A history of the Septuagint, the Greek Old Testament that the New Testament was written out of

Translated in Alexandria by men whose names are lost, sealed with a curse against altering a word of it, and then corrected for a thousand years by everybody who touched it.

A history in 15 parts · about 72 minutes · c. 280 BCE – today

Cairo · A leaf that was scraped clean and used again

A Page Somebody Wrote Over

In Cambridge University Library there is a single leaf of parchment, twenty-seven and a half centimeters by twenty-two, with a hole in it. It came out of the storeroom of a synagogue in Fustat, old Cairo, where for something like a thousand years the community put paper it could not throw away because the name of God might be on it. What is written on the leaf is Hebrew liturgical poetry, copied in about the tenth century. What is written under it is Greek, and it is the oldest surviving page of the greatest work of biblical scholarship the ancient world produced.

Somebody in the tenth century took a sixth- or seventh-century book, scraped the ink off it, cut it down and wrote something else on the skin. Under the poems, at an angle, are the remains of six columns: the Hebrew Bible in Hebrew letters, the Hebrew Bible spelled out in Greek letters so that a Greek could pronounce it, and then four different Greek translations set side by side so that a reader could see, line by line, where they disagreed. The passage is Psalm 22. Four of the columns can still be read in part. The first cannot be read at all, and whether it was ever there is a question two scholars have written articles about.

There is one more thing on the leaf worth stopping at. Where the Greek columns come to the name of God, they do not translate it. They print four letters that in Greek read pipi — a word that means nothing in any language. What happened is that a Greek scribe copying a Hebrew text came to four Hebrew letters he could not read, looked at their shapes, and drew the nearest Greek letters he could find. Jerome complains about the practice in a letter. The most careful piece of textual scholarship in Christian antiquity contains, at its center, a scribe copying a name he did not understand.

That leaf is a fair picture of the subject of this page. The Septuagint is the Greek translation of the Hebrew scriptures that Greek-speaking Jews made in Egypt beginning around the third century before Christ. It is the Old Testament that the writers of the New Testament quote. It is the Old Testament of the Greek and Slavonic churches to this day. Its book names are the ones on the spine of the Bible in your hand: Genesis, Exodus, Leviticus, Numbers, Deuteronomy, none of which are Hebrew words. For four hundred years it was the most widely read Bible on earth in any language.

And there is no such object. There is no manuscript of it from before the Christian era, no agreement about what books belong in it, no critical edition of it that is finished, and no ancient account of how it came to exist that is not a forgery. What survives is copies of revisions of revisions, and a name — the Seventy — taken from a story that says seventy-two.

This page is about how that happened. A good deal of what is commonly said about the Septuagint turns out, when you follow the figures back, to be a nineteenth-century estimate, a misread sentence, or a number that entered circulation because somebody multiplied one manuscript by five. Saying which is which is most of what a page like this can usefully do.

c. 130 BCE · The document that gave the book its name

The Book That Is Not a Letter

Everything anybody claims to know about the origin of the Greek Bible comes, in the end, from one text. It is conventionally called the Letter of Aristeas, and it is not a letter. There is no address, no greeting and no farewell; it opens by announcing that the story of an embassy is worth telling, and it closes by comparing itself favorably with the books of the romancers. Sylvie Honigman, who wrote the first book-length study of it, calls it the Book of Aristeas for that reason.

The story it tells is this. Demetrius of Phalerum, keeper of the royal library at Alexandria, tells King Ptolemy the library holds more than two hundred thousand books and that he hopes to reach half a million, and that the laws of the Jews ought to be in it. The king agrees, and first frees the Jewish slaves in Egypt at twenty drachmae a head. Letters go to Jerusalem; the high priest Eleazar sends back six elders from each of the twelve tribes, seventy-two men, with a Hebrew scroll written in gold. There is a week of banquets at which the king asks each of the seventy-two a question and is delighted with all seventy-two answers. Then Demetrius takes them along the causeway to the island of Pharos, to a house by the sea, where they work until the ninth hour each day, wash their hands in the sea, and produce a translation. It takes seventy-two days.

Then the part that matters most, and that almost nobody quotes. The finished book is read out, not to the king but to the assembled Jews of Alexandria. They ask for a copy. And the priests, the elders of the translators, and the leaders of the community stand up and resolve that since the work has been well and piously done, it should remain exactly as it is and no revision should ever be made — and they call down a curse on anyone who alters it by adding anything, transposing anything, or leaving anything out.

That is the founding document of the Greek Bible: a formal anathema against changing a word of it. What follows in this page is a thousand years of people changing it.

The demolition, and the man who did it at twenty-five

Nobody who works on this believes the frame. The reasons are worth having in order, because they are a small classic of how a document is broken open.

The fatal one is Demetrius of Phalerum. He was the great man of Ptolemy I’s court, and on the only ancient account we have — Diogenes Laertius, drawing on Hermippus — he advised the old king to leave the throne to his sons by Eurydice. The king left it to his son by Berenice instead. That son, Ptolemy II Philadelphus, took the throne about 282, had Demetrius detained in the countryside pending a decision about him, and Demetrius died there within a few years of a snakebite received, of all things, in his sleep. The man the Letter makes Philadelphus’s librarian and confidant was under house arrest at the moment Philadelphus became king, for backing the wrong heir. He was also, on the evidence of the surviving list of librarians, never the librarian at all. Henry St John Thackeray’s footnote on the point is one flat sentence: he was never the royal librarian.

The rest accumulates. The author twice forgets his mask, writing of “these kings” as a man looking back over a dynasty rather than serving inside one. He has Ptolemy celebrating the anniversary of a naval victory over Antigonus, which happened decades after Demetrius was dead. He puts the philosopher Menedemus of Eretria at the Alexandrian court, where no other source has ever placed him. He warns the reader off tampering with scripture by citing the historian Theopompus and the tragedian Theodectes, both of whom died long before Philadelphus reigned. And two of his phrases — a formula of polite deference and a court title in the plural — are not attested in the Egyptian papyri before 163 and 148 BCE respectively, which is the kind of evidence a date can actually be built on. Above all there is the courtier who cannot stop being Jewish: “we too, O king, though we address Him by other names, as Zeus and Dis.”

The dates proposed for the real composition run from about 200 BCE to the first century, and the spread is not narrowing. Schürer read the political geography of its Jerusalem section as pre-198. Elias Bickerman, working chancery formulae and greeting conventions against dated papyri in 1930, put it in the second half of the second century; Moses Hadas followed him and settled on shortly after 132. Honigman argues for shortly before 145. Thackeray himself said 120 to 80, Wendland later still. Bar-Kochva narrowed it to 116 or 118 down to 113, and was told by a reviewer that his method suffered from an excessive dose of optimism. What everybody agrees on is that it was written by an Alexandrian Jew, in Greek, at least a century after the events it describes.

The demolition is usually credited to Juan Luis Vives, in his 1522 commentary on Augustine. That credit is misplaced, and the correction is instructive: read the Latin and Vives is passing on Jerome’s objections and adding nothing of his own. The real work was done by Humphrey Hody, who published Contra historiam Aristeae at Oxford in 1684, at twenty-five, and dedicated it to the Bishop of Winchester as the first-fruits of his youth. Isaac Vossius, who had built a career on the Septuagint’s authority, replied in an appendix to a book on Pomponius Mela in 1686; the Dictionary of National Biography calls the reply angry and scurrilous. Hody answered him in the notes to a reissue nearly twenty years later, and he has held the field ever since.

The list of seventy-two names has seventy-one on it

Here is the detail that ought to be better known. The Letter, having said that six elders came from each of the twelve tribes, prints their names, tribe by tribe, and concludes: in all, seventy-two persons.

The manuscripts do not contain seventy-two names. Thackeray’s note on the passage records that one tribe’s sixth name has simply disappeared, and that a second name in the same group is corrupt. The roster is filled out to a round seventy-two only by importing what is missing from the Syriac version of a treatise by Epiphanius, a Christian bishop of the fourth century, writing five hundred years after the fact in a third language. The document that supplies the number is, as transmitted, one man short of it.

The transmission is worth a sentence of its own, because it explains why. Every surviving manuscript of Aristeas — about twenty-one of them, the oldest eleventh-century, the latest copied out by hand in 1586 — carries the text bound in front of a Christian commentary chain on the first eight books of the Bible. There is no Jewish manuscript tradition of it at all. A Jewish book about a Jewish translation survives only because Christian scribes used it as a preface.

What was it for? Victor Tcherikover argued in 1958, against most of his predecessors, that it is not propaganda aimed at Greeks at all but an Alexandrian Jew talking to other Alexandrian Jews about who they were. Honigman reads it as a charter myth, narrated on the model of the Homeric scholarship being done down the road at the Museum: the point is not that the Law was rendered into Greek but that it was given an authoritative text, the way Alexandrian editors gave Homer one. Maren Niehoff sharpens that: the curse at the end is not piety but polemic, aimed at Jewish colleagues who were applying Aristarchus’s critical marks — the obelus and the asterisk — to scripture. If she is right, the author was trying to ban a specific practice, and he failed so completely that within four centuries a Christian scholar would apply exactly those marks to this book.

And Swete made the oldest and simplest point a hundred and twenty-five years ago. In the story, the translation is read out to the Jewish community and approved by them, and only then presented to the king. An author inventing a royal commission does not stage it that way. He is reproducing, without noticing, what actually happened: the book was made by Greek-speaking Jews for their own use, and the king is a frame around it.

Before 205 BCE · What the evidence will actually carry

A Mistake That Proves the Date

Strip the king out and you still have a Greek Pentateuch that has to be dated. There are three independent lines of evidence, and the first of them is a mistranslation.

At Genesis 35:16 the Hebrew has an ordinary noun, a rough measure of distance — some way still to go. The Greek translator did not translate it. He heard a place name, spelled it out in Greek letters as Chabratha, and then repeated the error at Genesis 48:7, where the text speaks of the horse-course of Chabratha in the land of Canaan. It is the sort of thing a competent man does at the end of a long afternoon.

Now go to a Jewish writer called Demetrius, distinguished from the courtier by the label the Chronographer, who wrote a work on the kings of Judaea. He is quoted by Clement of Alexandria, who preserves a chronology running down to Ptolemy IV, which places him under Ptolemy IV, whose reign ended in 205 BCE. And when Demetrius comes to Jacob’s journey he writes that he came to Chaphratha and thence to Ephratha, which is Bethlehem. He is not reading Hebrew. He is reading a Greek book in which the translator’s slip has already happened, and treating it as a place he can put on a route.

