What the Ottomans gave the Balkans, and what they took (1385-1648)

Before: a peninsula of collapsing states

When Stefan Dušan died on 20 December 1355, aged forty-seven, he had spent nine years styling himself Emperor of the Serbs and Greeks, and his state ran from the Danube to the Gulf of Corinth. Within twenty years it had ceased to exist. His son Uroš held the title and nothing else; the great lords divided the empire and fought each other for the pieces — the Mrnjavčević brothers in Macedonia, the Balšići in Zeta, Lazar Hrebeljanović in the Morava valley.²² The Second Bulgarian Empire had split in the same decade into Tarnovo under Ivan Shishman, Vidin under Ivan Sratsimir, and the Dobrujan despotate on the Black Sea coast. Byzantium was a capital, a strip of Thrace, a despotate in the Morea and some islands; since John V's arrangement of 1373 it had paid tribute to the Ottoman sultan and sent him troops.²⁵ On the Adriatic the Venetians held the ports. In Bosnia Tvrtko I crowned himself king in 1377, over a kingdom whose church Rome regarded as heretical and Constantinople did not regard at all.

This is what the Ottoman advance found in the 1370s and 1380s: not a civilisation and not a nation, but a zone of successor polities that had spent two generations destroying one another, in which the units of political loyalty were a lord, a valley and a church, and in which nobody living had any use for the word "Balkans" — a term that would not acquire its present meaning until the nineteenth century, and would acquire it as an insult.²¹

What a peasant actually owed

The pre-Ottoman Balkan peasant was not free, and the obligations he carried were not written down anywhere he could consult. In Serbia, Dušan's own law code of 1349 and 1354 fixed the meropah — the dependent cultivator — to his land, specified two days of labour a week for his lord plus a day of mowing and a day of vineyard work, and provided that a peasant who fled be caught, returned, and have his ears slit and his face branded.²² In Byzantine Macedonia the paroikos held his plot under a monastery or a pronoia-holder, and the charters of Mount Athos record his dues in a language of custom that was elastic in exactly the direction the holder wanted. The obligations were heritable, unbounded at the top, and enforced by the man who collected them.

Above that lay a warrior aristocracy whose claims on land were becoming permanent. The Serbian baština was hereditary property; the Byzantine pronoia, originally a revocable life grant of revenue, had been made heritable in the fourteenth century, which is to say that the imperial government had lost the ability to take it back.¹ The structural consequence is the one that matters for this record: by 1370 the states of the Balkan peninsula could no longer count what they owned, could no longer reclaim what they had granted, and could no longer field armies except by asking lords who were now proprietors rather than officers.

Towns, churches and the Jews already there

The Balkan towns of the fourteenth century were small, walled and commercially dependent on outsiders — Ragusan merchants in the Serbian mining towns, Venetians and Genoese in the ports, Saxon miners at Novo Brdo and Srebrenica, where the silver came from. Thessaloniki, the second city of the Byzantine world, had perhaps 25,000 to 40,000 people and had spent 1342 to 1350 under the Zealot commune, a revolt of the urban poor that expelled the aristocracy and was put down with killings.

Jewish communities were long established and Greek-speaking: the Romaniote congregations of Thessaloniki, Ioannina, Ohrid, Kastoria and Constantinople itself, whose liturgy was Byzantine, whose vernacular was Greek written in Hebrew characters, and whose legal position under Christian rule had been narrowing for a century. Byzantine law had barred them from public office and from owning Christian slaves, which in a slave-owning agrarian economy barred them from agriculture; Venetian and Genoese quarters imposed their own restrictions; and the Fourth Lateran Council's badge requirements had been imported into Latin-ruled parts of the peninsula. They are the population most often left out of the account of what the Ottomans arrived into, and they are the population whose position changed most sharply and most measurably.

The point of this section is not that the pre-Ottoman Balkans were miserable. It is that they were unadministered. Nobody in 1370 could have told you how many households there were between the Danube and the Aegean, what they produced, or what they owed — and everybody with a claim on them was making it up as he went.

