2. The Bronze Age: where the paperwork starts

Slavery is older than writing. We know this because the earliest writing already takes it for granted.

The Laws of Hammurabi — 282 legal judgements cut into a two-and-a-quarter metre pillar of black basalt around 1750 BCE, now in the Louvre — are not a founding document of slavery. They are a mature regulatory framework for an institution already so established that it needs consumer-protection provisions.4 Law 278 gives a thirty-day warranty on a purchased slave who turns out to have the benu-disease. Law 231 provides that if a shoddily built house collapses and kills the owner’s slave, the builder “shall pay slave for slave.” Law 282: “If a slave say to his master: ‘You are not my master,’ if they convict him his master shall cut off his ear.”5

Two provisions are worth sitting with, because they establish patterns that recur for the next four millennia. Law 117 — debt bondage — provides that a man who cannot meet a debt and sells his wife, son and daughter into service sees them work three years in the buyer’s house and go free in the fourth.6 That is a time-limited institution with an exit. Compare Law 175: if a slave marries a free man’s daughter and they have children, “the master of the slave shall have no right to enslave the children of the free.” Babylonian law is already, in 1750 BCE, tracking status through the mother in a way that limits the slaveholder. Three and a half thousand years later, Virginia would run that same rule in reverse, and the difference is most of the story of the Atlantic world.

3. Greece and Rome: the ancestors everyone claims

The two civilisations Western democracies name themselves after ran on enslaved labour, and one of them enslaved an entire neighbouring people for three centuries.

Athens

How many enslaved people lived in classical Athens is the single most contested number in ancient history, and the honest answer is a range that spans a factor of nearly six. A. H. M. Jones put it at about 30,000 in the fifth century and 20,000 in the fourth; Walter Scheidel at 30,000–50,000 maximum; Michael Jameson at 50,000 as a minimum; Mogens Herman Hansen at 66,000–93,000; Moses Finley at 80,000–100,000 at peak; A. W. Gomme at 100,000–115,000 in 431 BCE. Ben Akrigg’s recent study rejects anything above 100,000 as unrealistic, and establishes a useful hard floor from the other direction: Athens needed at least 20,000 able-bodied adult male slaves for military purposes alone — hoplite attendants and trireme crews — before you count a single woman or a single non-military man. Nemanja Vujčić’s 2021 assessment lands at “hardly larger than ca. 100,000 and likely below that, though almost certainly not below 50,000” — somewhere between 15 and 30 per cent of the total population, fluctuating by decade.10

Give the range. Never give a point estimate. Anyone who tells you confidently that “a third of Athenians were slaves” is quoting the top of a spread they haven’t looked at.

What is not contested is the character of the thing. Xenophon, arguing in Ways and Means that the Athenian state should get into the slave-leasing business, mentions in passing that Nicias son of Niceratus owned a thousand men working the Laurion silver mines and rented them out at an obol a day each, that Hipponicus leased out six hundred, and Philemonides three hundred.11 Aristotle, in Politics Book I, works out the philosophical justification with a clarity that has embarrassed his admirers ever since: property is a collection of tools, some lifeless like a rudder and some living, and “a slave is a live article of property” — “one who is a human being belonging by nature not to himself but to another is by nature a slave.”12

Note carefully what Aristotle concedes at 1255a: that there is also slavery by law and convention, which is a different thing from his natural slavery, and that this distinction was already argued about in his own lifetime. The claim that nobody in the ancient world questioned slavery is false, and it is false on the testimony of the ancient world’s most famous defender of it.

Sparta: Greeks enslaving Greeks, as state policy

The helots of Laconia and Messenia were not foreign captives. They were Greeks, held collectively by the Spartan state, whose subjugation the classicist Paul Cartledge describes as being “enslaved as a nation.”13 Plutarch, writing on Lycurgus, reports the krypteia: Spartan magistrates would send the ablest young men into the countryside “armed only with their daggers” to kill helots by night and in the fields by day. He adds, on Aristotle’s authority, that the ephors formally declared war on the helots each year “that they might be massacred without a breach of religion” — a legal device to make the killing of a subject population ritually clean. Helots were also made to drink to excess and perform degrading songs in front of Spartan children, as a pedagogy of contempt.14

Plutarch himself doubts Lycurgus instituted this, suspecting it postdates the great helot revolt, and the modern scholarship is properly contested: Nino Luraghi argues the mass-enslavement narrative reflects ideologically selective reading of the sources; Hans van Wees situates the Spartan conquest within a broader wave of archaic Greek imperialism in which many city-states reduced their neighbours to serfdom between roughly 750 and 550 BCE.15 That correction makes the picture worse, not better: it means Sparta was not an aberration but an unusually well-documented instance of something several Greek states did. Messenian helotage ended in 369 BCE when Epaminondas of Thebes, having broken Sparta at Leuktra, marched into Messenia and founded the city of Messene — abolition by foreign invasion, three hundred years late.16

Rome

The estimates for Roman Italy diverge as widely as the Athenian ones and for the same reason. Karl Julius Beloch said two million out of six; P. A. Brunt said three million; Keith Hopkins said 800,000; Walter Scheidel’s bottom-up model gives 1.0–1.5 million and no more. Noel Lenski’s 2023 synthesis offers the currently safe formulation: 15 to 25 per cent of Italy’s population, roughly 1 to 1.5 million out of 6 to 7 million, in the High Empire. Census data from Roman Egypt independently suggests about 11 per cent there, which may be more representative of the provinces.17

Where did they come from? Lenski ranks the sources: first and largest, birth to enslaved mothers; then military capture — Caesar reportedly enslaved something like a million Gauls, Titus took 97,000 Jews after Jerusalem fell in 70 CE; then market purchase; then exposed infants reclaimed from abandonment; then child sale and debt bondage; then trafficking among frontier peoples, Irish, Scots, Frisians, Goths and Arabs among them.18 W. V. Harris argues child exposure was significantly underestimated by earlier scholars, and that self-sale was more common than assumed.19 Nobody in this literature thinks Roman slavery was racial. It could not have been: it drew on everyone within reach, including Romans.

Rome’s genuinely distinctive feature was the exit. Manumission was common, ritualised and legally structured — manumissio vindicta before a magistrate, manumissio censu by census enrolment, manumissio testamento by will — and a freed libertus became a Roman citizen with a vote. Crucially, the residual legal disabilities “were placed only on the former slaves themselves and did not apply to their sons.” In 2 BCE the lex Fufia Caninia had to be passed to cap the number of people a testator could free in his will, which tells you how routine it had become.21 Lenski estimates that perhaps more than 30 per cent of urban slaves over twenty-five could expect freedom.22 Christer Bruun, who works on the inscriptions, adds the necessary caution: these are models, not counts, and “on the very important question of the overall number of slaves in the Roman world, epigraphy cannot make a large contribution.”23

And the enslaved fought. The Third Servile War of 73–71 BCE began, according to Appian, when Spartacus “persuaded about seventy of his fellows to risk their lives for freedom” at a gladiatorial school in Capua. Within two years his army numbered in the tens of thousands. Crassus restored discipline in his own legions by decimation — drawing lots and executing every tenth man — and when it was over, six thousand survivors were crucified along the road from Rome to Capua.24 There had been two earlier servile wars in Sicily, in 135–132 and 104–100 BCE. Whatever people in the ancient world believed about the naturalness of slavery, the enslaved were not among them.