He does more than that. He works out from the figures that Dinah was sixteen years and four months old when Shechem took her, and Simeon twenty-one years and four months when he avenged her. He reconciles Moses’s Midianite wife with the Ethiopian woman of Numbers by genealogy. That is not how a man treats a draft. It is how he treats a scripture with a chronology in it that can be audited. And when you add up his own totals for the years from Adam, they come out on the Greek numbers and not the Hebrew ones, which is a second, independent proof that he had the Greek.

The caution is that we have none of this at first hand. Demetrius was excerpted by Alexander Polyhistor in the first century BCE, and Polyhistor was excerpted by Eusebius in the fourth century CE, and Clement quotes him separately; every figure has been through at least two copying traditions and some of them are visibly corrupt. What survives the corruption is the shape: a Greek Genesis and Exodus existed, and were being studied as authoritative, before 205 BCE.

The second line is the papyri, and they are thinner than their reputation. The famous one is Rylands Greek 458, in Manchester, acquired in 1917, published by C. H. Roberts in 1936 and dated by him to the middle of the second century BCE on the shape of the handwriting. Before the Dead Sea discoveries it was the oldest fragment of any Bible in any language. It is eight scraps, lettered a to h, carrying pieces of Deuteronomy 23 to 28 — perhaps fifteen partial verses in all, and the smallest of them a few letters. Its date is one scholar’s judgment about a bookhand, a method with a real margin that nothing found since has tightened. Papyrus Fouad 266, from the Fayyum, is first-century BCE and is really three rolls rather than one. Then Qumran and Naḥal Ḥever supply a handful more.

Put honestly: no surviving copy of the Greek Pentateuch is older than roughly a century after the date usually given for the translation. The papyri prove the translation is older than the second century BCE. They do not prove it is third-century. What proves that is Demetrius’s slip about Chabratha.

Why it was made: five answers, and no agreement

The third line of evidence is the language itself, and it points at Egypt. The Greek of the Pentateuch is the ordinary middle-register speech of the Ptolemaic papyri, not a sacred dialect — that is John Lee’s demonstration, and it has held for forty years. The translator of Genesis writes not Goshen but the land of Gesem of Arabia, Arabia being the name of the Ptolemaic administrative district in the eastern Delta; he has quietly located a patriarchal place on the map of his own tax office. He carries Egyptian words straight through: the reed-grass of Pharaoh’s dream, the dry measure called the oiphi, the papyrus basket Moses was put in. He calls Potiphar’s steward the one over the house, which is how the third-century papyri title a real official. The one caution, and it is Sofía Torallas Tovar’s, is that administrative Greek survives almost only from Egypt, because Egypt is where papyrus survives; calling a word Egyptian because it turns up in Egyptian documents is close to an argument from silence.

As for why anyone did it, there are five answers in the field and no consensus, which is worth stating plainly rather than picking one.

The royal account — the king wanted the book — has few but real defenders; Nina Collins argued in 2000 for a completed translation in 281 BCE got up as a spectacle for the start of Philadelphus’s reign, and Tessa Rajak grants the Letter more historical weight than most, calling it a historical myth that combines memory with story. The liturgical answer — Greek-speaking Jews could no longer follow the Hebrew read in synagogue — is the oldest and most intuitive and remains the default in general reference works. The educational answer is Sebastian Brock’s and Albert Pietersma’s: a translation can bring the original to the reader or the reader to the original, and this one does the second, functioning as a crib beside a Hebrew text rather than as a book to be read on its own. That is the interlinear paradigm, and the whole New English Translation of the Septuagint is built on it; it is also contested, and a 2023 paper argued from the actual syntax of Genesis and Exodus that the Greek is more idiomatic than the model needs it to be. The legal answer is Joseph Mélèze Modrzejewski’s: the Greek Torah was the civic law of the Jews inside the plural legal order of Ptolemaic Egypt, and there are twenty-odd papyri from a Jewish political community at Herakleopolis, dated between 144 and 133 BCE, in which petitioners argue from Deuteronomy about divorce. Peter Altmann has since gone through the same documents and reported largely negative findings. And the fifth answer, Rajak’s own, is that the translation was a tool for holding a group together in a Greek city — a claim to standing rather than an act of assimilation.

Notice what none of them is. None of them is a committee from Jerusalem.

132 BCE onward · The translator who apologized for his grandfather

Not One Book, and Not One Translator

In the 130s BCE a young Jew arrived in Egypt from Jerusalem carrying a book his grandfather had written, and set about putting it into Greek. When he finished he wrote a preface, and it is the single most valuable document in this whole subject, because it is a translator in the second century BCE talking, in the first person, about what translation costs.

He asks the reader’s indulgence in the places where the work seems flat despite great care taken over the wording, because what was originally said in Hebrew does not have the same force when it is turned into another language. The Greek word is isodynamei — it is not of equal power. And then he says the thing that matters: not only in this case, but with the Law itself and the Prophets and the rest of the books, the difference is not small when they are read in the original.

Two facts fall out of one sentence. First, by the 130s BCE there were Greek versions of the Law, the Prophets and other books already in circulation in Egypt. Second, the man who had them thought they were not equal to the Hebrew, and said so in the introduction to his own translation. The Alexandrian Jewish community’s own view of its Greek Bible, a century after the Letter of Aristeas puts a curse on altering it, is that it loses something.

He dates himself, too: he came to Egypt in the thirty-eighth year of King Euergetes, which has to be Ptolemy VIII, and works out to 132 BCE.

So the Greek Old Testament arrived in pieces over something like four hundred years, and the pieces were made by people who had never met and did not agree about what they were doing. Greek Esther ends with a colophon, a dated delivery note, naming the men who brought it: Dositheus, who said he was a priest and a Levite, and his son Ptolemy, and one Lysimachus son of Ptolemy, of Jerusalem, who made the translation. It is one of the very few Greek biblical books with a translator’s name attached, and even it cannot be dated closer than three candidate years a generation apart, because the regnal formula fits all three.

The shelf, from one end to the other

Thackeray printed the first real map of this in 1909, classifying the books by the kind of Greek they are written in, and the range he found runs from competent literary prose to something that is barely Greek at all.

At one end sits Job. The Old Greek of Job is about a sixth shorter than the Hebrew, and its translator writes real literary Greek with Homeric vocabulary in it. He was not working from a shorter Hebrew; the modern judgment, in the New English Translation of the Septuagint’s own introduction to the book, is that a man with so free and paraphrastic an approach is the likeliest explanation of his own omissions. He simply left out what he could not use, and the amount he left out rises steadily as the book goes on — four percent through chapter 15, and better than a third of chapters 32 to 37 — which is the profile of a translator getting tired of a long poem.

At the other end sits Ecclesiastes, whose translator renders the Hebrew word that marks a direct object — a particle with no meaning at all, a grammatical signpost — by the Greek preposition for “with,” some thirty-five times, from the first chapter to the last. It is not merely awkward. It is contrary to Greek grammar. Peter Gentry, who edited the book, calls the result at times almost incomprehensible to a native speaker. Whether it is in fact the work of Aquila, a second-century Jewish reviser we will come to, has been argued since 1904 and the current verdict is that it cannot be settled.

Between them: a Psalter so consistent in its equivalences that it carries Hebrew pronoun gender straight through into Greek where it makes no sense, and yet coins words like earthborn and moonless and throw-to-the-crows that appear nowhere else. A Song of Songs that invents Greek verbs when it is stuck and transliterates when invention fails. An Isaiah that quietly updates the geography and the politics to the translator’s own century. A Proverbs with material in it that is not in the Hebrew at all. A Jeremiah literal enough to be nonsense in places, and possibly the work of two hands.

They are bound in one volume, under one name, and the specialists have spent a century trying to get out of the name. The Society for Old Testament Study states that strictly speaking Septuagint refers only to the Pentateuch, and that the proper term for the earliest Greek rendering of any other book is Old Greek. The New English Translation of the Septuagint concedes in its own front matter that Old Greek would undoubtedly be the more suitable term and that it has bowed to the weight of tradition. That is a translation project admitting, in its introduction, that its title is wrong.

The ancient Jewish sources agree with the specialists. Aristeas, Philo, Josephus and the Talmud all take the Seventy’s work to be the Law and nothing else. On Edmon Gallagher’s reckoning, the first person to call a book outside the Torah part of the Septuagint is Justin Martyr, in the middle of the second century after Christ — a hundred years after Paul was dead. And the rabbinic tract Soferim preserves a quite different tradition in which the translators were five, one to a book, which Emanuel Tov calls more realistic than seventy-two.

Genesis, Samuel, Jeremiah, Daniel · Where the two Bibles part

The Places Where the Books Do Not Match

None of this would matter very much if the Greek said the same thing as the Hebrew. In most places it does. In some places it does not, and the differences are not small.

Jeremiah is the largest. The Greek book is substantially shorter than the Hebrew one, and the oracles against the foreign nations, which stand near the end in the Hebrew as chapters 46 to 51, sit in the middle of the Greek, dropped in after 25:13, and in a different order among themselves. More than twenty passages of at least a full verse have no Greek counterpart at all, and two of them are large: the thirteen verses of Jeremiah 33:14 to 26, promising a branch to David and an everlasting Levitical priesthood, and the ten verses of chapter 39 covering Zedekiah’s flight and Jeremiah’s release.

The figure everyone quotes for the size of the difference is that the Greek is about an eighth shorter, some 2,700 words. That number is on hundreds of pages and I could not find a source for it on any of them. What the specialists actually publish is a moving target: the original count, Y.-J. Min’s dissertation of 1977, gave a seventh; Tov has published a seventh, then some fifteen percent, then a sixth, in three different places; Georg Fischer says a sixth. The likeliest explanation of the popular figure is that an eighth and a seventh are the same count measured against two different books — 2,700 words is about an eighth of the Hebrew and about a seventh of what is left after you remove them. That is arithmetic, not testimony, and the underlying word count for Hebrew Jeremiah is itself a number nobody seems to have made.

What changed the argument about Jeremiah was Qumran. Two fragmentary Hebrew scrolls, 4QJeremiah-b and 4QJeremiah-d, are said to attest a short Hebrew edition — a Hebrew Jeremiah that looks like the Greek one. If that holds, then the Greek translator was not abbreviating; he was translating a different book, and the Hebrew we have is the expanded one. Most of the field, Tov at its head, holds that it does. Fischer does not, and his objection deserves to be stated because it is the sort that gets left out: the decisive verses are not actually on the surviving skin. The case rests on calculating how much writing would have fitted on the missing part of a line, and Fischer says the calculation does not work.

