Kiswahili took in the Indian Ocean (c. 1000) — and stayed African

Before: a Bantu coast that already traded

Around 800 CE, the eastern African shore from the Lamu archipelago to the Rufiji delta was lined with villages of people who fished the reef, farmed sorghum and millet behind the dunes, kept cattle and goats, worked iron, and made a distinctive comb-incised pottery that archaeologists call Tana ware or Triangular Incised Ware. They spoke closely related Bantu languages of the Sabaki group, whose speakers had reached the coast in the later first millennium. Their settlements, as Esther Brielle and her co-authors summarised the archaeological record in Nature in 2023, "arose out of fishing and agropastoral settlements on the eastern African coast during the late first millennium AD," and from the seventh century formed "a shared material culture and practice network across the eastern African region" ¹. These were not yet towns, not yet Muslim, and not yet called Swahili. They were the people into whose language an ocean's vocabulary would later be poured.

They were also not isolated. The coast had been on the map of Indian Ocean commerce for at least seven centuries. A Greek-speaking merchant's handbook of the mid-first century CE, the Periplus of the Erythraean Sea, describes "the very last market-town of the continent of Azania, which is called Rhapta," named for its "sewed boats," where "there is ivory in great quantity, and tortoise-shell" ². Rhapta was governed, the author says, by a chief from southern Arabia, "using Arab captains and agents, who are familiar with the natives and intermarry with them, and who know the whole coast and understand the language" ². Rhapta has never been located with certainty. But the sentence shows that by the first century foreign traders on this coast were already learning its language, and marrying into it, rather than the reverse.

Azania in the first millennium

What the Periplus lists as imports and exports is a fair outline of the coast's long-term position in the ocean's economy ²:

• Imports: iron lances, hatchets, daggers and awls made at Muza in Yemen "especially for this trade"; glass; "a little wine, and wheat, not for trade, but to serve for getting the good-will" of local chiefs.
• Exports: "a great quantity of ivory," rhinoceros horn, tortoise-shell and "a little palm-oil."
• Political form: "under separate chiefs for each place," with no paramount ruler on the coast itself.

That structure — raw materials out, manufactured goods in, many small polities rather than one kingdom — persisted for a thousand years. What changed between the eighth and twelfth centuries was not the existence of trade but its scale, its seasonality and, above all, the religion, writing and urban form of the people who conducted it.

The coastal communities' relationship with the sea also deepened slowly. A team led by Jeffrey Fleisher, Paul Lane, Adria LaViolette and Mark Horton argued in 2015 that the assumption that the Swahili were maritime "from their beginnings" is wrong: their review of fishing, boat-building and settlement evidence found that the coast's maritime orientation "increased greatly over time and was only fully realized in the early second millennium C.E." ³. The first-millennium villagers used the reef and the lagoons; the deep-water, long-distance seafaring that the later Swahili towns are famous for grew up alongside the transmission this record follows.

A language from the mainland

The language those villagers spoke is the hinge of the whole story. Derek Nurse and Thomas Spear's The Swahili: Reconstructing the History and Language of an African Society, 800–1500 (1985) used comparative linguistics to show that Kiswahili descends from a Sabaki parent language spoken on the northern Kenyan coast in the first millennium, and that it shares that parentage with the Mijikenda, Pokomo and Comorian languages of the coast and its hinterland ⁴. Its grammar — the system of noun classes, verb prefixes and agreement that defines a Bantu language — is entirely African.

Spear later quantified what that meant in everyday speech. Measured against standard 100-word lists of basic vocabulary, between 72 and 91 per cent of Swahili words are inherited from Sabaki, 4 to 17 per cent are loans from other African languages, and only 2 to 8 per cent come from outside Africa ⁵. The famous Arabic layer is real, but it sits on top of a Bantu foundation, concentrated in religion, law, commerce, abstraction and the vocabulary of the town.

What the coast had, and what it did not

Before the transmission, the coast had several of the ingredients a mercantile civilisation needed, and lacked others:

Within four centuries every item in the right-hand column had arrived — and the left-hand column had absorbed it rather than been replaced by it.

