le site de Dembeni (Mayotte Archipel des Comoros)," EtuedesOceCan Indien 11 (1990), 63-172.

EARLY SWAHILI HISTORY RECONSIDERED

265

Comorian towns then entered their "Classic" period in the 14"'-15"' centuries as mosques were elaborately remodeled, small hamlets grew up around the larger towns. and new building revealed marked social differentiation. Similar Swahili- and Comorian-speaking communities also developed along the northern Madagascar coasts during the same periods. Many of the earliest settlements were mere shelters established for the annual trading seasons, but percentury.24 manent trading towns were well established by the 12"' While developments in the Comoros and Madagascar have been largely neglected in earlier studies of the coast, the strong parallels between their development and that elsewhere along the coast make it clear that the islands played an importantrole in the overall history. Far from lying remotely off the coast, they lay astride the main sailing route from Mozambique (and the gold fields of Zimbabwe) to Kilwa and the northerncoast, as vessels followed the winds and currents flowing out to Madagascar and the Comoros before turning back toward the coast to Kilwa.25 Coast and Hinterland Archaeologists have augmented site reports detailing the internal development of individual communities with area surveys exploring the external relations among Swahili communities and their neighbors. Such surveys were initially inspired by the wide distribution and general similarities between two different traditions of pottery found along the coast: (I) Kwale ware and related styles associated with the earliest ironworking and farming sites in northeastern Tanzania and southeastern Kenya in the early centuries AD and (2) Tana tradition wares found ubiquitously on early Swahili sites and throughout the hinterland from the 8"'or 9"'century.26 These traditions have been explored within the context of the spread of ironworking, agriculture, and Bantu languages throughout eastern and southern Africa over the past two or three millennia. Indeed, the three are often linked in a single phenomenon, the Early Iron Age Industrialor Chifumbaze Complex,27 but it is now thought that the different elements spread separately and only came together in

24 Henry T. Wright et al., "Ihe Evolution of Settlement Systems in the Bay of Bueny and Azantia 31 (1996), 37-73; Robert E. the Mahavavy River Valley, North-WesternMadagascar," in Julian Reade, ed., The lIldiall Dewar, "The Archaeology of Early Settlement of Madagascar" OceantintAntiquity (London, 1996), 471-86; Pierre Verin (David Smith, tr.), The Iliston, of CivilisationtintNorth Madagascar (Rotterdam, 1986), 67-69, 143-47, 380-97; Marie de Chantal Radimilahy, Mahilaka (Uppsala, 1998). 25 Felix Chami, ThleTantzatialnt AD (Uppsala, 1994), 37. Coast intthe First Millenntliumii 26 Paul J. Sinclair et al., eds., UrbatnOrigitnsint EasterntAfrica (8 vols) (Stockholm, 1988-1993); Adria LaViolette, William Fawcett, and Peter R. Schmidt, "The Coast and the 32 (1989), 38-46; Peter R. Schmidt Hinterland:UDSM Field Schools, 1987-88," NyvaneAkiimla (Dar intthe Vicitnityof Mkiu, Kisarawe D)istrict, Tantznttia et al., Archaeological Itnvestigatiotns es Salaam, 1992). 27 David W. Phillipson, The Later Prehistory of Easterntatid SoutherntAfrica (London, Archaeology, 2nd ed. (Cambridge, 1993), 187-98. 1977), 102-39, 145-50; idem, Africatn

266

THOMAS SPEAR

differentareasat differenttimes.28 Jan Vansinahas hypothesizedthat the parallel andfarming mayhave led to dramatic population growth, expansionof ironworking trade,developmentof large-scale expansion of local industriesand long-distance in local culturaltraditionsthroughout territorial polities,and greaterdifferentiation eastern AfricabetweenAD 800 and 1100. This was precisely the time when the to earliestSwahilitowns were beginningto develop,thus linkingtheirdevelopment African as well as to those takingplace across the eventstakingplacewithineastern IndianOcean. At the same time,if one could establishcontinuouslinks betweenKwaleware of Swahilisocieties andTanaware,thenone could also link the initialdevelopment to earlierironworking gap betweenthe ones, and the problemof the chronological two would be lessened.Thatis tempting, but therehas been considerable disagreebetweenthe two ceramictraditions. For scholarslike David menton the relationship Collett,who has developeda systematicstylistic comparisonof the two traditions, the fate of early Bantu farmersand they are unrelated, leavingthe gap unbridged, of the Swahili a mystery.29 ironworkers Following unknown,and the antecedents Collett,others,such as George Abungu,James de Vere Allen, RichardWilding, RandallPouwels,and MarkHorton,have looked to the SouthernCushiticPastoral of Swahili Tana tradition than Kwale for the stylistic antecedents Neolithicrather (such as Felix Chamiand his colwares.But most scholarsworkingin Tanzania WilliamFawcett, Peter RandiHaaland, leagues at the Universityof Dar es Salaam, and RobertSoper) as well as Schmidt,KurtOdner, Carolyn Thorp,NevilleChittick, some working in Kenya (Chapurukha Kusimba,Athman Lali Omar, Thomas see Tanaand Kwalewaresas closely Wilson,Richard Helm,andDavidPhillipson), fromKwaIe.311 withTanaevolvingdirectly related, While general stylistic classificationof differentceramic traditionsremains of specific stylistic features,their distribufairlysubjective, systematiccomparison tions, and the sites where they are found in associationprovide more principled has a fairlylimited bases for analysis. Kwaleware,the earlierof the two traditions, distribution focusedon the northeastern highlands, extendingfrom Mt. Mwangeain further northin Barawa and on the TanaRiver) the north(withpossibleoccurrences Tanzania throughsoutheastern Kenya(ShimbaHills, TaitaHills) and northeastern and Ngulu) to the Rufiji Delta (Kilosa, Limbo,and (Kilimanjaro, Pare,Usambara, is more by contrast, Mafia)in the south betweenAD 100 and 500.31 Tanatradition,
28 Jan Vansina, "A Slow Revolution: Farming in SubequatorialAfrica," Azania 29-30 (1994-95), 15-26. See also David L. Schoenbrun,A Green Place, A Good Place (Portsmouth, NH, 1998); ChristopherEhret,An African Classical Age (Charlottesville, 1998). 29 David Collett, "'I'he Spread of Early Iron Producing Communities in Eastern and SouthernAfrica" (Ph.D. thesis, Cambridge, 1985). 30 See discussion and sources below. 31 Robert C. Soper, "Kwale: An Early Iron Age Site in Southeastern Kenya," Azania 2 Tanzania,"Azania 2 (1967), 19-36; Collett, (1967), 1-17; idem, "IronAge Sites in North-eastern "IronProducing";KurtOdner,"Usangi Hospital and other ArchaeologicalSites in the North Pare Mountain,North-easternTanzania,"Azania 6 (19071), 89-130; idem, "A PreliminaryReport on Azania 6 (1971), 131-49; Schmidt et an Archaeological Survey on the Slopes of Kilimanjaro,"

EARLY SWAHILI HISTORY RECONSIDERED

267

widely distributed. It is, as we have seen, the diagnostictraditionfor the earliest Swahili sites the length of the coast, but it is also found up the Tana River, in andin Taita,Pare,the Usambaras, on Kilimanjaro, andNgulu.32 Mijikenda, of Kwalethusformsthe core of the wider distribution The distribution of Tana, evolved from the earlier.In many cases the two suggesting thatthe latertradition traditionsare found on the same or nearbysites. In South Pare, for example, a related style to Tana,Maore,is closely relatedto Kwalestylisticallyand succeeds it, while in North Pareand southernKilimanjaro, Kwaleand Maore wares are found mixedon the same sites andsometimeson the same shards.On Kilimanjaro thereis a close stylisticresemblance betweenthe two anda clearevolutionary sequencefrom Kwaleto Maoreandlaterstyles. In Ngulu, Kwale,Tana,and a latertradition follow in logicalsequence.33 one another havenow greatlyexpandedthe Chamiandhis colleaguesworkingin Tanzania fromKwale (termedEarlyIron number anddistribution of sites wherethe transition Ware,or EIW,by Chami)to Tana(termed Triangular IncisedWare,or TIW) can be seen in areassouthinto the Rufiji Delta and inlandup to 200 km. from the coast.34 These include a numberof new sites in centralTanzania(Kisarawe,Bagamoyo, Lugoba,Kilosa,Limbo,Dakawa, Misasa,Dar es Salaarn, Mpiji, Kaole, Kiwangwa, KwaleIsland,Nkukutu, andKivinja)in additionto early TIW Masunguru, Mkadini, sites previously explored in Mozambique (Matola, Chibuene, Momapo, and Nampula),the Comoros and Madagascar(Dembeni, Sima, Mro Dewa, Mbeni, Mbashili, and Irodo), mainlandTanzania(Kilwa, Amboni, and the northeast), Zanzibarand Pemba (Unguja Ukuu and Mkokotoni),and Kenya (Kwale, Tana, Wenje, Ungwana, Manda, Shanga, and Mijikenda).Several of the new sites, includingLimbo,Misasa, Dakawa,Kwale Island, Nkukutu,and Kivinja,contain

al., Archaeological Investigations; Carolyn Thorp, "Archaeologyin the Ngulu Hills: Iron Age and EarlierCeramics,"Azania 27 (1992), 21-44; Chittick, Kilwa, I, 106, 110. Kwale may extend south into Mozambiqueand South Africa, where a similar style, known as Matola, dates from the 2nd-6th century;Paul Sinclairet al., "A Perspectiveon Archaeological Research in Mozambique," in Thurston Shaw et al., eds., The Archaeology of Africa (London, 1993), 417-18. 32 Henry Mutoro, "An Archaeological Study of the Mijikenda Kaya Settlements on the Soper, "Iron Hinterlandof the KenyaCoast" (Ph.D. thesis, UCLA, 1987); Odner,"Kilimanjaro"; Thorp, "Ngulu Hills"; Felix Chami, "Limbo: Early IronTanzania"; Age Sites in North-eastern working in South-easternTanzania,"Azania 27 (1992), 45-52; Sinclair, "Archaeologyin Eastern Africa";HenryT. Wright,"Tradeand Politics on the EasternLittoral of Africa, AD 800-1300 in ThurstanShaw et al., eds., The Archaeology of Africa (London, 1993), 658-72. 33 Odner,"Kilimanjaro"; Thorp, "Ngulu Tanzania"; Soper, "IronAge Sites in North-eastern Hills"; Wilson and Omar, "ArchaeologicalInvestigations at Pate," 62; Chami, Tanzanian Coast; idem, "A Review of Swahili Archaeology," African Archaeological Review, 15 (1998), 199-218; William Fawcett and Adria LaViolette, "Iron Age Settlement around Mkiu, SoutheasternTanzania,"Azania 25 (1990), 19-26. 34 Differences in nomenclature, frequentlysite-specific, furtherconfuse the overall situation. While Kwale ware and EIW are the same style, as are Tanaand TIW, those working in Kenya tend to retain the earlier geographic referents,while those in Tanzaniapreferthe less specific stylistic throughout,as shown in Figure 1. terms. I use both Kwale/EIWand Tana/TIWinterchangeably