Goliath is the famous one. The Hebrew of 1 Samuel 17:4 makes him six cubits and a span. The Greek makes him four cubits and a span, and so does Josephus, and so — this is the part that matters — does a Hebrew scroll from Qumran, 4QSamuel-a, where the passage is not fragmentary but clear. There is no Hebrew manuscript of 1 Samuel earlier than the tenth century of our era; the six-cubit reading is nonetheless old, since Symmachus, Origen’s fifth column and Jerome all have it by about the year 200.

The conversion is where the popular accounts go wrong, in both directions. You will see the two Goliaths given as nine feet nine and six feet nine. Both figures assume an eighteen-inch cubit, which is a convention rather than a measurement. The one place a biblical cubit can be checked against stone is the Siloam tunnel, whose inscription gives its length as twelve hundred cubits and which measures 533 meters, yielding a cubit of about seventeen and a half inches. Run the readings across the plausible range and the short Goliath stands somewhere between six foot two and seven foot nine, and the tall one between eight foot eleven and eleven foot two. Clyde Billington, arguing for the Hebrew reading, ends up with a shorter giant than the figure his opponents quote against him. Both of the round numbers in circulation are artifacts of a ruler nobody has.

The chronology of Genesis is the difference with the largest consequences and the worst arithmetic attached to it. In the ages of the patriarchs before the Flood, the Greek gives six of them — Adam, Seth, Enosh, Kenan, Mahalalel and Enoch — a hundred years more before the birth of the next man, and takes the hundred off the remainder so that the lifespans come out the same. Do the sums from creation to the Flood and the Hebrew gives 1,656 years, the Samaritan Pentateuch 1,307, and the Greek 2,242. After the Flood the Greek adds another hundred years apiece to six more men, fifty to a seventh, and inserts an extra generation, Kainan, who is in the Greek and in Luke’s genealogy of Jesus and in no Hebrew manuscript of Genesis at all.

The number in circulation is that the Greek runs 1,386 years longer than the Hebrew from Adam to Abraham. It is not a reading of any text. To get it you have to make two contested choices at once: take Methuselah’s age as 187 in the Greek, against the printed critical text, which has 167, and place Abraham’s birth in Terah’s hundred and thirtieth year rather than his seventieth. Make the same assumption about Terah on both sides and read the Göttingen Greek as it stands, and the gap is 1,366. The famous figure is 1,366 plus a twenty-year variant that is the whole substance of the dispute.

And the twenty years matter, because of what they do to Methuselah. On the Hebrew figures he dies in the year of the Flood, exactly. On the Samaritan figures he dies in the year of the Flood, exactly. On the Greek figures he dies fourteen years after it, having not been on the ark. Andrew Steinmann calls that the strongest single argument against the Greek chronology, and the counter-argument is that many Greek manuscripts read 187 like the Hebrew — to which Steinmann answers that those are late copies carrying readings back out of Origen’s work three centuries later. It is genuinely open, and the twenty years are why.

The seventy who went down into Egypt are seventy-five in the Greek, at Genesis 46:27 and Exodus 1:5, because the Greek list adds five of Joseph’s descendants. Stephen quotes seventy-five in the seventh chapter of Acts, with no variant anywhere in the manuscript tradition; Philo has seventy-five as well. And two Hebrew scrolls from Qumran read seventy-five, which means this is not a Greek invention. Which way it went is contested by the same volume’s own editors: James Davila, publishing one of the fragments, held that seventy was original and both the Greek and the scrolls inflated it; Itamar Kislev argues that seventy-five stood in Genesis and Exodus and the Masoretic scribes cut it to agree with Deuteronomy. The delicious part is that some Greek manuscripts of Deuteronomy read seventy-five, and some read seventy. Everybody was fixing the same contradiction, and they were fixing it in opposite directions.

Daniel is the strangest case of all, because there the church changed Bibles and did not record why. Daniel exists in three editions: the Hebrew and Aramaic, the Old Greek, and the version that goes under Theodotion’s name. In chapters 4 to 6 the Old Greek diverges wildly — two of those chapters are longer than the Hebrew, one substantially shorter — and less than a fifth of Theodotion’s vocabulary there agrees with it. The Old Greek Daniel was the one the New Testament writers knew. And then it disappeared from use. Jerome is the primary witness, and his candor is worth having: the churches of the Lord Savior, he writes, do not read the prophet Daniel according to the Seventy but use Theodotion’s edition instead, and he does not know why this happened. The Old Greek very nearly died. Until the eighteenth century it was known only from quotations, and a copy of it made before Origen did not surface until 1931.

c. 40 – 392 CE · How the miracle was assembled

Thirty-Six Cells, Built One Century at a Time

Ask anyone who knows the story what is remarkable about the Septuagint and you will be told that seventy-two translators were shut up separately and produced word-for-word identical Greek. It is the detail that survives when everything else has gone.

It is also the one thing the Letter of Aristeas explicitly denies. Its translators are in one house on the shore. They work until the ninth hour. And the operative sentence says they arrived at their text by comparing each other’s work and making it agree, whatever they agreed on being copied out under Demetrius’s direction. That is committee procedure. It is the opposite of the miracle.

The miracle can be watched arriving, author by author, over about three hundred years, and the sequence is the most useful single thing in this part of the subject.

Philo, writing in Alexandria around 40 CE, adds the identical wording and not the cells. He says the translators sat in seclusion — the phrase most probably describes their collective isolation on the island — and prophesied as men inspired, not one saying one thing and another another, but every one of them using the same nouns and the same verbs, as though an unseen prompter were speaking into each man’s ear. He calls them not translators but hierophants and prophets. It is worth knowing that Philo names no number at all, neither seventy nor seventy-two, and gives no count of days; every reference book that credits him with them is reporting Aristeas under his name. It is also worth knowing that Philo had little or no Hebrew, so the exact correspondence he asserts is something he was not in a position to check.

Philo does supply one thing nobody else does, and it is the best scene in the whole story. He reports an annual festival on Pharos, at which not only Jews but a great many people of other nations sail across, out of reverence for the place where the light of interpretation first shone; and some pitch tents on the shore and some simply lie down on the sand in the open air, and feast with their families and friends, thinking the beach a finer place to be that day than the furnishings of the palace. No other ancient writer mentions this festival. It rests on one witness — though a witness whose readers in Alexandria could have contradicted him.

Justin Martyr, about 155, has no cells, no miracle and no number, and has Ptolemy send for the books to Herod, king of the Jews, which is out by roughly two and a half centuries.

Irenaeus, about 180, is where the separation appears. Ptolemy — and for Irenaeus it is Ptolemy the son of Lagus, the wrong Ptolemy — wishing to test them individually and fearing that they might agree a story between them, separated them from each other and commanded them all to write the same translation. When they were read out, all seventy of them had the very same words and the very same names from beginning to end.

Then it acquires architecture. Pseudo-Justin puts the cells seven stadia from the city, one to each translator, and adds a claim of autopsy: having been in Alexandria ourselves, we saw the remains of the little huts on Pharos still preserved, and heard the story from the inhabitants. A miracle with a tourist attraction attached to it.

Epiphanius, about 392, gives the full apparatus. Thirty-six cells, two men to a cell, no opening in the walls but skylights in the roof; thirty-six little boats to carry them to dinner with the king; thirty-six separate bedchambers so that they could not confer at night; one book to each pair; and at the collation, thirty-six readers sitting below with thirty-six copies, finding no disagreement — wherever one pair had added a word, all of them had added the same word.

Read that sequence again and notice what is happening. Each retelling adds a control condition. Over three centuries a story about a committee is progressively rebuilt into a laboratory experiment, with isolation, replication, blind conditions and a review panel. Nobody is lying, exactly; each writer is supplying the safeguard the previous version obviously lacked.

Augustine accepted it, and we will come to what he built on it. Jerome did not, and his sentence is the best in the literature: I do not know who was the first lying author to construct the seventy cells at Alexandria, in which they were divided and yet all wrote the same words, whereas Aristeas and, long after him, Josephus report nothing of the kind, but say they were gathered in one hall and conferred together, and did not prophesy. And then the distinction the whole quarrel turns on: it is one thing to be a prophet, another to be an interpreter.

The story the rabbis tell, which is the same story inside out

There is a Jewish version of all this and it runs the other way entirely.

Tractate Soferim: it once happened that five elders wrote the Torah in Greek for King Ptolemy, and that day was as hard for Israel as the day on which the golden calf was made. A calendar of fast days records that on the eighth of Tevet the Law was written in Greek in the days of King Ptolemy, and darkness came upon the world for three days.

Then, two paragraphs later, the same tract tells the seventy-two-elders story, with the cells, and adds something the Christian versions do not have: a list of passages that all the translators, independently, altered in the same way. The count is given as thirteen in one place and fifteen in the Babylonian Talmud, and fourteen are actually listed. Most of them are places where a literal Greek rendering would have embarrassed a Jew in front of a Greek king — the plural verb at the beginning of Genesis, the animals of Exodus 4:20, and, most famously, the hare of the dietary laws, because Ptolemy’s wife was supposedly named Hare and the elders wrote “the small-footed animal” instead.

Almost none of it survives checking. Thackeray reports that only about four of the listed passages appear in that form in our Greek manuscripts, two more in African versions, and the rest may simply be Hebrew variants; Zacharias Frankel suggested they were originally cited as readings in the Hebrew scroll sent to Egypt and not as Greek renderings at all. The hare has been taken apart by David Wasserstein: Ptolemy’s wives were Arsinoe and his mother was Berenice, and none of them was called Hare. What is behind it is that the Greek word for hare is all but identical to Lagos, the name of the founder of the dynasty — and the Greek Bible in fact avoids the word, writing rough-foot instead. The rabbinic explanation is a later rationalization of a real translation decision, and a good one.

So the same tradition, within two paragraphs, calls the translation a catastrophe on the scale of the golden calf and a miracle in which seventy-two men in seventy-two rooms made the same fifteen corrections. It is as coherent as the Christian tradition, which within twenty years had Augustine defending the cells and Jerome calling their inventor a liar.

1st century BCE – 2nd century CE · The correcting begins, and it is Jewish

Two Words No Greek Would Write

Around 1951 or 1952, Bedouin selling antiquities in East Jerusalem, then under Jordanian administration, brought in fragments of a Greek scroll of the Minor Prophets and said they had come from Wadi Seiyal. That was untrue, and it was untrue for a commercial reason: Seiyal lay on the Jordanian side of the line, and material from the Israeli side could not be sold there. The Palestine Archaeological Museum bought the fragments as the Seiyal collection over the following two years. Because they had been dug out clandestinely, whatever archaeological context they had was gone.