The transmission: monsoon, marriage and the mosque

The mechanism was a wind. From roughly November to March the north-east monsoon carries ships from the Persian Gulf, Arabia and western India down to the eastern African coast; from roughly May to October the south-west monsoon carries them home ¹. A merchant who sailed south on one monsoon could not leave until the next. He spent months in a coastal town, needed a host, a house, credit, a wife or a household, and a place to pray. The calendar of the wind produced the social form of the transmission: not conquest and not migration in bulk, but long seasonal residence, repeated year after year by men who often never settled permanently.

The Sirafi ships

The best tenth-century witness is the historian and geographer al-Masʿūdī, who sailed the coast himself. In The Meadows of Gold, finished in 947, he wrote that the sailors of Oman, "Arabs, of the tribe of el-Azd," sailed "as far as the island of Kanbalu and the Sofalah," and that "the merchants of Siraf are also in the habit of sailing on this sea" ⁶. Sīrāf, on the Iranian shore of the Gulf, was the great entrepôt of the ninth and tenth centuries; al-Masʿūdī made one voyage from Ṣuḥār in Oman "with a crew of Sirafians," and "in 304 A.H." — 916–17 CE — returned "from the island of Kanbalu to 'Oman, in a vessel of Ahmad and 'Abd es-Samad, brothers of 'Abd er-Rahim Ben Ja'fer of Siraf," both of whom later "perished with their whole crew" ⁶. He thought the sea of the Zanj "the most dangerous of all" he had sailed ⁶.

His description of Kanbalu, usually identified with Pemba, is the single most important sentence in the early record of this transmission. The island, he wrote, "is well cultivated, and the inhabitants are Moslims but they speak the Zanjee language" ⁶. By the early tenth century, then, there were Muslims on the coast whose language was African. Al-Masʿūdī also records the title of the paramount ruler of the Zanj as Wafliman ⁶ — recognisably the Bantu word that survives in modern Kiswahili as mfalme, "king." The religion was arriving from the ocean; the words for power were still the coast's own.

Shanga: a timber mosque in the eighth century

Archaeology pushes the beginning back further. In the 1980s Mark Horton excavated Shanga, a fifteen-hectare town site on Pate Island in the Lamu archipelago, and found beneath its stone Friday mosque a sequence of earlier mosques, the first of them built of timber and thatch ⁷. The earliest radiocarbon date from the mosque sequence, a burnt tree stump, calibrates to 703–848 CE ⁷. Horton has since summarised the evidence for the coast as a whole: "Accurate dating for Islam on the East African coast can be obtained through Islamic burials, early mosques, and locally minted coins with Islamic inscriptions from the late eighth century" ⁸.

Shanga's small silver coins, struck with Arabic inscriptions between roughly 800 and 1100, are not known from any other Islamic mint; they were made on the coast, for the coast ⁷. Horton's interpretation of the early timber mosques is that they served a local community, not a colony of foreigners. They were small, flimsy, and rebuilt repeatedly on the same spot, in the same way that East African communities still build temporary mosques — "they did not simply adopt Islamic forms as slavish copies of Middle Eastern forms but often developed their own styles" ⁸. Brielle and her co-authors put the consensus plainly: "Muslims were present from the eighth century AD, probably as a minority" ¹.

Persian fathers, African mothers

For most of the twentieth century the question of who these early Muslims were — immigrants or converts — was argued from pottery, chronicles and prejudice. In 2023 it acquired genetic evidence. Brielle, Fleisher, Stephanie Wynne-Jones, David Reich, Chapurukha Kusimba and more than forty colleagues sequenced ancient DNA from 80 individuals buried at six medieval and early-modern coastal towns — Mtwapa, Manda, Faza, Kilwa, Songo Mnara and Lindi — dated 1250–1800, and from an inland site, Makwasinyi ¹. Their findings reshape the transmission:

