Greek science made Arabic: Baghdad's translation century (830)

Before: an empire with everything except this library

In the year 762 the caliph al-Manṣūr laid out a round city on the west bank of the Tigris and called it the City of Peace. Within forty years it was the largest city west of Chang'an. It had a mint, a canal system, a postal intelligence service inherited from the Sasanians, a judiciary, a standing army recruited out of Khurāsān, and a court that could pay for anything it wanted. What it did not have was a single book of geometry, medicine, astronomy or logic in Arabic. Not one. The language of the Qurʾān, of the greatest lyric poetry then being produced anywhere on earth, and of an imperial chancery running a territory from the Indus to the Atlantic, had no word for hypotenuse, no word for syllogism, no word for artery, and no prose in which such words could have been used ¹.

This is the fact the record turns on, and it is easy to state badly. It does not mean the Arabs of 762 were unlearned. They were, in their own sciences, formidably learned — more systematically so than any Latin-reading population of the same date. It means something narrower and stranger: that a specific library existed, in Greek and Syriac, a few hundred kilometres to the north and west, and that the people running the largest empire in the world had not read it.

The sciences the Arabs already had

The indigenous Arabic sciences of the eighth century were complete disciplines with their own methods, their own controversies and their own professional class. They were not a preparation for anything. They were finished work.

• Grammar and lexicography. Al-Khalīl ibn Aḥmad al-Farāhīdī (d. c. 791) at Basra compiled the Kitāb al-ʿAyn, the first Arabic dictionary, organised by phonetic point of articulation rather than alphabetically, and codified the quantitative metres of Arabic verse. His student Sībawayh, a Persian who died around 796, wrote al-Kitāb — some nine hundred pages of syntax, morphology and phonology, edited after his early death by his own student al-Akhfash al-Awsaṭ. It is one of the great analytical achievements in the history of linguistics and it was produced with no Greek input whatever ².
• Ḥadīth criticism. A discipline of source evaluation built on the isnād, the chain of transmitters, requiring biographical dictionaries of tens of thousands of named individuals with dates, teachers, places of residence and assessments of reliability. Nothing in the Greek corpus resembles it.
• Jurisprudence. Abū Ḥanīfa died in 767, Mālik ibn Anas in 795, al-Shāfiʿī in 820. The formal apparatus of Islamic legal reasoning — analogy, consensus, the hierarchy of sources — was worked out in exactly the decades in which the first Greek books were being commissioned, and by men who mostly never read them.
• Genealogy, historical report, poetry and the vast miscellany later called adab. Baghdad's literary culture was in no sense provincial.

M. G. Carter's point about Sībawayh is the one to keep: Arabic grammatical science arose from methods developed inside Islamic legal thought, not from any Greek model, and its technical vocabulary is native ². When Greek logic did arrive, it arrived as a competitor to an existing science of language, and it was received as one. That collision produced the most famous intellectual set-piece of the tenth century, and we will come to it.

The libraries next door

What Baghdad lacked, its neighbours had. Three of them, in three languages.

The Syriac world had been reading Aristotle for two hundred years. Sergius of Reshʿaynā, a priest and physician who died in 536 and who had studied under Ammonius in the Neoplatonic school at Alexandria, wrote the first exposition of Aristotle in a Semitic language and translated some thirty treatises of Galen into Syriac, along with the pseudo-Aristotelian De mundo and Alexander of Aphrodisias's On the Principles of the All ³. The monastic school of Qenneshre on the Euphrates taught the Alexandrian logical curriculum straight through the Persian wars, the Arab conquest and the Umayyad century. Sebastian Brock's formula for what happened in those centuries is the standard one: Syriac Christian attitudes to Greek learning moved "from antagonism to assimilation" ⁴. By 750 there was a functioning Semitic-language technical vocabulary for logic, and men who used it professionally.

The Sasanian world had a medical school and hospital tradition at Jundīshāpūr in Khūzistān whose physicians — the Bukhtīshūʿ family above all — moved to Baghdad and became the caliphs' doctors, a dynasty of court physicians lasting eight generations ⁸. It also had a substantial Middle Persian secular literature, of which almost nothing now survives in Middle Persian.