In the spring of 1961 a Hebrew University expedition under Yohanan Aharoni went along the southern side of Naḥal Ḥever, working caves eighty meters below the clifftop, reached by rope. In the one called the Cave of Horror they found more fragments of the same scroll. That is how the lie was caught: the roll had been in Israel all along. The definitive edition took another twenty-nine years to appear.

Long before that, in 1953, a young Dominican named Dominique Barthélemy had looked at the fragments and noticed something about the Greek. Hebrew has a small word meaning “also.” Any competent translator renders it with the ordinary Greek word for “and.” The man who wrote this scroll rendered it with two words, kai ge, “and indeed” — every time, in defiance of idiom, because he wanted one piece of Greek for each piece of Hebrew. It is not a translation choice. It is a policy, and it produces Greek nobody would write.

Barthélemy published a note on it that year and the full study, Les Devanciers d’Aquila, in 1963; essays routinely date the discovery to the book. What he had found was not a translation but a revision: somebody had taken an existing Greek text of the Twelve Prophets and gone through it dragging it toward the Hebrew, word by word. The same fingerprints turn up in identifiable stretches of the Greek books of Samuel and Kings, which is why the last third of Reigns reads differently from the first — a fact Thackeray had noticed in 1907 and explained wrongly, supposing that the Old Greek had never translated those chapters out of embarrassment at the fall of the monarchy. Barthélemy showed it was not a fresh translation but a coat of revision laid over one.

Two things follow, and both are large.

The first is chronological, and it reorganized the field. The scroll is pre-Christian — the palaeographer Peter Parsons put it at the end of the first century BCE, and its date is disputed by others but not by centuries. So the systematic correction of the Greek Bible toward the Hebrew was under way before there was a church to be provoked by it. The popular story is that Jews abandoned the Septuagint because Christians took it over. The physical evidence says the pull toward the Hebrew had begun for reasons entirely internal to Judaism, and that Christian appropriation intensified and moralized a process already running. It also means the curse in Aristeas had failed within a century or two of being written, and failed in Jewish hands.

The second is that a revision presupposes something revised. That single implication settled a methodological quarrel we will come back to at the end of this page.

Aquila, Symmachus, and a man quoted before he was born

Three named revisers stand behind the Greek columns Origen would later set out, and all three are less solid than their entries in the reference books.

Aquila is dated to about 128 or 130. Follow that date back and you arrive at one source: Epiphanius of Salamis, writing in 392, more than two and a half centuries later, who says Aquila came from Sinope in Pontus, was related to Hadrian by marriage, was put in charge of the rebuilding of Jerusalem, became a Christian, was expelled for refusing to give up astrology, went over to Judaism out of spite, and made his translation in Hadrian’s twelfth year — four hundred and thirty years and four months, he says, after the Seventy. Epiphanius also says, in as many words, that Aquila translated in order to make the testimonies to Christ come out otherwise. There is no independent attestation of the date. Irenaeus knows Aquila as a Jewish translator and gives no date; the rabbinic sources that mention an Aquila the proselyte give none either. The date every handbook prints is borrowed from one late enemy.

What Aquila did is not in doubt, because enough of it survives to be mocked. He rendered every Hebrew element with a Greek one, whether or not the result was Greek. His signature is the little Hebrew particle marking a direct object, which he renders with the preposition “with,” so that the first verse of Genesis becomes, in his hands, God created with the heaven and with the earth. That is not stupidity. Rabbi Akiva had taught that the particle could be read as the identically spelled word for “with,” and Aquila is following a rabbinic reading rule into Greek. Whether he was as closely tied to Akiva’s school as Barthélemy argued has been doubted since — Lester Grabbe and James Barr both pushed back — and Jenny Labendz has argued that the rabbis who praise particular renderings of his are showing individual appreciation rather than institutional endorsement, and that no rabbinic text uses his translation as a whole.

Symmachus wrote the best Greek of the three and is the least known as a person; Eusebius calls him an Ebionite, and the Talmudic Sumkhos has been proposed as the same man, and nothing can be settled. Jerome used him heavily.

Theodotion is the hardest case, because readings attributed to him appear in books written before he lived. He is placed around 180. His Daniel is quoted in the New Testament; his readings show up in Justin. The field’s answer is that Theodotion inherited and continued a revising tradition already a century or two old — the tradition the Naḥal Ḥever scroll belongs to — which is why the whole stratum is now written with a hyphen, kaige-Theodotion, and why a man of the second century after Christ can be quoted by an author of the first.

235 – 245 CE · Caesarea, and the largest book ever made to that date

Six Columns, and the Marks That Fell Off Them

By the third century a Christian arguing with a Jew about scripture had a practical problem: they were not reading the same book. Origen’s answer was to build an instrument that would let anyone see exactly where and how far the texts differed.

He assembled it at Caesarea, on the coast of Palestine, most probably between 235 and 245. Eusebius, our only near-contemporary witness, says he learned Hebrew, got hold of the Hebrew scriptures the Jews were using, hunted down Greek versions beyond the three famous ones, discovered some that had been hidden from remote times in out-of-the-way corners — one at Nicopolis near Actium and one in a jar at Jericho — collected them all, divided them into sections, and set them opposite one another with the Hebrew itself.

Notice how little that is. Eusebius never numbers the columns and never lists their order. The familiar six — the Hebrew in Hebrew letters, the Hebrew spelled out in Greek letters, Aquila, Symmachus, the Septuagint, Theodotion — are a reconstruction, secured from Epiphanius, from Jerome, and from the surviving palimpsests. It is not seriously in doubt, but it is not in our earliest source either. And whether the first column ever existed at all is a live question: no fragment of it survives anywhere, and Pierre Nautin concluded there had never been one. He has been answered on measurement — the Cambridge genizah leaf still shows the fold at the center of its opening, and the surviving columns can be measured, and the arithmetic wants a column before Aquila’s.

The second column is the one that matters most to scholarship now, and for an unexpected reason. It transliterated the Hebrew into Greek letters, which as written Greek means nothing but which, read aloud, reproduces Hebrew. Because Greek writes vowels and unpointed Hebrew does not, it is the largest single body of evidence for how Hebrew was pronounced — roughly seven hundred years before the Masoretes of Tiberias fixed the vowel points, and centuries before any pointed manuscript. It may not even be Origen’s own work. Greek pronunciation was shifting in his day in ways the column does not reflect, which is an argument that he inherited it.

Every number attached to it, and where each one comes from

The Hexapla is famous for having been enormous. Four figures circulate for how enormous, and only one of them has any arithmetic behind it.

Fifty volumes. This is Bernard de Montfaucon, in 1713, and he flags it himself as a guess: he uses the word he should think. His reasoning is entirely impressionistic — Greek copyists had to reproduce Hebrew they could not read, then Hebrew in Greek letters, then four to seven versions in exact horizontal register, in a large script, with wide margins for notes; therefore, scarcely in fifty volumes of ample bulk could the whole thing have been contained. No count, no measurement.

Six thousand five hundred pages. This is Swete in 1900, and to his credit he shows his working. Codex Vaticanus has 759 leaves; a complete Old Testament in that format would run about 650; the Hexapla’s bulk he judges nearly five times as great; 650 times five is 3,250 leaves, or 6,500 pages. That is one arbitrary multiplier applied to one manuscript’s leaf count. Six columns ought to give six, not five, and the discount is not defended. Swete’s own footnote then undercuts him: Nestle had pointed out that in the Cairo palimpsest even two-letter words take a line to themselves, and if the Hexapla was written like that the estimate is far too low.

Forty codices. This is Anthony Grafton and Megan Williams in 2006, and it is the only estimate in the tradition built on an external price series rather than a feeling. They invite the reader to imagine a cupboard loaded with forty matching codices of about eight hundred pages each, and they cost the parchment and copying of one edition at something like 150,000 denarii, working from the rates in Diocletian’s price edict. The edict is about sixty years later than the Hexapla, which is the obvious caveat.

Twenty-eight years. This is the figure you will see for how long the work took, and it is a misread sentence. It comes from Epiphanius, and in the standard English of the Syriac it runs: twenty-eight years, as the story goes, he devoted to ascetic practices, and he set forth the Scripture, placing the six columns and the two columns of the Hebrew side by side. The twenty-eight years belong to the asceticism. They are hearsay even in Epiphanius, who says so, and they sit in a sentence his own translator’s editor flags as confused about where Origen was living. There is no reliable figure for how long the Hexapla took.

The marks, and the disaster

Origen’s real invention was not the columns but two small marks, and he took them from the Alexandrian editors of Homer — from Aristarchus, who had used the obelus to flag lines of the Iliad he thought spurious.

An obelus, in Origen’s fifth column, marks something present in the Greek and absent from the Hebrew. An asterisk marks something present in the Hebrew and absent from the Greek, which he has supplied from one of the other versions, usually Theodotion. A metobelus shows where the marked passage ends. He explains himself in the commentary on Matthew, and the explanation is careful: he did not dare remove the doubtful matter, only mark it; and he added what was missing under an asterisk so that it should be clear the addition was his. Epiphanius’s gloss on the asterisk is lovely — the words it marks are fixed in the Hebrew scriptures like stars in the firmament, and have been hidden by the Seventy’s translation as stars are hidden by cloud.

Then Pamphilus and Eusebius published the fifth column on its own, with the signs, and it was so successful that within Palestine it displaced the ordinary text altogether.

And the signs came off. They are marginal marks; they mean nothing to a copyist in a hurry; they cost ink and attention; and a scribe who drops them has produced a perfectly clean, perfectly readable Greek Bible in which passages imported from Theodotion are now indistinguishable from the original translation, and passages with no Hebrew behind them carry no warning that they have none. That is what textual critics mean by hexaplaric contamination, and it is in the bloodstream of the Greek manuscript tradition permanently. The most scrupulous editorial project of the ancient church did more damage to the text it loved than any careless scribe ever managed, because its scrupulousness lived in a layer that copying discards first.

S. R. Driver’s verdict on the underlying assumption is the one to keep. Origen assumed that the original Septuagint was the one that agreed most closely with the Hebrew text as he knew it — which was, Driver says, a step in the wrong direction.

What happened to it

Jerome saw something at Caesarea in the fourth century and mentions it three times, once describing an old Hexaplaric psalter corrected in Origen’s own hand. A colophon in a sixth-century manuscript of the Pauline epistles says the book was collated against the copy in the library of holy Pamphilus at Caesarea. After that, silence.