4. Asia: the largest systems nobody mentions

If you want a single case that breaks the “slavery is what one race did to another” frame, it is Korea — and it is not obscure. It is the best-documented mass slave society in East Asian history.

Korea: the nobi

Sun Joo Kim of Harvard opens her survey with the flat statement that “Chosŏn Korea (1392–1910) was one of the most enduring slave societies in world history.” Between the fifteenth and eighteenth centuries, enslaved people — nobi — comprised over 30 per cent of Korea’s population.25 They were Korean. Their owners were Korean. The system ran for five hundred years.

The economic historians Rhee Young-hoon and Yang Donghyu compared it directly to the antebellum United States and found that each institution held “about one-third share in population,” and that both were legally chattel — subject to sale and to inheritance.26 Their own conclusion, which I include because leaving it out would be dishonest, is that it is nonetheless “inappropriate to call the Chosŏn nobi slaves in general sense”: some nobi held property rights and civil protections, particularly on larger holdings, and the Korean system lacked the racial permanence of the American one. There is a genuine and unresolved “slaves or serfs?” literature here. Both halves of that finding matter. The share of population is not in dispute.

Regional variation was extreme. Figures reported in the secondary literature — which I flag as not verified at source — put the southern half of the country near 40 per cent in the fifteenth to seventeenth centuries, while in the north the status was not even hereditary, and record one surveyed district of Seoul in 1663 in which 1,729 of 2,374 residents, 73 per cent, were nobi.27 Abolition came in stages: government-held nobi were emancipated in 1801, roughly 66,000 people, and the registers were burned; hereditary slavery ended in 1886; and the Gabo Reform of 1894 abolished the status system entirely.28 That last date is worth holding onto. Korea abolished slavery twenty-nine years after the United States.

China

The Chinese legal category was nubi (奴婢), and its distinctive feature was status pollution: enslavement placed a person permanently in the jianmin (賤民), the “mean” or debased people, severed from liangmin, honourable society — barred from the imperial examinations, deprived of normal kinship privileges, treated as household members but permanently inferior.29 Claude Chevaleyre’s work argues against the widespread assumption that slavery did not really exist in late imperial China, finding “unprecedented numbers” of nubi by the late Ming, driven by commercialisation, with entry routes running from sale contracts through penal enslavement to outright trafficking and debt seizure. Nubi revolts erupted in more than thirty districts during the crisis of the 1630s to 1660s.30

There is no reliable aggregate count of nubi for any Chinese dynasty, and I will not invent one. What is documentable is the legal chronology. In 1723 the Yongzheng emperor began emancipating debased status groups — the yuehu musician households of Shanxi and Shaanxi first, then others by province — and by 1729 had to rename the state Bureau of Instruction because recruitment had collapsed once the category was abolished.31 The nubi category itself survived until 1909–10, when the Qing Commission of the Constitutional Government drafted a “Prohibition of Buying and Selling Human Beings” and incorporated it into the criminal code promulgated on 15 May 1910. Chevaleyre notes it was done under Western pressure, as part of the New Policies era, “without widespread public support or prior debates.”32 The legal chain then runs on for decades: an explicit offence of enslavement added in 1918, Chinese signature of the International Slavery Convention in 1926, a nationwide prohibition on keeping domestic slaves in 1932, further ordinances into the later 1930s.33 Formal abolition and actual extinction were about forty years apart.

India and South Asia

Under Firuz Shah Tughlaq of the Delhi Sultanate (r. 1351–88), the court chronicler records 180,000 slaves in Delhi; under Ala al-Din Khalji (r. 1296–1316), 50,000.34 Court chroniclers inflate, and these should be read as impressionistic rather than counted. More useful is the scholar Fouzia Farooq Ahmed’s conclusion applying Finley’s criteria: the Delhi Sultanate was “a society with slaves rather than a slave society” — huge slave numbers, but no large-scale slave peasantry, with agriculture running on free peasant labour and enslaved people filling urban roles as soldiers, artisans and builders.

The Mughal period saw domestic slavery and concubinage as the commonest forms, agrarian slavery significant in the south, and eunuch slavery in elite households. Akbar issued orders against enslaving the relatives of enemy combatants and banned slave sales; Jahāngīr banned the castration of eunuchs.35

The British case is where it gets instructive. The Slavery Abolition Act 1833 excluded British India. India got its own instrument ten years later: the Indian Slavery Act, Act V of 1843. Read what it actually does. It provides that no public officer shall sell a person on the ground of slave status; that “no rights arising out of an alleged property in the person and services of another as a slave shall be enforced by any Civil or Criminal Court”; and that any act which would be a penal offence against a free man is equally an offence against a person held as a slave.36 Notice what is missing: it does not criminalise slaveholding. It withdraws state enforcement and leaves the institution standing. Emma Kalb characterises the colonial approach as “abolition by denial”: the Act proved ineffective, enslaved people were simply re-designated “servants,” and agrarian slaves became “agrestic labourers” with conditions unchanged.37

India’s substantive instrument arrived in 1976: the Bonded Labour System (Abolition) Act, No. 19 of 1976, whose long title is worth quoting because of what it does not say out loud — “An Act to provide for the abolition of bonded labour system with a view to preventing the economic and physical exploitation of the weaker sections of the people.”38 “Weaker sections of the people” is the statutory euphemism for caste-linked debt bondage. That system is still running; see section 13.

Southeast Asia and Japan

The Sulu sultanate’s market at Jolo sold between 200,000 and 300,000 people between 1770 and 1870. Batavia housed roughly 27,000 people by 1673, about half of them enslaved.39 In Siam, society distinguished phrai — freemen owing corvée — from that, the enslaved, and the majority of the enslaved appear to have been Siamese themselves. Abolition under Chulalongkorn was gradual by design: a royal act in 1874 freeing those born into slavery since 1868 at age twenty-one; clarifying laws in 1884, 1890, 1897 and 1900; the Slave Abolition Act of 1905 introducing declining caps on redemption prices; explicit criminalisation in the Penal Code of 1908; and full legal cessation in 1915.40 The nationalist framing of Chulalongkorn as a Thai Lincoln is a later construction, and Thai historians have said so.

Japan’s ritsuryō codes ranked nuhi at the bottom of the “base people of five colours” — the finer details usually quoted about the system, including the ages at which government-held kunuhi were freed, rest on sources I could not verify and are left out here. And in 1587, in the eleven-article memorandum preceding his Bateren Edict, Toyotomi Hideyoshi banned the sale of Japanese people abroad: “It is unreasonable to sell Japanese to China, Nanban, and the Korean Peninsula. Therefore, in Japan, the buying and selling of people is prohibited.”41 He was banning something specific. Portuguese merchants had been buying Japanese people; a contemporary document records roughly a thousand loaded onto a single vessel bound for Macao in 1588.42 The first European slave trade in East Asia ran in the opposite direction from the one everybody knows about, and a Japanese warlord shut it down two hundred and twenty years before the British Parliament got round to the Atlantic.

5. Europe enslaving Europe

The English word for the institution is probably an ethnic slur, and the people it names were the commodity.

The standard lexicographic position is that slave descends from Slav: Middle English sclave, first attested around 1290, from Old French esclave, from Medieval Latin sclavus, from Byzantine Greek sklabos — a word that meant both “Slav” and “slave.” The American Heritage Dictionary puts it bluntly: “they are, in fact, historically identical.”43 The ethnonym itself is thought to derive from a Proto-Slavic root meaning renown or fame — the same element in names like Boh-slav — so the semantic collapse from “the famous people” to “property” happened entirely inside the medieval slave trade.