The transmission: how a raid became a register

The Ottomans crossed into Europe as auxiliaries. Süleyman, son of Orhan, took Gallipoli in 1354 after an earthquake brought its walls down, and the frontier lords who followed him were raiders operating for plunder and slaves, not administrators. What turned the raid into a state happened over the following sixty years, and it happened in Thrace: Edirne fell around 1369 and became the Ottoman capital for the next eighty-four years, the Serbian lords were broken at the Maritsa near Chernomen on 26 September 1371, Sofia fell in 1385, the Serbian coalition was destroyed and both Sultan Murad I and Prince Lazar killed at Kosovo Polje on 15 June 1389, and Tarnovo, the Bulgarian capital, was stormed in July 1393.²²

Behind the raid came paperwork. This is the transmission.

The tahrir: government by survey

The instrument was the tahrir defteri, the cadastral register. Ottoman officials went into a newly conquered district, walked it, and wrote down every village, every household head by name, every adult unmarried male, every widow, every mill, every vineyard, every hearth-tax liability and the estimated annual revenue of each — then assigned that revenue, in specified amounts, to named individuals.¹ The surveys were redone every twenty to thirty years. They survive in their thousands in Istanbul.

The revenue assignment was the timar: a cavalryman received the tax income of a village or part of one, in return for appearing on campaign with a specified number of armed retainers. He did not own the land. He could not sell it, bequeath it or subdivide it. The peasant's holding — the çift, notionally the area a pair of oxen could work — was the peasant's to cultivate and to pass to his son so long as he cultivated it, and the sipahi could not evict him from it. Halil İnalcık named the arrangement the çift-hane system and argued that it was not an Ottoman invention but a synthesis: Byzantine fiscal categories, Seljuk practice and Islamic land law assembled into a single unit of account in which "the peasant family farm" was the object the state taxed and therefore the object the state protected.¹²

What the peasant paid, and what it replaced

The Ottoman fiscal settlement in Rumelia looked like this:

The comparison with what came before is the record's most uncomfortable fact and it has to be stated at full strength. A registered poll tax and a cash commutation of labour service, assessed on a survey a judge could be shown, imposed by a cavalryman who held his revenue at the sultan's pleasure, was a lighter and far more predictable burden than the open-ended dues of a Serbian or Byzantine proprietor who answered to nobody.² Peasants in fifteenth-century Rumelia had access to a kadı court that was not their lord's court, and the surviving registers of complaints show them using it. This is not an apology for the Ottoman conquest. It is the reason the Ottoman conquest stuck.

The predictability had a shelf life, and it is worth naming the mechanism that ended it. The Ottoman fiscal system was denominated in the akçe, a small silver coin whose weight the treasury controlled; Şevket Pamuk's monetary history tracks its debasements, the sharpest of which came in 1585–86, when the silver content was cut by roughly forty-four per cent in a single stroke.²⁴ Fixed cash dues like the ispençe lost value; the state responded by making the avarız war levy annual and by farming taxes out to contractors who bought the right to collect and profited from what they could extract beyond it. By the mid-seventeenth century the tax-farmer had replaced the surveyed timar across much of Rumelia, and the peasant's burden was no longer capped by anything a court could be shown. The Balkan countryside that rose in the eighteenth and nineteenth centuries was rising against that system, not against the one described above — a distinction the national histories have every reason to blur.

The Christian sipahis of Albania, 1431

The Ottoman state also did not immediately dispossess the class it had defeated. The cadastral register of the sanjak of Arvanid — Albania — compiled in AH 835 (1431–32) and published by İnalcık in 1954, lists the timar-holders of the province by name.¹¹ A substantial proportion of them are Christians: men with names like Yorgi, Dimitri and Pavlo holding revenue assignments on exactly the same terms as their Muslim colleagues, several of them the sons of the lords whose fathers had fought the Ottomans a generation earlier. The Albanian rebel Skanderbeg — Gjergj Kastrioti — was himself the son of a timar-holding Ottoman vassal, raised at the Ottoman court, who defected in 1443 and held Krujë against two sultans for twenty-five years.