268

THOMAS SPEAR

EIW from the ISt to 4th centuryand early TIW from the 4IhK5rh on the same or adjacent sites, thuslinkingthe two temporally as well as stylistically.35 Early TIW sites are characterized by ironworking,beadmaking,weaving, extensivefishing and gatheringshell fish, raisinggoats and chickens,farming,and of the early TIW phases is that similarstyle trading.What is most characteristic potteryis foundall alongthe coast,thatthesewere often precededby EIW, and that EIWfeatures continued to appearon earlyTIW, includinglines manycharacteristic of punctates,comb-stamping, cross-hatchedbands, zigzag double incisions, and bevels, flutes,and thickenedrims,while new TIW elements,such as incised triangles, beginto appear on late EIWbeforeit was eclipsedand its featureslost on later TIW.Chamihas now classifiedthese sites to posit a clearevolutionthroughout the from EIW to TIW and laterforms: Plain Ware (PW), Neck coast and hinterland Punctating (NP), andSwahili,as shownin Figure1. Figure 1: Classification of Early Coastal Pottery (after Chami) Tradition EarlyIronWare(EIW) early/transitional middle late Plain Ware(PW) Neck Punctating (NP)
Swahili

Names Alternative Kwale

Dates l-5h c.
4-6h c.

Area NE high

Triangular IncisedWare(TIW) Tana,Wenje,Maore Dembeni Kitchen, EarlyKitchen

all south north

8-lOlh c.

10-13'hc. 11-14h c.
14-15"' c. all

TIWbeganto succeedEIW,retaining betweenthe 4th and manyof its features,

Dakawa,Unguja Ukuu, Pemba,and the Rufiji Delta), sites associated with iron and some pre-Islamicgreen- and blue-glazed Sassanian smelting, beadmaking, as the pottery.By the 8'hcentury,mostof the EIW featureshad been lost, however,
35 Chami, Tanzanian Coast, 13-23, 43-69, 90-93; Felix Chami and Paul Msemwa, "A New Look at Culture and Trade on the Azanian Coast," Current Anthropology, 38 (1997), 673-77; Chami, "Review of Swahili Archaeology";Fawcett and LaViolette, "IronAge Settlement aroundMkiu"; Schmidt, Archaeological Investigations; Randi Haaland,"Dakawa:An Early Iron Age Site in the Tanzanian Hinterland," Azania 29-30 (1994-95), 238-47; Wright, 'Trade and Politics"; Felix Chami and Emanuel Kessy, "ArchaeologicalWork at Kisiju, Tanzania, 1994," Nyame Akuma43 (1995), 38-45; Adria LaViolette and Jeffrey Fleisher, "Reconnaissanceof Sites Bearing TriangularIncised (Tana Tradition)Wareon Pemba Island, Zanzibar," Nyame Akuma 44 (1995), 59-65; Emanuel Kessy, "ArchaeologicalSites Survey from Kisiju to Dar es Salaam," Nyame Akuma 48 (1997), 57-69; Felix Chami and Paul Msemwa, "The Excavation at Kwale Island, South of Dar es Salaam," Nyame Akuma 48 (1997), 45-56; Felix Chami and Bertram Mapunda,"The 1996 Archaeological ReconnaissanceNorthof the Rufiji Delta,"Nyame Akuma 49 (1998), 62-78; Kusimba, Rise and Fall of Swahili States, 90-100.

EARLY SWAHILI HISTORY RECONSIDERED

269

of the coastaltowns began. TIW continuedto formative phase in the development developinternally duringthe succeedingPW phaseon the southerncoast (at Kilwa, Kaole, Changwechela,Kwale Island, Kivinja) and NP phase on the northern (Horton's Tana B & C; Wilson and Omar's Tana II & III), where it ultimately came to define a new "Swahili" phaseas a style all alongthe coast.36 RichardHelm has now affirmedsimilarlyclose relationsbetween Kwale/EIW and Tana/TIW on the Kenyancoast as well throughcarefulstratigraphic analysis and meticulousstylistic comparisonof individual featuresof the two. After conductinga surveyof sites associatedwith inlandMijikenda, he excavatedfive sites. contained One,Mgombani, EIW,transitional EIW-TIW,and TIW, revealing a clear transitionfrom EIW to TIW and a laterphase of TIW betweenthe 7th and gth centuries. Two othersites in Kwale(Chomboand Mteza)were late EIW/early TIW sites (8'h_IOthcentury)with some PW (9-1Othcentury)in the laterstages. Mtsengo in Kilifi displayeda transition from lateTIW to PW in the 14th_ 15'hcentury.And Mbuyuni in Kwale containedlate TIW datingfrom the 15'h to the 17'h century. Helmconcludesthattherewas continuous of the coastalhinterland from occupation the early stone age throughthe modernperiod,with considerablemutualinfluence and development betweenLSA and EIW as well as betweenEIW and TIW sites. EIWsites were clusteredin the most fertilezones aroundMombasaand date from andearlyTIWsites occupythe same areafrom 100BCto AD 600, while transitional AD 600 to 1000; laterTIW sites extendnorthand west into the drier uplandsand from 1000to 1650; and modernsites coverthe hinterland, hinterland from 1650 to the present.37 Given the overlapping distributions and stylistic patternsof Kwale/EIWand Chami's classificatoryschemaestablishinga clear transitionfrom the Tana/TIW, earliestironproducers on the coastto laterSwahilisites now appearsfundamentally correct.If so, Swahiliis now firmlylinkedwithearlierEIW and TIW traditions and backintothe earlierhistoryof the Bantu-speaking ultimately peoples. Scholarsworkingin northern Kenyahavechallengedthese conclusions,howto the PastoralNeolithictraditions ever,and look for Swahili's antecedents of central Kenya associatedwith SouthernCushiticspeakersinstead.Based largely on what they see as stylistic correspondences between PastoralNeolithic and Tana tradition andthe earlier pottery presenceof SouthernCushitic-speaking pastoralists along the TanaRiverand northern coast,they posit that SouthernCushites traded along the coast long before the first Bantuspeakersarrived early in the 1St millen36 Chami, Tanzanian Coast, 11-18, 90-93; Chami and Msemwa, "Culture and Trade"; Chami, "Review of Swahili Archaeology";Wright, 'Trade and Politics." Chami's classification parallels that for Shanga by Horton:Shanga, 243-264. 37 RichardHelm, "Conflicting Histories: The Archaeology of the Iron-working, Farming Communities in the Centraland SouthernCoast Region of Kenya"(Ph.D. thesis, Bristol, 2000), 115-39, 146-208, 224-27, 275-94; idem, Conflicting Histories: Archaeology and Identityon the Swahili Coast (forthcoming). In earlier surveys, Henry Mutoro and ChapurukhaKusimba also found early Tana/W throughoutthe area, dating perhapsfrom the 7th or 9th century, but their radiocarbondates are problematic and neither found any evidence of earlier EIW. Mutoro, "ArchaeologicalSurvey of the MijikendaKaya"; Chapurukha M. Kusimba, "The Social Context of Iron Forging on the Kenya Coast,"Africa 66 (1996), 386-410.

270

THOMAS SPEAR

multilingual society niumandtherebybecamethe nucleusof a dynamicmultiethnic, they claimthata PastoralNeolithic thateventually gave rise to the Swahili.Further, Swahilisites, as seen in the reputedpresenceof cattle cultural substratum underlies kraalsand bones at their lowest levels and the putativeprevalenceof Southern in Swahili.38 Cushiticvocabulary coast before the first While Southern Cushiticspeakersdid inhabitthe northern cultural influenceon early Swahili. Bantuspeakers appeared, theyhadlittleapparent is 300 km. from the coast and dates The closest Pastoral Neolithicsite, Marangishu, from the ISt century,too far awayand too early to influenceSwahili significantly. Swahiliis a Bantu languageand contains little SouthernCushitic vocabFurther, at its earlier stages,andthereis no evidenceof its emergenceas a ulary,particularly Nor is thereany archaeological evidenceof a pastoral mixed languageor pidgin.39 substratum in early coastal sites; there are no kraals,very few bones of goats or the early Rather, noneof cattlebeforeAD 1000 at the earliest.40 sheep,andvirtually to contemporary Swahili subsisted largely on fish and shellfish, both abhorrent Cushites. In sum, Southern Cushitic influence on early Swahili societies was CushiticOromoandSomalisince.41 minimal,as has been thatof Eastern evidenceare limof potterywithoutdirectarchaeological Stylisticcomparisons the claims for morethanlimitedSouthern ited,andwe can now probablydisregard Cushiticinfluenceon the development of Swahiliin the light of clearand directeviIf so, we can see a continuouspattern denceof the evolutionof TIW from EIW.42 from the early iron age to the present,but thatleaves further of development problems in its wake. What do we make,for example,of the emergenceof the earliest at opposite ends of the coast? As a Swahili communitiesnearlysimultaneously
38 Abungu, "River Tana," 114-75, 248-55; James deVereAllen, Swahili Origins: Swahili Culture and the Shungwaya Phenomenon (London, 1993), 28-35, 43-52, 99-112, 120-132; Richard Wilding, The Shorefolk:Aspects of the Early Development of Swahili Communities(Fort Jesus Occasional Papers No. 2, Mombasa, 1987), 43; Horton, "Early Settlement," 315; idem, "EarlyMuslim TradingSettlements," 315; idem, "EarlyMaritime Trade,"443-46, 454; Randall L. Pouwels, Horn and Crescent: Cultural Change and Traditional Islam on the East African Coast, 800-1900 (Cambridge, 1987), 9-22; but see Horton, Shanga, 243-70, 407-11 and Pouwels, "EastAfrican Coastal History,"Journal of African History 40 (1999), 290-91, both of whom have now become critical of this view. 39 For a brcakdownof Southern Cushitic vocabularyin NEC, Sabaki, Swahili, and other Sabaki languages, sec Dcrek Nurse and Thomas J. Hinnebusch, Swahili and Sabaki: A Linguistic History (Berkeley, 1993), 306-29, 577-691. 40 Horton, Shanga, 243-70, 407-11. 41 Chami, Tanzanian Coast, 97; Wilson and Omar, "Archaeological Investigations at Pate," 63-64; Sinclair, "Archaeologyin EasternAfrica."While Horton now accepts that therc is littlc direct archaeologicalor linguisticevidence of Southern Cushitic influence on early Swahili and is critical of Allen's arguments,he still stresses the multiethnic,multilingual natureof earlier coastal society: Horton,Shanga, 410-11. 42 In fact, accordingto Abungu's comparativeschema, Tana/TIWshares only one less motif with Kwale/E1W(15/21) than it does with PN (16/21), hardlyenough to claim exclusive PN-Tana connections: Abungu, "Rivcr Tana," 148. Kusimba, Rise and Fall of Swahili States, 90-140, provides a good ovcrview of this transition.