Frederick Field, who spent his life collecting the surviving readings, would not name a year: not long after, he wrote, but in what year and by what misfortune is uncertain. Montfaucon offered two guesses, the Persian capture of Caesarea in 614 or the Arab conquest a generation later. Swete supplied a date — in 638 Caesarea fell into the hands of the Saracens, and from that time the Library was heard of no more — and because Swete’s was the standard English handbook, 638 is the year you will read everywhere.

It is worth being clear about what that date is. There is no evidence connecting the Hexapla’s destruction to any event. It is an inference from a sixth-century colophon plus a general sense that something bad must have happened. And it is very probably the wrong year even for the conquest: the historians of the city put the fall of Caesarea Maritima in 640 or 641, after operations beginning in December 640. Nobody who works on Caesarea says 638. Meanwhile Peter Gentry holds that the Hexapla may not have survived 350, and Alison Salvesen’s formulation is the honest one: the turmoil of the seventh century is probably responsible, and we have no evidence of what happened to it.

What is left is four or five objects. The Cambridge genizah leaf this page opened on. A palimpsest in Milan, the lower text of which is a Hexaplaric Psalter in five columns, spotted by Giovanni Mercati in 1896 and not published for another sixty-two years. Two further scraps. A Syriac translation of Origen’s revised column made in Egypt about 616, which is why we can still read a good deal of it at one remove. And an enormous scholarly literature reconstructing, from quotations, a book that nobody has seen for something over a thousand years.

The first century · What the Greek Bible made possible

A Body You Prepared for Me

The New Testament is written in Greek and quotes the Old Testament in Greek. That much is not in dispute. Almost everything else usually said about it is a number, and the numbers are a swamp.

How many Old Testament quotations are there in the New? Swete counted about 160 formal quotations in 1900. A Victorian one-volume dictionary said 283, and that figure is still what the English Wikipedia article gives, which means it is the first number most readers will meet. Roger Nicole, defining his terms carefully, counted 295 references occupying 352 verses. C. H. Toy counted 613. Dittmar counted 1,640. Hühn counted 4,105. The spread is not a matter of a few percent; it is a factor of twenty-five, and it is not carelessness. Nicole counts introductory formulae; Hühn counts reminiscence. Revelation contains no formal quotation at all and is nevertheless soaked in the language of the Greek Old Testament, which no counting method can turn into an integer.

Then the figure that actually gets used: that the New Testament cites the Septuagint in 340 places and the Hebrew in only 33. It is on the apologetics sites and in the study Bible introductions, credited to Archer and Chirichigno’s survey of 1983. Go to their table and you find six categories, not two. Their category A — 268 entries — is the cases where the Hebrew and the Greek say the same thing, so the quotation cannot be assigned to either. Category B, where the New Testament follows the Greek against the Hebrew, has 50. Category D, close to the Greek with minor deviations, has 22. Add A to B and D and you get 340. Category C, closer to the Hebrew, has 33.

The most-quoted statistic in the field is manufactured by counting the mute evidence as testimony for one side. Two thirds of it proves nothing at all, because in two thirds of cases the two Bibles agree.

The older pair of numbers is more honest and much older than anyone realizes. Swete quotes a calculation that the New Testament differs from the Masoretic text in 212 citations and departs from the Greek in 185. Those figures are David Turpie’s, made in 1868, on a nineteenth-century critical text and a nineteenth-century understanding of the Greek Bible’s own history. They reappear, unattributed, in a standard modern introduction to the Septuagint, and from there on to the internet. Nobody appears to have redone the count in a hundred and fifty-eight years.

The one modern breakdown worth having is Fernández Marcos’s of Paul: 93 quotations, of which 51 agree with the Greek, 22 of those against the Hebrew, 4 follow the Hebrew against the Greek — and 38 agree with neither. That last number is the honest headline. Paul quoted something, freely, from memory or from a collection of proof texts or from a Greek Bible we no longer possess, and in more than a third of cases it matches nothing we have.

Four places where the argument is the Greek

Statistics aside, there are passages where the point being made in the New Testament cannot be made from the Hebrew at all. Four are worth setting out.

Hebrews 10:5. The writer puts Psalm 40 into the mouth of Christ coming into the world: sacrifice and offering thou wouldest not, but a body hast thou prepared me. The Hebrew of that psalm says ears — ears thou hast dug for me. And the writer then builds the entire argument of the chapter on the word body: the many offerings against the one body, and we are sanctified through the offering of the body of Jesus Christ once for all. Remove the word and the paragraph falls down.

The manuscript position here is the reverse of what you might expect. Nearly every Greek manuscript of the psalm reads “body,” including Vaticanus, Sinaiticus and Alexandrinus; the witnesses for “ears” are Jerome’s Gallican psalter, itself revised toward the Hebrew, one Old Latin witness and two late minuscules. And yet Rahlfs’s critical edition prints “ears,” and the standard English translation of the Greek Psalter follows it. Martin Karrer has argued the edition should be corrected. If he is right, then “a body you prepared for me” was being read in Greek-speaking synagogues for two centuries before anybody applied it to Jesus, and the Greek Bible handed the letter to the Hebrews its central image ready made. If the other camp is right, the change was made inside Greek by accident, or by the writer of Hebrews himself. It is genuinely open.

Acts 15. James, in Jerusalem, settles whether Gentile converts must be circumcised by quoting Amos: that the rest of mankind may seek the Lord, and all the Gentiles upon whom my name is called. The Hebrew of Amos 9:12 says something else entirely — that they may possess the remnant of Edom — a territorial promise about a restored Davidic kingdom taking back a neighbor. The distance between the two is the distance between a nationalist oracle and a charter for the Gentile mission, and it is produced by three very small things: reading one Hebrew verb as another that differs by a single letter, reading Edom as adam, mankind, which differs by one letter more, and taking the particle that marks the object as introducing a subject instead.

Two honest complications. Acts supplies an object — the Lord — that the Greek Amos does not have, so Acts is not quoting the Greek Bible we possess either; and the opening words of the citation are not Amos at all but resemble Jeremiah. Richard Bauckham has argued that James’s exegesis works from the Hebrew as well, since the phrase he actually leans on stands in both. And the first half of the proof text was already being read as a messianic prophecy in Hebrew at Qumran. Only the second half needs the Greek.

Hebrews 1:6 rests on a sentence that is not in the Hebrew Bible. Let all God’s angels worship him is the pivot of the letter’s opening argument that the Son is greater than the angels. It comes from the end of the Song of Moses in Deuteronomy 32, where the Hebrew has four lines and the Greek has eight. A copy of the Song from Qumran has six — longer than the Hebrew we print, shorter than the Greek — and it does contain a line about bowing down, addressed to gods rather than angels. So the sentence has some Hebrew ancestry, but the form the letter quotes is Greek, and the Masoretic text simply does not contain it.

Romans 15:12 closes the argument of the letter with Isaiah: there shall be a root of Jesse, and he that shall rise to reign over the Gentiles; in him shall the Gentiles trust. In the Hebrew the root of Jesse stands as a signal to the peoples, and the nations inquire of him. That the Messiah rules the Gentiles and that the Gentiles hope in him are both Greek. So is the promise Paul uses twice, that whoever believes on him shall not be ashamed, where the Hebrew of Isaiah 28:16 has neither “on him” nor “ashamed.”

Isaiah 7:14, done properly

The most argued-over word in this subject is the one the Greek translator of Isaiah chose for the young woman who would conceive. The Hebrew is almah. The Greek is parthenos. Matthew quotes the Greek. Both sides of the argument overreach, and the evidence that settles the lexical question is in neither of the places people look.

Almah is a rare word, occurring nine times in all its forms. It means a young woman of childbearing age; virginity is not part of its meaning, and Hebrew has a specific word for virgin, which occurs about fifty times, and Isaiah did not use it. The Greek translators render almah with the plain word for young woman four times, and with parthenos only twice, at Genesis 24:43 and here.

But parthenos was not a technical term either, and the proof of that is a verse nobody cites. At Genesis 34:3, in the same Pentateuch translated in Alexandria in the third century BCE, the Greek calls Dinah parthenos twice — in the sentence immediately after the one saying that Shechem lay with her and humbled her. Whatever the word carried, it was not an assertion about anatomy.

Two further points of honesty. The Pentateuch was translated in the third century BCE; Isaiah was not, and is generally placed in the middle of the second, so what parthenos meant to the men who translated Genesis is evidence about them and not necessarily about him. And the man who translated Isaiah is not a mechanical substituter: he loses his bearings on rare words, invents whole clauses when stuck, and at one point inserts a negative that reverses his source. He cannot be assumed to have chosen a standard gloss, and he cannot be made to carry a doctrine on one lexical decision.

What can be said is that the reading is pre-Christian and Jewish, made centuries before anybody had a stake in it, and that Aquila, Symmachus and Theodotion all replaced it with the plain word for young woman in the second century after Christ. That replacement is the first recorded fight between Christians and Jews about a translation, and we have a transcript of it.

The vocabulary · Words a Greek already had, put to a new use

The Words It Handed Over

The larger debt is not any one verse. It is that a body of Greek-speaking Jews, over three centuries, had to decide what Greek word to use for every Hebrew idea in the Bible, and that when Christians began writing in Greek they inherited the decisions whole. The New Testament’s theological vocabulary is, to a degree that is easy to miss, a set of choices made in Alexandria by men whose names are gone.

The figures below were counted, for this page, in the parallel-aligned Hebrew and Greek text maintained by the Tov and Kraft project, which sets each Hebrew word beside the Greek that stands opposite it. The matching is by word stem and it is crude; treat the numbers as accurate to within a few, and as checkable, which is the point of giving them at all.

Covenant. The Hebrew berit occurs 349 times in the aligned text. The Greek that stands opposite it is diatheke 311 times. Ordinary Greek had a perfectly good word for a compact between two parties — syntheke — and in the whole aligned corpus syntheke appears five times, in Isaiah and in Daniel, and not once does it render berit. Diatheke means something different: a unilateral disposition, a settlement made by one party, and in ordinary usage a will. The standard explanation is that a covenant God imposes is not a treaty between equals. That explanation is old and has been argued about; the distribution has not. And the payoff is a paragraph in Hebrews arguing that a diatheke requires the death of the one who made it — an argument that works only because the Greek word also means a will, and that cannot be translated back into Hebrew at all.