1. Most ancestry was African. African proportions averaged about 57 per cent at Mtwapa and Faza, 32 per cent at Manda, 67 per cent at Songo Mnara and 74 per cent at Kilwa ¹.
2. The Asian ancestry was mostly Persian, with an Indian component. Persian-related ancestry made up 84–100 per cent of the Asian share at the three sites where it could be measured; the rest was Indian-related ¹.
3. It was sharply sex-biased. Of 62 individuals with confidently determined mitochondrial lineages, 59 carried African L haplogroups; most men carried Y-chromosome lineages typical of south-west Asia. "The Southwest Asian ancestors of the Mtwapa and Faza, Manda and Kilwa individuals were almost entirely male, whereas the African ancestors were almost entirely female" ¹.
4. It began around 1000 CE. Admixture dates fell at 795–1085 CE for the northern towns and 708–1219 CE for Kilwa and Songo Mnara — "coinciding with the large-scale adoption of Islam" ¹.
5. After about 1500 the Asian source shifted towards Arabian-related ancestry, consistent with the growing role of southern Arabia and later Oman ¹.

The linguistic consequence follows directly. "Kiswahili is a Bantu language," the authors wrote, "and since most ancestry in medieval Swahili people derives from African people, our results suggest that the children of immigrant men of Asian origin adopted the languages of their mothers, a common pattern in matrilocal cultures" ¹. The transmission ran through households. Men came on the monsoon with a religion, a script and a commercial vocabulary; their children grew up speaking their mothers' language, and carried their fathers' words into it.

Richard Mortel. Great Mosque of Kilwa Kisiwani, 11th–18th centuries, Kilwa Kisiwani, Tanzania (UNESCO World Heritage Site), 2016. CC BY 2.0 via Wikimedia Commons. · CC BY 2.0

The study's limits matter, and its authors stated them. The sampled burials were mostly elite graves near mosques; the geographical coverage is skewed towards Kenya; and "such genetic data cannot reveal the processes contributing to these patterns" ¹. The genome cannot tell marriage from concubinage, or a free bride from an enslaved one.

Kilwa and the Shirazi story

The coast's own account of the transmission is the Kilwa Chronicle, the Kitāb al-Sulwa, written down in Arabic in the sixteenth century and preserved also in a Portuguese version by the historian João de Barros ¹¹⁰. It tells how ʿAlī ibn al-Ḥasan, a son of a ruler of Shiraz by an Abyssinian mother, sailed from the Persian Gulf with his household, bought the island of Kilwa from a local king, and founded a dynasty. "Shirazi" became, all along the coast, the name of a prestigious origin.

Neville Chittick, who excavated Kilwa between 1958 and 1965, tested the story against the coins. He showed that the coins of ʿAlī ibn al-Ḥasan were the earliest Kilwa issues, and argued that the "Shirazi" settlement "consisted not of a migration of people from the Persian Gulf direct to Kilwa and other places, but rather a movement of settlers from the Banadir coast" — the Somali coast around Mogadishu — about two centuries later than the chronicle implied ¹⁰. The latest radiocarbon programme at Kilwa, published in 2022 by Horton, Jesper Olsen, Fleisher and Wynne-Jones, places "the first coral buildings and coins" there "at the end of the tenth century" ⁹. The Nature team noted that coin evidence dates ʿAlī ibn al-Ḥasan to the mid-eleventh century, and that their genetic results are "consistent with the narrative of the Kilwa Chronicle" in that the early Asian ancestry was Persian ¹.

The dated evidence for the transmission, assembled from these sources, runs as follows:

The chronicle is best read as myth-history: a true memory of Persian-connected men arriving and marrying in, compressed into a single founding voyage and an aristocratic lineage. Its usefulness to Kilwa's rulers is obvious. Its usefulness to colonial historians, as the next section shows, was greater still.