Heath Lowry built the strongest version of the argument this evidence supports, that the early Ottoman polity expanded less by extermination than by istimâlet — accommodation, the deliberate purchase of local elites with a place in the new order.¹⁹ The argument is correct about the mechanism and is regularly overstated about the experience. Accommodation is what the register records because the register records the accommodated. The men who refused are in a different kind of document.

The devşirme: the state takes children

The second mechanism has no precedent in the region and no straightforward parallel anywhere.

Every few years — the interval was irregular, sometimes three years, sometimes seven — the Ottoman government sent an officer with a warrant into the Christian villages of Rumelia and later Anatolia, accompanied by a scribe and a local kadı to witness. The village priest produced the baptismal register. The boys were lined up. The officer picked, at a notional ratio of one per forty households, and wrote each chosen boy into two copies of a register: name, father's name, village, age, height, complexion, distinguishing marks, so that the boy could not be substituted en route.⁸ The chosen were marched to Edirne or Istanbul in convoys, circumcised, converted, given Muslim names, and distributed.

The Ottoman term is devşirme, from devşirmek, "to gather, to collect." The phrase "blood tax" — kan vergisi — is not an Ottoman term and appears in the Balkan national historiographies of the nineteenth and twentieth centuries rather than in the fifteenth-century record; it is worth using the Ottoman word precisely because the euphemism and the counter-euphemism are both doing work.⁴

Ottoman court painter. Sokollu Mehmed Pasha exploring the surroundings of Szigetvar, Nuzhetu'l-esrar, fol. 22a, 1568. Public domain via Wikimedia Commons. · Public domain

Basilike Papoulia's monograph of 1963, still the fullest treatment of the institution's origin and legal character, argued against the Ottoman jurists' own justification — that the levy was a lawful extension of the pençik, the sultan's one-fifth share of war captives — on the ground that the boys were not captives of war but the children of protected subjects, whose status under Islamic law was precisely that they could not be enslaved.⁴ The institution was, on her reading, a sustained legal irregularity that the empire's own jurisprudence never comfortably absorbed. Vasilis Demetriades, working from the Ottoman side, argued that the levy grew instead out of frontier practice in the Thracian marches and only later acquired the legal dress.⁷ The debate is live and the record should not pretend otherwise; what both sides agree on is that it was happening by the 1390s.

What a levied boy became

The boys were sorted. The great majority — the acemi oğlan, the "foreign boys" — were placed with Turkish farming families in Anatolia for some years to learn the language and the religion, then enrolled in the corps of novices, then in the janissary infantry: a standing, salaried, barracked, celibate professional army at a time when European states had none. The janissary payroll registers of the seventeenth century, kept in the Topkapı archive, list them by pay, company and name.

Janissary pay register (Yeniceri Ulufe Defteri), first page, 1652. Topkapi Palace Archive, book no. 795, fol. 1. Photograph by Wikimedia user SAIT71. Public domain via Wikimedia Commons. · Public domain

A minority — the ablest, the best-looking, the ones the sorting officer marked — went instead into the palace schools at Galatasaray, Edirne and the Topkapı itself, where they were taught Ottoman Turkish, Arabic, Persian, calligraphy, horsemanship, archery and an art or trade, and from which they emerged into the household and provincial service of the sultan. They were kul: the sultan's slaves in law, without heritable property or standing lineage, entirely dependent on his favour and entirely disposable by it.²⁰ They also ran the empire.

Sokollu Mehmed Pasha is the case that contains the whole argument. Taken from a Bosnian village as a boy, educated in the palace, married to a sultan's granddaughter, grand vizier for fourteen years and effective ruler of the empire for most of them, he used his position in 1557 to obtain a decree restoring the autocephalous Serbian Patriarchate of Peć, abolished in 1463, and installed his own kinsman Makarije as its first patriarch.²²²³ A boy taken by force from an Orthodox village grew up to refound, by imperial fiat, the Orthodox church of the country he had been taken from. He was stabbed to death by a dervish in his own palace in October 1579.

