Shungwaya Myths as History

The earliest text to claim Shungwaya as the Mijikenda homeland was the Kitab al-Zunuj “Book of the Zanj.” Though likely written in the late nineteenth century, it was first made available to European scholars in a few Arabic manuscripts between 1912 and 1923.Footnote ²⁹ The anonymous author located Shungwaya at Bur Gao near the Juba River in southern Somalia and described it as the realm of the kings of the Kashur people. He also identified the Kashur as the ancestors of the Wanyika (“bush people,” an epithet later deprecated in favor of Mijikenda). The author emphasized that Kashur refugees became the clients of Arab communities after settling near Kenya’s coast.

This written version contradicted stories that Wanyika communities had told missionaries in the mid-nineteenth century. Some claimed to come south from Mount Mwangea, east from the Taita Hills, or even north from Rombo, near Mount Kilimanjaro. They agreed that they shared ancestry with the Pokomo and sometimes mentioned Shungwaya; but they described it as the ancestral homeland of Kiswahili speakers rather than their own.Footnote ³⁰

During the twentieth century, Mijikenda speakers began circulating oral traditions about migration from Shungwaya that aligned better with the Kitab al-Zunuj. Mijikenda elders told Spear in 1971 that their history began with a frantic flight southward from Shungwaya. They claimed a man with two wives and nine sons as their common ancestors, who lived alongside Oromo-speaking pastoralists whom they called “Galla.” Most narrators blamed their precipitous departure on an ancestor who ritually sacrificed the son of their Oromo-speaking neighbor. The sons separated as they fled retaliatory raids, and their families expanded into lineages and clans that formed nine confederations. They finally escaped their adversaries when foragers showed them forest clearings along Kenya’s coastal escarpment where they would be safe. Each confederation founded a kaya (“fortified town,” pl. makaya) on separate hilltops. The Digo settled furthest south, the Giriama settled in the north, and the other seven (Duruma, Rabai, Jibana, Chonyi, Ribe, Kauma, and Kambe) fell between.

These traditions served as a constitution for the emerging Mijikenda ethnicity. Claiming a distant location as an earlier origin extended the boundaries of ethnic belonging around all communities derided as Wanyika. The Shungwaya myth provided a schema for ordering people, places, rituals, and customs in historical memory. It explained how all the Mijikenda peoples related to one another as well as their neighbors. As Mijikenda historiologists recycled the Shungwaya myth to craft a shared ethnic ideology, they arranged and adapted knowledge of the past to suit contemporary political arguments; but they also assembled stories that their audiences could accept as true. Their efforts to balance the relevance of their narratives to modern audiences with accurate representations of the past were not so different from the work of scholars trained in Euro-American historiologies.

The Shungwaya Debates

Debates over Shungwaya narratives between 1955 and 2000 offer insights into some of the first interdisciplinary research conducted in Africa. In theory, scholars were supposed to seek confirmation only after reaching conclusions through their own disciplines. In practice, ambiguities of interpretation in each discipline often led scholars “to spurious agreement” as they fit evidence into prevailing hypotheses.Footnote ³¹ So, scholars often advanced debate by correcting misrepresentations of data from their respective disciplines. Though these corrections are common in interdisciplinary research, the Shungwaya debates generated exceptionally acerbic recriminations whenever scholars questioned one another’s commitment to the value of African perspectives as presented in oral traditions.

The first phase of research on Shungwaya narratives was fragmented and opportunistic. Missionaries and travelers recorded the origins of communities they proselytized and visited in the nineteenth century, but only the Kilindini and Segeju communities claimed to come from Shungwaya.Footnote ³² By the time anthropologists and colonial officials began recording oral traditions in the early twentieth century, Mijikenda, Taita, and other communities had begun claiming origins in Shungwaya as well.Footnote ³³ Scholars decided that the Kashur of the Kitab al-Zunuj were likely the Mozungullos encountered by Portuguese travelers in the sixteenth century, who became the Wanyika in the eighteenth century, and eventually adopted the Mijikenda ethnonym in the mid-twentieth century. By 1955, A. H. J. Prins was describing Shungwaya as a linguistic homeland (urheimat) for Bantu speakers in the region, and V. L. Grottanelli evocatively called it a “lost African metropolis.”Footnote ³⁴ These anthropologists buttressed the migration narratives with catalogs of cultural similarities among Bantu-speaking communities in littoral East Africa.Footnote ³⁵

However, scholars sometimes conflated oral traditions with only slight similarities. For example, linguist H. E. Lambert extrapolated a slender claim about crossing a river near “the great water” in Meru traditions to suggest that they and linguistically related communities in central Kenya also migrated from Shungwaya.Footnote ³⁶ By the time J. Forbes Munro and Iris Berger (separately) pointed out the improbability of this logic in 1967, historians had already integrated Lambert’s claims about migration into survey texts on East African history.Footnote ³⁷ Their critiques invited a closer look at what scholars began to call the “Shungwaya problem.”