What changed and what was replaced

By the time Ibn Baṭṭūṭa sailed south from Mogadishu in 1331, the coast had a name. He "embarked at Maqdashaw for the Sawahil country, with the object of visiting the town of Kulwa in the land of the Zanj" ¹². Sawāḥil is the Arabic plural of sāḥil, "coast" or "shore"; Swahili is the people of the coasts. At Mombasa he found the inhabitants "pious, honourable, and upright" with "well-built wooden mosques"; Kilwa was "a very fine and substantially built town" whose people were mostly "Zanj, jet-black in colour, and with tattoo marks on their faces" ¹². The visitor from Tangier saw no contradiction between African and Muslim, and neither did the people he met. What had changed in the four centuries since the timber mosque at Shanga can be traced in language, writing, building, money and identity.

A Bantu grammar with an ocean's vocabulary

Kiswahili kept its grammar and remade its lexicon. The layers can be dated roughly by the history of contact ¹⁴¹³¹⁴:

Two points in the table cut against the popular picture. The Arabic share is large in dictionaries but small in the core vocabulary, as Spear's word-list figures show ⁵. And much of it is late: Brielle and her co-authors note that the Arabic loans "may be primarily due to incorporations after AD 1500," when the coast's Asian connections shifted from Persia towards Arabia ¹. The language of the medieval stone towns was more Bantu, not less, than the language of the nineteenth-century Zanzibar sultanate.

What the loanwords displaced is harder to see, because displacement in a lexicon is silent. Where Arabic supplied the vocabulary of God, law, the calendar and literacy, it pushed aside whatever Sabaki terms had served those purposes; the older religious vocabulary of the coast is now recoverable mainly from related languages of the hinterland, such as Mijikenda and Pokomo ⁴.

Writing: Arabic for law, Swahili for verse — six centuries late

Islam brought a script. For a very long time the coast did not use it to write its own language. Arabic was the language of the mosque, the court, the contract and the chronicle; the Kitāb al-Sulwa itself is in Arabic ¹⁰. The first datable texts in Kiswahili, written in an adapted Arabic script (ajami), are centuries later. Jan Knappert treated a Swahili rendering of the Hamziyya, a devotional ode by the thirteenth-century Egyptian poet al-Būṣīrī, as the earliest, assigning it to 1652 ¹⁵; the dating is debated, and a surviving manuscript at SOAS gives a later date for the translation. The earliest securely dated Swahili epic is the Utendi wa Tambuka, composed by Bwana Mwengo wa Athman at the Yunga palace of Pate and dated 1141 AH — 1728 CE — which retells the wars of the early Muslims against the Byzantine emperor Heraclius in the eight-syllable lines of the utendi form ¹⁵.

That gap — roughly eight hundred years between the first mosque at Shanga and the first dated Swahili poem — is itself evidence about the transmission. The script arrived as the vehicle of a sacred language, not as a tool for writing the vernacular, and the vernacular took centuries to claim it. When it did, it produced one of Africa's great written literatures: the utendi epic, the shairi lyric, the homiletic poem, all in Arabic letters, all in a Bantu language.

Coral towns and a new urban order

The most visible change was in building. Brielle's team summarised the archaeological consensus: "A major archaeological transition is evident during the eleventh century, with the establishment of new settlements and the elaboration of older ones with coral-built mosques and tombs, a set of changes generally understood as coinciding with the widespread adoption of Islam" ¹. Coral rag set in lime mortar, and finely dressed reef coral cut while still soft, became the material of mosques, tombs, wells and, from the thirteenth and fourteenth centuries, the houses of the merchant elite.

At Kilwa the change can be followed in a single building. The Great Mosque, whose northern prayer hall Chittick dated to the eleventh or twelfth century, was enlarged in the early fourteenth century by Sultan al-Ḥasan ibn Sulaymān — the Abu'l-Muzaffar Hasan whom Ibn Baṭṭūṭa met — with a domed southern extension; the same ruler built the palace and warehouse complex of Husuni Kubwa on the headland east of the town ¹¹. In 1331 he was minting coins, raiding the mainland, and giving away "the fifth part of the booty made on his expeditions to pious and charitable purposes, as is prescribed in the Koran" ¹².