EARLY SWAHILI HISTORY RECONSIDERED

271

greatdistancesand easily makethe society,Swahilicouldquicklytraverse maritime their home, but such a sudden developmentdemands whole coastal environment explanation.And it reopens the question of where Swahili's antecedentslie: in on the southernKenyacoast,as Walsh and Kenya,as we claimedearlier; northern coast,as suggestedby Chami?We need a Helm argue;or on the centralTanzanian clearerdelineation of the early phases of TIW to answerthese questions,but the from one culturalphase into anothermakes sense in terms of the slow transition changein the sameway. linguisticevidence.Languages Historical Linguistics While the linguistichistoryof earlySwahilihas become well known since 1973, it It is not clearwhy thisis so, but Derek Nurse continuesto be widelymisinterpreted. new study,Swahiliand Sabaki,now makes monumental andThomasHinnebusch's method,Swahiliand of the comparative it difficultto ignore.A rigorousapplication reconstructs protoSabakicarefullyclassifiesSwahiliandrelatedSabakilanguages, Swahili, traces its genetic development,and explores a wide range of outside influences on it. While much of the book is technicallinguistics,there is much as well.43 herefor historians valuablematerial the linguistichistoryof Swahiliis fairly Accordingto Nurse and Hinnebusch, Far from being a mixed Afro-Arab pidginor creole,as commonly straightforward. believed,Swahiliis a Bantulanguagewhose linguistichistorywas thoroughlyconventional.It is an Eastern(or Savanna)Bantu language,closely relatedto other Bantulanguagesof the coastalarea.Its closest relativesare the Sabakilanguagesof the coast and Kenyan hinterland(Comoro, Swahili, Ilwana, Mijikenda, and relatedto the NortheastCoast (NEC) languages Pokomo),while it is moredistantly as shownin Figure2.44 coastandits hinterland, downthe Tanzanian scattered Figure 2: Northeast Coastal Bantu (after Nurse and Hinnebusch) Ilwana Swahili/Mwani Sabaki NE Coast Seuta Pare Ruvu Pokomo Mijikenda Comorian

43 Nurse and Hinnebusch, Swahili and Sabaki. Nurse has also provided an excellent introductionto historical linguistics for historians:Derek Nurse, "The Contribution of Linguistics to the Study of History in Africa,"Journal of African History, 38 (1997), 359-91. 44 In additionto Sabaki, the NEC languages include: Seuta (Bondei, Shambaa, Ngulu, and Zigula), Ruvu (Gogo, Kaguru, Kami, Kutu, Doe, Nhwele, Luguru, Sagara, Viduna, and Zaramo), and Pare (Pate and TuBeta):Nurse and Hinnebusch,Swahili and Sabaki, 271-83.

272

THOMAS SPEAR

of any sort in NEC, Sabaki,or Swahilithatset them Therearefew innovations apartfrom one another,showing that these languages experiencedfew external influencesand flowed fairly rapidlyfrom one into anotherin the course of their and grammar, vocabulary, Swahiliretainsa high degreeof inherited development.45 for example,72-91 percent is sounds. Of a 100 word list of basic vocabulary, while only 4-17 percentare loans from otherAfricanlanguagesand 2-8 inherited, limitedand Arabinfluencewas particularly languages.46 percentfrom non-African relativelyrecent.There has been little Arabic impacton Swahili morphologyor phonologyover 1,000yearsof contact,andwhile Swahilihas adopteda largeset of Arabicloan words,they are mostly fairlyrecentand limitedto fields where Arabic influence was greatest during the 17th_ 19gh centuries, such as religion, law, thereare only and kinship.By contrast, trade,sailing,measurement, administration, for proto-Sabakiand 16 for proto12 Arab loan words presentlyreconstructed Swahili.47 Swahili,then,must havedevelopedin place in associationwith the other NEC The NEC languagestodaycenteron the highlandsof northand Sabakilanguages. and Pare,Usambara, and southeastern Kenya-Taita, Kilimanjaro, easternTanzania Ngulu-and extend south into the Rufiji Delta. This is both a relic area of great linguisticdiversityand a highlandarea suitablefor early Bantu farming.Only the extendoutsideit, makingit probablethatthe earliestNEC speakSabakilanguages ers initiallysettledheresome 2,000 yearsago.48 of NEC and linguisticevidence,the distribution the archaeological Comparing and both date from the correlatesclosely with that of the Kwale/EIWtradition, and farmersin this ISL-5th centuries AD, thusmakingit likely thatearly ironworkers of the TanafITWtradition areawere NEC speakers.The subsequentdevelopment with the betweenthe 4h and 8h centurieswould then haveoccurredsimultaneously of NEC into Sabaki,Pare,Seuta,and Ruvu.Localculturaldifslow differentiation and as both the TIW tradition minorduringthis stage,however, ferencesremained Throughit all, individual remainedremarkably homogeneous.49 the NEC languages cultures and languages slowly changed and gained new elements from others, leading to the developmentof new cultures,dialects,and languages. No sharp and the changes were no doubt imperceptimarked thesedevelopments, boundaries differentiate when we artificially ble to people at the time. It is only in retrospect, or languages, thatsuchchangesappeardisjointedor cultures betweenarchaeological discontinuous. of Swahili with the earlierNECandSabakiphases,the development In contrast and relatedlanguages extended over a longer period of time to produce more as one This periodis also moredifficultto reconstruct, definedlanguages. markedly broughtaboutby population musttakeaccountof successivegeneticdevelopments
45 Ibid., 22-23, 61, 203-15, 463-78, 501-03. 46 Ibid., 324-28.
47 Ibid., 308-28.

48 Ibid., 33, 491. 49 Ibid., 490-549. See also Helm, "Conflicting Histories,"275-94.

EARLY SWAHILI HISTORY RECONSIDERED

273

with others, and the shifts, numerousexternalinfluences caused by interactions distribution of the languagestoday. Accordingly, early Sabaki speakersprobably lived fairly compactlyalong the narrowKenyan-Somali coastlands.By the middle of the V millenniumAD, a core group of linguistic innovators(consisting of Comorian/Mijikenda/Pokomo) were probablysurrounded by a number of more conservative languages (Swahili,Uwana,and Mwani),and by AD 800, that group probablysplit betweenMijikenda/Pokomo speakersand Comoro speakers.About the same time, Swahili speakersbegan to disperse along the coast to the Lamu Archipelago,Zanzibar,Kilwa, and Chibuene,and subsequentlyto Barawa and andsouthern dialectsin the process. At the same Mogadishu, dividinginto northern time, Mombasa/Chifundi, Mwiini,and Tikuu progressivelybroke away from the and stronglydefineddialect northern group,leavingSiu/Pate/Amu as a conservative and Vumba/Mtang'ata cluster.In the south,Pemba/Unguja, Tumbatu/Makunduchi, all divergedfrom one another, but the subsequentsoutherndialectswere isolated, not strongly defined, and experienced considerable influence from adjacent languages.Subsequently,Iwanaspeakersmovedup the Tana River,Mwani down the coastto southern Pokomoalongthe Tana,and Mijikenda south of the Tanzania, SabakiRiver.50 The associationof successivegroups of Bantuspeakerswith the development and subsequenttraditionsdoes not exclude other of the Kwale/EIW, Tana/TIW, influences, however.These were probably numerous,judging from widespread phonologicaland morphologicalchanges and the presence of loan words from andArabic,as well Cushiticlanguages, Southern Dahalo,Oromo,Somali,Malagasy, as extensive influences from other Bantu languages.51Such influences were facilitated by continualcontactamongthe NEC languages;the small size of most and languagecommunities; the presenceof larger,more prestigiouscommunities; that made similarities structural among languagesand widespread multilingualism linguistictransfers easy.52 and Tana/TIW; of Kwale/EIW thatof The apparently continuousdevelopment NortheastCoast, Sabaki,and Swahili;and the probableassociation of individual ironphasesof each with those of the other,all makeit likely thatBantu-speaking, and andfarmingpeoplessettlednearthe coastin the early 1St millennium producing, subsequentlydevelopedduringthe 2nd millenniuminto the peoples, cultures,and and islands from southernSomaliato Mozamlanguagesof the coast, hinterland, biquetoday.

50 Nurse and Hinnebusch,Swahili and Sabaki,, 490-549. See also Wilhelm J. G. Mohlig, "The Swahili Dialects of Kenya in Relation to Mijikenda and the Bantu Idioms of the Tana Valley," SUGIA 6 (1984/85), 296-303. 51 Nurse and Hinnebusch, Swahili and Sabaki, 288-89, 329-33; Mohlig, "Swahili Dialects," 296-303; Ehret, African Classical Age, 75-97, 172-73; Derek Nurse, "South Meets North: Ilwana = Bantu + Cushitic on Kenya's Tana River," in Peter Bakkerand Maarten Mous, eds., Mixed Languages (Amsterdam, 1994), 213-22; idem, Inheritance, Contact, and Change in Two East African Languages (Cologne, 1999). 52 Nurse and Hinnebusch,Swahili and Sabaki, 33-34.