The Greek world, in the person of the Byzantine state, had the books but had largely stopped reading the technical ones. This is the point at which the popular story usually inserts a compliment to Islam and an insult to Byzantium; the truth is more mundane. Greek secular science in the eighth century was not destroyed, it was dormant, kept in monastic and imperial libraries by people whose intellectual energy was going into theology and law.

The Umayyads, who ruled from Damascus until 750 and who governed a population that was still overwhelmingly Christian and Greek-speaking, did almost nothing with any of this. A few pseudo-Aristotelian letters on kingship were put into Arabic for the chancery of Hishām ibn ʿAbd al-Malik (r. 724–743) by his secretary Sālim Abū l-ʿAlāʾ, and administrative documents were converted from Greek to Arabic as the tax registers were arabised. That is close to the whole of it. There is a persistent story that the Umayyad prince Khālid ibn Yazīd (d. c. 704) commissioned the first translations of Greek alchemy; Gutas treats it as a later fabrication, and it does not survive contact with the sources ¹. The point is worth making because it disposes of the idea that proximity to Greek books was sufficient. The Umayyads had proximity. They had no reason.

Paper, and the collapse in the price of a book

None of this would have produced a translation movement without a change in the cost of writing. Papyrus was an Egyptian monopoly and fragile; parchment required a herd. Chinese papermaking technique reached the eastern Islamic lands in the eighth century and papermakers were at work in Baghdad by the 790s, under the patronage of the Barmakid vizieral family. Jonathan Bloom's demonstration is that paper did not merely make books cheaper; it made a certain kind of book — long, prose, technical, heavily revised, produced in multiple copies — economically thinkable for the first time in the Near East ⁵. One of the earliest surviving specimens of Arab-made paper is, appropriately, a Greek manuscript of patristic texts copied at Damascus around 800.

A translation movement is an industrial process. It needs a supply of source texts, a supply of trained bilinguals, a supply of paying customers, and a cheap substrate. By about 790 Baghdad had all four.

The transmission: two centuries, three languages, one payroll

The Sasanian argument for reading Greek

The reason the Abbasids wanted Greek books was not that they admired Greece. It was that they had inherited an argument from the dynasty they had displaced.

Sasanian imperial ideology held that the ancient Persian kings had possessed the totality of knowledge; that Alexander, in destroying their empire, had scattered and stolen their books; that the Greeks had merely translated what they had looted; and that a legitimate Iranian empire was therefore obliged to gather the material back. Al-Manṣūr's court astrologer Abū Sahl ibn Nawbakht set this argument out explicitly, and Dimitri Gutas's reconstruction shows the whole early Abbasid translation programme running on it ¹. Translating Aristotle into Arabic was not an act of cultural humility. It was, in the terms the court used, repatriation.

Al-Maʾmūn (r. 813–833) added a second layer, aimed at Constantinople. The Byzantines, on this account, had abandoned their own philosophical inheritance out of Christian hostility to it, and had thereby forfeited it; the caliphate was the rightful heir of Greek reason. Ibn al-Nadīm preserves the story of al-Maʾmūn dreaming of Aristotle, who tells him that reason and revelation do not conflict, whereupon the caliph sends to the Byzantine emperor for books ⁷. Read as biography, it is a pleasant anecdote. Read as what it is — an official narrative circulated to justify expenditure — it is a policy document.

The House of Wisdom, and what it was not

Here the record parts company with the popular account, and with the one-line description in this atlas's own editorial calendar.

Almost every general history says that the Greek works were translated at the Bayt al-Ḥikma, the House of Wisdom, a great research academy founded by al-Maʾmūn. The evidence does not support it. Gutas and Kevin van Bladel, in the Encyclopaedia of Islam THREE, reconstruct the bayt al-ḥikma as a palace library, most probably established as an administrative bureau under al-Manṣūr on the Sasanian model, whose function was to house translations from Middle Persian of Sasanian history and culture. Gutas's sentence is blunt: it "was certainly not a center for the translation of Greek works into Arabic," and the Graeco-Arabic translation movement was essentially unrelated to its activities ¹⁶.