Being honest about this requires flagging a real dissent. Georg Korth argued in 1970 that the Greek ethnonym merely coincided with a pre-existing Greek word for unfree persons, producing a folk etymology rather than a genuine descent; Elmar Seebold adopted that view in Kluge’s Etymologisches Wörterbuch, the standard German etymological dictionary, and the Oxford etymologist Anatoly Liberman endorses it and advises abandoning the Slav–slave equation.44 The mainstream position remains the first one, but it is not unanimous, and it is worth saying so rather than deploying a contested etymology as a rhetorical trump card.

The trade underneath the word is not in dispute. Marek Jankowiak of Oxford estimates the tenth-century Slavic slave trade into the Islamic world ran “in the thousands of slaves per year,” inferred from roughly 400,000 Islamic silver dirhams recovered in Scandinavia and Slavic lands — itself, he notes, a small fraction of the total inflow. The route ran from Iraq via Khazaria in the ninth century, shifting after 900 to Samanid Central Asia by way of Bulgar.45 Eastern Europeans were being sold south and east in industrial quantity, for silver, six hundred years before Columbus.

The Vikings

Norse raiding was, substantially, a business in people. Poul Holm’s foundational study of Dublin found slave-taking integral to Viking warfare from the ninth century, peaking in the ninth and tenth, and — the detail that matters — that by the eleventh century Irish kings themselves were selling Irish war prisoners through Dublin.46 The Annals of Ulster record the plundering of Howth in 821 and the raiders “carried off a great number of women into captivity”; in the 940s an Irish bishop held on Dalkey Island died attempting escape; in 980 Máel Sechnaill stormed Dublin to release Irish captives; in 1098 Dublin ships in the Anglesey campaign were paid in “captives… of young men and maidens.”47

Proportions are genuinely uncertain and I will not pretend otherwise. The highest published figure — up to 25 per cent of Viking-Age Norway’s population held as thralls — comes from Jón Viðar Sigurðsson of Oslo.48 Stefan Brink’s Thraldom (Oxford, 2021), the standard modern treatment, deliberately declines to give a proportion at all, on the grounds that archaeological remains are scarce and ambiguous, that the written sources postdate the period, and that it is frequently impossible to distinguish unfree from free tenant farmers.49 Brink even prefers “thraldom” to “slavery” to avoid importing a Roman or modern frame. The right formulation is: estimates range widely, the highest is around a quarter for Norway, and the leading specialist declines to quantify.

England: 10 per cent, and then gone

The Domesday Book of 1086 records servi as the fourth-largest group among the English peasantry — over 10 per cent of the recorded population, and, as the Hull Domesday Project notes, significantly higher if you allow for their near-total omission from the northern circuit.50 Two caveats belong in the same breath: Domesday counts heads of household, not individuals, so this is 10 per cent of households; and the translation is itself disputed, with servus rendered “slave” in some editions and “serf” in others. John S. Moore is emphatic on the point, following Frank Stenton: “The servi and ancillae of Domesday Book are undoubtedly male and female slaves,” and mistranslating them as serfs collapses two opposite processes — slavery declining, serfdom rising — into one.51

English slavery then vanished within about two generations. Moore dates the completion of the decline to the reign of Henry I, 1100–1135. It was helped along by clerics: Wulfstan II, Bishop of Worcester, is credited with ending the Bristol slave trade to Ireland by repeated preaching — William of Malmesbury says he “gradually destroyed this long-standing custom, handed down from remote ancestors to their descendants” — and in 1102 Anselm obtained a resolution from the Council of London against the trade, aimed chiefly at the sale of English people into Ireland.52 That resolution also broke Dublin’s exchange network. In 1171 the Council of Armagh decreed that English slaves in Ireland be freed.

It did not vanish cleanly. Worcester Cathedral’s records carry more than 140 mentions of nativi, servi and ancillae, over 120 manumissions in the Liber Albus between 1301 and 1450, and a document of 1370 that reads as the outright sale of a man named Nicholas Salewy.53 And what replaced slavery was villeinage — unfreedom of a different legal shape, which the same peasants experienced as continuity.

Russia

Russia ran both institutions, in sequence. Hellie’s account is that medieval Russia had no serfdom, because land was abundant — but slavery, kholopstvo, was ancient, and about 10 per cent of the population were slaves before enserfment took hold.54 A large share of Muscovite slaves were self-sold: people selling themselves into slavery to escape destitution, which is a fact worth sitting with when you consider what the alternative must have been. Peter I effectively abolished chattel slavery by accident, in the 1720s: his soul tax, introduced 1719 and first collected 1724, made lords liable for their people’s taxes and thereby merged the legal categories, absorbing kholopy into the serf population.

By the Tenth Revision of 1857–58, roughly 22 million serfs lived in European Russia out of a population above 60 million — about 36 per cent.55 Alexander II’s Emancipation Manifesto of 19 February 1861 (Old Style) freed them, and immediately mortgaged them: redemption loans totalling over 860 million nominal rubles, structured as 49-year mortgages at 6 per cent, at roughly 95 rubles per former serf when mean incomes were probably under 50 rubles a year.56 Redemption payments were not cancelled until 1907. This is the first appearance in this article of a pattern you will see repeatedly: emancipation as a financial event, in which the freed party is billed.

The Mediterranean

Caffa in the Crimea and Tana on the Don were the hinge of the late medieval slave trade, and Venice and then Genoa fought over the privileges. Hannah Barker’s work supplies the one solid proportion in this area: enslaved people were 1–2 per cent of Genoa’s population in the thirteenth century, rising to 4–5 per cent in the fifteenth.57 Barker deliberately declines to estimate for Venice or the Mamluk sultanate because the sources will not bear it, and that restraint is worth imitating. The Genoese levied a head tax on slave sales at Caffa, generating annual revenues ranging from 13,666 aspers in 1465 to 219,332 in 1446 — fiscal evidence for a volume nobody can count.58 The Ottomans took Caffa in 1475 and excluded the Italians from the trade.

And then there is Malta. The Knights of St John compelled enslaved, convicted and indebted men to row their galleys for nearly three centuries, from 1530 to 1798 — described in the archaeological literature as “a rare archaeological example of modern slavery within a European context.”59 The English traveller Philip Skippon recorded roughly 2,000 slaves belonging to the Order in 1663, noting that because escape from an island was near-impossible they wore only “an iron ring or foot-lock.” Around 2,000 were still held in 1798 when Napoleon took the island and ended it. A Catholic military order, headquartered in Europe, held galley slaves until the year Wordsworth published Lyrical Ballads.

6. The Sahara, the Indian Ocean and the Barbary coast

Three trades, running for twelve centuries, moving people in numbers comparable to the Atlantic — and with none of the Atlantic’s paperwork, which is why every figure in this section is a reconstruction.