274

THOMAS SPEAR

Such conclusionshavebeen questionedby some linguistically influencedhistorians, however.In a recentsurveyof earlyEast Africanhistory,Christopher Ehret sees the development of the Kwale/E1W and Tana/TIW traditionsas separateand with the Kwaletraditions discontinuous, associatedwith an earlierpresence at the coast of Upland Bantu speakers (Chaga/Dabida/Thagichu), who were then supplantedby NEC speakers responsible for the Tanatradition. Ehret'sinterpretation is basedon five mainpoints:(1) thattheTanatradition was "much too differentto be a separatelinguistic directlyderivablefrom Kwale," (2) thatUpland represented sub-group,(3) thatUplandspeakersprecededNEC speakerson the coast, (4) that the distribution of Kwalewarecorresponded withthatof Uplandspeakerswhile that of NortheastCoast speakersdid not, and (5) thatNortheastCoast speakersshould thusbe associated withTanaware,not Kwale.53 None of these assertionscan be sustained,however.While the relationship betweenKwaleandTanawareshasbeenthe subjectof muchdebateamong archaewereindeed conologists, as we have seen, most now agreethatthe two traditions tinuous. Second, the sub-groupingof Upland is neitherjustified by Ehret nor supportedby others.54 Third,the assumptionof an earlierpresence of Upland speakerson the coast is basedon a few Thagiculoans in Northeast Coast,but Nurse and Hinnebuschinterpret these loans as post-1 5th century;Tana now appearsto from the 4th century;and there remainsa sizable unsucceed Kwale everywhere of Uplandspeakers explainedchronological gapbetweenEhret'spositedsettlement along the coast and the earliestpresenceof Thagicuin centralKenya in the 12th of Kwale/E1W from the Sabakito the the distribution century.55 Fourth, (stretching RufljiRiver)is muchlargerthanthatof Upland(cloisteredin the Pare-Kilimanjaromoreclosely with that Taita-Central Kenyanhighands).Its distribution corresponds while that of Tana of the NEC languages,makingtheirassociationmoreprobable, matchesthatof the successorlanguagesto NortheastCoast. The limitedevidence Ehretadducesthusdoes not sustainhis widerconclusions. to interpret Noor Sheriffgo further Swahilias an AlaminMazruiand Ibrahim AD and subsequently evolved Afro-Arab pidginthatfirstdevelopedin the V century into a local creole that ultimatelyshed its Arabic forms and replacedthem with Bantu ones.56While they cite the Periplus as evidence for their early dating of Swahili,it does not, in fact,identifyeitherthe people living along the coast at the time or the languagethey spoke. Nor is there any linguisticevidencethat Swahili first developedas a pidgin or creole,as Mazruiand Sheriffthemselvesadmit.If it was originallya pidgin(basedon an Arabicacrolectand a Bantubasilect),its oldest andmostbasic vocabulary would be derivedfrom Arabicand reflectall subsequent soundchanges,while wordsof Sabakioriginwouldbe later,not reflectearliersound
53 Ehret, African Classical Age, 183-92. 54 Nurse and Hinnebusch,Swahili and Sabaki, 19-20. 55 Ehret, Classical Age, 206n; Nurse and Hinnebusch, Swahili and Sabaki, 281-83, 287, 293, 329, 332-33, 541, 577-655; Chami, "Review of Swahili Archaeology." 56 Alamin M. Mazrui and Ibrahim Noor Sheriff, The Swahili: Idiom and Identity of an African People (Trenton, 1994), 64-67.

EARLY SWAHILI HISTORY RECONSIDERED

275

cultural changes,andlimitedto non-core vocabulary. In fact,just the reverseis true. Arabicderivedvocabulary is mostlylate,does not reflectearliersound changes,and is limitedto selectivecultural fields, while Sabakilexis is early,reflects all sound changes, and constitutesmost core vocabulary. Swahili is thus a Bantu language whose historicaldevelopmentas a genetic member of the Sabaki family was as alreadynoted. Its grammaris not "universal" as Mazrui and conventional, Sheriff claim;phonologicalchanges in Swahili are representative of Sabaki and NortheastCoast generally;and its lexis is largely inherited.Arabic never was a dominantinfluenceon otherlanguagesalong the coast, and most Arabicspeakers were absorbed locally.57 While few historianshave attempted to reinterpret the linguistic evidence so its implications. Allen dismisses comparative broadly, manyhavefailedto appreciate linguistics out of hand as "insufficientlyevolved ... to enable us to unravelthe linguisticcomplexities... of peoples such as the historicalSwahilis" and "largely to non-specialists," incomprehensible while at the same time constructing his own model of a complex multilingual language-switching society that defies linguistic logic to assertSouthernCushiticorigins for a Bantu language.58 Following Allen, Abunguacceptsthe reputedrole playedby SouthernCushiticspeakersin the formationof Swahili.59 Pouwels prefersEhret'sclaims over Nurse and Hinnebusch's bettersubstantiated ones, andhe wronglyclaims that"virtuallyall linguists" accept And Willis neglects to consider linguistic Tanzania as the natalzone for Sabaki.60
data.6'

Middlemen Societies Justas earlierwritersviewedthe stone towns and the Swahililanguageas evidence Muslim Swahilisociety as a of foreignorigins,so they saw contemporary, urbane, immigrant "civilization,"divorcedfrom the "primitive"societies of the coast, a view oftenendorsed by Swahilithemselves.Again,we arguedthatSwahilisocieties moreeconomicallydifferentiated, developedalongthe coast,becomingprogressively and Muslim with the expansionof international socially stratified, trade,increasing fromArabia andIndia.62 wealth,andimmigration has been a weak point in Swahili historiograDetailedhistorical ethnography and to seek to phy, however,forcing us to rely on contemporary ethnographies to dynamicelementsin Swahilicultural historicize themby payingcarefulattention The World of the practice.John Middleton's detailedprocessualethnography,

57 Thomas J. Hinnebusch, "What Kind of Language is Swahili?" Afrikanische Arbeitspapiere 47 (1996), 73-95; Nurse, pers. comm. 58 Allen, Swahili Origins, 17. 59 Abungu, "TanaRiver," 150-61. 60 Pouwels, "EastAfrican Coastal History,"287-88, 292-96. 61 Justin Willis, Mombasa, the Swahili, and the Making of the Mijikenda (Oxford, 1993), 28-36. 62 Nurse and Spear, The Swahili, 68-98.

276

THOMAS SPEAR

Swahili,is thusespeciallywelcome. Middletonviews Swahilisociety as comprising with marked regional and temporal a single oikumene,or maritimecivilization, variations broughtabout throughdiffering historicalexperiences,trade relations, and religious beliefs. It was a middlemansociety patterns, descent and marriage betweenthe commerand cultural brokerswho mediated composed of commercial port Like precapitalist one of the hinterland. cial worldof the sea andtheproductive frontiers,composed of Swahilitowns were polyglot,multiethnic cities everywhere, Arab merchantsand ship owners; Indian financiers;Swahili middlemen,traders, ship builders,sailors, iron and leatherworkers, weavers, furnituremakers, and fishermen;slave laborers;and neighboringfarmers,herders,hunters,and traders. and exchange economicresources,specializations, Eachtown hadits own particular systems within the overallecology of the coast. Marketrelationswere based on princes;lineagesactedas corporate personaltrustandkinship;kingswere merchant and status rested tradinghouses; social identitywas forged in intensecompetition; of wealth,exchange,honor,andprestige.63 on shiftingfoundations social to Swahilihistoricaltraditions, Middleton'sis a dynamicmodel,attuned processes,and the differencesbetweeninclusivecountrytowns and the exclusive mercantile stonetownsthatemergedfromthem.Most Swahililived in impermanent houses in small countrytowns where they workedas subsistence mud-and-wattle Therewas a townswereopen,inclusive,and egalitarian. or fishers.Country farmers Marriagewas exogamous, minimumof a divisionof labor or social stratification. descent.All who cognaticbilaterial extendingkin linksout to othersin wide-ranging as its ownersand were presidedover by egalitarian residedin a town wereregarded councilsof elders.64 trade Stone towns, by contrast,were complex, economically differentiated emporiawhose residentsperformeda wide rangeof differentfunctions-hosting andprovisioningforeigntraders, bulkingexportcargo and breakingimports,wareand building transactions, housing tradegoods and supplies,financingcommercial workedas farmersand fishers, sailors, house and and repairing ships. Individuals ship builders;wood, iron, and leatherworkers;and financiersand traders.Surroundedby stone walls,stone towns wereclosed to outsidersand people of lesser status,and putativedescent rank. Internally, they were subdividedby occupation, stone houses turnedinward into distinctmoitiesand wards(mitaa),and individual in the communitywas andthemselves.Membership inhabitants on theirupper-class was endogamous;kinship was restricted and rankedby origin;marriage restricted andexclusive;andhereditary kingsandeldersruled.65 to them on the Historically,stone towns rose from ruralroots and returned elites often struggledto maintaintheir posivagariesof trade,while theirpatrician

63 John Middleton, The World of the Swahili: An African Mercantile Civilization (New Haven, 1992), viii-ix, 1-26. Cf. Michael N. Pearson, Port Cities and Intruders: The Swahili Coast, India, and Portugal in the Early ModernEra (Baltimore, 1998), 38-45, and Middletonand Horton, The Swahili. 64 Middleton, Worldof the Swahili, 58-73, 83-88. 65 Ibid., 21-25, 41-44, 54-80, 90-95, 198-200.

EARLY SWAHILI HISTORY RECONSIDERED

277

tionsagainstnewcomers. Eachstone town servedas the focal point of an exchange network comprising a number of countrytownsandhinterland areaslinkedwith one anotherthroughethnicity,language,kinship, and patron-client relations.Country towns and hinterland areasproduced basic foodstuffsand exportcommoditiesfor a stonetownentrepot in exchangefor locallyproduced craftsandimports. Coastal peoples were thus immersed in dense webs of production and exchange,ethnicity, kinshipand descent,and residence.They were also joined by Islamic (dini) and customary(mila) religiouspracticesthat linked them with the widerMuslim community(umma),on the one hand,and with local Africancomon the other.Hadrami munities, scholarssettledin East Africa,while local scholars studiedin the Middle East, and local Muslim scholars devised their own maulidi of the Prophet'sbirth,which immigrant celebrations scholars often condemnedas unwarranted innovation.66 Nowhereis this dynamicbetterdemonstrated than in LindaGiles' insightful andwide ranginganalysisof spiritpossession amongSwahiliand theirneighbors. Swahili embracean arrayof spirits-Arab, Swahili,and African-each of which I"represents a differentsymbolic universe, which is refracted into a numberof variant combinations... to form complex symbolic interactions."Each has its own symbolic language,music, colors, dance,and food, reflectingdifferentaspects of thathistoryin a distinctiveperformative local historical experience andencapsulating tradition.The prototypicalkipemba spirit world "'providesthe most powerful statement of Swahiliself-identity," giving "symbolic expression to a Swahili historyandidentitythatreachesfarinto the past and out into the fields and villages.It the ideology of the urbanelite and the historical transcends periodof Arabcultural more indigenoustradition.It is, in ascendancyto reclaiman older,more syncretic, of what is 'truly Swahili'." Different"Swahili statement fact,a counter-hegemonic which "thecult compossessioncults generate... differentsymbolictexts"through elements which plex gives dialectical expressionto a whole rangeof contradictory have formed the socio-cultural universeof Swahili coastal society throughoutits history." Possession cults are thus both productsof the diverse arrayof Swahili to thoseexperiences.67 as well as testaments historical experiences Middleton'sand Giles' comparative ethnographic surveys are complemented by a numberof detailed studies of individualSwahili communitiesthat further of theircomplexityand diversity.Pamelaand Leif Landenrichour understanding accountsof Kigombe,a small,obscureMtang'ata town that bergprovidefascinating is complemented by new studies of the Comoros,Mafia, Mombasa,Malindi,and

66 Ibid., 162-81. See also Pamela Landberg, "Kinship and Community in a Tanzania Coastal Village" (Ph.D. thesis, University of California at Davis, 1977), 532-68; David Parkin, "Swahili Mijikenda:Facing Both Ways in Kenya,"Africa 59 (1989), 161-75. 67 Linda L. Giles, "Spirit Possession and the Swahili Coast: PeripheralCults or Primary Texts" (Ph.D. thesis, Texas, 1989), 493-97; idem, "Possession Cults on the Swahili Coast: A Reexamination of Theories of Marginality,"Africa 57 (1987), 234-58. Cf. Michael Lambek, Knowledge and Practice in Mayotte: Local Discourses of Islam, Sorcery and Spirit Possession (Toronto, 1993).