Church. Two Hebrew words for a gathering, qahal and edah, run through the Old Testament. In the aligned text the Greek ekklesia stands opposite a form of qahal 97 times out of 106 occurrences, and opposite edah not once; synagoge stands opposite edah 148 times. And there is a geography to it. In Genesis, Exodus, Leviticus and Numbers, qahal is rendered synagoge every time, and the noun ekklesia does not appear at all. From Deuteronomy onward ekklesia takes over — twelve times in that book, twenty-four in 2 Chronicles alone — and then Jeremiah and Ezekiel go back to synagoge.

Which means that the two words the first century would use to distinguish a church from a synagogue had already been pulled apart, book by book, by different translators, for reasons of Greek style rather than theology, two or three centuries before there was anything to distinguish. In the world outside, ekklesia was the voting assembly of a Greek city.

Mercy seat. In the Greek Old Testament hilasterion is a piece of furniture: the gold cover over the ark, nine of its occurrences in the tabernacle chapters of Exodus, plus the ledge of Ezekiel’s altar. When Paul writes that God set forth Christ as a hilasterion, the word’s biblical pedigree is an object in a room. Whether it means the place where sin is dealt with or the means by which God is appeased is the substance of a quarrel between C. H. Dodd and Leon Morris that ran for a generation and put the word “expiation” into the Revised Standard Version at Romans 3:25.

Christ. The Greek adjective christos means smeared, rubbed with oil. In the Greek Old Testament it renders the Hebrew for anointed, of priests and of kings — of Saul asleep in the cave, of David, of Zedekiah. The Greek Bible turned an adjective about oil into a noun about a person. Christianity then turned the noun into a name.

Paradise. The Hebrew of Genesis 2 says garden. The Greek says paradeisos, which is a Persian loanword that Xenophon had used for the walled hunting parks of Persian kings. Eden is a royal game park because a translator reached for the grandest enclosure he knew.

And the counter-example, which is worth as much as the rest. The verb baptizo occurs three times in the whole aligned Greek Old Testament: Naaman dipping himself in the Jordan, one metaphorical use in Isaiah, and one in Sirach. It is not a technical term and it carries no ritual. Christianity did not inherit its word for baptism; it made one. Not everything in the New Testament’s vocabulary came out of this book, and saying so is what makes the case for the words that did.

And the name. The Greek Bible we can read renders the divine name kyrios, Lord, several thousand times, and Romans and Philippians build their highest claims about Jesus on that word. But every complete manuscript that gives us kyrios is a Christian book of the fourth century or later. What survives from Jewish hands before Christianity does something else: a Deuteronomy roll from the Fayyum where the first scribe left blank spaces, marked with a raised dot on each side, into which the four Hebrew letters were later written; the Naḥal Ḥever scroll, with the name in archaic Hebrew script inside the Greek; a Leviticus fragment from Qumran that writes it out in Greek letters as Iao, which is also what Diodorus Siculus says the Jews called him. No surviving pre-Christian Jewish manuscript of the Greek Bible writes kyrios for the name. That is a fact about what has survived and not a proof about what was written — Albert Pietersma and Martin Rösel argue vigorously that kyrios was there from the start and the Hebrew-name copies are archaizing revisions — and it is one of the genuinely open questions in the field.

c. 150 – 405 CE · Two churches, two Old Testaments, one riot

The Congregation That Corrected the Bishop

Once Christians had the Greek Bible and Jews had the Hebrew one, each side reached the same conclusion about the other: that the text had been tampered with.

Justin Martyr made the accusation first and most bluntly, in the Dialogue with Trypho, around the middle of the second century. Your teachers, he says, refuse to admit that the interpretation made by the seventy elders who were with Ptolemy is correct, and they attempt to frame another; and they have taken passages away altogether. Trypho asks him to name some, and he names three: a sentence about the passover attributed to Esdras, a sentence in Jeremiah about the Lord remembering his dead who lay in the graves, and, most famously, three words in a psalm, so that “the Lord hath reigned from the tree” has been reduced to “the Lord hath reigned.”

Check them. The Esdras sentence exists in no Hebrew or Greek Ezra, no Nehemiah, no book of Esdras and no known Jewish text of any kind; it turns up once more, in Lactantius, a century and a half later, and never in a Bible. The Jeremiah sentence about the graves is in no Jeremiah either, and Irenaeus, who quotes it repeatedly, attributes it to Jeremiah in one place, to Isaiah in another, and to unnamed others in a third — which is the signature of a floating proof text rather than a verse. As for “from the tree,” it appears in no Hebrew manuscript at all, and its one appearance in a Greek biblical manuscript is in a sixth-century bilingual psalter where the Greek is ungrammatical in a way that looks like it has been dragged into shape by the Latin column beside it. In Latin it is common: Tertullian has it, Cyprian has it, Augustine has it. It is a Christian addition that Justin mistook for a Jewish deletion.

Justin also names a fourth passage that he says the Jews have cut out, which is Jeremiah 11:19 about the lamb led to the slaughter. That one is still there, in the Hebrew and in the Greek. He half-notices the problem himself and explains that it is still in some synagogue copies, for it is only a short time since they were cut out.

The best sentence in the chapter is the one Justin gives to his opponent. Whether the rulers of the people have erased any portion of the scriptures, Trypho says, God knows; but it seems incredible.

The scholarly explanation, Oskar Skarsaune’s, is that Justin was not comparing manuscripts at all. He had inherited collections of proof texts, already assembled and already Christian, of the kind that physically survive from Qumran, and he was comparing his collection against a synagogue’s Bible and concluding that the Bible had been cut. It is also worth remembering that the whole Dialogue reaches us in one Greek manuscript finished on 11 September 1364, roughly twelve hundred years after it was written, with a hole in it a chapter after the passage that matters. And that the same author, in another book, has Ptolemy of Egypt write to Herod, king of the Jews, for the books — an error of some two hundred and fifty years, made by the man making a forensic claim about deleted verses.

Augustine, Jerome, and a shrub

The other quarrel was inside the church, and it is the one that decided what Western Christendom would read for a thousand years.

Jerome, at Bethlehem, was translating the Old Testament from the Hebrew. Augustine, in Africa, wrote to tell him not to. His reasons are worth separating, because only one of them is theological. He was puzzled that anything could be found in the Hebrew that had escaped so many translators. He wanted, instead, a careful Latin version made from the Greek, because the Latin copies in circulation differed intolerably from one another. He was afraid that if Jerome’s version spread, the Latin churches and the Greek churches would end up reading different Bibles. And he made the practical point that a reader disturbed by an unfamiliar rendering would find it difficult or impossible to get at the Hebrew and check.

Then he told Jerome what had happened at Oea, which is modern Tripoli.

A bishop there had introduced the new version, and came in his reading to a word in the fourth chapter of Jonah that Jerome had rendered differently from what had been familiar to the ears and memory of every worshipper and had been chanted in that church for generations. And such a tumult arose in the congregation, Augustine writes, especially among the Greeks, correcting what was read and denouncing the translation as false, that the bishop was compelled to send for the testimony of the Jews of the town. The Jews said that the Hebrew was correctly rendered by the Greek version and by the Latin made from it. And the bishop was compelled to correct the passage as if it had been falsely translated, since he did not wish to be left without a congregation.

Look at what is physically happening in that room. A congregation that knows the lesson by ear overrules a book. A bishop sends out into the town for a second opinion and gets it from the synagogue. The new codex is emended back to the old reading in front of everybody. And a translation project backed by the most learned man in the Latin church loses to people who simply knew how the sentence went.

Jerome’s reply is characteristically unbothered. The Greek has a gourd, he writes, and Aquila and the others have ivy; the Hebrew has a word which is in Syriac a kind of shrub with large leaves like a vine, growing fast and standing upright on its own stem without needing canes or poles, as gourds and ivy both do. If I had transliterated it nobody would have known what it meant; if I had said gourd I would have said what is not in the Hebrew; so I put ivy, in order not to differ from every other translator. And then the sneer: the Jews of Oea either did not know Hebrew or were lying, to make sport of the gourd-planters.

Neither man was right. The plant is now identified as the castor-oil plant, whose Egyptian and Hebrew names Jerome is in fact reporting accurately. The Greek gourd is wrong and Jerome’s ivy is wrong, and Jerome’s own description of a self-supporting shrub with vine-like leaves is a good description of a castor plant, which is to say he had the right information and printed the wrong word out of deference to the tradition he spent his life attacking. The congregation at Oea rioted in defense of a mistranslation, and won.

The doctrine Augustine built to make the quarrel go away

Augustine’s position hardened as he aged. In the City of God he takes the full miracle — seventy-two men placed apart to test their fidelity, differing in no word that had the same meaning and force, because the one Spirit was in them all — and then does something genuinely audacious with it. The same Spirit that was in the prophets when they spoke, he argues, was in the seventy when they translated. So whatever stands in the Hebrew and not in the Greek, the Spirit chose to say through the prophets; and whatever stands in the Greek and not in the Hebrew, the Spirit chose to say through the translators. One Spirit, speaking twice, in two books, both of them true.

He then takes the hardest case he can find. Jonah gives Nineveh forty days in the Hebrew and three in the Greek, and Augustine declines to choose: the three days signify the resurrection and the forty the ministry between the resurrection and the ascension, and Christ is signified by both. It is the first sustained attempt to detach an authoritative Bible from dependence on the Hebrew altogether, and Jerome’s answer to it is one line long. It is one thing to be a prophet, another to be an interpreter.

The joke at the end of it

Jerome won the argument and lost the book. The Vulgate that Western Christendom actually used is not one man’s work. Wisdom, Sirach, Baruch and the books of Maccabees in it are unrevised Old Latin, translated from the Greek and never touched by him. Most of the New Testament in it is somebody else’s revision.

And the Psalms in it are the Gallican psalter, which Jerome made from the Greek of Origen’s fifth column between 386 and 389. His own psalter translated from the Hebrew was never used in the liturgy at all. Alcuin installed the Gallican under Charlemagne, and from the ninth century it was the psalter of the Divine Office across the West. So the psalms that Latin Christendom sang, chanted, memorized and set to music for well over a thousand years — every Miserere, every De profundis, the whole psalm text underneath Western sacred music from plainchant to the eighteenth century — came through the Greek. By exactly the route Jerome spent his life arguing against.

The last act is modern. In 1945 Pius XII permitted a new Latin psalter made from the Hebrew for the Divine Office. It was not loved and it did not stick, and the Nova Vulgata went back to a Gallican-shaped text corrected against the Hebrew. Sixteen hundred years after Oea, the church tried Jerome’s argument again, and the congregation won again.