The counter-position deserves a fair hearing, because it is not frivolous. Marie-Geneviève Balty-Guesdon's 1992 study in Arabica assembles the evidence for a real institution with a real staff and argues for a broader remit, including astronomical work under al-Maʾmūn ⁹. What is agreed on both sides is narrow and worth stating precisely: an office called the bayt al-ḥikma existed; it had a director (Salm, who appears in the Fihrist as ṣāḥib bayt al-ḥikma and who was sent on al-Maʾmūn's book-buying embassy to Byzantium); it held manuscripts. What is not attested anywhere is a body of salaried Greek-to-Arabic translators working inside it.

The correction matters for more than pedantry. The academy story makes the translation movement a state project with a building and an opening date — something a caliph decided. What actually happened was a market: two centuries of individual commissions, from named patrons, for named books, at negotiated prices, sustained across at least eight caliphates and through civil war, precisely because it was not dependent on any single institution or any single ruler's enthusiasm. That is why it lasted.

The workshops

The work was done by families and circles, most of them Christian, one of them pagan.

Aristotle, Organon in Arabic. Paris, Bibliothèque nationale de France, MS Arabe 2346, copied by al-Ḥasan ibn Suwār, c. 1000 CE. Digitised by the BnF (Gallica). Public domain via Wikimedia Commons. · Public domain

Four things are visible in that table which the academy story hides.

The first is that the Christian share is overwhelming. Ḥunayn ibn Isḥāq (808–873), the movement's central figure, was a Nestorian from al-Ḥīra, of the Arab Christian tribe of the ʿIbād. His son Isḥāq, his nephew Ḥubaysh ibn al-Ḥasan and his collaborator ʿĪsā ibn Yaḥyā were Nestorians. Qusṭā ibn Lūqā, who came up to Baghdad from Baalbek, was a Melkite. Abū Bishr Mattā was a Nestorian trained at the monastery of Dayr Qunnā. Yaḥyā ibn ʿAdī was a Jacobite. Ibn Zurʿa, Ibn al-Samḥ, Ibn al-Ṭayyib — all Christians. Thābit ibn Qurra (c. 826–901), who did the hardest mathematics, was a Ḥarrānian Sabian: a member of a Hellenised pagan star-cult that survived inside the caliphate by identifying itself with the "Sabians" named in the Qurʾān and so claiming protected status ¹⁰.

The second is that a great deal of the Arabic came through Syriac rather than from Greek. Abū Bishr Mattā, who gave Arabic its Posterior Analytics and its Poetics, did not read Greek at all; he worked from Isḥāq ibn Ḥunayn's Syriac ¹¹. Ḥunayn himself routinely made a Syriac version for his Christian colleagues and an Arabic one for his Muslim patrons, and his Risāla — an annotated bibliography of his own working life, addressed to the patron ʿAlī ibn Yaḥyā ibn al-Munajjim around 856 — records which of the 129 Galenic titles he had done in which language, which he had revised, and whose earlier attempts he thought were bad ¹².

The third is method. Ḥunayn's practice, as he describes it, was to collect multiple Greek manuscripts, collate them into a corrected exemplar, and only then translate — sense for sense, not word for word. It is recognisably textual criticism, three centuries before the phrase existed, and it is the reason the Arabic Galen is in places a better witness to the Greek than the surviving Greek.

The fourth is succession. The Parisian manuscript Arabe 2346, copied by al-Ḥasan ibn Suwār around the year 1000, carries three colophons on each text. The first names the original translator — Isḥāq ibn Ḥunayn for the Categories. The second records that Ibn Suwār copied from Yaḥyā ibn ʿAdī's autograph, which Yaḥyā had himself collated against Isḥāq's original. The third records a further collation against Ibn Zurʿa's copy. Henri Hugonnard-Roche calls the codex the memory of four centuries of philological labour, holding « la mémoire de cette assimilation en éléments juxtaposés » — the memory of that assimilation in juxtaposed layers ¹³. It is a single book that shows you the whole chain of custody.

What a translation cost

The money is not a detail; it is the mechanism.

• The Banū Mūsā brothers — Muḥammad, Aḥmad and al-Ḥasan, sons of a converted highway robber turned astronomer, and the richest private patrons of science in the ninth century — are reported to have kept translators on 500 dinars a month.
• Ibn Abī Uṣaybiʿa, writing in Damascus in the thirteenth century, reports that Ḥunayn was paid the weight of his finished books in gold. He adds a detail no legend-maker would invent: Ḥunayn's scribe al-Azraq therefore wrote in a broad hand, with thick strokes and widely separated lines, on sheets two or three times the standard thickness. Ibn Abī Uṣaybiʿa says he had handled such volumes himself and owned some ¹⁴.
• Al-Maʾmūn is reported to have sent an embassy to the Byzantine emperor — including al-Ḥajjāj ibn Maṭar, Yūḥannā ibn al-Biṭrīq and Salm of the bayt al-ḥikma — to select old books from the Byzantine libraries and bring them back ⁷.