Trans-Saharan

The two standard series are Ralph Austen’s and Paul Lovejoy’s. Austen gives about 5.15 million for 1300–1880; Lovejoy gives about 7.42 million for 650–1900.63 Most of the gap is not disagreement — it is start date. Austen begins in 1300; Lovejoy begins in 650, and those pre-1400 rows supply roughly half his total while resting on almost no documentary base. Raymond Mauny put it as high as 10.6 million; Rönnbäck and Nunn, taking all three non-Atlantic trades together, land nearer 5 to 6 million — though Rönnbäck’s figure covers the sixteenth to nineteenth centuries only, so it is not directly comparable to Lovejoy’s 650–1900 span. Austen himself revised his figures for arrivals in the Maghrib specifically downward — 683,000 for 1700–1800 and 580,990 for 1800–1880 — to account for desert-crossing mortality and for captives retained elsewhere along the routes. Exports are not arrivals, and the two revisions are not on a like-for-like basis.

One qualitative contrast is well established and important: the Atlantic trade ran roughly two-thirds male, with about a quarter of all captives recorded as children, while the Saharan, Red Sea and Indian Ocean trades ran the other way, with demand concentrated on women and girls for domestic labour and concubinage.64 The hard ratios are thinner than the generalisation implies — nineteenth-century trading-post registers from southern Morocco show 15 women to 3 men across nineteen transactions at Iligh, but near parity at Bairouk — so treat “two women for every man” as a directional consensus rather than a measurement.65 Castration of male captives to produce eunuchs was practised, with very high mortality, which is a large part of why these trades left a smaller demographic footprint in receiving societies than volume alone would predict.

The Indian Ocean and Zanzibar

Red Sea and East African exports between 800 and 1896 are estimated at roughly 3.26 to 4.04 million.66 The chronology is the interesting part: perhaps two-thirds of everyone trafficked across the Indian Ocean moved in the nineteenth century.67 This trade peaked as the Atlantic trade was being suppressed. Whatever ended the Atlantic system, it did not end the global one.

Zanzibar is the case study. Sayyid Said moved his court there in the 1830s–40s and built a clove economy on enslaved labour; by the 1850s roughly two-thirds of the island’s population is estimated to have been enslaved.68 Suppression came in stages under British pressure: the Moresby Treaty of 1822, the Hamerton Treaty of 1845, and then the Frere Treaty of 1873, under which Sultan Barghash prohibited the seaborne transport of slaves and closed the Zanzibar market. Legal abolition of the status came in 1897 under Sultan Hamoud — and it freed nobody automatically. Freedom had to be claimed through a court, and 12,000 people applied between 1897 and 1909. Concubinage was not brought within the abolition until 1909, a deliberate twelve-year carve-out.69

How badly was this counted? Estimates of the number of people legally held in Zanzibar at the moment of the 1897 decree “varied between 7,000 and 70,000.”70 That is a tenfold spread, inside a single peer-reviewed article, about the population being emancipated, at the moment of emancipation. Keep it in mind the next time someone hands you a confident historical number.

One figure worth naming for the complication it causes: Tippu Tip — Hamed bin Muhammed el-Murjebi, born 1837 — the dominant Zanzibari-Swahili ivory and slave trader of the eastern Congo, who by the late 1860s was leading caravans of four thousand men. In February 1887 he was appointed governor of Stanley Falls district for the Congo Free State.71 Leopold II’s regime, which justified itself to Europe as an anti-slavery mission, put the region’s most successful slave trader on its payroll as an administrator.

The Ottomans: the child levy and the Crimean raids

The devshirme conscripted Christian boys from the Ottoman Balkans and Anatolia into the sultan’s service — the levy rate given as one boy per forty households, taken every four to seven years, aged between seven and twenty, first documented in writing in 1438.72 The Janissary corps stood at around 30,000 under Suleiman and 50,000 by the 1650s. The system decayed from the late sixteenth century — a 1666 recruitment target for the central and western Balkans of only 300 to 320; in 1687, just 130 inductees graduated — and was formally ended somewhere between 1638 and the early reign of Ahmet III.

Whether devshirme is properly called slavery is genuinely disputed and I am not going to resolve it here. Recruits were legally kul, servants of the sultan, and could rise to grand vizier; Ottomanists often resist the chattel framing, while Balkan national historiographies treat it as a defining atrocity — the “blood tax,” Greek paidomazoma. Both readings are describing the same practice: children taken from their parents by state compulsion, converted, and never returned. No reliable cumulative total exists; figures of 200,000 to 500,000 circulate without support, and I am not publishing one.

Further north, the Crimean Khanate and the Nogai raided the East Slavic lands for captives on a scale that depopulated regions. Bohdan Baranowski estimated up to a million people taken from the Polish–Lithuanian Commonwealth between 1474 and 1694; Mikhail Khodarkovsky puts 150,000–200,000 from Muscovy in the first half of the seventeenth century alone; Orest Subtelny counts 86 raids between 1450 and 1586 and 70 more between 1600 and 1647, averaging around 3,000 captives each and reaching 30,000 in a single raid.73 In the 1570s close to 20,000 people a year went on sale at Caffa, which occasionally held as many as 30,000 at once. These estimates overlap in geography and period and must not be added together. The scale is nonetheless not seriously disputed: eastern Europe was a captive-export region for two and a half centuries.

The Ottoman Empire never issued a general abolition of slavery as an institution. It issued successive, geographically limited, largely unenforced prohibitions on the trade — a firman in 1830 freeing certain Christian captives, the closure of the Istanbul market in 1847, prohibition of the Circassian and Georgian trade in 1854–55, and the 1857 firman of Abdülmecid I banning the import of African slaves, issued under British pressure, with the Hejaz explicitly exempted because the earlier prohibition had provoked rebellion there. The 1857 firman was never enforced in practice.74 Slavery persisted in Ottoman lands to the end of the empire in 1922.

The Barbary corsairs — handled properly

This is the single most weaponised topic in this entire article, so let me be careful in both directions.

The raiding was real, large and long-running. Corsairs from Tunis, Algiers, Salé and Tripoli took captives from Italy, Spain, France, the British Isles and further. On 20 June 1631, a raiding party under Murad Reis the Younger — Jan Janszoon of Haarlem, a Dutch convert — sacked the village of Baltimore in West Cork and carried off 107 people: 20 men, 33 women and 54 children and youths. Two women, Joan Broadbroke and Ellen Hawkins, are confirmed ransomed after fifteen years. Most of the rest died, converted, or vanish from the record. The village was substantially abandoned.75 In the summer of 1627, corsairs raided Iceland twice — the Tyrkjaránið, the Turkish Abductions — striking the southwest in late June and the southeast and the Westman Islands about a fortnight later, and taking over 400 Icelanders. About ten per cent were successfully ransomed, through protracted Danish negotiations hampered by Denmark having no relations with the North African ports.76

Now the number everyone quotes. Robert C. Davis, in Christian Slaves, Muslim Masters (2003), estimated 1 million to 1.25 million European captives taken to North Africa between 1530 and 1780.77 His method, which he describes candidly, was to work backwards from population stability: identify recorded slave populations at particular places and times, then calculate the replacement rate needed to hold those numbers steady against death, escape, ransom and conversion. He concluded about a quarter of the enslaved had to be replaced each year, implying roughly 8,500 new captures annually at peak. His own caveat: “It is not the best way to make population estimates, but it is the only way with the limited records available.”