What changed, and what was replaced

Confession becomes legal personhood

On 6 January 1454, five months after his soldiers took Constantinople, Mehmed II enthroned the monk Gennadios Scholarios as Ecumenical Patriarch and gave him a berat — a written patent of office — defining his rights and duties.¹³ Gennadios was an unpromising choice for a sultan seeking accommodation with the West and a very shrewd one for a sultan who was not: he was the leader of the anti-unionist party, the churchman who had most fiercely opposed the union with Rome agreed at Florence in 1439.

What the berat did was make a bishop into an official. The patriarch became responsible for the Orthodox population's taxes, its internal discipline, its marriage and inheritance law, and its good behaviour, and he became removable by the sultan for failure in any of those. He paid for the appointment. His successors paid more: the peşkeş, the accession payment to the imperial treasury, became a standing charge that the patriarchate financed by selling bishoprics, and the chair changed hands with a frequency that had nothing to do with theology — between Gennadios's enthronement in 1454 and 1662 the throne turned over some sixty times, and Cyril Loukaris alone held it on six separate occasions, deposed and restored as the money moved.¹³

Something similar, less tidily, happened to the other confessional communities. The word later attached to the arrangement is millet, and here the record has to be careful, because the neat administrative diagram found in most textbooks — three or four chartered nations, each with a head, each with defined autonomy, established by Mehmed II in 1453 — is not what the fifteenth-century evidence shows. Benjamin Braude's 1982 essay, the most consequential single piece of scholarship on this subject, showed that the charters do not exist, that the word millet is not used in this sense in Ottoman documents until much later, and that the tidy system is a nineteenth-century construction retrojected onto the conquest.³ Braude's conclusion is blunt: what the Ottomans built in the fifteenth century was "a series of ad hoc arrangements," improvised locally and differently for each community, and only codified into a system in the reform era of the 1830s onward.

This matters editorially, not pedantically. The ad hoc arrangements did almost everything the mythical system is credited with. They just were not a design, and nobody in 1454 was administering a policy of tolerance. They were administering a tax base.

Salonica, and the city the expulsions built

The clearest single demonstration of what the arrangement made possible is a city.

Thessaloniki fell to Murad II on 29 March 1430, after the Venetians who had bought it from its Byzantine despot failed to hold it; the city was sacked and a large part of its population enslaved.¹⁴ It was a diminished town for sixty years. Then, in 1492, Ferdinand and Isabella expelled the Jews of Castile and Aragon, and Bayezid II's government let them in — not out of sentiment but out of policy, because the empire wanted urban artisans, capital and the Iberian textile trade, and because a population with nowhere else to go is a reliable population. Sephardi refugees went to Istanbul, Edirne, Sarajevo, Safed and above all to Salonica.

By the Ottoman tax register of 1519 the city held some 3,143 Jewish households and around 930 unmarried Jewish men — a Jewish population of roughly 15,700, more than half the town's total.¹⁴ Salonica became, and remained for four centuries, the only city of any size in Europe with a Jewish majority. Its congregations were named for the places they had fled — Catalan, Aragon, Castile, Mallorca, Lisbon, Sicilia, Provincia — and kept those names for four hundred years. Its people spoke Judeo-Spanish. Its presses printed Hebrew books from 1512. Its woollen cloth clothed the janissary corps, and the community paid part of its tax obligation in bolts of it.

Elkan Nathan Adler. Catalan Synagogue at Salonica, from Jews in Many Lands, before 1905. Public domain via Wikimedia Commons. · Public domain

Mark Mazower's account of the city puts the point without softening either end of it: the Ottoman state offered these communities not equality but a place, and the place was real, durable and preferable to the alternative on offer in Christian Europe, which was the stake.