The gold-by-weight story is a thirteenth-century report about a ninth-century practice and should be held at arm's length as a quantity. As a description of the trade it is exactly right. Translation in ninth-century Baghdad was a lucrative profession with a competitive market, quality disputes, plagiarism accusations and price negotiation. That, and not a royal foundation, is what sustained it for two hundred years.

What changed and what was replaced

An Arabic that could say "hypotenuse"

The first and least visible change is lexical. There was no Arabic technical vocabulary for any of this material; the translators built one, and it is still in use.

They had three options and used all three. They calqued: falsafa for φιλοσοφία, hayūlā for ὕλη (matter), usṭuqus for στοιχεῖον (element), muwsīqā for μουσική. They extended native roots: ʿaql for intellect, jawhar for substance, ʿaraḍ for accident, mawḍūʿ for subject, maḥmūl for predicate, qiyās — a term already load-bearing in Islamic legal reasoning, where it meant analogy — for the syllogism. And they invented, most consequentially in mathematics: al-jabr, the restoration of a subtracted quantity to both sides of an equation, from the title of al-Khwārizmī's Kitāb al-jabr wa-l-muqābala of the 820s.

The vocabulary then travelled onward, and a surprising amount of it is in the reader's mouth already:

• algebra (al-jabr), algorithm (from al-Khwārizmī's own name), alkali (al-qily), alcohol (al-kuḥl), alchemy, cipher and zero (both from ṣifr)
• zenith (samt al-raʾs), nadir (naẓīr), azimuth (al-sumūt)
• Aldebaran (al-dabarān), Altair (al-ṭāʾir), Betelgeuse (a mangling of yad al-jawzāʾ), Rigel (rijl), Vega (al-nasr al-wāqiʿ), Deneb (dhanab)
• Almagest — Ptolemy's Mathematical Syntaxis is known to the whole world by an Arabic corruption, al-majisṭī, of a Greek superlative the Greeks themselves did not use as a title

That last item is a small monument. The most influential astronomical work ever written carries, in every European language, the fingerprints of the Baghdad translators ¹⁵.

The lexical work was not mechanical, and the translators knew they were legislating. Ḥunayn's generation argued about whether qiyās could bear the weight of the Aristotelian syllogism when it already meant legal analogy, and about whether to transliterate a Greek term or find an Arabic root for it. Gerhard Endress's survey of the scientific literature traces the standardisation across two centuries: an unstable vocabulary in the al-Kindī circle of the 840s, a largely settled one by the time of the Baghdad Aristotelians a hundred years later ²⁷. What was being built was not a glossary but the capacity of a language to carry an argument it had never had to carry.

Falsafa, and the sciences that outgrew their sources

What the translations produced was not a commentary tradition. It was a research programme that started correcting its sources within two generations.

• Al-Kindī (d. c. 870), the first philosopher to write in Arabic, argued explicitly that truth has no nationality and should be accepted from whatever people supplied it — a position that had to be argued because it was not obvious to his readers ¹⁶.
• Al-Khwārizmī (d. c. 850) wrote a treatise on solving equations that is neither Greek nor Indian in method, and a set of astronomical tables that fused Indian, Persian and Ptolemaic material.
• Thābit ibn Qurra produced work on the composition of ratios and on statics that goes past Euclid and Archimedes rather than glossing them.
• Al-Rāzī (d. c. 925) wrote a book titled Doubts about Galen. The title is the argument.
• Ibn al-Haytham (d. c. 1040) demolished the Greek consensus that vision works by rays emitted from the eye — held by Euclid, Ptolemy and Galen alike — and replaced it with an intromission theory grounded in controlled experiment.
• Ibn Sīnā (d. 1037) wrote the Qānūn fī l-ṭibb, which was still a set text in European medical faculties six hundred years later.