The criticism is substantial and belongs in the same paragraph as the figure. Earle — the historian of the Malta and Barbary corsairs — told the Guardian in 2004 that “his figures sound a bit dodgy and I think he may be exaggerating,” the substantive objection being that Davis extrapolated from 1580–1680, the most intense slaving period, across a much longer span.78 Ehud Toledano, reviewing in the Journal of African History, faulted Davis for relying exclusively on European documents without consulting Ottoman sources — a different objection from Earle’s arithmetic one, and both should be cited. John Wright cautions that all modern estimates in this field are back-calculations from human observation. On the other side, the economic historian Ian Blanchard defended Davis, saying he did not find “that absolute figure of one-million at all surprising.” The criticism is not unanimous either.

A useful independent benchmark, though one I flag as reported rather than verified at source: contemporary observers are said to have put roughly 35,000 European slaves across the Barbary coast at any given time in the seventeenth century, with more than 30,000 in Algiers alone by 1650.79 A standing population of 35,000 is compatible with Davis’s cumulative total only under his high 25 per cent annual replacement assumption — which is precisely what Earle disputes. That is where the argument actually turns, and laying it out lets you judge it yourself rather than picking a team.

7. Africa: the hardest section to write honestly

Two claims are both true and are constantly used to cancel each other out. They don’t.

Claim one: slavery existed across West and Central Africa before Europeans arrived, some African states organised warfare around captive-taking and grew rich on it, and European traders overwhelmingly bought rather than raided. Claim two: the demand, the capital, the shipping, the insurance, the plantations and the profits were European, and the trade at Atlantic scale did not and could not exist before that demand did. Both are supported. Neither cancels the other. What follows is an attempt to hold both at once.

Internal African slavery: the Sokoto case

The best-verified figure in African slavery scholarship comes from Paul Lovejoy on the Sokoto Caliphate: two to three million enslaved people by the 1860s, constituting “at least a quarter of the total population and perhaps as much as half,” and still numbering several million at the time of colonial conquest in 1897–1903.80 Lovejoy’s own comparative phrasing is precise and I am going to quote it exactly, because the internet routinely inflates it: the Caliphate’s enslaved population was “on the same order as the United States and Brazil and much larger than Cuba.” He does not say larger than the United States. Anyone who tells you Sokoto held more slaves than the American South is overstating a real scholar’s real finding.

Elsewhere the numbers thin out fast. Boniface Obichere’s study of Asante and Dahomey states plainly that “hard statistical data is difficult to come by if not non-existent,” and the single quantitative datum he offers is roughly 25,000 war captives labouring in the Lake Bosomtwi region of Asante by 1820.81 What he can establish is character: the majority of enslaved people in both states worked in domestic service; others served militarily, on royal plantations — notably Dahomean palm oil — and in skilled crafts. Manumission and absorption into free lineages were common. In Dahomey, children born to free men and enslaved women were automatically free. Asante maintained a specific category, akyere, designated for ritual sacrifice at chiefly funerals; Dahomey practised human sacrifice but kept no equivalent category. Obichere reads this slavery as primarily social and political rather than economic — a position that sits in real tension with Lovejoy’s evidence for productive plantation slavery and with Robin Law’s work on Dahomean palm-oil estates. Present both; the field does.

African elites in the Atlantic trade

John Thornton’s Africa and Africans in the Making of the Atlantic World is the standard statement of African agency: Europeans generally purchased captives taken in warfare between African states rather than raiding for them; peaceful trade became the norm on African coasts by around 1494; African states set the terms of coastal commerce and Europeans held factories, not territory, until the nineteenth century.82

The institutional detail is well documented. The Aro Confederacy, a merchant network without a centralised state, drove the expansion of the Bight of Biafra trade: 182,066 people embarked from Biafran ports between 1750 and 1800, making the Bight of Biafra the third most important supply region after West Central Africa and the Bight of Benin, and by the mid-nineteenth century the Aro had over 150 trading settlements inland. (A percentage-of-all-departures figure is often quoted alongside that count; the two do not reconcile against any single denominator, so it is left out.)83 At Old Calabar, the Efik trader Antera Duke kept a diary in trade English between 1785 and 1788 documenting an African–British slave-trading network in which he was a principal, not an intermediary — recording commercial disputes with British captains over fee structures and the practice of “pawning,” human beings pledged as collateral against trade credit.84 That is sophisticated commercial agency and pretending otherwise is condescension.

Now the other half, which is equally documented.

The phrase “Africans sold their own people” imports a category that did not exist. Pre-colonial African political identity was ethnic, lineage-based and political, not continental. A Fon captive-taker selling an Igbo captive was not, by any category available to him, selling “his own people” — any more than a Genoese selling a Circassian at Caffa was. The racial category that makes the sentence feel damning was substantially forged by the Atlantic system and its diaspora, not before it.

The gun-slave dynamic was structurally coercive. Warren Whatley’s econometric work finds gunpowder imports and slave exports co-integrated in a long-run equilibrium, with a log correlation around 0.80 — guns and captives moved together, tightly, over the whole period. British gun exports to West Africa rose roughly tenfold, from about 20,000 a year in the 1680s to around 200,000 a year by the 1780s. Whatley’s conclusion is the one that matters: states that abstained “found themselves on the losing side of an arms race.”85 “Nobody forced them” is not a serious description of that equilibrium.

Africans resisted the trade at every level. Sylviane Diouf’s edited volume Fighting the Slave Trade catalogues defensive, protective and offensive West African strategies against it.86 Named rulers protested: Afonso I of Kongo wrote to João III of Portugal in 1526 about the depopulating effect of Portuguese slaving on his kingdom.87 That letter also needs its complication stated: Afonso was himself a participant in and beneficiary of slave exports, and his objection was to unregulated Portuguese slaving and to the enslavement of his own free subjects and nobles, not to slaving as such. Flattening that in either direction is propaganda.

African rulers also resisted abolition, and the documentation is excellent. King Gezo of Dahomey (r. 1818–1858) held out against a decade of British missions, a naval blockade and a treaty. He signed a treaty of friendship and commerce with Lieutenant-Governor Winniett on 3 April 1847 but refused the anti-slave-trade provisions; refused abolition again to Consul Beecroft’s embassy in 1850; and when Palmerston ordered a blockade in September 1851, accepted only a limited treaty on 13 January 1852, rejecting the standard form’s clauses on ending human sacrifice, protecting missionaries, expelling slave traders and British enforcement powers. The blockade was lifted in June 1852 once London accepted the limited version. The last transatlantic slave shipment from Dahomey left in 1863.88 Whether Gezo’s resistance was ideological or a rational judgement that palm oil could not replace slave revenue fast enough is genuinely contested by historians.

The synthesis I would defend: local bargaining power was real and should not be denied. Structural control over the Atlantic system was not African — Anne Bailey’s critique of Thornton is precise on this, that African elites held continental power but not intercontinental power over capital, shipping, insurance and plantations.89 Participation was concentrated in specific coastal states and merchant elites, not general. And Lovejoy’s central thesis — the reason his book is called Transformations in Slavery — is that the external trade did not merely tap an existing African institution but reshaped it into something far more extractive, and that internal African slavery expanded in the nineteenth century as Atlantic exports were suppressed.90

Which is the pattern this whole article keeps running into. Slavery was everywhere. What the Atlantic economy did was industrialise it.

8. The Indigenous Americas, before and after

This section is easy to misuse, so here is the frame up front: every specialist cited below insists these systems were structurally different from Atlantic chattel slavery, and several of them show colonial demand transforming those systems into something far more lethal. The transformation is the story. The comparison is not a defence of anything.