In the second phase of debate, scholars identified problems with the migration narrative. Archeologist Neville Chittick expressed some of the first doubts in 1969, but he limited his critique to the issue of Shungwaya’s location.Footnote ³⁸ His survey of three sites near Bur Gao did not meet his expectations for an abandoned capital city. Historian Fred Morton went much further in his 1972 critique of the Kitab al-Zunuj by questioning whether the Mijikenda had ever migrated.

Morton argued that members of the colonial Arab Administration of British East Africa inserted passages about the “famous laws” of the Kashur (i.e., Mijikenda forebears) into the Kitab al-Zunuj before sharing the manuscript with administrators. These passages validated practices that British officials were then litigating as violations of anti-slavery laws in their colonial courts.Footnote ³⁹ Morton also suspected that Arab administrators inserted passages that characterized the Kashur as dependents of their “Arab tribes.” The Mijikenda, he argued, borrowed these migration narratives from court hearings to unify their nine communities as a single tribe. Turning to the wider set of Shungwaya myths, Morton questioned why scholars presumed that Portuguese descriptions of Mozungullos in the late sixteenth century were evidence for recent immigration rather than a belated acknowledgement of their presence.

In addition to challenging the migration hypothesis, Morton recycled some archeological evidence. Archeologist James Kirkman had previously interpreted a style of fingernail ornamentation on pottery that he recovered at Gedi, Kenya, to suggest that the Kiswahili-speaking Kilindini had lived there for two centuries before fleeing Oromo raids in the mid-sixteenth century. Since the ChiMijikenda-speaking Giriama were the only group in the region that used fingernail ornamentation in the 1960s, Kirkman reasoned that they must have learned the technique from the Kilindini after joining them along the coast. Morton suggested instead that Kirkman was misled by a desire to confirm the Shungwaya myth. He offered the less convoluted possibility that this style of ornamentation indicated the earlier and continuing presence of the Giriama (rather than the Kilindini) around Gedi. He also criticized archeologists’ common practice (at the time) of associating material culture with discrete ethnic groups as too “risky” to rely on over long periods of time.Footnote ⁴⁰

Morton’s hypothesis prompted a series of rebukes.Footnote ⁴¹ First, Spear pointedly reminded him that arguments from negative evidence are fallacious; hence, missionaries’ failure to note a connection between Shungwaya and the Mijikenda was unconvincing. He suggested that missionaries’ informants were unreliable because many of them were formerly enslaved refugees who lacked knowledge of their adoptive communities’ histories. Spear also disputed Morton’s specific argument that the tradition spread among Mijikenda communities through colonial courts. He considered it unlikely that the Shungwaya narrative could have acquired several variants among the Mijikenda within a few decades of the circulation of Kitab al-Zunuj, especially since the movement to unify all nine Mijikenda communities seemed to begin after World War II. This kind of criticism about sources and sequences was standard fare for arguments among historians.

However, Spear then disparaged Morton’s hypothesis as a “historian’s myth” because it dismissed Africans’ traditional histories as fictitious. The article was the first time Spear published his interpretation of Mijikenda oral traditions, collected in 1971. After presenting a detailed version of the migration narrative, he suggested that interdisciplinary research confirmed the history as presented by Mijikenda narrators, with some modifications. Moreover, oral traditions performed a necessary role in reconstructing African pasts, which lacked documentary evidence. He accepted the sociological role of the Shungwaya myth as a charter for key institutions like age-sets, clans, and makaya, which he would later name the “kaya complex.” However, he claimed these institutions had not been functioning since the nineteenth century, so they would be most helpful in reconstructing Mijikenda life prior to their disintegration rather than contemporary Mijikenda culture.

Spear’s argument is a clear example of how Africanists used the confirmation approach to validate oral traditions as valid historical sources. He offered four corroborating forms of evidence to support the migration. First, the broad correspondence of details in the myth among seven of nine Mijikenda groups supported their veracity. This logic required the (improbable) situation that these nine communities were isolated from one another in their separate makaya for nearly four centuries. Allowing the possibility that they did not develop distinct traditions, he suggested that Morton’s hypothesis would still unjustifiably disregard the consonance of migration myths among Mijikenda, Pokomo, Segeju, Taita, and Galla (i.e., Oromo). Second, he noted that the distribution of cultural features among these communities (descent patterns, age-set names, etc.) aligned well with the traditions. Third, he drew on Malcolm Guthrie’s classification of Bantu languages to claim that Taita, Mijikenda, and Pokomo shared a common linguistic origin. Finally, he turned to Portuguese documents between 1569 and 1624 that positively identified Segeju, Mijikenda (as Mozungullos), and Oromo in coastal Kenya. Spear concluded that the latter three bodies of evidence confirmed the validity of the Shungwaya traditions. Morton would need more than negative evidence to dismiss them.