Conversion: the arithmetic of the poll tax

Islam spread in the Balkans unevenly, slowly, and along lines that were not primarily military. There was no general forced conversion; there was a standing financial incentive, a closed career ladder, and centuries of time.

The incentive was the cizye, which a convert stopped paying. Because the poll tax was assessed in bands but fell as a much larger share of a poor household's income than a rich one's, the incentive was strongest exactly where people had least. Leonard Kukić and Yasin Arslantaş tested this in 2026 against village-level Ottoman tax registers for Bosnia from 1468 and 1604 — the two surveys that bracket the province's conversion — and found that settlements with lower income per household were significantly more likely to have converted.¹⁷ Bosnia, with its thin soils and poor yields, converted more thoroughly than anywhere else in Ottoman Europe; Albania followed; the fertile Bulgarian and Serbian plains largely did not.

Conversion was also a narrated, negotiated, contested process rather than a switch. Tijana Krstić's study of early modern conversion narratives shows converts producing accounts of their own change that argued with Christian polemic, drew on Sufi and messianic idioms, and were shaped by the Ottoman state's own sixteenth-century project of defining Sunni orthodoxy against the Safavids.¹⁵ Anton Minkov's work on the kisve bahası petitions — the applications converts filed for a cash grant and a suit of clothes, of which thousands survive from 1670–1730 — shows the material texture of it: who converted, at what age, from what trade, and with what expectation of the state.¹⁶

What was displaced

The Ottoman settlement removed or hollowed out a set of institutions, and it is worth listing them rather than gesturing at them:

• The independent Balkan monarchies. Serbia's despotate ended in 1459, Bosnia's kingdom in 1463 with the execution of King Stefan Tomašević at Jajce, Byzantium in 1453 and the Morea in 1460, Albania's resistance with the fall of Shkodër in 1479.
• The indigenous nobility, as a class. Those who accepted timars were absorbed and, over two or three generations, converted; those who did not were dispossessed or killed. By 1550 there was no Christian aristocracy in Ottoman Europe.
• The autocephalous churches. The Bulgarian Patriarchate of Tarnovo was suppressed after 1393 and its territory placed under Constantinople; the Serbian Patriarchate of Peć was abolished in 1463, restored in 1557 through Sokollu Mehmed Pasha, and abolished again in 1766. Ecclesiastical Hellenisation of the Slav churches is a real and documented process, and it was carried out by the Greek hierarchy under Ottoman licence, which is why nineteenth-century Bulgarian nationalism fought the Patriarchate before it fought the sultan.
• Urban Latin commerce. Ragusan and Venetian merchant colonies survived but under capitulation, on terms set elsewhere.
• The written vernaculars of the Slav chanceries, which lost their courts and shrank to monastic and liturgical use.

What it cost, and who paid

The three registers

This subject is fought over by five national historiographies and every number in it carries a flag. The discipline this record applies is the one it applied to the Mongol conquest of Iran: keep three registers of evidence visibly separate.

1. Documentary. Counts derived from Ottoman registers: 2,604 boys levied empire-wide in 1603–4; 530 of them from the Bursa region in six months; an average age of 16.5 in that levy, with 42 per cent aged eighteen or over.⁹ 3,143 Jewish households at Salonica in 1519.¹⁴ These are administrative counts and they are as good as the administration that made them, which was good.
2. Estimate. The frequently cited total of around 200,000 boys taken across the institution's life is an extrapolation from levy intervals and ratios, not a count, and figures up to 300,000 circulate on the same basis. It should always be stated as what it is.
3. Rhetoric. The casualty and abduction figures in the nineteenth- and twentieth-century national literatures — the ones that make the levy annual, universal and total — are political claims, and so, in the opposite direction, is the twentieth-century apologetic that presents the levy as an opportunity families competed for.