The Northwest Coast

The peoples of the Pacific Northwest — Tlingit, Haida, Tsimshian, Nuu-chah-nulth, Chinook and others — ran slaveholding societies substantial enough that Leland Donald’s standard study argues some communities “possessed enough slaves to merit consideration as large-scale slave systems,” with slavery “central to these hunting-fishing-gathering societies” economically and ritually.91 Two trading networks operated: a northern one linking Tlingit, Haida, Tsimshian and Heiltsuk, and a Columbia River network to the south.

The village-level proportions are striking. Donald Mitchell’s demographic profile puts it at up to a quarter of the coast-wide population at any given time. Among the Makah at Cape Flattery, 20–40 per cent of a village’s population “at many places.” James Douglas’s 1853 census gives roughly 25–30 per cent for the Mowachaht at Nootka Sound. Yvonne Hajda’s work on the Lower Columbia records Fort George village under Concomly at 47 per cent enslaved and Kiesno’s village at 31 per cent. Maquinna held perhaps fifty slaves; Haida chiefs typically ten to twelve; most titleholders two or three.92 David Graeber and David Wengrow note these proportions “rival what could be found on the cotton plantations of the colonial South and are in line with estimates for household slavery in classical Athens.”93

The Comanche

Pekka Hämäläinen’s The Comanche Empire (2008) argues Comanches ran an expansionist power that pulled Europeans and other Native groups into interaction spheres organised around trade in goods and captives.95 I am not going to attach a captive-proportion figure to Hämäläinen, because the number that circulates online is not one I can find in him, and because his population estimates are themselves under sustained attack — Joaquín Rivaya-Martínez rejects them as “overinflated” (Hämäläinen’s roughly 40,000 around 1780 against Rivaya-Martínez’s archivally grounded 12,000–14,000) and in 2023 called the Comanche-empire hypothesis a “historiographic mirage.”96

Use Rivaya-Martínez instead, because his figures are archival and explicit. Analysing more than 800 individuals captured by Comanches between 1820 and 1875, he finds that over 61 per cent of captives were children aged five to twelve; that over 66 per cent were male; that boys aged five to twelve alone were over 46 per cent of abductees; and that by the 1900s, captives and people of captive descent amounted to over 40 per cent of the Comanche population.97

And here is the crucial nuance, which is the whole reason this section needs care. Comanches incorporated captives by adoption — young children taken in by childless families — by marriage, and by co-optation as equals. They enslaved older captives they did not adopt, designating them tʍrʍʔaiwapɨ: forced labour, corporal punishment, no inherent rights. But that enslavement was “neither hereditary nor necessarily permanent.” That single clause is the difference between this institution and the one being built simultaneously in Virginia.

Mesoamerica and the Haudenosaunee

Aztec tlacotin status was not hereditary — children of enslaved people were born free. The enslaved could marry, have free children, and own property. The principal route in was self-sale during famine; spouses could sell each other and parents could sell children, with the possibility of redemption by repaying the price; a change of status had to be formally witnessed by four officials. In the marketplace, those for sale wore wooden or metal collars to prevent escape.98 No reliable estimate of the tlacotin share of the population exists, and tlacotin were a distinct category from war captives destined for sacrifice — conflating the two is a common error.

Haudenosaunee mourning wars operated on a different logic again: captives were taken to fill the social positions of the dead and restore spiritual power lost to the community, with bereaved matrons deciding whether a given captive was adopted into the household or subjected to ritual torture and execution. By the mid-1660s, Jesuit missionaries estimated that two-thirds or more of the people in many Iroquois villages were adoptees; in 1657 Paul Le Jeune believed there were “more Foreigners than natives of the country” in Iroquoia.99 That figure is a contemporary impression, not a count, and should always be attributed as such. The mechanism behind it is the important part: European epidemics had halved Iroquois populations by the 1640s, converting sporadic, targeted mourning wars into constant, desperate campaigns for ever more captives. The colonial encounter did not create the institution. It put it under a pressure it was never designed to bear.

The other slavery

And then colonists enslaved Indigenous people, on a scale most people have never heard a number for. Andrés Reséndez’s The Other Slavery — Bancroft Prize, 2017 — estimates between 2.5 and 5 million Indigenous people enslaved across the Americas, from Columbus’s arrival through the early twentieth century, and argues that enslavement, not epidemics alone, was primarily responsible for Indigenous population collapse, through a synergistic relationship between the conditions of enslavement and disease mortality.100 The range is very wide, and it has to be: much of this enslavement was formally illegal and left no paper trail. Brett Rushforth’s independent estimate is 2 to 4 million, with the striking observation that through roughly 1680–1690 there had been more enslaved Indians than enslaved Africans in the Americas.101

The English colonial case is documented tightly. Alan Gallay’s work on the Carolina trade finds between 30,000 and 51,000 Indians enslaved by the English in the American South between 1670 and 1715 — and, the single most quotable fact in the field, that between 1670 and 1715 more Indians were exported into slavery through Charles Town than Africans were imported.102 Destinations included Virginia, New England and the Caribbean. In some colonial regions American Indians made up as much as 25 per cent of the enslaved population.

The Spanish systems ran in parallel: encomienda, formalised in 1503, a crown grant of a specified number of Indios in exchange for nominal protection and Christian instruction, which “in practice became a form of enslavement”; and repartimiento, operating as early as 1499 and formalised around 1575, called mita in Peru and cuatequil in New Spain, conscripting an initial 5 per cent of an Indigenous community for mining and 10 per cent for seasonal agriculture — and, after the reforms of 1601–1609, requiring 25 per cent of the Indigenous population to work for Spaniards.103 New grants of encomienda were prohibited in 1721; the system was officially abolished in the late eighteenth century.

9. The Atlantic system: what actually made it different

Having spent seven sections establishing that slavery was universal, this is the section where I explain why the Atlantic version is not just one more entry on the list.

Start with what is not the answer. It is not scale alone — the Saharan and Indian Ocean trades together moved comparable numbers over a longer span. It is not cruelty — the Spartan krypteia and Caffa’s galleys will hold their own against anything. It is not that other people didn’t do it — they did, on every continent.

Four things made it structurally distinct, and all four are legal rather than moral facts.

1. Heritability through the mother

In December 1662 the Virginia General Assembly passed a statute that is, more than any other single document, the hinge of the Atlantic world:

“WHEREAS some doubts have arrisen whether children got by any Englishman upon a negro woman should be slave or ffree, Be it therefore enacted and declared by this present grand assembly, that all children borne in this country shalbe held bond or free only according to the condition of the mother…” Hening, Statutes at Large, 2:170, December 1662104

Partus sequitur ventrem — the offspring follows the belly. Jennifer Morgan’s analysis shows this was not a straightforward borrowing from Roman law: English common law and bastardy codes would have assigned paternal responsibility. Virginia’s legislators instead reached for property law — the principle that if a man sows on another’s land the produce belongs to the owner — converting a question of sexual morality into one of commodity ownership.105

The consequences compound. It makes slavery self-perpetuating without importation. It forecloses kinship as a route out. It makes the rape of an enslaved woman by her owner a profitable act. And it makes reproduction an asset class, which is why, after 1808, the American domestic slave trade became economically central.