Before Morton could respond, Neville Chittick published a “note” focused solely on debunking his argument. Curiously, the archeologist made no comments on Morton’s treatment of pottery evidence. Instead, he tried his own hand at recycling evidence by interpreting the text of Kitab al-Zunuj. He outlined each section to demonstrate that the Kashur were integrated throughout much of the manuscript. He then claimed that if these portions were late additions, “they would surely have been distinguishable as separate sections” – disregarding the possibility of more subtle manipulations.Footnote ⁴² While he accepted Morton’s suggestion that lines related to enslavement may have been added, he denied that it had any bearing on the migration hypothesis. Unfortunately, both Morton and Chittick relied on English translations of an Italian translation of the single published version of one Arabic manuscript. Without detailed comparisons across multiple manuscripts, Chittick’s critiques were unconvincing.

Linguist Thomas Hinnebusch joined the debate in 1976 to offer a sobering view of the pitfalls that historians encounter when they use linguistic data without specific training in linguistic methods.Footnote ⁴³ In particular, he noted that Spear could not confirm Shungwaya as a linguistic homeland for Taita, Pokomo, and Mijikenda because he misunderstood the data: Guthrie’s Comparative Bantu cataloged typological resemblances and geographic proximity but not reconstructions of shared innovations that indicate “genetic” relationships to linguists. Hinnebusch’s own research supported a classification of Pokomo, Mijikenda, and Kiswahili as a Sabaki subgroup of Northeast Coast Bantu but excluded Segeju and Taita languages. Thus, communities with oral traditions of Shungwaya did not correspond with those that shared language origins. Hinnebusch also noted that the two successive migrations proposed by Spear – a northward movement through Kenya to Southern Somalia, followed by staggered migrations back south – violated the “least-moves” rule of parsimony by which historical linguists account for the sequential flowering of languages.Footnote ⁴⁴

These interdisciplinary interventions inspired both Morton and Spear to revisit their hypotheses in 1977. Morton refuted Spear’s aspersions on his sources by presenting a slew of documents that showed that missionaries, travelers, and their informants were both knowledgeable and interested in the question of “origins.”Footnote ⁴⁵ He also reviewed Pokomo oral traditions instead of focusing solely on the text of Kitab al-Zunuj. He found no specific evidence to support Alice Werner’s oft-cited assumption that Pokomo claimed origins in Shungwaya. As a new hypothesis, he asserted that Mijikenda origins “will be discovered, not at any distant location, but on the Kenya Coast.”Footnote ⁴⁶

In Spear’s response to Hinnebusch that same year, he accepted the linguistic corrections and offered a reconsideration of his earlier article as “a case study in the application of linguistic evidence to historical data.”Footnote ⁴⁷ The first salient point from this revision was the acknowledgment that time depths derived from linguistic classifications did not align precisely with historical events recalled in oral traditions. He therefore conceded that Shungwaya could not be the homeland of all Northeast Coast Bantu, but he still felt that four centuries was sufficient time to account for the more recent division of its Sabaki subgroup into Pokomo and Mijikenda; he pointed to archaic features in Kiswahili to suggest its speakers separated themselves earlier.Footnote ⁴⁸ Second, since the Taita and Segeju languages could not be classified as Sabaki or even Northeast Coast Bantu, he agreed that they had adopted the myth later from groups that had migrated from Shungwaya. Third, he accepted the grouping of Swahili with the Mijikenda and Pokomo, which allowed him to apply evidence in debates about Swahili origins to the Shungwaya problem. However, his new interpretation relied on what he misread as finds of Kwale ware pottery between the eleventh and fifteenth centuries in southern Somalia, as well as assumptions that Kiswahili was a mixture of Bantu and immigrant “Shirazi” settlers from Persia to that area around the tenth century CE.Footnote ⁴⁹

This second phase of debate thus illustrated the hazards Africanist scholars faced as they felt their way toward interdisciplinary approaches by educating one another about their respective disciplines. When Morton reinterpreted pottery and Chittick analyzed the Kitab al-Zunuj, they made what seemed to be valid conclusions but also overlooked historiological assumptions of the discipline that generated the data. Morton’s suggestion that the ease of learning a specific pottery decoration made it unsuitable for identifying past peoples unwittingly confirmed why pottery decorations are a useful measure of contact among historical populations. Chittick’s analysis of a single manuscript was similarly unsophisticated in his reliance on a single manuscript filtered through two translations, as well as his failure to discuss the complex relationships in East Africa between written texts and oral narratives.Footnote ⁵⁰ While these two scholars seemed to talk past one another, Hinnebusch and Spear’s engagement enabled scholars to integrate discussions on Mijikenda migrations into wider debates about Swahili origins.

Scholars’ disagreements over how to understand relationships among different communities in coastal Kenya dominated the third phase of research. Historians focused on reconciling the Mijikenda migration narratives with the oral traditions of Swahili communities, which had claimed descent from immigrants hailing from either Shiraz (in Iran) or various sites in Arabia since the sixteenth century. Scholars and laypeople had long considered the Swahili a creole population formed through the mixture of Africans with Arab or Persian migrants.Footnote ⁵¹ However, surveys and excavations by George Abungu in the Tana River estuary and Mark Horton at the seaside settlement of Shanga indicated that East Africa’s indigenous residents established dense coastal settlements prior to significant Arab or Persian influence.Footnote ⁵² Archeologist Henry Mutoro also began the first surveys and excavations along Kenya’s coastal escarpment that indicated earlier occupation than the Shungwaya migration narrative anticipated.Footnote ⁵³ These findings led historians to consider the possibility of relationships among Mijikenda, Swahili, and other communities in Kenya prior to the purported migration from Shungwaya.