The collision with the indigenous sciences was not always peaceful, and it produced one superb scene. In or about 932, at the majlis of the vizier Ibn al-Furāt in Baghdad, the Christian logician Abū Bishr Mattā ibn Yūnus debated the Arabic grammarian Abū Saʿīd al-Sīrāfī on whether Greek logic was a universal instrument of thought or merely the grammar of one particular dead language, dressed up. Al-Sīrāfī, by the account preserved in al-Tawḥīdī, won on points before a partisan crowd, and made a case that has never entirely gone away. It is the earliest recorded argument about whether a borrowed intellectual system is a universal or a parochialism wearing a crown, and the borrowing culture staged it itself.

Institutions built out of translated books

Vocabulary and argument are the visible changes. The structural change is that Baghdad built permanent institutions whose entire reason for existing was the imported material, and those institutions outlasted the translating.

• The hospital. A bīmāristān was established at Baghdad under Hārūn al-Rashīd, staffed on the Jundīshāpūr model and run by physicians of the Bukhtīshūʿ family, with teaching, wards, and a pharmacy. The medical theory it taught was Galen's, in Ḥunayn's Arabic. Pormann and Savage-Smith trace the institution from there across the Islamic world and, by the twelfth century, into Latin Europe ²⁸.
• The observatory. Al-Maʾmūn founded observing programmes in the Shammāsiyya quarter of Baghdad from about 828 and on Mount Qāsiyūn above Damascus a few years later. Their product, the Zīj al-mumtaḥan — the "verified tables," compiled under Yaḥyā ibn Abī Manṣūr — is exactly what the name says: Ptolemy's parameters checked against fresh observation, and corrected where they failed. Translating the Almagest was not the end of the project. It was the instrument specification.
• The measured earth. Around 830 al-Maʾmūn's astronomers went out onto the plain of Sinjār in northern Mesopotamia and measured the length of one degree of the meridian by direct survey, obtaining a figure in the region of fifty-six and two-thirds Arabic miles. This is an experiment that Ptolemy's text makes possible and does not perform.
• The bibliography. In 987 a Baghdad bookseller named Ibn al-Nadīm compiled the Fihrist, ten discourses attempting to catalogue every book written in Arabic, with what he could learn of each author, including systematic notice of what had been translated, by whom, from which language, and how well ⁷. A universal bibliography is only conceivable in a city with a book market large enough to need one, and it remains the single most important source for everything in this record.

Falsafa's opposition, and what the receiving culture did with it

A borrowed system generates its own critics, and the Arabic ones were formidable. Al-Sīrāfī's case against logic in 932 was the opening move; the sustained critique came from theologians who had read the material properly. Al-Ghazālī (d. 1111) wrote a scrupulous summary of the philosophers' positions before writing the refutation, on the principle that you may not attack what you cannot restate. Ibn Rushd (d. 1198) then wrote a refutation of the refutation. This is not a culture rejecting a foreign import; it is a culture that has absorbed an import thoroughly enough to conduct a first-rate argument about it in its own terms — the same argument Latin Europe would begin having, in translated form, a century later.

The library that outlived its originals

Some of the Greek corpus now exists only because Baghdad paid for it.

Nathan Sidoli and J. L. Berggren's edition of the Arabic Planisphere makes the general point concrete: for a work like this, the Arabic is not a translation of the text, it is the text, and the editorial problems are those of a primary witness ¹⁷. The same is true of the last three books of the Conics, where Roshdi Rashed's critical edition works from the Arabic because there is nothing else to work from ¹⁸.

Toledo, three hundred years late

Europe enters this story at the end, and as a customer.

From about 1120 Latin scholars went to Spain and Sicily to get at the Arabic material. Adelard of Bath translated al-Ḥajjāj's Euclid. Gerard of Cremona (c. 1114–1187) went to Toledo specifically because he wanted the Almagest and there was no Latin one; he stayed for the rest of his life, finished the Almagest in 1175, and left about seventy translations attributed to him and his team, including Ibn Sīnā's Qānūn as the Canon medicinae ¹⁹. The Aristotle that Thomas Aquinas argued with in Paris in the 1260s had, in many cases, come from Greek into Syriac in a Mesopotamian monastery, from Syriac into Arabic in Baghdad, from Arabic into Latin in Toledo, and only then into a Dominican's hands.