2. A racial basis

Roman slavery was explicitly non-racial: enslavement came from warfare, piracy, kidnapping and child abandonment, and “people of any race, ethnicity, or place of origin might become slaves, including freeborn Romans.”106 Atlantic slavery mapped bondage onto a hereditary phenotypic category. This is why emancipation did not end the caste: when the legal status is abolished, the marker of the status remains visible on every descendant, and the social machinery built to justify the status keeps running without it. No other major slave system had that property, and it is the single best explanation for why the aftermath in the Americas looks the way it does and the aftermath in, say, Korea does not.

3. The manumission pathways were closed

Rome had three ratified mechanisms plus informal manumission, and the freed libertus got citizenship and a vote, with disabilities that did not pass to his sons. Islamic law built in exits: mukātaba, contracted self-purchase; the umm al-walad, a concubine bearing her master’s acknowledged child, who could not be sold and was freed at his death, her child free; manumission as an act of piety; and military slaves rising through the ranks, with Mamluks ruling Egypt from 1250 to 1517 and janissaries reaching the grand vizierate. Spanish America had coartación, self-purchase in instalments, formalised in the Código Negro of 1789, producing large free populations of colour.107

Against that: Virginia in 1723 forbade manumission “except for some meritorious services to be adjudged and allowed by the governor and council.” South Carolina required a special act of the legislature. Florida prohibited it outright. By the 1850s most Southern states had made manumission all but impossible. A slave system with exits produces a free population of the same descent; a system without them produces a permanent caste. That was a policy choice, made repeatedly, by legislatures.

4. Legal personhood as property, demonstrated

The cleanest single illustration is the Zong. On 29 November 1781 the crew of the slave ship Zong threw more than 130 enslaved Africans into the sea, reportedly to conserve fresh water. The litigation that followed — Gregson v Gilbert — was an insurance dispute. The owners sued the underwriters for the value of the cargo.108 That is what chattel status means, expressed in the only language a court could hear. Granville Sharp pursued the case; Olaudah Equiano is credited with bringing it to his attention; it became a rallying point for abolition and, sixty years later, Turner’s Slave Ship.

Now the numbers — and the surprise

That last figure is the one that stops arguments in both directions, so let me give it precisely and with its ambiguity intact. The editors of the Trans-Atlantic Slave Trade Database state that “the Caribbean and South America received 95 percent of the slaves arriving in the Americas” and that “less than 4 percent disembarked in North America, and only just over 10,000 in Europe.”109 David Eltis’s overview essay on the same site says the United States “absorbed only 5 percent of the slaves arriving in the Americas.”110 The two figures are probably measuring slightly different things — direct trans-Atlantic arrivals in North America versus captives who eventually reached the United States including those transshipped from the Caribbean — but I could not find a source that says so, so I am giving you both, with both wordings, rather than harmonising them myself. Brazil alone took roughly 40 per cent of everyone embarked — about 45 per cent of everyone who arrived alive; Brazil and the British Americas together took more than seven of every ten captives. Watch that denominator: a great many arguments about this trade are conducted with one side using embarkations and the other using arrivals.111

The demographic paradox

Fewer than 400,000 people arrived directly in mainland North America. By 1860 the enslaved population of the American South was 3,950,511.112 A tenfold increase, almost entirely by natural increase — and growth that continued undiminished after the external trade closed in 1808.

No other major New World slave society did this as a whole — though, as below, non-sugar regions within Brazil did. The Caribbean and Brazilian sugar regimes consumed lives faster than they reproduced them and required constant resupply. The pivotal scholarly treatment is Michael Tadman’s “The Demographic Cost of Sugar” (American Historical Review, 2000), which argues the decisive variable was the crop, not national character or slaveholder benevolence: heavy sugar regimes produced sustained net natural decrease, while non-sugar regimes — the US South, and non-sugar zones within Brazil — produced increase.113 The mechanisms generally cited are gang labour and mill-season mortality, heavily male-skewed cargoes, suppressed fertility, the disease environment, and planters’ calculation that buying replacements was cheaper than rearing children while the trade stayed open.

10. Canada’s blind spot

Two facts most Canadians have never encountered: the majority of enslaved people in New France were Indigenous, not Black; and the famous 1793 anti-slavery Act freed nobody who was already enslaved.

Slavery existed in what is now Canada for roughly two centuries, and it was legal. On 13 April 1709, Intendant Jacques Raudot formalised what had been informal practice by ordinance: “Tous les Panis et nègres qui ont été achetés… appartiendront en pleine propriété à ceux qui les ont achetés comme étant leurs esclaves” — all Panis and Negroes who have been purchased, or will be, belong in full property to those who have purchased them, as their slaves.115

“Panis” derives from Pawnee, among the earliest important suppliers of Indigenous captives, and became the generic French term for any enslaved Indigenous person — drawn from Plains Apache villages in the southwestern Plains, Sioux settlements in present-day Minnesota, Fox and Sauk communities near the Great Lakes.116

Marcel Trudel’s exhaustive archival work documented 4,185 enslaved people in New France and later Upper and Lower Canada: 2,683 Indigenous, 1,443 Black, 59 of unknown origin.117 Roughly two-thirds Indigenous. That is a documented floor rather than a population — Trudel found about 3,000 in Catholic parish registers and roughly a thousand more only in lacunary records — and Brett Rushforth estimates the true figure much higher, suggesting colonists in New France held “thousands of American Indians as slaves, perhaps as many as ten thousand over the course of the century.”118 Give both, and say which is which.

The demographic profile is the detail that lands hardest. The average age of enslaved Panis was fourteen; of enslaved Black people, eighteen. Fifty-seven per cent of enslaved Indigenous people were girls or young women; 57 per cent of enslaved Black people were male. In Louisiana, the average age at death for enslaved Indigenous people was seventeen.119

Rushforth’s analytical contribution matters here too, because it prevents the story from being a simple one. Indigenous nations used captive-giving as diplomacy — captive enemies given as gifts to create or affirm alliances — and used it to steer French colonial expansion, blocking a French–Sioux alliance for decades. French demand then fused that practice to Atlantic slaving, and Martinique’s difficulty obtaining African captives during the Saint-Domingue sugar boom “facilitated slaving raids on the Great Plains.” A Caribbean sugar shortage produced raids in what is now Manitoba.

Chloe Cooley, and what the 1793 Act really did

On 14 March 1793, at Queenston in Upper Canada, Sergeant Adam Vrooman bound an enslaved Black woman named Chloe Cooley with rope, put her in a boat and took her across the Niagara River to sell her in the United States, assisted by his brother and another man. She fought and screamed for help. It did not save her.120

On 21 March, Peter Martin — a Black Loyalist and army veteran — and William Grisley, a white labourer who had worked for Vrooman, reported the incident to the Executive Council. Lieutenant Governor John Graves Simcoe, Chief Justice William Osgoode and Peter Russell directed Attorney General John White to prosecute. The charges were dropped, because what Vrooman had done was legal.

Simcoe and White used the case to push abolition legislation. It met opposition from slaveholders inside the government, and what emerged on 9 July 1793 was a compromise: “An Act to prevent the further introduction of slaves.” Here is exactly what it did:

1. Banned importation of any Black person into Upper Canada as a slave or indentured servant. Anyone enslaved arriving in the province thereafter was free.
2. Freed nobody already enslaved. Existing enslaved people remained the property of their owners for life.
3. Children born to enslaved mothers after 1793 remained enslaved until age twenty-five.