For Randall Pouwels, Shungwaya denoted a pre-Islamic period when Sabaki speakers (the forebears of the Swahili and Mijikenda) worked out strategies for managing the diverse ecologies of littoral East Africa. He cited the archeological research of Mutoro, Abungu, and Horton, as well as linguistic research by Derek Nurse, to suggest that pastoralists speaking Cushitic languages must have played some role in their success, at least in Kenya.Footnote ⁵⁴ Pouwels presented the formation of Shirazi societies as the next era, when established “Shungwaya” communities converted to Islam, followed by a third era of “Swahili” communities who encountered Portuguese visitors and forged political relationships with Omani Arabs. Pouwels’s work treated hinterland communities as marginal to (because they were marginalized by) Swahili communities. However, his Shungwaya epoch also incorporated Mijikenda, other Sabaki speakers, and Cushitic speakers into the broader sweep of Swahili history in order to dismantle the prevailing theory of Asiatic colonization.

Conversely, Spear’s publications reinforced the division of Swahili and Mijikenda communities into different ethnicized historiographies. Having corrected his linguistic interpretations, he presented the history of the Mijikenda in three publications: The Kaya Complex (1978), Kenya’s Past (1981), and Traditions of Origin and their Interpretation: The Mijikenda of Kenya (1981).Footnote ⁵⁵ These works presented Shungwaya as the origin of the Mijikenda and Pokomo, while acknowledging the presence of some Swahili communities. He also collaborated with linguist Derek Nurse to publish The Swahili: Reconstructing the History and Language of an African Society in 1984.Footnote ⁵⁶ This book soundly repudiated myths of Asiatic colonization by presenting the Swahili as an African people rather than a creole community. However, it also categorically distinguished the Swahili from their nearest neighbors and linguistic cousins.Footnote ⁵⁷

Pouwels’s and Spear’s work complemented the work of other Africanists by demonstrating how the confirmation approach of multidisciplinary research effectively dismantled colonial historiography. However, historian James de Vere Allen offered some doubts. In particular, he dismissed linguistic and archeological evidence that seemed to contradict his interpretations of oral traditions he had collected during his long residence in East Africa. He favored a fully integrated narrative of East Africa’s littoral history that recast Shirazi myths as Islamized variants of Shungwaya myths and emphasized multilingual collaborations in East Africa that accorded with the findings of other scholars. However, he also advanced extravagant claims about the occupation, collapse, and legacy of a Shungwaya state. Allen’s monograph was published posthumously in 1993, but he first outlined his hypothesis in a 1983 review of the Mijikenda traditions published by Spear.Footnote ⁵⁸

This review is worth detailing because it illustrates how historians’ entangled their disagreements over interdisciplinary evidence with their commitment to magnifying African voices. Allen began by admiring the tremendous collection of oral traditions that Spear had published; but he worried that historians would subsequently disregard less systematic collections of oral traditions as deficient. He also accused Spear of prematurely dismissing Segeju oral traditions in “his haste to accommodate linguistic data.”Footnote ⁵⁹ For Allen, the Segeju were essential to his hypothesis that Shungwaya was a multilinguistic state in northern Kenya with Cushitic rulers and Bantu subjects, which broke up into several successor states shortly before the arrival of Portuguese travelers. In his interpretation, the Mijikenda lived on the outskirts of this state, but they adopted Shungwaya’s traditions and institutions from Segeju migrants as the state collapsed. Allen’s elaborate description of an expansive Shungwaya state, a Shirazi “mode of dominance,” and loose associations of ethnonyms with languages overshadowed his keen observation that Segeju influence needed more attention.

Spear declined to dispute the details of Allen’s critique when the journal editors requested a response before publishing the review. Instead, he characterized Allen’s hypothesis as a “conflation of fact and fiction,” derided him for collecting facts like “children collect shells, pebbles, and bits of seaweed on the beach,” and described him as disdainful of the “lessons of historical linguistics, anthropology, history, or archeology.”Footnote ⁶⁰ He concluded by characterizing Allen’s work as “parahistory.” Though Allen and Spear disagreed sharply about how to evaluate the information they provided or confirm these narratives with other evidence, they both stalwartly defended the potential of oral traditions to elevate Africans’ knowledge of the past. Their dispute cemented Spear’s interpretation of migration as conventional, even as archeological and linguistic research complicated its underlying assumption that communities moved great distances in relative isolation from one another.