Charles Burnett's study of the Toledo programme shows it was not a scatter of individual enthusiasms but a coherent curriculum-building exercise: the translators worked through the Arabic corpus in something close to the order in which a student would need it, logic and mathematics before astronomy, astronomy before medicine ¹⁹. They were, in other words, doing in Latin in the twelfth century what Baghdad had done in Arabic in the ninth, and using Baghdad's reading order to do it.

The "gave it back to Europe" framing gets the direction right and everything else wrong. Nobody in Baghdad was keeping anything warm for Paris. Europe went and fetched it, three centuries later, because by then it wanted what Baghdad had already built. And a good deal of what it fetched was not Greek at all: the Canon medicinae that governed European medical teaching until the seventeenth century is Ibn Sīnā's, written in Persia around 1025, and the algebra that Latin Europe learned was al-Khwārizmī's.

What the cost was

Nobody was conquered into this transmission. No population was displaced by it, no city sacked for it, no one enslaved to produce it. The participants were among the best-paid intellectuals who have ever lived. An honest cost accounting for this record therefore has to be careful about causal distance — and it still finds a bill.

The miḥna, 833–851

In Rabīʿ I of the year 218 — the spring of 833, four months before his death — al-Maʾmūn wrote to Isḥāq ibn Ibrāhīm, his governor in Baghdad, ordering that judges and traditionists be interrogated on a point of doctrine: whether the Qurʾān was created in time or uncreated and eternal. Those who would not affirm createdness were to be removed from office, and worse. The policy is known as the miḥna, the testing. It ran under al-Muʿtaṣim and al-Wāthiq and was abandoned by al-Mutawakkil from about 848, formally ended by 851 ²⁹²⁰.

• Aḥmad ibn Ḥanbal was interrogated, flogged and imprisoned under al-Muʿtaṣim for refusing to affirm the created Qurʾān. He did not recant; the refusal made him the most authoritative figure in Sunni traditionism and gave his name to a legal school that still exists ²¹.
• Aḥmad ibn Naṣr al-Khuzāʿī was arrested in 846 over a planned rising in Baghdad, sent to al-Wāthiq at Samarra, questioned on the same doctrine, and executed by the caliph in person. His head was displayed in Baghdad and his body left on a gibbet at Samarra for years. He is the one unambiguous corpse in this record.
• Careers were ended, judgeships vacated, testimony refused, and a generation of scholars learned that a doctrinal position could be an offence.

How much of this belongs on this record's bill is a real question and should not be fudged. The rationalist theology that the miḥna enforced — the Muʿtazilī doctrine of a created Qurʾān — was not lifted out of Aristotle; it had its own indigenous development, and John Nawas's review of the competing explanations for al-Maʾmūn's decision concludes that the standard accounts, including the Muʿtazilī one, do not survive scrutiny as sole causes ²⁰. What can be said with confidence is narrower and still substantial: the translation programme and the miḥna came out of the same court, the same reign, the same conviction that demonstrative reason was the caliph's to define and impose, and the same year. Gutas's reading of al-Maʾmūn's philhellenism as an instrument of caliphal authority makes them two expressions of one policy ¹. The atlas bills the miḥna to this transmission not as its product but as its shadow: the cost of a state that has decided reason is on its side.

Ḥunayn ibn Isḥāq, flogged and dispossessed

The Syriac Galen Palimpsest, folios 010r–011v. Imaging hosted by the University of Pennsylvania Libraries (digitalgalen.net). CC BY-SA 3.0 via Wikimedia Commons. · CC BY-SA 3.0

The single clearest measure of what a Christian translator's standing was actually worth is what happened to the best of them.

Around 856, rival physicians at al-Mutawakkil's court moved against Ḥunayn ibn Isḥāq. Knowing he held iconoclast views, they induced him to spit on an icon in the caliph's presence. The manoeuvre worked perfectly: it enraged the Nestorian catholicos, who excommunicated him, and it enraged al-Mutawakkil, who dismissed him from his post, had him flogged and imprisoned, and confiscated his entire estate — including the library, the working collection of collated Greek manuscripts that was the physical basis of his method ²². He was released after six months, restored to office, and held it until he died in 873.

He was the most valuable intellectual asset in the caliphate and it took one court intrigue to destroy him for half a year. Nothing in the record suggests anyone was punished for it.