The canon · Three great books that do not agree

What Is Actually In It

Every Christian who has wondered why a Catholic Bible has books a Protestant Bible does not has been offered the same explanation, and for about seventy-five years it was the standard one. Greek-speaking Jews in Alexandria, it ran, had a wider canon than Jews in Palestine, including roughly the books we call the Apocrypha; the church, being Greek-speaking, inherited it whole; the Reformers cut back to the Hebrew list.

The hypothesis is not ancient. Albert Sundberg traced it to the confessional quarrels of the sixteenth and seventeenth centuries and followed it forward through a series of Christian controversialists, one of whom invented a Jerusalem council that canonized Tobit and Sirach, and another of whom replaced it with an Alexandrian Sanhedrin. In 1964 Sundberg published The Old Testament of the Early Church and took it apart. There is no primary evidence for a distinct Alexandrian canon; the alleged list is a modern reconstruction read backwards out of fourth-century Christian codices; most of the disputed books were written in Hebrew or Aramaic in Palestine rather than in Greek in Egypt; Josephus says the Jews have twenty-two books and does not qualify it by geography; Philo cites only the Law and the Prophets as scripture; and Egyptian Jews plainly knew the threefold division, since Ben Sira’s grandson names it in Egypt after 132 BCE.

The negative half of that argument has held. The positive half has not, and the reason is a coincidence worth savoring. Sundberg’s replacement account required the Writings to have been closed at Jamnia around 90 CE. In 1964 — the same year — Jack Lewis published the article that dismantled the Council of Jamnia. The book everyone still cites won its argument and lost its foundation within twelve months.

The decisive evidence, though, is not an argument at all. It is three books, and they disagree with each other.

Codex Vaticanus, from about the middle of the fourth century, has a nearly complete Greek Old Testament with no Maccabees in it at all — none of the four — and no Prayer of Manasseh. Codex Sinaiticus, of the same generation, has 1 Maccabees and 4 Maccabees but not 2 or 3, and appends the Epistle of Barnabas and the Shepherd of Hermas after Revelation. Codex Alexandrinus, from the fifth century, has all four books of Maccabees, and fourteen Odes after the Psalms, and an index announcing an appendix containing the Psalms of Solomon — which has been torn off, so that a canon survives there only as an entry in its own table of contents. Even the order of books differs in all three, and none of them follows the order of the modern printed Septuagint.

E. Earle Ellis put it in a sentence that gets quoted everywhere without his name on it: no two Septuagint codices contain the same apocrypha.

And the methodological point behind that is simple and easy to miss. A codex is a bookbinding decision, not a canon list. Athanasius, in the year 367, sorts books into three classes — canonical, to be read, and apocryphal — which means the people who commissioned Sinaiticus could put canonical and merely useful books between the same covers without any confusion at all. The thing scholars have been calling the Septuagint canon since 1935 is, substantially, the table of contents of Rahlfs’s printed edition.

8 February 553 · An emperor legislating on what may be read aloud

Read Aloud, by Order

The usual account of what happened next is that Jews abandoned the Septuagint because Christians had taken it. That is too tidy in three places.

It is wrong about the timing, because the correcting had started before Christianity, as the Naḥal Ḥever scroll shows. It is wrong about the completeness, because Greek Bibles went on being used by Jews for a very long time: two of the oldest objects in the Cairo genizah are palimpsests of Aquila’s Kings and Aquila’s Psalms, from the fifth or early sixth century, with the divine name written in archaic Hebrew letters inside the Greek — which is one of the arguments that they were made by Jews and not by Christians. Nicholas de Lange holds that the genizah evidence shows Jewish use of Greek Bible translations as late as the twelfth century and perhaps beyond. And it is wrong about the agency, because at least once the matter was settled by an emperor.

On 8 February 553 Justinian issued the law known as Novella 146, addressed to his praetorian prefect, on the subject of the Hebrews. Some Jewish congregations, it appears, wanted the scriptures read in synagogue in Greek and others insisted on Hebrew alone, and the dispute had reached the palace. The emperor ruled that those who read in Greek should use the translation of the seventy interpreters, as the most accurate of all — and gave his reason. It happened during the interpretation, he wrote, that the interpreters, being divided into twos and placed in different places, nevertheless all produced the same translation.

That is Epiphanius’s thirty-six cells, invented within Christian polemic in the fourth century, functioning as the stated premise of Roman law in the sixth, in a statute regulating what Jews may read in their own synagogues. He also permitted Aquila’s version, which is a small oddity worth noticing: the emperor licensing the Christian miracle story also licensed the translation the Christians had spent four centuries attacking.

International Septuagint Day falls on 8 February, and it falls there because of this law. In the words of the scholarly society that chose the date, it is the one date we know of from late antiquity on which the Septuagint received special attention. A modern learned society dates its feast from a piece of Byzantine coercive legislation, which is as fair a summary of the whole afterlife as anything on this page.

The other half of the afterlife is that the book never stopped being anybody’s Bible. For the Greek and the Slavonic and the other Orthodox churches the Septuagint is not a historical curiosity or a witness to something better; it is the Old Testament, read in church this morning. And it seeded a whole family of further translations, made from the Greek and not from the Hebrew: Old Latin, Coptic in two dialects, Ethiopic, Armenian, Georgian, Old Church Slavonic, the Syriac of Paul of Tella, Arabic. Those daughters matter for more than piety. A translation of a translation, made in the fourth century, is evidence about what the Greek said in the fourth century, and sometimes a Coptic or an Old Latin reading preserves a Greek text older than any Greek manuscript we have.

1514 – about 2040 · The edition that is not finished

The Septuagint You Can Buy

The Greek Old Testament first went to press at Alcalá de Henares in Spain, in the Complutensian Polyglot, whose Old Testament volumes were finished in July 1517 and which then sat for years waiting on a papal license — or, on another account, waiting out a printing monopoly Erasmus had been granted. Because of the delay, the first complete Greek Bible actually to reach readers was the Aldine, printed at Venice in February 1518 by the family of Aldus Manutius, who had died three years earlier. Its editor claimed many very ancient copies and named none of them.

Then, in 1587, the Sixtine edition, printed at Rome under Cardinal Carafa to help the revisers of the Vulgate, and based on Codex Vaticanus. That book held the field for something like three and a half centuries. Holmes and Parsons reprinted its text and hung their apparatus underneath it. Tischendorf’s seven editions were revisions of it. It was the received text of the Greek Old Testament until Swete’s Cambridge manual in the 1880s and, for the ordinary reader’s shelf, until 1935.

Two of the enterprises in between are worth pausing on for their scale.

John Ernest Grabe’s Oxford edition, begun in 1707, was based on Codex Alexandrinus, and its real innovation was typographic: he printed Origen’s asterisks and obeli, and used different sizes of type, so that a reader could see on the page exactly where the Greek and the Hebrew parted company. It is the first printed Septuagint that tries to make its own editorial surgery visible. He died in 1711 and the last two volumes were finished by other men, the fourth appearing in 1720.

Robert Holmes and James Parsons took twenty-nine years, from 1798 to 1827, and collated 311 Greek manuscripts. Holmes died a quarter of the way through and Parsons finished it. The detail that makes the thing tangible is that not less than a hundred and sixty-four volumes of manuscript collations prepared for that work are still in the Bodleian. Swete’s verdict a century later was that they had accumulated materials on which a critical revision might eventually rest, and that the collations themselves were of uncertain value; every editor since has had to do them again.

The great diplomatic edition of the twentieth century, Brooke and McLean’s larger Cambridge Septuagint, is a lesson in optimism. They took the work up in 1895 hoping to finish the first eight books in five years. The final part of that first volume appeared in 1917. Publication stopped in 1940, having reached Esther. The poetic and prophetic books — Job, Psalms, Proverbs, Isaiah, Jeremiah, Ezekiel and the Twelve — were never done at all.

The book everybody quotes is a pocket edition

Which brings us to the single most common error in writing about this subject.

When somebody says “the Septuagint reads” and gives you Greek, the Greek is almost certainly Rahlfs. Alfred Rahlfs’s two-volume Septuaginta appeared in 1935, the year he died. It is a Handausgabe — a manual edition, semi-critical — and its text was built chiefly from three codices: Vaticanus in the lead, with Sinaiticus and Alexandrinus. For the handful of books where he had already made a fully critical edition of his own, he took that text over, so it is a hybrid. It is complete, it is cheap, it is what every electronic Septuagint is built on, and it is not the critical edition. Robert Hanhart’s revision of 2006 changed the accents and breathings and almost nothing else; anyone claiming a large number of textual changes in it is wrong.

The real critical edition was begun in Göttingen on 1 April 1908. It is an eclectic edition: rather than printing one manuscript and recording the others, the editor reconstructs, reading by reading, the earliest form of the text that can reasonably be recovered — a text that stands before all the recensions. The apparatus is enormous and comes in two layers, one for the Greek witnesses and one for the hexaplaric material.

By 2015 it had published twenty-four volumes, about two thirds of the corpus. It is now somewhere around three quarters done. What is missing is almost exactly the historical books — Joshua, Judges, Samuel, Kings, 1 Chronicles — plus Proverbs, the Song of Songs and 4 Maccabees. And the Psalms, the most-read book in the Bible, are represented in the series by Rahlfs’s own volume of 1931; a genuinely new critical Psalter became a separate long-term project of its own on 1 January 2020, whose first stage was given three and a half years simply to describe the thirteen hundred manuscripts involved, and which is expected to run about twenty-one years.

I could find no published completion date for the Göttingen Septuagint anywhere, and any figure quoted for one should be treated as invented until somebody traces it. What can be said honestly is this: a hundred and eighteen years in, roughly three quarters done, with the hardest books left.

Whether there was ever one book to recover

Underneath the edition is a question about whether the edition is possible.

Paul de Lagarde, who died in 1891, held that there had been one original translation of each book and that it could be recovered, by reconstructing the three fourth-century recensions Jerome names and working back behind them. He got as far as one attempted volume, and handed the program to his pupil Alfred Rahlfs, which is the direct line from him to Göttingen. He was also a vicious antisemite whose writing fed straight into what came after, and there is no reason to leave that out of a paragraph about him.

Paul Kahle argued from 1915 that there had never been one Greek Bible. There had been many independent Greek renderings made for the synagogue, standing to the Hebrew roughly as the Aramaic targums stand to it, and only later reduced to something like a standard. On his account the early papyri are members of a plurality rather than witnesses to a lost archetype.

Lagarde won, decisively, in practice, because the Göttingen edition exists and could not exist on Kahle’s premises. And what settled it was not an argument but a manuscript: Barthélemy’s scroll from the Cave of Horror. A revision presupposes something revised. Kahle’s plurality is now generally read as a plurality of revision rather than a plurality of origin.