The fourth, most recent, phase of sustained research on Shungwaya dismantled the migration hypothesis from two different directions. Scholars reassessed how the migration narrative reflected colonial identity politics and increased the archeological evidence available on Kenya’s coastal escarpment (rather than Somalia or Tanzania). Anthropologist Martin Walsh offered an opening salvo in a 1992 literature review of the Shungwaya debate that supported Morton’s hypothesis for a Kenyan origin of the Mijikenda.Footnote ⁶¹ He expanded on previous critiques of Spear’s research, such as the inconsistency of the age-set lists that Spear used to date the migration and his undue confidence in the precision of Portuguese maps. Walsh also stridently critiqued Spear’s method for reducing the myth to a “single historical truth” by ignoring variants, especially in the widely read Kaya Complex. He also suggested that Spear “suffered from a ‘kaya complex’” that made him overlook incongruities between the “number and organization” of makaya described in the traditions and the empirically observed pattern of Mijikenda settlements.Footnote ⁶²

Historian Justin Willis was already thinking along similar lines. His (1993) monograph examined how the Swahili and Mijikenda ethnic groups mutually constituted one another as their proponents negotiated with the British colonial government in the early twentieth century.Footnote ⁶³ He criticized Spear and other scholars for projecting the modern Mijikenda ethnic identity and Wanyika epithet onto sixteenth-century Mozungullos without examining the distinct contexts in which Africans created these successive identities. By focusing on the formation of the most recent escarpment identity, he showed that Mijikenda speakers began circulating the Shungwaya myth to unify their interests decades before World War II, when Swahili communities expelled those who sought assimilation in colonial Mombasa. Hence, the variability of Shungwaya traditions among Mijikenda communities was not an erosion of a single ancient tradition, as presumed by Spear, but the result of recent, uneven adoption.Footnote ⁶⁴ Willis also followed through on Walsh’s suggestion to survey Mijikenda settlements. Even excluding Digo sites, he identified more than forty makaya. The complex landscape could not accommodate Spear’s hypothesis of a staggered migration into nine settlements that remained isolated from one another until the nineteenth century.Footnote ⁶⁵

In 2000, Richard Helm’s report on his archeological survey of settlement sites along Kenya’s escarpment similarly challenged this migration hypothesis; but he also convincingly showed that Mijikenda knowledge about the past extended much deeper in time than previously considered.Footnote ⁶⁶ Helm applied the seriation of local pottery wares that archeologists had developed for coastal settlements in the previous decade to sequence and date dozens of settlement sites in the escarpment where most Mijikenda live. He found that the sequence derived from this objective data aligned with Mijikenda informants’ independent classification of the sites as pre-kaya, kaya, or post-kaya settlements. They had retained knowledge about the occupation sequence of settlement sites for over a thousand years, perhaps because the physical features of the landscapes served as effective mnemonics.Footnote ⁶⁷

Helm also presented a remarkable plot twist: escarpment residents abandoned their makaya right around the time that Spear claimed they were founded. As anticipated in Mutoro’s excavations, his survey showed Kenya’s coastal escarpment had been occupied continuously by (likely) Bantu speakers since at least 100 CE, as opposed to 1500 CE. In addition, the changes in size, density, and expanding ecological distribution of these settlements matched those of “Swahili” coastal settlements. These settlement patterns did not diverge until the seventeenth century. At that point, the number of large, dense settlements (i.e., makaya) decreased and scattered homesteads increased until the escarpment resembled the modern rural landscape.Footnote ⁶⁸

As archeologists Peter Schmidt and Jonathan Walz noted over a decade ago, Helm’s work “overturns how historians have used oral tradition to represent the history of the region.”Footnote ⁶⁹ Yet his findings barely percolated beyond his discipline. Anthropologists Janet McIntosh (Reference McIntosh2009) and Jeong Kyung Park (Reference Park2012) built on David Parkin’s foundational work of Giriama ethnography to examine the roles of makaya in contemporary Mijikenda societies but limited their analysis to colonial and postcolonial times.Footnote ⁷⁰ A multidisciplinary collection in 2014 similarly sidestepped discussions of Helm’s archeological evidence, even though the volume intended to recenter Mijikenda perspectives in Kenya’s coastal history.Footnote ⁷¹ The sole archeological chapter presented evidence for connections between the coast and the Taita Hills without discussing the escarpment between them. Walsh’s contribution to the volume made just a passing reference to Helm to support his hypothesis that the Mijikenda adopted the Shungwaya myth (and many loanwords) from Segeju migrants rather than migrate themselves.Footnote ⁷² Walsh disclaimed assumptions that he was a linguist and invited other scholars to evaluate the data. For now, he has had the last word. Scholars have largely abandoned the search for origins and migration paths that animated Shungwaya debates in the twentieth century.

The Shungwaya debates became increasingly acrimonious as participants accused one another of peddling myths, engaging in parahistory, or suffering from a “complex.” When historians borrowed evidence from linguistics or archeology in the Shungwaya debates, they often received advice professionally and modified their hypotheses. However, when they suspected their colleagues of dismissing Africans’ knowledge about the past, they peppered their critiques with biting rhetoric. Chittick, Spear, and others did not pile on their criticisms of Morton’s hypothesis simply because they disagreed with his analysis of Kitab al-Zunuj. Rather, they objected to the presumptuous extension of his dismissal to all Shungwaya traditions. Allen seemed less concerned about Spear’s use of linguistic data than his wholesale rejection of Segeju traditions; and Spear’s belittling retort was prompted by his concern that oral traditions needed systematic analysis to establish them as verifiable historical evidence.