The conquest itself

The Ottoman conquest of the Balkans killed people, enslaved people, and destroyed towns, and the administrative story does not cancel it. Thessaloniki was sacked in 1430. Constantinople was sacked over three days from 29 May 1453 and a large part of the surviving population enslaved. Novo Brdo, the great Serbian silver town, surrendered in June 1455, and what happened there is described by a man who was in it:

"They drove us all out into a field, and, having separated the women to one side and the men to the other, they took 320 boys and 700 women, and the remainder they let go into the city." — Konstantin Mihailović, taken from Novo Brdo at about twenty and enrolled as a janissary, writing some forty years later.¹²

Mihailović served in the sultan's campaigns from 1456 to 1463, was taken prisoner in Bosnia, and ended in Poland writing, in a Slavic vernacular, the earliest insider account of the janissary corps by one of its own. He is the record's most valuable witness precisely because he is neither a victim testifying nor an apologist: he describes the corps with professional admiration and describes his own taking without one word of gratitude.

The children

This is the cost centre, and the fullest contemporary statement of it is a sermon.

On 28 February 1395 — the Sunday of the Prodigal Son — Isidore Glabas, metropolitan of Thessaloniki, preached to a congregation from which children had just been taken. His homily is the earliest datable Christian testimony to the levy and the document that pushed the institution's established start back from 1438 to the 1390s.⁵⁶ Speros Vryonis, who identified and translated it in 1956, observed that "Glabas' sermon thus pushes back the date of the institution of the devshirme from 1438 at least to the year 1395."⁵

Isidore does not argue about law. He argues about a father:

"What would a man not suffer were he to see a child, whom he had begotten and raised … carried off by the hands of foreigners, suddenly and by force, and forced to change over to alien customs and to become a vessel of barbaric garb, speech, impiety, and other contaminations, all in a moment?"⁶

And then, the image that makes the homily worth quoting six hundred years later — the parent as a body cut in half, one part taken, the other part still standing in the church:

"if a man were to find himself as if cut into two parts; and if he were to see the one dismembered section of his body, his son, become a substance of baseness and not used for any good purpose; and to see the remaining section, himself that is, not only useless but dead and full of lamentation and agony? … Shall he lament himself because he has been deprived of the staff of his old age? Because the light of his eyes has perished?"⁶

The institution he was describing ran for roughly two and a half centuries after that sermon. The levies became less frequent from the 1590s as janissary recruitment opened to Muslim volunteers and the corps became hereditary in practice; the last large regular levies belong to the 1630s and 1640s, with isolated instances into the eighteenth century.

The honest accounting of the cost has to include four things the polemics on both sides leave out.

• The exemptions were real and documented. Only sons, married boys, boys from Istanbul, artisans' apprentices in some periods, Jews, Armenians in some periods, and — with one standing exception — Muslims were not liable. The exemptions mean the levy was not a universal terror; they also mean it fell hardest on the villages with the least ability to resist it.
• The exception was Bosnia. Bosnian Muslim boys, recorded in Ottoman documents as Poturoğulları, were levied — the one case in which Muslims were taken — on the basis, stated in the janissary corps' own seventeenth-century law book, that the Bosnians had converted en masse at the conquest and had petitioned Mehmed II for inclusion.¹⁰ Aşkın Koyuncu's re-examination of that text treats the petition story with proper suspicion while confirming the practice. Whatever its origin, it is evidence that by the sixteenth century the levy had become, for some families, a career rather than a catastrophe.
• The Anatolian levies. Gülay Yılmaz's register shows 530 of the 2,604 boys taken in 1603–4 coming from the Bursa region.⁹ The devşirme was never exclusively Balkan, and the "Balkan boys only" version of the story is wrong.
• They did not come back. Individual careers aside, the levy was permanent removal. The boy was converted, renamed, and legally the sultan's property. Sokollu Mehmed Pasha, at the height of his power, could restore a patriarchate for the country he came from; he could not go home to it.