Richard Mortel. Mihrab, Great Mosque of Kilwa Kisiwani, 11th–18th centuries, Kilwa Kisiwani, Tanzania (UNESCO World Heritage Site), 2016. CC BY 2.0 via Wikimedia Commons. · CC BY 2.0

The new urban order had several defining features:

• Coinage. Shanga's silver coins (c. 800–1100) and Kilwa's copper and silver issues from the late tenth century onwards made the coast one of the few parts of sub-Saharan Africa to mint its own money ⁷⁹¹⁰.
• A shared civic religion. Friday mosques anchored each town; Islamic burial, oriented towards Mecca, replaced earlier practice ¹⁸.
• Writing for law and lineage. Arabic contracts, tomb inscriptions and dynastic chronicles gave elites a written past ¹⁰.
• A distinct coastal material culture. The coast's ceramics and burial practices diverged from the hinterland's from the eleventh century, "even as many aspects of material culture remained deeply linked with inland African groups" ¹.
• Long-distance seafaring. The maritime orientation of the coast was "only fully realized in the early second millennium" ³.

Gold from the south

The money and the masonry were paid for by trade, and from the twelfth century the most valuable item in that trade was gold. It did not come from the coast. Al-Masʿūdī already knew the Omani sailors' farthest point as "the Sofalah (low country), of the Demdemah, which is on the extremity of the country of the Zanj" ⁶. Four centuries later a merchant told Ibn Baṭṭūṭa at Kilwa "that the town of Sufala lies a fortnight's journey [south] from Kulwa and that gold dust is brought to Sufala from Yufi in the country of the Limis, which is a month's journey distant from it" ¹². Kilwa sat on the sea route between that gold and the monsoon ships of the north, and its rulers' fortunes in the thirteenth and fourteenth centuries — the domed extension of the Great Mosque, the palace of Husuni Kubwa, the coinage — followed their grip on the Sofala trade ¹¹¹⁷.

The pattern matters for the transmission. The coast's synthesis was not produced by a single exchange of African raw materials for Asian manufactures; it was produced by African towns positioning themselves as intermediaries between an African interior and an Asian ocean. The towns needed the religion, the law and the script of their trading partners in the north to deal with them as equals, and they needed the language of their neighbours in the south and west to deal with the interior at all. Kiswahili — Bantu in structure, oceanic in vocabulary — is the linguistic shape of that double position.

What the synthesis displaced

The transmission was additive in language and largely additive in material culture: imports at most coastal sites "typically comprise less than 5% of total assemblages," and crops, livestock, craft styles and local ceramics continued ¹. What it displaced was chiefly religious and social. Pre-Islamic burial and ritual practice disappeared from the towns. The old pattern of small chiefships "for each place" gave way to sultanates and town councils whose legitimacy was expressed in Islamic and, increasingly, in foreign-genealogical terms. And a new boundary was drawn between the coast and its own hinterland.

That boundary is the record's most important displacement. In the towns, to be fully a person of standing — mwungwana, plural waungwana, the freeborn and "civilised" — came to mean being Muslim, town-dwelling, literate in the religious sense, and able to claim an honourable descent, often a foreign one ¹⁷. The Nature team observed that "Swahili social 'elites' used claims of foreign origin and rejection of cultural connections within Africa to establish their social status and to signal their religious and cultural affinities" ¹. Kinsmen who spoke closely related Sabaki languages a day's walk inland, and who had not converted, became outsiders. The synthesis that made the coast cosmopolitan also made it exclusive.

The colonial misreading

For most of the twentieth century European scholarship read the stone towns as foreign achievements. The Kilwa Chronicle's Shirazi founders, the coins with Arabic legends and the Chinese porcelain set into mosque vaults were taken as proof that the towns were colonies of Arabs and Persians on an African shore. James Kirkman titled his 1954 excavation report The Arab City of Gedi ¹⁶; Chittick's Kilwa monographs presented "an Islamic trading city" whose impulse came from outside ¹¹. The Nature team's verdict on that era is unsparing: "archaeology during colonial times tended to ignore the evidence of deep African roots, emphasizing foreign objects at medieval Swahili sites rather than providing a balanced picture of the archaeological record," and the Shirazi accounts "were central to the narrative constructed by mid-twentieth century colonialist archaeologists, who interpreted second millennium coastal eastern African sites as having been built by Persian and Arab settlers" ¹.