Reimagining the Settlement Histories of Littoral East Africa

First, linking uganga to clanship explains why cross-linguistic collaborations became more common as speakers of Sabaki Bantu languages moved into more diverse environments between 500 CE and 1250 CE in Kenya. Though foragers, Bantu speakers, and Southern Cushitic speakers occasionally interacted in the region before then, they usually confined their activities to different ecological niches that suited their respective subsistence strategies. This pattern began to change around 500 CE, just as Proto-Sabaki Bantu began diverging into its daughter languages: Kiswahili, Elwana, Comorian, Pokomo, and Mijikenda.Footnote ⁷⁷ Archaeologists and linguists usually explain the increase in cross-linguistic collaboration during this period as a consequence of intermarriage. However, interrogating this explanation from Govi’s theoretical perspective on clanship and uganga suggests that speakers of Sabaki languages began organizing clans around distinctive knowledge as well as ancestry. This innovation enabled them to integrate communities who spoke unrelated languages and pursued different subsistence strategies into a shared society without the burdens of intermarriage.

Oral traditions emphasize that Bantu speakers in littoral East Africa depended on foragers for their specialized knowledge of the forests, where they could find well-watered areas for cultivating gardens.Footnote ⁷⁸ Helm’s initial survey of the coastal escarpment confirmed that Bantu speakers settled in places previously occupied by foragers.Footnote ⁷⁹ Yet, other surveys of contemporary forager and farmer sites in the early first millennium CE show no signs that they traded items or otherwise shared a common material culture, even when located within walking distance of one another. Archeologists have thus concluded that foragers and iron-working farmers did not interact frequently in the “pioneer phase” of Bantu settlement.Footnote ⁸⁰

Their collaborations changed between 500 and 1250 CE. Speakers of Sabaki languages in Kenya’s coastal escarpment started organizing a three-tier settlement hierarchy (hamlet, village, town) that accommodated larger, more complex settlements as they stretched into new environments.Footnote ⁸¹ This expansion almost certainly displaced foragers; but archeologists also found oceanic imports and pottery (which they associate with Bantu speakers) at foraging sites in the later first millennium CE. These finds suggest the two populations had begun trading frequently with one another.

Instead of detailing foragers’ displacement or their roles in trade, Mijikenda oral traditions focus on the knowledge that would have motivated Bantu-speaking communities to integrate them into their communities. For example, Toya wa Kiti told Spear, “I can’t explain exactly who they [the foragers] were, but they were waganga [medicine-men].”Footnote ⁸² Some of Spear’s other informants emphasized that foragers taught them rituals, war strategies, and how to hunt (and fight) with bows and arrows. Mijikenda traditions thus describe their ancestors’ relationships with foragers in positive terms because of the uganga they contributed.

Sabaki speakers also intensified their engagements with neighboring agro-pastoralists who spoke Southern Cushitic languages.Footnote ⁸³ These neighbors provided knowledge about cultivating rain-fed crops (sorghum, finger millet, and pearl millet) that made it possible to live in drier environments. Archeologists have recovered traces of these crops in (likely) Bantu settlements starting as early as 200 CE but certainly by 600 CE.Footnote ⁸⁴ Govi made the relationship between uganga and cultivation techniques explicit. He explained: “The Chonyi [i.e., his Mijikenda confederation] were magicians and they tested the land by using uganga… . They examined the soil and decided it was fertile and suitable for their crops.”Footnote ⁸⁵ In addition, ethnographic descriptions of agricultural practices in modern Swahili communities suggest that their forebears would have regarded rain-fed agriculture as a new kind of uganga because they select “guardians of the soil” who protect and oversee annual agricultural rituals.Footnote ⁸⁶ Speakers of Sabaki languages thus regarded seemingly mundane knowledge – including where and when to plant – as uganga mediated through the unseen world.

Scholars conventionally use the generic process of intermarriage among Southern Cushitic- and Bantu-speaking populations to explain the transfer of knowledge that made settlement in more diverse ecologies possible in littoral Kenya. Dental consonants that Southern Cushitic languages share with Elwana, Upper Pokomo, and northern Kiswahili dialects, but not other Sabaki languages, support this hypothesis.Footnote ⁸⁷ The distinction between dental and alveolar consonants is subtle enough that linguists assume Southern Cushitic-speaking parents introduced the sounds to children they raised with Sabaki speakers.