Sürgün

Alongside the levy ran a second form of coerced movement, less dramatic and much larger: sürgün, forced resettlement as an instrument of policy. Ömer Lütfi Barkan's studies of the 1940s and 1950s established it as a routine administrative technique rather than an episode — populations moved to settle emptied frontier districts, to break the local power of a conquered elite, or to fill a city.¹⁸ After 1453 Mehmed II repopulated Constantinople by order, transferring Muslim, Christian and Jewish households from across the empire into a capital that his own conquest had emptied, with quotas issued to provincial governors.¹

Sürgün was not a massacre and should not be narrated as one. It was the compulsory relocation of households by administrative order, applied for centuries, and it is the sort of cost that leaves no martyrs and therefore no memory.

Going back was capital

The exit was one-way in law. Conversion to Islam was irreversible: apostasy from Islam carried the death penalty, and it was applied. The Orthodox churches accumulated, from the fifteenth century to the nineteenth, a body of neomartyrs — men and women executed for reverting to Christianity or for publicly renouncing an Islam they had adopted under pressure, as children, or while drunk. The Athonite monk Nikodemos compiled their lives into the New Martyrologion in 1799, at the moment the Greek national movement was forming and had a use for them.

The neomartyr corpus is hagiography and must be read as such: it was compiled to be edifying, its numbers are unverifiable, and the eighteenth-century Kollyvades monks who assembled it were engaged in a deliberate campaign to stiffen a population they judged to be drifting. Several of the lives describe men who had converted as children under the levy, or as adults under pressure, and who sought out execution by publicly renouncing Islam in a market square — a form of witness the Ottoman authorities usually tried to avoid providing, offering the accused repeated chances to retract before carrying out the sentence.

It is nonetheless a register of a specific, enforced, lethal asymmetry. Conversion into Islam was encouraged, subsidised with a suit of clothes and a cash grant, and recorded with satisfaction; conversion out of it was a capital offence for four hundred years. The empire's confessional architecture had a door that opened in one direction, and the neomartyrs are the people who tried to walk back through it.

The ledger

The deepest cost of the Ottoman settlement in the Balkans is the hardest to date, because it is a category rather than an event.

For four centuries the Ottoman state administered south-eastern Europe by confession. It taxed by confession, judged by confession, conscripted by confession, and recorded every household in a register that said which confession it belonged to. It did not create the region's religious diversity — that long predates it — but it made confession the master fact of legal personhood and it wrote the result down, comprehensively, for four hundred years.

When the national movements arrived in the nineteenth century they inherited that ledger and used it as an ethnic map. The Ottoman census categories became the basis of claims to territory; the Bulgarian Exarchate's 1870 fight with the Patriarchate was a fight over which villages counted as which nation, conducted through the machinery of the millet; the 1923 Convention concerning the Exchange of Greek and Turkish Populations sorted roughly 1.2 million Orthodox Christians out of Anatolia and roughly 400,000 Muslims out of Greece on the basis of religion rather than language, because religion was the category the records held. Salonica's Jews — who had been let in when Christian Europe expelled them, and who had been the city's largest community for four hundred years — were deported to Auschwitz in 1943, some 46,000 of them, by a German occupation that found the community already enumerated.

That last sequence is not an Ottoman crime and this record does not assert that it is. It is the observation that an administrative category, kept long enough and written down carefully enough, becomes a fact that later states can act on — and that the men who designed the category in the 1450s were solving a tax problem.

The reckoning

What the Balkans received was a state of a kind the region had not had: solvent, surveyed, judicially accessible, capable of taxing without ruining, and for two centuries the strongest military power in Europe. What the Balkans paid was a conquest, an aristocracy, a set of independent churches, the permanent removal of somewhere on the order of two hundred thousand children, centuries of a one-way confessional door, and a ledger that outlived the empire that kept it.

Both halves are true at once. The nineteenth century's "Turkish yoke" is not an adequate description of the first two centuries of Ottoman rule, and Maria Todorova has shown in detail what political work that phrase was invented to do and is still doing.²¹ But the correction is not that the empire was benign. The correction is that its cruelty was specific, legally documented, aimed at children, and administered by a scribe.