The correction came from three directions:

• Linguistics. Nurse and Spear showed that Kiswahili descends from a Sabaki parent spoken on the coast itself, and that its non-African vocabulary is a later layer on an African core ⁴⁵.
• Archaeology. Horton's excavations at Shanga showed Islam in an African community from the eighth century, in local building forms, with local coins; imports formed a small fraction of what the towns used ¹⁷⁸.
• Ancient DNA. The 2023 genomes showed that even the elite buried beside the mosques were mostly African in ancestry, descended from African mothers who passed on their language: "the elite inhabitants of Mtwapa and other sites developed from admixed populations and were not foreign migrants or colonists" ¹.

The coast's synthesis was African, and it was a synthesis. Both halves of that sentence are now established.

What the cost was

The transmission of Islam, script and vocabulary to the Swahili coast was largely peaceful and largely voluntary at the point of contact: it ran through trade, residence and marriage, not conquest. That is why this record holds the cost below the atlas's higher grades. But the same monsoon ships that carried the mosque and the book also carried people south to north as merchandise, and the society the synthesis produced defined itself in part against those it was willing to enslave.

The Zanj in Iraq, 869–883

The Arabic word Zanj, used for the people of the eastern African coast, became in Iraq a word for a class of enslaved labourers. In the ninth century landowners around Basra put gangs of enslaved Africans to work removing the salt crust from the marshlands of the lower Tigris and Euphrates. In September 869 ʿAlī ibn Muḥammad, a claimant to descent from ʿAlī, raised them in revolt; for fourteen years the rebels held the marshes, built a capital at al-Mukhtāra, and sacked Basra in 871, before the ʿAbbāsid regent al-Muwaffaq crushed them in 883 ¹⁸. Alexandre Popovic's study, first published in Paris in 1976 as La révolte des esclaves en Iraq au IIIe/IXe siècle, remains the only full-length treatment ¹⁸.

How many of the rebels came from the Swahili coast is genuinely disputed. Ghada Talhami argued in 1977 that the rebellion drew heavily on the free poor and Bedouin of the Basra region, and that many of the enslaved were from north-east Africa rather than the coast ¹⁹. The atlas names the debate rather than resolving it. What is not disputed is that by the ninth century "Zanj" was a category of enslaved person in the ʿAbbāsid heartland, and that the ships of Sīrāf and Oman that carried merchants to the coast carried captives back.

A king sold in Oman, 922

The clearest single episode comes from the Book of the Wonders of India by Buzurg ibn Shahriyār, a tenth-century sea captain of Rāmhurmuz. He tells how in 310 AH (922 CE) a ship's crew, driven onto the Sofala coast, was received hospitably by a local king, traded, and then seized the king and his companions as they came aboard to bid them farewell. "When the day came, the king and his companions were put with the other slaves whose number reached about 200 head," and in Oman "the slaves were sold, and the king with them" ²⁰. Years later the same crew was driven onto the same coast and found the king restored to his throne; he had escaped, made the pilgrimage, learned the Qurʾān and Arabic, and returned home a Muslim. He greeted them — "Ah! here are my old friends!" — and let them go ²⁰.

The story is a sailors' tale, and Buzurg's book mixes observation with marvel. But its incidental details are what matter: a cargo of some two hundred captives was unremarkable, enslavement in the story recurs more than once, and conversion to Islam happened, in this telling, in captivity ²⁰.