Although intermarriage provided a way to acquire knowledge over many generations, the link that Govi emphasized between uganga and clanship suggest speakers of Sabaki Bantu languages adopted a strategy that made the benefits of diverse knowledge available more rapidly. They integrated entire communities into their settlements as co-residential, but not intermarrying, groups. Since these groups focused on producing different resources, they could use the surrounding lands, rivers, forests, and ocean more effectively than a village with a single lineage or even a set of intermarrying lineages. These collaborations also enabled settlements in environmental transition zones that would otherwise be difficult for communities following a single subsistence strategy.Footnote ⁸⁸

Roderick McIntosh modeled a similar scenario for the Inland Niger Delta of West Africa, where residents increased their capacity to support larger populations and a more complex society by sharing the fruits of their respective knowledge.Footnote ⁸⁹ This strategy depended on heterarchical relationships that maintained each community’s internal coherence so that they could preserve hyper-specialized knowledge. McIntosh used modern occupational castes in West Africa as ethnographic analogies for how these communities would have supported rather than dominated one another. Govi’s description of uganga as corporate knowledge suggests that the relevant ethnographic analogy in littoral Kenya would be clans.

A conventional definition of clan is a group of lineages who claim descent from a common ancestor but cannot trace precise genealogical links with one another. The most widespread word for this concept in Bantu-speaking East Africa is *lukólò. Footnote ⁹⁰ However, it is rare among Sabaki speakers: only some Kiswahili speakers use the reflex ukoo to mean a cognatic descent group.Footnote ⁹¹ Most Kiswahili and some ChiMijikenda speakers use their reflexes of *lukólò to denote generic kinship instead.Footnote ⁹² Derivations from *lukólò are absent from other Sabaki languages. Instead, the speakers of each Sabaki language innovated different words for clan as they diverged from one another.Footnote ⁹³

The link that Govi posited between clanship and uganga suggests they needed a new word to emphasize the knowledge that defined clans rather than common ancestry. As Neil Kodesh has suggested for early Buganda, these clans would have used strategies for forming sets of intermarrying lineages to distinguish and protect their networks of knowledge.Footnote ⁹⁴ The earlier *lukólò clans (i.e., defined only by intermarriage) could connect lineages from different settlements (and speech communities). However, articulating new words for clan (i.e., defined by distinctive knowledge) helped manage relationships with others in the same settlement without the burdens of assimilating people through marriage. Sabaki speakers thus innovated a cosmopolitan ethic that enabled rapid and sustained cross-linguistic collaboration.

Since Sabaki speakers materialized these social divisions in the organization of their towns, Govi’s association of uganga with clanship also clarifies the innovation of multi-component settlements in the latter half of the first millennium CE. Archeologists identify these sites as the antecedents of Swahili “stone towns.” However, Helm’s survey of escarpment communities showed that they also joined in this centuries-long transition to denser (if not quite urban) living prior to 1650 CE. Although escarpment communities did not build with coral, the novel clan structure that speakers of most Sabaki languages innovated is visible in coastal towns like Shanga.

Shanga is a sea-side settlement on the island of Pate in the Lamu Archipelago near the Kenya-Somali border. It was founded in the eighth century CE and abandoned around 1400 CE. When Horton excavated Shanga, he drew directly on Mijikenda descriptions of makaya settlements as an ethnographic analogy to imagine its spaces.Footnote ⁹⁵ He interpreted the town quarters as the domains of different clans while the central enclosure served as a communal space shared by the entire town. Residents later added mosques to this space, which Horton used in support of his interpretation that a large “hall” in the enclosure was a ritual building similar to moro shrines where Mijikenda elders stored sacred objects in their makaya. He also suggested that clans established different cemeteries outside of town that suggest a concern for descent that would be expected of *lukólò clans.

Govi’s observation that communities need a lot of uganga to build a “peaceful home” does not substantially alter Horton’s interpretation. However, narratives in Mijikenda oral tradition about learning new uganga from foragers, or clan members who had been “lost” in the migration from Shungwaya, suggests that Sabaki speakers consciously developed new knowledge to distinguish their clans from others in their shared settlements. Compared to the relatively dispersed, small settlements of Bantu-speaking communities prior to 500 CE, multi-component towns created a new dimension of public rivalry. One of the ways that clans could assert themselves in negotiations over space and influence was by developing useful uganga. As Govi noted, every clan acquired “two or three different kinds of uganga.” At Shanga, for instance, residents innovated new subsistence techniques (e.g., deep-sea fishing and cattle husbandry) and craft specialties (e.g., carving coral basins and casting silver coins) over several centuries. Horton’s excavations show that the products of these specialized activities were often dispersed to different residential areas of the town.

Entrusting the stewardship of knowledge and space to clans appears to be the key innovation that enabled Sabaki speakers to stretch out their settlements from Somalia to Mozambique, as well as the Comoros Islands, the inland estuary of the Tana River, and Kenya’s littoral escarpment. The other Northeast Coast Bantu speech communities (Ruvu, Seuta, and Pare in Tanzania) remained relatively stationary in comparison.Footnote ⁹⁶ Early scholars’ theory of Asiatic colonization had suggested that immigrants were responsible for the rapid growth of coastal communities during this era. Interpreting these innovations as uganga instead suggests speakers of Sabaki languages developed new innovations at a relatively rapid pace because they were motivated by inter-clan competition. As in the Inland Niger Delta, they were well on their way to urbanism before intensifying their participation in trans-regional commerce.