278

THOMAS SPEAR

Lamu.68And processualstudies of "Intermediary Swahili" and "Swahilization" by David Parkin and others are demonstrating the complex ways Mijikenda, Hadrami, andotheroutsiders negotiate ethnicity to becomeSwahili.69 Kigombe is merely one of eighteen small ruraltowns between Tanga and Panganithatonce formedthe ShiraziMtang'ata confederacyin the 13"hto 15"hcenturies. As such, it is representative of many coastal fishing and farmingvillages, especially those of the lesser-knownMrimacoast. Kigombe was a rathertypical fishing village,whose inhabitants lived in mud-and-wattle houses dividedinto local neighborhoods,or wards, surroundingthe local mosque. While its inhabitants thoughtof themselvesas an egalitarian communityof kinsmen,"brothers under Islam," who sharedmoralvaluesand operatedby consensus, the communitywas dividedbetweenproprietors (wageni),men and women,kin (wenyeji)and strangers and non-kin,and free and ex-slave,all of whom competedwith one anotherfor dancecompetitions, and malulidis statusandpowerin roundsof weddings,funerals, the year.70 throughout New ethnographic studies thus provide us with a range of possibilities for interpreting Swahili traditions, cultures, and historicalevents in the context of dynamicsocial and economicuniverses.And yet our pictureis far from complete, most notablyin the lack of detailedanalysesof the roles of slaveryand gender in of economic differentiaSwahilisocieties and how these fit into historicalpatterns While the rise and fall of plantation tion andsocial stratification. slaveryin the 19"h centuryand the ensuing strugglesover new forms of labor have been extensively we haveverylittleideaof the roles slaves coveredby Frederick Cooperandothers,7'
68 Landberg, "Kinship and Community"; Leif Landberg, "Men of Kigombe: Ngalana Fishermen of NortheasternTanzania"(Ph.D. thesis, UC-Davis, 1975); Martin Ottenheimer, Marriage in Donioni (Prospect Heights, 1985); Martin and Harriet Ottenheimer, Historical Dictionary of the Conioro Islands (Metuchen, 1994); Lambek, Knowledge and Practice in Mayotte; Pat Caplan, African Voices, African Lives: Personal Narratives from a Swahili Village (London, 1997); Marc J. Swartz, The Way the World Is: Cultural Processes and Social Relations among the Monibasa Swahili (Berkeley, 1991); Susan F. Hirsch, Pronouncing and Persevering: Gender and the Discourses of Disputing in an African Islamic Court (Chicago, 1998); Franqoise Le Guennec-Coppens, Les femmes voilees du Lamu (Paris, 1983). See also Susan Beckerleg, "Maintaining Order,CreatingChaos: Swahili Medicine in Kenya"(Ph.D. thesis, London, 1989); Allyson Purpura, "Knowledge and Agency: The Social Relations of Islamic Expertise in ZanzibarTown" (Ph.D. thesis, City College of New York, 1997). 69 David Parkin,"Being and Selfhood among Intermediary Swahili," in Joan Maw and David Parkin,eds., Swahili Language and Society (Vienna, 1985), 247-60; idem, "Swahili Mijikenda"; David Parkin and Franqois Constantin, eds., "Social Stratification in Swahili Society," special issue, Africa 59 (1989), 143-220. 70 Landberg,"Kinshipand Community,"407-29, 353-54, 391-404, 483-519, 532-68. 71 FrederickCooper, Plantation Slavery on the East Coast of Africa (New Haven, 1977); idem, From Slaves to Squatters: Plantation Labor and Agriculture in Zanzibar and Coastal Kenya, 1890-1925 (New Haven, 1980); idem, On the African Waterfront:Urban Disorder and the Transformationof Work in Colonial Mombasa (New Haven, 1987); Abdul Sheriff, Slaves, Spices and Ivory in Zanzibar:Integrationof an East African Commercial Empire into the World Economy, 1770-1873 (London, 1987); FredMorton,Childrenof Ham: Freed Slaves and Fugitive Slaves on the Kenya Coast, 1873 to 1907 (Boulder, 1990); and Willis, Mombasa.

EARLY SWAHILI HISTORY RECONSIDERED

279

Similarly, while there are some good playedwithinSwahilisocieties themselves.72 few suggest how such roles might have earlierstudies of modem gender roles,73 workedhistorically, especiallyin thecase of womenslavesor ex-slaves.74 Historical Documents datais of little historicaluse withouthistoricaldocuContemporary ethnographic While documentsfrom as early as the I" centuryAD have long been mentation. editionsanda historiographic commentary remind us of two new scholarly available, the importance of readingsuch sources critically.Lionel Casson's new edition of with informedcomcombinesmeticuloustranslation the PeriplusMaris Erythraei A first-hand to makethis often obscuresource more instructive.75 account mentary now datedto AD 40-55, of the East Africancoast by an EgyptianGreekmerchant, the Periplusdetailsthe main tradingsites and theirplace withinIndianOcean trade of oratoi (previously as faras the Azanian(Tanzanian) coast.Casson'sretranslation as "pirates") as "tillers of the soil" alone has already translated enigmatically vibrant discussionfromHorton, Vansina,andWrigleyin the continuing encouraged of thecoast at the time.76 But we still lack debateover the identifyof the inhabitants firm archaeological, linguistic,or social correlatesfor "Rhapta" and other places in the text thatwouldenableus to makemoreeffectiveuse of this source. mentioned and Tolmacheva In another new scholarly collects,retranslates, edition,Mariana discussesseven versionsof ThePate Chronicle.In the process,she establishesthat of Muhammad bin Fumo consciousness all seven emergedfromthe fertilehistorical knownas BwanaKitini,in the early 1900s,thus limiting 'UmarNabahani, popularly sources.77 theirusefulnessas independent Finally, John Shen has provided an admirableanalysis of seven Chinese includingthe famous StarryRaft accounts,datingfromthe 8'h to the 17th centuries, expeditionto EastAfricain 1405-33. Most,he finds,aresuccessivecourtredactions
72 But see Jonathon Glassman, Feasts and Riot: Revelry, Rebellion, and Popular Consciousness on the Swahili Coast (Portsmouth, N.H., 1995) for an excellent historical study of slavery and class on the northernTanzaniancoast in the 19thcentury. 73 Margaret Strobel, Muslim Women in Mombasa, 1890-1973 (New Haven, 1979); Franqoise Le Guennec-Coppens, Les femmes voilees de Lamu: variations culturelles et dynamiques sociales (Paris, 1983); Ottenheimer, Marriage in Domoni; Pamela Landberg, "Widows and Divorced Women in Swahili Society," in Betty Potash, ed., Widows in African Societies (Stanford, 1986), 107-30; Sarah Mirza and MargaretStrobel, Three Swahili Women: Life Histories from Mombasa, Kenya (Bloomington, 1989). 74 But see MargaretStrobel, "Slavery and ReproductiveLabor in Mombasa," in Claire C. Robertson and Martin A. Klein, eds., Womenand Slavery in Africa (Madison, 1983), 111-29; Giles, "Spirit Possession"; idem, "Possession Cults." 75 Lionel Casson, trans. and ed., The Periplus Maris Erythraei (Princeton, 1989). 76 Mark Horton, "The Periplus and East Africa,"Azania 25 (1990), 95-99; Jan Vansina, "Slender Evidence, Weighty Consequences: On One Word in the Periplus Maris Erythraei," History in Africa 24 (1997), 393-97; ChristopherWrigley, "The Periplus and Related Matters," Azania 32 (1997), 112-17. 77 MarianaTolmacheva, trans. and ed., The Pate Chronicle (East Lansing, Mich., 1993).