Raids into the "heathen Zanj"

Inside the coastal world, the line between the converted and the unconverted became a line of enslaveability. Islamic law forbade a Muslim to enslave a fellow Muslim; it permitted the enslavement of non-Muslims taken in war. Al-Masʿūdī, describing Kanbalu, reported that "the Mahomedans have conquered this island and made the inhabitants prisoners," which he dated to "the beginning of the 'Abbaside and end of the Omaiyide dynasties" — around 750 ⁶. Four centuries later Ibn Baṭṭūṭa found Kilwa's people "constantly engaged in military expeditions, for their country is contiguous to the heathen Zanj," and its sultan setting aside the canonical fifth of the booty ¹². In the fuller translation of the passage, the sultan "frequently makes raids into the Zanj country, attacks them and carries off booty," and gifts at his court were made in enslaved people as well as ivory ²⁰.

Who was raided? The text says only "the Zanj" — the unconverted Africans of the mainland, many of whom spoke languages closely related to the raiders' own. The synthesis did not abolish the older kinship between coast and hinterland. It converted part of it into a frontier.

Sex, status and the genome

The ancient DNA adds a further, harder question that it cannot answer. The almost entirely male Asian and almost entirely female African ancestry of the medieval elite is, as Brielle and her co-authors wrote, "evidence for asymmetric social interactions between groups" ¹. In a society that both married and enslaved, and in which Buzurg's ships carried captives of both sexes, some of those African mothers were surely wives with families and standing, and some may not have been. The genetic data cannot distinguish them, and the atlas does not pretend that it can. The asymmetry is recorded here as part of the bill, unquantified.

1505: the end of the medieval coast

The synthesis was ended as an independent order not from within but by a new arrival on the monsoon. On 24 July 1505 a Portuguese fleet under Francisco de Almeida entered the harbour of Kilwa. Hans Mayr, a German who sailed with it, recorded a town of 4,000 people; the Portuguese marched to the palace, and "on the way only those Moors who did not fight were granted their lives," after which "everyone started to plunder the town of all its merchandise and provisions" ²¹. At Mombasa in August the fleet met resistance. Mayr counted "more than 10,000 people," set fire to the town, and afterwards "the streets and houses were full of dead, who were estimated to be about 1,500"; the Portuguese took "many prisoners, and white women among them and children," and "a large quantity of rich silk and gold embroidered clothes … and carpets also" ²¹.

The Portuguese century reshaped the coast's politics but did not unmake its language. Kiswahili took from the occupiers a small vocabulary, including the word gereza, "prison," from igreja, "church" ¹³. When Omani power rose in the seventeenth century, Thomas Vernet has shown, some Swahili cities treated it as an ally against the Portuguese between 1650 and 1698, and turned against it as a new occupier once the coast passed under Omani control after 1698 ²². The nineteenth-century Omani-Zanzibari economy of clove plantations and the inland caravan trade in people and ivory — a far larger slave system than anything documented for the medieval towns — belongs to a later record; the Nature team notes it as one of the population movements that changed the coast's ancestry again ¹.

What the record cannot count

The documented costs of this transmission are a floor, not a total. The sources are silent on:

• The number of people exported from the coast before 1500. No medieval source gives a figure; Buzurg's cargo of 200 is one ship, in one story ²⁰.
• The share of East Africans among the Zanj of Iraq. Popovic and Talhami disagree, and the sources do not settle it ¹⁸¹⁹.
• The status of the African mothers of the medieval elite. The genome shows the asymmetry, not its causes ¹.
• What the coast's pre-Islamic religion was. The towns that adopted Islam left almost no record of what they gave up, and the hinterland languages preserve only fragments ⁴.
• The lives of non-elite townspeople. Nearly all the excavated and sequenced burials are from prominent graves near mosques ¹.

The bill

The coast took in a religion, a script and an ocean's vocabulary, and stayed African. It built one of the world's oldest continuously documented mercantile cultures, and a language now spoken by more than 200 million people, which UNESCO in 2021 made the first African language with its own international day ²³. The bill for that synthesis was paid, on the documented evidence, mainly by people the synthesis did not include: the captives carried north on the same ships, and the unconverted neighbours on the other side of the coast's new line.