The third transition illuminated by Govi’s interpretive framework is the integration of Islamic knowledge by coastal towns. Specifically, it reveals the conceptual framework into which early Kiswahili speakers would have adopted specific Islamic practices and explains why Muslim communities on the coast did not proselytize to their neighbors. Scholars often suggest that Africans adopted Islamic practices to engender trust with traveling Muslim merchants. However, the residents of littoral East Africa also sought instruction in Islamic techniques for healing maladies, praying for rain, and guarding against malevolent spirits. Just as clans relied on one another to provide a wide variety of food and other resources from the surrounding landscapes and seascapes, they valued multiple forms of spiritual power, including Islamic ones, to protect and benefit the entire town.

They elaborated their understanding of Islamic practices through familiar concepts. The testimony of Muslim writers from the time shows that it may have been relatively simple for Kiswahili speakers to extend their notions of uganga to include Islam. Bantu speakers had long ago developed the notion of a Creator God, albeit one far removed from the daily concerns of humans.Footnote ⁹⁷ In addition, Muslim writers described many of the activities they observed in East Africa as religious. In the tenth century, Abū Zayd al-Ḥasan al-Sīrāfī reported that men “who devote themselves to a life of piety [preach] to them, calling on them to keep God in their minds … and describing to them the fate of their people who have died.”Footnote ⁹⁸ Al-Masʿudi wrote around the same time that East African communities “have an elegant language and men who preach [obedience to God] in it.”Footnote ⁹⁹ Al-Idrisi reported that non-Muslim East Africans called these people mganga as late as the twelfth century.Footnote ¹⁰⁰ This last example emphasizes that in the minds of Muslim observers, uganga was a religious practice comparable to Islam. The early Kiswahili speakers at Shanga may have made the obverse conclusion that Islam was a kind of uganga, or perhaps many kinds of uganga.

This view of Islam as a collection of spiritual techniques rather than a single, abstract religion is precisely how Buzurg ibn Shahriyar described the conversion of an East African community to Islam in the tenth century. Allegedly told in the voice of an East African ruler whom Arab traders enslaved, the tale described conversion to Islam as a gradual unfolding of ritual knowledge rather than a spontaneous, all-encompassing conversion that required new Muslims to reject previous practices and beliefs. After Arab merchants kidnapped the ruler in East Africa, they sold him to a slaveholder in Oman, who took him to Basrah. There, the ruler learned to pray and fast, as well as read parts of the Qur’an. After being sold to a man in Baghdad, he “completed [his] knowledge of the Qur’an and prayed with the men in the mosque.” Later, the enslaved ruler absconded with a caravan headed for Mecca, and the pilgrims taught him how to perform the ceremonies of hajj. After completing his pilgrimage, the ruler followed the banks of the Nile to return home, resumed his place as king, and taught his people the “precepts of Islam,” which he summarized as “true faith, prayers, fasting, the pilgrimage, and what is permitted and what is forbidden.”Footnote ¹⁰¹

Though this apocryphal tale may not represent the authentic voice of any particular ruler in East Africa, it demonstrates that the tenth-century Arabs who told the story regarded Islam as a repertoire of spiritual techniques, just as Mijikenda orators describe uganga in their oral traditions. Scholars sometimes attempt to identify when various sects of Islam arrived in different parts of Africa, but examining Islam from the perspective of uganga (or other local systems of knowledge) emphasizes instead the specific Islamic practices that Africans integrated into their respective spiritual and technical schemas.Footnote ¹⁰² Coastal communities would have been motivated to seek proprietary Islamic knowledge for their clans as relationships with Muslim traders increased in importance – though no clan could monopolize Islamic knowledge.

The relationship between uganga and clan affiliation also answers a question that scholars have often asked about Islam in East Africa: why was conversion to Islam among Mijikenda and other inland communities limited to individuals rather than communities, until the late nineteenth century? Since Sabaki speakers regarded Islamic knowledge as a kind of uganga, their clans were loath to share their knowledge. They protected their Islamic knowledge and the coherence of their communities by not sharing it outside their clans. However, anyone could benefit from Islamic forms of uganga by allying with a Muslim clan or seeking the aid of a Muslim healer. Non-Muslims could also learn Islamic knowledge themselves, but only if they relinquished affiliations with their natal clan and joined a Muslim clan, which was the conventional process of conversion well into the nineteenth century.Footnote ¹⁰³ The communal ownership of knowledge and the value that non-Muslims placed on their own uganga discouraged conversion efforts but also encouraged alliances among communities. It was only after Muslims in the Zanzibar Sultanate committed to exclusionary religious identities of Arab orthodoxy that ChiMijikenda-speaking wanyika became marginalized enough that people who aspired to lead inland communities rather than escape them sought conversion to Islam.

Govi’s historiological intervention of linking uganga to clan affiliation clarifies how speakers of Sabaki languages integrated autochthonous communities in the lands they entered as well as newcomers to their shores and hills. Elevating his interpretation allows the historical knowledge that African communities have curated about their past to interrogate the historiologies of Euro-American disciplines.