280

THOMAS SPEAR

of originalaccountssince lost and many have suffered from poor and misleading translations,as he notes tellingly in criticizing Allen's reliance on FreemanGrenville'sflawedcollectionof documents.78 While we thusneedto approach the existingdocumentsmore criticallythanwe havein the past,thereis also a vast troveof Portugueseand other documentsthat butcontaina vastwealthof information, havebarelybeen explored especiallyfor the southern in his detailedstudy of textile coast, as demonstrated by JeremyPrestholdt
trade in the
15th

and 16th centuries.79

East Africa and the Indian Ocean The widespread existence of Egyptian,Arab,Chinese,and Portuguesedocuments roles the easternAfricancoast playedin the widerIndian alertsus to the important in Ocean,but these roles have only been rarelyexploredin the Swahili literature, in the development of the coastaltowns. Egypt dominated spiteof theirsignificance Red Sea tradefrom 2500 BC to 200 BC, acquiringmyrrh,frankincense, cinnamon, laudanum, andebony;slaves;cattle;gold and electrum; elephantsand ivory;leopard skins; tortoiseshell; and rhinohorn from Ethiopiaand the Horn via Punt, DM'T, and Aksum.Aksumbecamethe focus for such tradewith the transitionto GrecoRoman dominance from the 2nd century BCto the 3rdcentury AD, and it continued to

in the Red Sea followingthe declineof Romeuntilthe 6" century.This be dominant with the tradingsites at Ras Hafun,and it was also the period when the correlates sites appeared first Kwale wareironworking along the coast, suggesting,perhaps, as describedin the Periplus and thatit was visits of early Greco-Roman traders, Ptolemy's Geography,thatinitiallydrewNortheastCoast speakersto the coast to lay the foundations for Swahilitownsto come.80 Persiaplayedan increasingrole in Red Sea and IndianOceantradefollowing the declineof Rome andthe Persian conquestof Yemenand Egypt in 570 and 616. flowed from East Africato Siraf Mangrove poles,gold,ivory,slaves,and ambergris and Sohar and on to Chinain returnfor PersianSassanian-Islamic, opaque white andbeads,cloth,and glaze, andsgraffiato wares;ChineseQingbaiglazedstoneware; of the initialpreironwares.It was probably thistradethatsparked the development Muslim coastal towns that emergedbetweenthe 8th and 10th centuriesat Manda, andthe Comoros. Shanga,Pate,UngujaUkuu,Kilwa,Chibuene, FatimidEgypt then gained controlover the Red Sea from the 10th century, rockcrystal,ivory,andgold fromEast Africa.At the same time,southern importing for a shift role in IndianOceantrade,accounting Arabiabeganto play an increasing
78 John Shen, "New Thoughts on the Use or Chinese Documents in the Reconstructionof Early Swahili History," History in Africa 22 (1995), 349-58. 79 Jeremy Prestholdt, "As Artistry Permits and Custom May Ordain:The Social Fabric of Material Consumption in the Swahili World, circa 1450-1600" (Program in African Studies Working Papers, No. 3, Evanston, III., 1998). 80 Jacke Phillips, "Punt and Aksum: Egypt and the Horn of Africa," Journal of African History 38 (1997), 423-58; StuartC. Munro-Hay,"AksumiteOverseas Interests"in J. Reade, ed., The Indian Ocean in Antiquity(London, 1996), 403-16.

EARLY SWAHILI HISTORY RECONSIDERED

281

to Yemeni black-on-yellow ceramics and an increase in Islamic influence on the East Tradealso increasedmarkedly, African coast from the 12th century. especiallyalong

the southern of Muslimtownsfrom the 12th coast,leadingto the dramatic expansion

MichaelPearsonin his insightfuland broad-ranging book, Port Cities and Intruders. Here we see the peasantsof Gujarat producingthe cotton and cloth on which much of IndianOcean tradewas based; the Indian brokers and financierswho facilitated trade; the Muslimtraders, religiousscholars,and pilgrimswho integrated the Indian Ocean into an "Islamic World System"; the littoralsocieties of the coaststhatmediated betweenthe sea andthe interior; and the peoples of the African interiorwho producedthe ivory and gold that fueled the trade. Together with Middleton's analysis of middleman societies,we now see Swahili towns not as nor as solely local developments, exclusive foreign transplants but as dynamic culturaland commercial entrepotsin an IndianOcean world stretchingfrom East The debateoverthe riseof Swahilitowns can thus no longer be Africato Malaya.82 seen in exclusively foreign or local terms,but must take accountof the dynamic interaction betweenlocal andinternational forcesin an expansivemaritime world. Rise of the Swahili Towns AdoptingHorton's detailedchronologyfor Shanga,RandallPouwels has periodized Swahili history in terms of the dominantmotifs of Swahili traditions.The "Shungwaya" period,extendingfrom 800 to 1100, saw the culturalgenesis of Swahiliin the north,the beginningsof tradewith the PersianGulf, and the emergence of the first pre-Muslimtowns along the coast. As trade and wealth grew between 1100 and 1300 and tradingcontactsshifted to the Red Sea a numberof new townswere founded,foremost amongthemMogadishu;coastaldwellersbegan in to convertto Islamin increasing and Yemenisharifs becameprominent numbers; Mogadishu and Kilwa in a period characterized by increasing"Islamization"of Swahilisocieties. Tradeand wealthreacheda climax between 1300 and 1600, the familiesalong the coast to claim prestigious "Shirazi" period,causingprominent Arabsand mainland Shiraziorigins to distinguishthemselvesfrom both immigrant and dress,build elaborately decorated stone Africans, adoptexclusiveparaphernalia houses and pillartombs,and endow new mosques.The coast then declinedunder came underincreasingArabinfluencein the Porguguese rule,while it subsequently
16th and 17th centuries and Omani in the 18th and 19th, as evidenced by the rise of

new Sufi orders; the adoption of Arab-style nisbas, architecture, dress, and
81 Horton, "Early Muslim Trading Settlements"; Mark Horton, "The Swahili Corridor," Scientific American 257/3 (Sept. 1987), 86-93; Horton, et. al, "The Mtambwe Hoard." 82 Pearson, Port Cities and Intruders.For an innovative study of labor in the IndianOcean, see Janet Ewald, "Crossersof the Sea: Slaves, Freedmen,and Other Migrants in the Northwestern Indian Ocean, c. 1750-1914," American Historical Review 105 (2000), 69-91. Malagasy taders and immigrantswere also active in the southern IndianOcean world from the 7th or 8th century, but we know little of their activities.

282

THOMAS SPEAR

vocabulary; andthe eventualshift fromSwahiliuungwana to Arabustaraabuas the in the 1870s.83 markof civilizedbehavior This broadoverviewhasbeendevelopedin fascinating detailby HowardBrown in his historyof Siyu, a town on Pate Island,from the 1Ilh to the 19h centuries. While not extensivelyinvolvedin trade,Siyu was a thriving communityof weavers and embroiderers, woodworkers, leatherworkers, metalworkers, copyists, and book binders.It was, in short,a prosperousthoughordinary town, overshadowed by its wealthierneighbors,Pate and Lamu. As neithera local fishing village nor an international trading center,Siyu was a middlingmiddleman society of the sort that probablypredominated along the coast and can tell us a greatdeal about Swahili societies in general. While Siyu's historicaldevelopment followed the general Swahili pattern,it was also intenselylocal,reflectingits own particular social compositionand neighbors. Siyu was probablyfoundedin the 11thcenturyas a mud and thatchvillage, butby the 16chcenturyit hadgrownintoa largestonetown,perhaps the largesttown mainland. The local Waswahili on PateIsland,withextensivefarmson the adjacent Famao and Somali clan was dominant, but soon it was displaced by immigrant Katwa,who came to constitutethe two moitiesthatvied for hegemony in the 18th century.Such internalcompetitionwas also reflectedin externalpolitics as Siyu allied with the Turks against the Portuguesein the 1580s, but then joined the PortugueseagainstPate in the 1630s and 1640s until the Portuguesewere finally defeatedby combinedPate and Omani forces in 1652. At the same time, on the mainlandSiyu alliedwith GarreSomaliagainstOromoand Pate,on the one hand, and Bajuni and Faza, on the other,during which Famao agreed to share power domesticallywith Katwa,who had been increasingin power,wealth,and prestige and were gradually being absorbedinto the waungwana.But conflict betweenthe and Famaoeventually alliedwith the exiled Nabahanirulersof Pate two continued, to displaceKatwaalliedwith the OmanisandPatein the 19"h century.84 towns of A number of studieshave also expandedour view of the little-known the Mozambiquecoast, where small, autonomousSwahili towns scatteredamong off-shore islands, inlets,and inlandtradefairs were easily overwhelmed by their neighborsand,after 1505,by the Portuguese.Both Swahili and latterPortuguese small and dependedon extensiveintermarriage with were numerically communities While we still lack Africansto maintaintheireconomic and politicalpositions.85
83 Horton, "Early Settlement"; idem, Shanga; Pouwels, Horn and Crescent, 72-73, 129-32; idem, "The East African Coast, c 780 to 1900 CE"in Nehemia Levtzion and Randall L. Pouwels, eds., The History of Islam in Africa (Athens, 2000), 251-7 1. 84 HowardBrown, "History of Siyu: The Development and Decline of a Swahili Town on the Northern Kenya Coast" (Ph.D. thesis, Indiana University, 1985); idem, "Siyu: Town of Craftsmen: A Swahili Cultural Centre in the Eighteenth and Nineteenth Centuries," Azania 23 (1988), 101-13. 85 Malyn Newitt, A History of Mozambique(Bloomington, 1995); RicardoTeixeira Duarte, Northern Mozambiquein the Swahili World(Stockholm, 1993); Joseph F. Mbwiliza, A History of CommodityProduction in Makuani, 1600-1900 (Dar es Salaam, 1991); Nancy Hafkin, 'Trade, Society and Politics in NorthernMozambique,c. 1753-1913" (Ph.D. thesis, Boston University, 1973).

EARLY SWAHILI HISTORY RECONSIDERED

283

should towns,the extensivePortuguesedocumentation detailedstudiesof individual yield rich results and allow us to follow up Pearson's intriguingsuggestion of coastthanalong the on the southern greater integration betweencoastandhinterland northern one. towns, especiallyJonathonGlassFinally,a numberof historiesof individual analysesof more man's magisterial Feastsand Riot, provideinsightfulcomparative Takentogetherwith the detailedarchaeological studies of recentSwahilihistory.86 Shanga, Manda,Ungwana,Gedi, Kilwa,and the Comoros, and ethnographiesof Kigombeand the Mrimacoast, Mafia,and the Comoros, Lamu,Malindi,Mombasa, we probablynow havemoreextensive,subtle,and deeper historicalknowledgeof the EastAfricancoastthanwe haveforanyotherareain Africasouth of the Sahara. not all new accountsinspireconfidence,especiallyJames de Vere Unfortunately Allen's long-awaitedand ultimatelyposthumous book, Swahili Origins. Allen work reorienting the historiography of the deservesmajorcreditfor his pioneering of domestic historical coast from concernsover foreignorigins to the exploration an ambitious attempt to consolidatethis earlierwork, processes.His bookrepresents but unfortunately it strays far from the historicalevidenceto erect new myths in place of older ones. Allen's cavalieruse of evidence conflates references far removedin space and time and piles conjectureon conjectureto create wholly multiethnic states and cultures.When his totteringedifice imaginary multilingual, to fall in the face of linguisticevidenceto the contrary,he simply denies threatens sites fail to confirmhis andwhenarchaeological the validityof historical linguistics, he imputes new evidenceto them.87 expectations, If Allen was concernedlargely with the views of foreigners,two eminent Swahilischolars,AlaminMazruiand IbrahimNoor Sheriff, have been moved by In East Africa.88 ongoing debates regardingSwahili identitywithincontemporary for the indigehistorical the process of buildingan eminentlyreasonable argument of a dynamic,urbane,Muslim Swahili culture,however,they nous development the case for an unbrokenline of Swahilicultural historyfrom the time of overstate the Periplus in the 1St centuryAD to the present.The referencesfrom the Periplus are simply too vagueto assertthatpeople specificallyknown as Swahili inhabited evidenceforthe development thecoast at the time,andtheyignorethe archaeological Swahilitownsandcultureonly fromthe gth century. of recognizably They also deny of the Swahili language from earlier the linguisticevidence for the development
86 MargaretYlvisaker, Lamu in the NineteenthCentury:Land, Trade and Politics (Boston, 1979); Peter Koffsky, "History of Takaungu,East Africa, 1830-1896" (Ph.D. thesis, Wisconsin, 1977); FrederickJ. Berg, "Mombasa under the Busaidi Sultanate"(Ph.D. thesis, University of Wisconsin, 1971); William McKay, "A Precolonial History of the Southern Kenya Coast" (Ph.D. thesis, Boston University, 1975); and Glassman, Feasts and Riot. 87 These problemshave been extensively airedin reviews of Swahili Origins by Randall L. Pouwels, Journal of African History 34 (1993), 518-20; Martin Walsh, Azania 28 (1993), 143-47; Justin Willis, African Affairs 93 (1994), 147-49; Adria LaViolette and Thomas H. Wilson, International Journal of African Historical Studies 27 (1994), 439-40; and Mervyn Hiskett, Journal of Islamic Studies 5 (1994), 139-41. 88 Mazruiand Shariff, The Swahili.