It applied only to Upper Canada — not Lower Canada, not Nova Scotia, not New Brunswick.121 It is routinely described in Canadian public memory as “abolishing slavery in Upper Canada.” It was a gradual-abolition importation ban, and it was the first such legislation in the British Empire, and both of those things are true simultaneously.

Actual abolition in Canada came from Westminster: the Slavery Abolition Act 1833, in force 1 August 1834, which ended slavery throughout the British Empire including all of British North America, freeing over 800,000 people empire-wide.122 Afua Cooper frames it as concluding “206 years of enslavement of African people in Canada.” By then relatively few enslaved people remained in Upper Canada, because the 1793 Act had done its slow work — so 1834 matters more legally than demographically there. Cooper adds a demographic note that completes the picture: by the British period, 1760–1834, nearly 100 per cent of the enslaved in Canada were of African descent. The Indigenous majority was a feature of the French regime.

12. How it ended, country by country

Five mechanisms, not one: revolution or coup, legislation, war, colonial imposition, and external diplomatic pressure.

The only case where the enslaved abolished it themselves

The largest slave uprising since Spartacus began in Saint-Domingue on 21 August 1791; 100,000 people joined within weeks. French abolition followed, it did not precede: Léger-Félicité Sonthonax proclaimed abolition in the northern province on 29 August 1793 as an emergency wartime measure, freedom offered to enslaved men who would fight for the Republic; the National Convention decreed general abolition on 4 February 1794. Napoleon sent Leclerc in 1802 with secret instructions to restore slavery once French military dominance was re-established; Guadeloupe was re-enslaved by force, Saint-Domingue was not. Dessalines declared independence at Gonaïves on 1 January 1804.139 France would abolish slavery again, permanently, on 27 April 1848 — and in Martinique it took effect on 23 May, ahead of the decree’s own timetable, because the enslaved rose and forced the governor’s hand.

Britain: the argument that never ended

The Slave Trade Act 1807 received royal assent on 25 March 1807 and took effect 1 May. The Slavery Abolition Act 1833 received assent on 28 August 1833 and came into force 1 August 1834, exempting East India Company territories, Ceylon and St Helena — a gap closed only by the Indian Slavery Act 1843.140

Why did Britain do it? This is one of the genuinely unresolved debates in economic history and I am going to leave it unresolved, because it is.

Eric Williams’s Capitalism and Slavery (1944) advanced two claims: a decline thesis, that abolition followed falling profitability, and a capitalism thesis, that slave-trade profits financed British industrialisation. British publishers refused the book for twenty years. Seymour Drescher’s Econocide (1977) attacked the decline thesis directly: in 1807 the British slave system was expanding and profitable, and abolition was economic suicide driven by moral mobilisation among people who could vote. On the capitalism thesis, Stanley Engerman calculated slave-trade and plantation profits at under 5 per cent of the British economy, and David Richardson at under 1 per cent of domestic investment. On the other side, David Beck Ryden finds evidence of an emerging profitability crisis by the early nineteenth century, and Joseph Inikori challenges the revisionists by widening the frame to whole-Atlantic trade rather than the sugar islands alone.141

The fair summary: Drescher refuted the narrow decline claim about 1807. The profits-financed-industrialisation claim remains genuinely contested. And neither econometric result settles the moral and political question of whether Britain’s humanitarianism was self-interested — which is a different kind of question.

Resistance as cause

The claim that white people ended slavery survives contact with the Baptist War badly. On Christmas Day 1831, up to 60,000 of Jamaica’s roughly 300,000 enslaved people rose under the Baptist deacon Samuel Sharpe — the largest slave revolt in the British Caribbean. Fourteen white people were killed. Over 300 rebels died in the fighting and roughly 310 to 340 were judicially executed afterwards, Sharpe among them. The causal link to 1833 runs through the plantocracy’s reprisals, which extended to white missionaries: when Burchell and Knibb testified to the House of Commons about their treatment, MPs were outraged, and Parliament passed the Abolition Act the following year.142

Add Nat Turner in Southampton County, Virginia, on 21 August 1831 — which precipitated Virginia’s 1831–32 legislative debate on emancipation, and, when that failed, a wave of harsher slave codes. Add the maroon communities that won territorial concessions by force decades or centuries before general emancipation: Palmares in Brazil, reportedly 30,000-plus people at its height, from around 1600 to 1694; the Jamaican Maroon treaties of 1739–40 granting 2,500 acres and freedom nearly a century before abolition; the Ndyuka treaty in Suriname of 10 October 1760.143 Add the Black abolitionists who were central rather than decorative: the Sons of Africa, founded 1787, Britain’s first Black political organisation, whose lobbying via Equiano produced the Slave Trade Act 1788; Quobna Ottobah Cugoano, whose Thoughts and Sentiments (1787) demanded immediate and total emancipation — more radical than most white abolitionists of the day — and argued the enslaved had a duty to escape and a right to use force.

The Royal Navy paradox

From 1808 the West Africa Squadron patrolled to suppress the trade Britain had recently dominated. Over roughly sixty years it seized around 1,600 ships and freed about 150,000 people, intercepting perhaps 6 per cent of transatlantic slave voyages. Almost 1,600 British sailors died on Squadron duty between 1830 and 1865, principally of disease.144 Kaufmann and Pape’s study characterises the campaign as the most expensive international humanitarian intervention in modern history — I have seen a specific percentage-of-national-income figure attached to that in a hundred places, and could not verify it against the article, so it is not here.145

Hold that alongside this: in the decade 1791–1800 alone, British vessels made around 1,340 slaving voyages landing nearly 400,000 people, and Britain’s all-time total was 3,259,441 embarked — the second-largest carrier in history.146 Both facts are true. Britain built the largest slave-trading operation of the eighteenth century, then spent sixty years and 1,600 sailors’ lives suppressing what it had built, and paid £20 million to the owners and nothing to the freed. Any account that keeps only one of those three is propaganda, whichever one it keeps.

The long tail

The dates most people never learn:

Two of those entries deserve a second look. Ethiopia shows what a decree is worth without enforcement: Menilek abolished slavery in April 1889 — while exempting war captives, who owed seven years’ service; the Treaty of Wuchale the following month abolished the slave trade but not slavery, on the stated ground that the trade was “against the principles of the Christian religion”; a further proclamation in 1906 punished slave raiders by enslaving them for life; Empress Zawditu and Ras Tafari reconfirmed the prohibitions in September 1923, the same month Ethiopia joined the League of Nations; comprehensive regulations were submitted to the League in 1924, including a fifty-thaler fine for calling an emancipated person a “slave”; and then the 1931 amendments to the Penal Code moved backwards, delaying manumission on an owner’s death and requiring the enslaved to “render what is due” to heirs before receiving freedom certificates.147 Fifty-three years of decrees, reissued because they kept not working. Italy then used Ethiopian slavery as a justification for the 1935 invasion, and its occupation laws claimed to free 420,000 people — an occupation-era figure with no peer-reviewed corroboration, and a genuine abolition deployed as the pretext for a war of conquest.

Mauritania shows the same gap in the present tense. Slavery was abolished by Ordonnance n° 81-234 of 9 November 1981 — and the ordinance created no criminal offence, no penalty and no mechanism, which is why it changed almost nothing for twenty-six years.148 Criminalisation came in 2007. The first successful prosecution came in 2011. A 2015 law reclassified slavery as a crime against humanity, with ten to twenty years for slavery itself and five to seven for servile marriage, serfdom and debt bondage, and created special anti-slavery courts.