The languages that were used up

The deepest cost is the one with no victim to name, and it fell on the intermediaries.

Syriac had been the scientific language of the Near East for three centuries. It had the vocabulary, the schools, the personnel, and a two-hundred-year head start on Arabic in reading Aristotle. Its scholars made themselves indispensable to the Arabic project — and within about a century of that project's peak, Syriac had ceased to be a language anyone wrote science in. The translators' own grandchildren wrote in Arabic. Syriac survived as a liturgical and communal language and had a literary revival in the twelfth and thirteenth centuries, but as a vehicle of natural philosophy and medicine it was finished, and it was finished by its own success.

The emblem is a physical object. The Syriac Galen Palimpsest is a manuscript of Melkite liturgical hymns written in about the eleventh century on parchment that had been scraped clean. The text that was scraped off was a ninth-century copy of Sergius of Reshʿaynā's Syriac translation of Galen's On Simple Drugs. Somebody in the eleventh century looked at Galen in Syriac, decided the parchment was worth more than the text, and washed it. Books VI to IX of that undertext have been recovered since 2009 by multispectral imaging, in a project led by Peter Pormann and Siam Bhayro across 231 folios; parts of the Greek original are lost, so what is being read off the erasure is in places the only witness there is ²³.

Middle Persian fared worse. The Sasanian secular literature was translated into Arabic and then, in its original language, largely ceased to be copied. The Pahlavi Kalīla wa-Dimna — itself a translation from Sanskrit, made for Khusraw I in the sixth century — is simply gone; every version in the world, in every language including modern Persian, descends from Ibn al-Muqaffaʿ's Arabic of the 750s. The Arabic translation movement is one of the great acts of preservation in history and one of the great acts of replacement, and it is the same act.

There is a third erasure, less material and still running. The men in the table above were Christians and a pagan. Their work is filed, in nearly every popular account and a good many scholarly ones, under "Islamic science" — a phrase that is accurate about the civilisation and misleading about the personnel. The communities they came from — Church of the East, Syriac Orthodox, Melkite, Mandaean, and the vanished Sabians of Ḥarrān, whose star-cult disappears from the record after the eleventh century — are now, in Iraq and Syria, a small fraction of what they were.

What this record declines to bill

Cost honesty requires saying what is not on the account.

1. The Mongol sack of Baghdad in 1258. A catastrophe by any measure, and unrelated in causation to a translation programme that had finished two and a half centuries earlier. The frequently repeated image of the Tigris running black with the ink of drowned books is a literary topos found in later sources, not a datum, and this record does not use it.
2. The Zanj rebellion, 869–883. East African agricultural slaves in the salt flats of lower Iraq rose against the same caliphate, in the same decades, twenty days' march from the libraries, and held their own state for fourteen years before it was destroyed. The surplus that paid 500 dinars a month to a translator came out of that economy. It is named here because a record about the intellectual glory of ninth-century Iraq that does not mention it is lying by omission — and it is not billed here, because the causal link runs through the whole Abbasid fiscal system and not through the translations.
3. The end of the movement. There is a popular story in which Arabic science was strangled by religious reaction, usually with al-Ghazālī cast as the executioner. It does not fit the evidence. Translation slowed after about 1000 because it had run out of Greek books worth translating and because Arabic science had moved past its sources — Ibn al-Haytham on optics and Ibn Sīnā on medicine are not glosses on Greek authorities, they are replacements for them. Gutas argues this directly; George Saliba, who disagrees with Gutas about a great deal else, argues it too ²⁴. The best scientific work in Arabic in astronomy came after the translations ended, not before.
4. The "Alexandria to Baghdad" migration. Max Meyerhof's celebrated 1930 paper traced a physical relocation of the Alexandrian philosophical school to Antioch, then Ḥarrān, then Baghdad, on the strength of a report attributed to al-Fārābī ²⁵. It is a lovely story and it is not history; Gutas has taken the narrative apart as a piece of tenth-century Baghdad self-fashioning, and John Watt's work on the Syriac Aristotelians shows what actually carried the curriculum — monasteries in northern Mesopotamia, working continuously, no migration required ²⁶¹¹. What the record loses in romance it gains in accuracy: the transmission was not a caravan of exiled professors. It was two hundred years of paid, ordinary, extremely skilled work.