284

THOMAS SPEAR

Bantulanguages spokenon the coastduringthe I"millennium, as we haveseen, and of internal the significance divisionswithinSwahilisocieties thatare they underplay often expressedby Swahilithemselvesin ethnicterms,thus neglectingthe internal bases for Swahiliexclusion. While earlierdebates over Swahili identity and history usually turned on of culture, racialized interpretations contemporary ones often focus on constructivist views of the "inventionof tradition" or "creationof tribalism."Rightly pointing out thatthe people labeled"Swahili" by scholarsrarelyidentifythemselvesas such, choosing identitiesbased on their own town, putativeorigins, status, or descent group instead,criticshavegone on to argueagainstthe development of a common coastal culture.Such a view informsJustin Willis's Mombasa, Swahili-speaking
the Swahili, and the Making of the Mijikenda. A perceptive study of changing forms

of labor and social relations in 20h century Mombasa,Willis views modern of colonial rule,thus playingdown local and foreignhistoriethnicitiesas products cal accountsof earlieridentities. While identitiesare certainlysocially constructed and changeover time,a narrowly instrumentalist view underestimates the limits of inventionby neglectinglocal agency,overlookingthe complexways traditionsand from earlierhistoricalartifacts,and denying the very identitiesare reconstructed historicity thatgives ethnicityits rawsocial and politicalpower.89 Such a view thus begs the very historicalquestions it asks and invalidatesall but contemporary historical analysis. While the debatesoverearly Swahilihistoryare a welcome sign of historical are uncritically vitality, we must be carefullest new interpretations acceptedas the wisdom. An innocentexample is providedby Pearson's Port new conventional to EastAfricanhistory,Pearson discussedearlier.A newcomer Citiesand Intruders, and gaineda sophisticated of readwidely in the specializedliterature understanding it, but he often refers to secondary and tertiaryaccounts instead of original monographs,thus missing the subtleties of the original studies and risking This is not an idle risk. Several becominga victimof others' readingsof them.90 of Shungwritershave too readilyacceptedAllen's deeply flawed interpretation while otherscontinueto criticizean earlierflawedanalysisof the linguistic waya,91
89 See my review of Willis in International Journal of African Historical Studies 27 (1994), 630-34. Willis (Mombasa, 28-36) confines his analysis of the invention of the Shungwaya traditionto the Mijikenda, while Pouwels ("East AfricanCoastal History," 293-294) in extends it to the Swahili. On the limits of invention more generally, see idem, "Introduction" Thomas Spear and RichardWaller, eds., Being Maasai (London, 1991), 1-24; T.O. Ranger, "The Inventionof Tradition Revisited: The Case of Colonial Africa"in T.O. Ranger and 0. Vaughan, eds., Legitimacy and the State in TwentiethCenturyAfrica (Houndmills, 1993), 79-83; Carolyn Hamilton, Terrific Majesty: The Powers of Shaka Zulu and the Limits of Historical Invention (Cambridge, 1998); JonathonGlassman, "Sortingout the Tribes; The Creation of Racial Identities in Colonial Zanzibar'sNewspaperWars,"Journal of African History 41 (2000), 395-428; and Jamie Monson, "Memory, Migration and the Authority of History in Southern Tanzania, 1860-1960," Journal of African History 41 (2000), 347-72. 90 See, e.g., Pearson, Port Cities and Intruders, 19-21 (fns 47, 52, 55), 73 (fns 34-35), or 82 (fns 72-74). 91 See fn. 38 above.

EARLY SWAHILI HISTORY RECONSIDERED

285

or while ignoring subsequent corrections,92 evidence for Shungwayatraditions, acceptearlierargumentsregardingthe alleged inventionof Shungwayatraditions of them.93 while neglectinglatercritiques can and do havehistoriMazruiandSheriffarguethatacademicinterpretations cal consequencesfor the peopleconcerned.The Swahilihave long been affectedby of colonial what was writtenabout them, whetherin the differentialapplication and law to those deemed"Arabs" and "'Swahili"or in theirpoliticaland taxation Swahili themselveshave a economicdisplacement by ethnicpolitics. Furthermore, deep sense of history and evoke it constantlyin internalstruggles for status and power. Such internalhistories have frequentlybeen influencedby externalones. Swahiliclaims to "Shirazi" (i.e., Persian),"Arab," and now "African" identities politicalforces on domestic by the playof external haveall beenstronglyinfluenced Thus, historians'debatesare not, and never have been, politics and vice versa.94 strictlyacademic. Toward a New Historical Synthesis of the new work as a whole,we are presentedwith the contributions In appraising images of early Swahilihistory.The one, continuousand inclusive, two contrasting to Tana/TIW andlinguisticanalysesthatlink Kwale/EIW flows fromarchaeological as well as NEC to Swahiliand relatedlanguages.Seen in traditions andsubsequent these terms,Swahili appearto share a common culturaltradition,language,and the length of the coast. The civilization historyand to constitutea single maritime other image, however,is discontinuousand exclusive, conveyed in the diverse urbaneport commuof polyglot,stratified, histories,and ethnographies traditions, learnedhow to the coastalenvironment; groupsof peopleadapted nities,as different with others; to work iron, fish, raise new crops, and build in stone; interacted and overseasvisitorsalike;and slowly adoptedIslam. Seen neighbors incorporated

92 E.g. Willis, Momnbasa, 29; and Pouwels, "East African Coastal History," 287, 294. For Myths and Historians' Myths: Variations of the the full debate, see Thomas Spear, "Traditional Singwaya Theme of Mijikenda Origins," Hi.story in Africa 1 (1974), 67-84; Thomas J. Hinnebusch, "The Shungwaya Hypothesis: A Linguistic Reappraisal,"in J. T. Gallagher, ed., East African Culture History (Syracuse, 1976), 1-41; Thomas Spear, "TraditionalMyths and Linguistic Analysis: Singwaya Revisited," History in Africa 4 (1977), 229-46; idem, The Kaya Complex (Nairobi, 1978), 16-43; idem, Traditions of Origin and their Interpretation (Athens, 1982), 3-17; Nurse and Hinnebusch,Swahili and Sabaki, 490-96, 542-43, 547-49. 93 E.g. Willis, Momnbasa,28-36; Pouwels, "East AfricanCoastal History," 293-294. For the overall discussion, see R. F. Morton, "The Shungwaya Myth of Mijikenda Origins: A Problem in Late Nineteenth Century Kenya Coastal History," International Journal of African Historical Studies 5 (1973), 397-423; Spear, "TraditionalMyths and Historians' Myths"; H. Neville Chittick, "The Book of Zenj and the Mijikenda," International Journal of African Historical Studies 9 (1976), 68-73; R. F. Morton, "New Evidence Regarding the Shungwaya Myth of Mijikenda Origins," International Jourmalof African Historical Studies 10 (1977), 628-643; Spear, Traditionsof Origin; Spear, review of Willis, Mombasa. 94 Mazrui and Sheriff, The Swahili, 17-53, 131-63. Cf. Nurse and Spear, The Swahili, 30, 74-75.

286

THOMAS SPEAR

and autonomous, thus,each town was distinctive the productof its own particular andsocial influences. history,environment, Of course,these two images are not irreconcilable. Just as Swahili all speak different dialectsof a commonlanguage, so eachcommunity represents variations of commoncultural patterns andhistorical experiences. In the transition from Kwaleto Tana traditions,from Sabaki to Swahili speech, or from one town to another, distinctivelocal variationsemerged that later led to different cultural traditions, of a highly mobileculture, dialects,andsocieties.Members Swahilitraveled up and in a maritimeculturethat embraceda diversityof down the coast, participating dialects,ethnicities,and occupationsin the culturalpracticesof each town and village. Over the longue duree, the fundamental outlines of Swahili history are now well established.Takingthe archaeological and linguisticevidencetogether,it now seems probablethat the transitionfrom Kwale/EIWto Tana/TIWthroughout the coast andhinterland was accompanied by a parallel transition from NortheastCoast millennium AD. Thatlaterceramic to Sabakiand Swahiliover the course of the 1"S traditions and languagesremained fairly homogenoussuggests that these cultures remainedcompact,but in fact both TIW and the Swahili language spanned the lengthof the coast by the end of the millennium, the productof a maritime culture remained that,thoughwidelyscattered, tightlyknitandremarkably homogeneous. While coastingtradewas conductedfromearly in the 1st millennium, permanent settlementsonly began to develop duringthe 8'K-10thcenturies,when such towns as Mogadishu,Manda,Shanga,Pate, Ungwana,Unguja Ukuu, Mkokotoni, Kilwa, Dembeni, Sima, and Chibuene all emerged. Small fishing and farming communities,they obtained ivory, timber,tortoise shell, ambergris,rock crystal, slaves, gum copal, and iron from their mainlandneighbors, which they then for Persianand Chinese pottery,cloth, and iron exchangedwith Gulf merchants ware.By the 1Ih century, the townswereprospering, peoplewere buildingelaborate stone houses, and Islam was beginning to establish a presence in coastal communities.Mogadishu,Kilwa,and then Mombasasubsequentlyrose to wealthand butnew townswereemergingall along the coast, fueled by fameas majorentrepots, the expansion of tradewith southernArabiaand India,bringingincreasedimmiin its wake as the and cultural socioeconomic diversification gration, differentiation, Oceanworld. coast participated fully in the widerIndian If the broad outlines of Swahili developmentare now clear, however,they theirown history.For if, as we continueto conflictwith Swahilitraditions regarding posited earlier,proto-Sabaki speakershad expandednorth and were clustered in borderby the beginningof the 6th century, "Shungwaya" along the Kenya-Somali proto-Swahili speakershad begun to moveback down the coast by the 9h century, andMijikendaand Pokomo speakersonly occupiedtheircurrenthomelandsin the 16th century,the Kenya coast and hinterland would have been devoid of known in spite of now abundant inhabitants evidenceto for 6-13 centuries, archaeological the contrary.9-5

95 Helm, "Conflicting Histories," 275-94.

EARLY SWAHILI HISTORY RECONSIDERED

287

This is not a new problem, thoughit is still a perplexing one. Earlier, we found the putative thattraditions regarding establishment of coastaltowns by immigrants fromShirazto be historicalmetaphors for the establishment of hegemonyby leading waungwana familiesfromthe north,or Shungwaya.96 But how can one account for the alleged dispersal earlier of Swahili, Mijikenda, and Pokomo from Shungwaya betweenthe 9gh and 16thcenturies, long afterthe settlement of the coast andhinterland by peoples makingKwale/EIW and TanaITIW and speakingSabaki andSwahililanguages? This lackof congruence betweenarchaeological dataand the has led many to dismiss the Shungwayatraditions traditions as inventions, but that fails to takeaccountof the linguisticdataand possibly fails to appreciate the deeper meaningof the traditions. Accordingto Nurse and Hinnebusch,proto-Sabakifirst developedas some NortheastCoast speakersexpandedto the north and began to differentiatefrom theirfellows. Channeled along the narrowKenyancoastalplain,they occupied the coastalareaof Kenya,whereby the middleof the 15 millennium they were beginning to differentiate into Swahili,Ilwana, and Comorian/Mijikenda/Pokomo. By ca. AD 800, the latergroup also beganto split,with Comorian speakersmoving south along the coast beforesettlingin the ComoroIslands.Aboutthe same time,Swahili speakers also began to disperse along the coast, eventuallydifferentiatinginto northern and southerndialectclusters,leavingllwana speakerson the upperTana southof the SabakiRiver.97 River,Pokomoalongthe lowerTana,andMijikenda Such a hypothetical migratorypattern,based on the theory of least moves, makes the most logical sense of the genetic affiliationsand mutualinteractions As but it is not the only possibleexplanation. observedamongthe Sabakilanguages, Jan Vansina has found, positing migrationsto account for the developmentof in thisway maynot represent the best historical languagesdepictedin treediagrams solution.TakingEasternBantuas an example,Vansinahypothesizesthat it develBantuoriginallycoveringmost of easternand oped in waves, with proto-Eastern Africa.Overtime,it beganto be adoptedby autochthons and to differentiate central at the southernand northernextremes into Southern,Makuan,and Kilimanjaro, GreatLakes,and (North) while laterthecentral respectively, groupsplitintoCentral, Bantulanguagesand EastCoast,thusaccounting for the development of the Eastern theirpresentdistribution.98

96 Nurse and Spear, The Swahili, 70-79; Thomas Spear, "The Shirazi in Swahili Traditions, Culture, and History," History in Africa, 11 (1984), 291-305. See also Randall Pouwels, "Oral Historiographyand the Problemof the Shirazi on the East African Coast," History in Africa, I1 (1984), 23747. 97 Nurse and Hinnebusch,Swahili and Sabaki, 490-549. 98 Vansina,"New LinguisticEvidence and 'The BantuExpansion'."

288

THOMAS SPEAR

Such a wave model might also be used to accountfor the developmentand of Northeast distribution Coast,Sabaki,and Swahiliin an economicalway.99Under such a scenario,following the splits withinEasternBantunoted above,Northeast Coast speakerswould have been distributed throughout the coast and hinterland from the Rufiji River in the south to southernSomalia in the north.The earliest withinthisgroupwould then haveoccurredat its northern differentiation extremity, resultingin the development of Sabaki,with subsequentdifferentiation producing Pare,Seuta,and Ruvu in the south. Sabaki-speaking communities would then have been relativelystrung out and isolated along the Kenya coast, where the more peripheralgroups-Swahili and Ilwana-differentiated first, followed by the of the Comorian/Mijikenda/Pokomo disintegration cluster.'(( Accordingto the wave model,then,the Kenyan coastandhinterland wouldhave been inhabited throughout, in placeand theirlanguagesdeveloped most Sabakispeakerswould haveremained in situ, and only some Swahili,Comorian, and Mwanispeakers-all highly mobile seafarers-would havehadto moveaway. While differentfrom the previousmodel,the wave model still requiresthatthe earliestSwahili (and Comorian)speakersemergedin the northand subsequently spreadalongthe coast fromthere.Given Swahili's and Comorian'sclose linguistic with TIwana, it would be hard to envision a relationships Pokomo,and Mijikenda, in any case. Thus, it would seem that we still have to accord plausiblealtemative temporal priorityto northernSwahili sites over southernones, but not much, for Zanzibar, Kilwa,and Chibuene(along with the Comoros)were virtually contemporaneouswith the earliestnorthernsettlementsand dialects.The rapiddispersalof from other Swahilispeakerswould havebeen a factorin Swahili's differentiation Sabakilanguages,while the close maritime connectionsmaintained among coastal as separate townswouldhavekeptSwahilirelatively dialect homogeneous internally clusters slowly emerged among the northernand southern communities. The for northern vs. southernorigins of Swahilimay thus be less significant argument than previously thought,as new Swahili communitiesdeveloped more or less at widely scattered simultaneously pointsalongthe coastafterthe 8'h century.While of earlySwahililanguageand cultureprobablylay in the antecedents the immediate of Swahili speakers, local peoples, and overseas north, then, the incorporation in individual immigrants towns,each with its own physicaland social environment, meantthateach also developedalongits own linesthereafter. we choosethe treeor wave model,then,it seems obvious of whether Regardless the Kenyacoast continuouslyfrom early in thatEasternBantuspeakersinhabited of the Kwale and Tana with the development the I1" millennium, simultaneously thus causing us to reassess earlierdebates regardingthe Shungwaya traditions,
99 Alternatively,Sabaki speakers may have simply continued to slowly expand up the coast, differentiatingas they went, as envisioned by Walsh, "MijikendaOrigins" and Helm, "Conflicting Histories." 100 Nurse and Hinnebusch,Swahili anidSabaki, 476-80.

EARLY SWAHILI HISTORY RECONSIDERED

289

and Pokomoorigins. In analyzingthe Shirazitradiof Swahili,Mijikenda, tradition Swahili,formedof the prototypical thatthey represented we concluded tionsearlier, with people among whom they movementfrom the northerntowns, interaction into their societies. The centralmetaphorswere those of settled,and integration fromthe north, exchangedcloth as Shiraziarrived exchange,andmarriage migration, local elders' daughters for the rightto settle,and married withthe local inhabitants to establish prestigious lineages thatjoined the wealth and powers of each.'0' as expressing local social structures Similarly,Pouwels sees the Shirazitraditions andprocesses,especiallythe conflictsbetweenlocal farmersand herders,fishermen and Swahili and Arabs.'02And Middleton and traders,locals and immigrants, concludesthatSwahilitraditionsrepresent"the shaping of Swahili identity" and of that identityover the centuries"as people sought to "thecontinualreformations understandand resolve "the basic structuralprocesses and ambiguitieswithin is the main of the Swahilipositionas middlemen Swahilisociety.""Thevalidation moralsuperiority, anddivinewill to show the legitimacyof message,usingethnicity, The Shirazitraditionsthus represent and aspirations."'03 activities theirmercantile into what which diverse groups were incorporated historical processes by complex becameSwahilicommunities. may lie, as Pouwels has noted, not in the traditions The key to understanding to be aboutbut in the periodand condipurport the historicalperiodthe traditions themselves.While it is not possible to establish tionsthatgave riseto the traditions thatthey relatenot to Pouwelshypothesizes of the Shirazitraditions, the provenance were active and from Shiraz centuries when traders from the 8' to the 11"' the period 17'h centurieswhen from the to to the 13'h but period towns were formed, the earliest Swahili and local sought to Swahili hegemony challenged Arab immigrants Persian origins.104 themselvesby claimingprestigious distinguish tradition mayalso haveplayeda centralrole Viewed in thisway, the Shungwaya in Swahilimythology. Seen as the areawhere Swahili identitywas forged in the of the northeastern coast, from which the first Swahili earliesttradingsettlements the coast, it encapsulates the towns other to establish along trading spread in series of social processes and interaca Swahili of nature society incorporative thus in other settlements. The be Shungwayatradition were to tions that repeated while the and of the as a mainland, culturally the Swahili socially establishes people

1I01Nurse and Spear, The Swvahili, 70-79; Spear, "The Shirazi,"291-305.

102 Pouwcls, Horn and Crescent, 32-37; Pouwcls, "Oral Historiography,"242-54. See also Horton,Shanga, 423-26. 103 Middleton, Worldof the Swahili, 27-35. 104 Pouwcls, Horn and Crescent, 10-21, 35-37; Pouwcls, "OralHistoriography,"242-54.

290

THOMAS SPEAR

becameidentifiedreligiously laterShirazitraditions explainhow they subsequently with the widerworldof the Indian OceanandIslam.105 andhistorically an early IS millennium century,'06 Farfrom being an invention of the late 19"' of or the homeland the Sabaki-speaking peoples,'08 then, ShungCushiticstate,107 waya may connote the northernfocus of early Swahili society and the complex social processes by which that society emergedfrom its TIW and Sabaki roots. on the one hand, Reconcilingthe differenthistoriesconveyedby Swahilitraditions, and archaeology, linguistics,and history,on the other,thus requiresus to interpret the complex mythopoeicidioms that Swahili themselvesemploy to resolve the of theirown history.Only when these are analyzedrigorouslyin the contradictions data will we achieve a light of all the archaeological, linguistic,and documentary of earlySwahilihistory. convincingnew understanding

105 Middlcton, World of the Swvahili, 14-15; Pouwcls, Horn and Crescent, 54; Steven Feerman, "Economy, Society and Languagein Early East Africa," in Philip D. Curtin et al., African History (London, 1995), 126-27. 106 See fn. 93 above. 107 Allen, Swvahili Origins. 108 See fn. 92 above.