He was not simply wrong, though, and the residue is worth stating. Judges circulates in two forms so different that the digital corpora keep them as separate files. Daniel exists in two Greek editions. Joshua and Samuel and Kings show recensional layering deep enough that “the original” is a question rather than an object. Tov’s own formulation is that most scholars think there was one original translation of each book and a minority accept Kahle’s view, and he declines to close the matter.

What the Greek left behind

The Names on Your Own Bible

It is easy to finish a page like this with the impression that the Septuagint is a specialist’s subject, recoverable only in fragments and arguments. So it is worth ending on what is in front of you.

Open an English Bible. The first five books are called Genesis, Exodus, Leviticus, Numbers and Deuteronomy. Not one of those is a Hebrew word. In Hebrew the books are named from their opening phrases — in the beginning, names, and he called, in the wilderness, words. The Greek titles were taken by the translators from sentences inside the books they had just rendered: Genesis from the book of the origin of heaven and earth at Genesis 2:4, so that the book is named after its second chapter; Exodus from the going out of the sons of Israel in chapter 19; Numbers from the counting in chapter 1; Deuteronomy from the instruction that the king shall write himself a copy of this law.

That last one is the celebrated case. The Hebrew tells the king to have a duplicate of the Teaching written out. The Greek read it as this second law, deuteronomion, and a whole book of the Bible has been named after the reading ever since. Ecclesiastes is Greek too, from the opening words of its own first verse.

Three of the Greek titles were later taken back. Where the Greek Bible has four books of Kingdoms and a book called Paraleipomena — the things left out — we have Samuel, Kings and Chronicles, and we have them because Jerome argued for the Hebrew names and suggested that the last of them be called a chronicle of the whole of sacred history. The Greek bookkeeping survived him anyway in the places you would least expect: the Council of Trent, defining the canon in 1546, is still listing four books of Kings and two of Paralipomena.

And if you have ever wondered why a Catholic or Orthodox prayer book numbers a psalm differently from a Protestant one, that is the Greek as well. Because Psalms 9 and 10 are one psalm in the Greek, and Greek Psalm 113 is two psalms in the Hebrew, the Greek numbering runs one behind the Hebrew between those points. The Miserere is Psalm 50 in the Greek and 51 in the Hebrew. De profundis is 129 or 130 depending which Bible is on the shelf.

Chapter and verse, for what it is worth, are neither Greek nor ancient. The chapters are Stephen Langton’s, about 1205. The New Testament verse numbers are Robert Estienne’s, printed at Geneva in 1551. The story that he made the divisions on horseback between Paris and Lyon, and that the jolting explains the awkward breaks, is repeated everywhere and I could not find a source for it in anything I read.

Which is a small last example of the thing this whole page has been about. A book that no longer exists in any single copy gave the Bible in your hands the names of its parts, the numbering of its psalms, the vocabulary of its theology and, for the first Christians, its Old Testament entire — and around it there has accumulated a body of stories, figures and confident sentences, most of them repeated in good faith, a surprising number of which turn out on inspection to have been made up by somebody.

The men who made it left no names, no colophons and, with one exception in Esther, no signatures. The document that claims to name them is a forgery whose own list is one man short. And the last thing they are recorded as doing, in that forgery, is standing up in an assembly and cursing anyone who would ever change a word of what they had written.

c. 331 BCE – today

The whole thing, in order

1. c. 331 BCEAlexandria is laid out in barley meal for want of chalk, and the birds eat the lines.
2. 282 BCEDemetrius of Phalerum, having backed the wrong heir, is detained by the king a later story will make his patron.
3. c. 280 BCEThe five books of Moses are put into Greek at Alexandria. Nothing else is.
4. before 205 BCEDemetrius the Chronographer quotes Greek Genesis, and repeats the translator’s mistake about Chabratha.
5. c. 150 BCEA roll of Deuteronomy is copied in the Fayyum. Eight scraps of it will survive.
6. 132 BCEBen Sira’s grandson reaches Egypt and apologizes in his preface for what translation costs.
7. c. 130 BCEThe Letter of Aristeas is written, and curses anyone who alters a word of the translation.
8. late 1st c. BCEThe Twelve Prophets are revised toward the Hebrew, in Greek, by Jews. The curse has already failed.
9. c. 40 CEPhilo reports an annual festival on the island, and says the translators used the same words. He gives no number and no cells.
10. c. 93 CEJosephus retells the story and replaces the curse with a standing procedure for correction.
11. 115–117 CEThe Jewish revolt in Egypt. The tax receipts at Edfu stop in May 116 and do not resume.
12. c. 128 CEAquila’s version, dated on the sole authority of a bishop writing 264 years later.
13. c. 155 CEJustin accuses the Jewish teachers of deleting passages. Two of the three are in no Bible at all.
14. c. 180 CEIrenaeus separates the translators from one another. The cells have arrived.
15. c. 235–245 CEOrigen sets six Bibles side by side at Caesarea and marks every difference between them.
16. c. 392 CEEpiphanius gives the miracle thirty-six cells, thirty-six boats and thirty-six bedchambers.
17. c. 403 CEThe congregation at Oea shouts down the new Jonah, and the bishop corrects it back at the lectern.
18. c. 405 CEJerome: I do not know who was the first lying author to construct the seventy cells at Alexandria.
19. c. 426 CEAugustine rules that where the Greek and the Hebrew differ, both are the same Spirit speaking.
20. 553Justinian tells the Jews which Greek Bible they may read aloud, and cites the cells as his reason.
21. c. 616Paul of Tella turns Origen’s fifth column into Syriac, which is why much of it can still be read.
22. 640–641Caesarea falls. Nothing more is heard of its library, and no source connects the two.
23. 10th centuryA scribe at Fustat scrapes down a leaf of the Hexapla and writes liturgical poems on it.
24. 1517The Complutensian Old Testament is finished at Alcalá, and then waits years for permission to appear.
25. 1518The Aldine Bible reaches readers first, edited from manuscripts its editor declines to name.
26. 1587The Sixtine edition prints Codex Vaticanus, and holds the field for three and a half centuries.
27. 1684Humphrey Hody, aged twenty-five, demonstrates that the Letter of Aristeas is a forgery.
28. 1707–1720Grabe prints Origen’s asterisks and obeli, so a reader can see where the two Bibles part.
29. 1798–1827Holmes and Parsons collate 311 manuscripts in twenty-nine years, and leave 164 volumes of notes.
30. 1897Burkitt publishes Aquila’s Kings out of a palimpsest from the Cairo genizah.
31. 1908The Göttingen critical edition begins, on 1 April.
32. 1935Rahlfs’s pocket edition appears, and becomes what almost everybody means by the Septuagint.
33. 1940The larger Cambridge edition stops at Esther. The prophets are never done.
34. 1961Israeli archaeologists rope down into the Cave of Horror and find the rest of a Greek scroll bought from Bedouin.
35. 1963Barthélemy shows that the correcting of the Greek Bible began before Christianity.
36. 2020A genuinely new critical Psalter is begun, and given twenty-one years.

Who says so

A note on sources

This is a secondhand history, assembled from the works named below, and where they disagree the disagreement has been left standing in the prose rather than tidied out of it. On this subject that is not a stylistic preference. A great many of the figures everybody quotes turn out to be estimates, misread sentences, or numbers with nothing behind them at all, and saying which is which is most of what a page like this can usefully do.

Several things here rest on a single witness, and the sentences name them on purpose. Everything about the origin of the Greek Pentateuch comes from the Letter of Aristeas, which is a forgery; everything about Aquila’s date comes from Epiphanius, writing more than two and a half centuries later with a declared motive; the annual festival on Pharos is Philo’s and nobody else’s; the survival of the Caesarean library into the sixth century rests on one colophon; and the whole of Justin’s accusation reaches us through one manuscript finished in 1364.

The figures that are somebody’s reckoning rather than a reading are marked where they occur, and there are more of them than is comfortable. The Hexapla’s fifty volumes is Montfaucon guessing in 1713 and saying so; its 6,500 pages is Swete multiplying one manuscript’s leaf count by five; its twenty-eight years is a sentence about somebody’s asceticism, misread. The 1,386-year gap between the Greek and Hebrew chronologies requires two contested choices to produce, and is 1,366 without them. The New Testament’s “340 citations of the Septuagint against 33 of the Hebrew” is made by counting 268 cases in which the two agree as evidence for one of them. The often-quoted 212 and 185 are David Turpie’s, from 1868, and nobody appears to have redone the count since. Greek Jeremiah’s “2,700 words” I could not source at all. And 638, the year usually given for the Hexapla’s destruction, is Swete’s sentence, is attached to no evidence about the Hexapla, and is probably the wrong year for the conquest as well.

One set of figures here is a measurement rather than a report. The counts of diatheke against syntheke, of ekklesia against synagoge, and of baptizo were made for this page in the parallel-aligned Hebrew and Greek text of the Tov and Kraft project, which sets each Hebrew word against the Greek standing opposite it. The matching is by stem and it is crude; the numbers are given so that somebody can check them, which is the only thing that makes a number worth printing.

Several matters above are contested and are marked as contested: whether the Greek Jeremiah translates a shorter Hebrew or abbreviates a longer one, whether the extra generation in Genesis 11 was inserted into the Greek or dropped from the Hebrew, whether “a body you prepared for me” is the original Greek of the psalm, whether the divine name stood in Greek letters or Hebrew ones in the oldest copies, and what the men of Alexandria thought they were making. Where a claim rests on a late or hostile witness, the page says so instead of repeating it.

Further reading

• Karen H. Jobes and Moisés Silva, Invitation to the Septuagint (2015) — The book to start with, in its second edition. It explains what the discipline actually does, and it is honest about how much of the standard account of the origins is a story.
• Natalio Fernández Marcos, The Septuagint in Context: Introduction to the Greek Version of the Bible (2000) — The standard one-volume scholarly introduction in English, and the source most of the Pauline citation figures in circulation actually come from.
• Alison Salvesen and Timothy Michael Law, editors, The Oxford Handbook of the Septuagint (2021) — Where to see the current state of every question on this page, argued by the people arguing it. Long, and worth the length.
• Timothy Michael Law, When God Spoke Greek: The Septuagint and the Making of the Christian Bible (2013) — The readable narrative account, and the one to hand somebody who wants the story rather than the apparatus.
• Abraham Wasserstein and David J. Wasserstein, The Legend of the Septuagint: From Classical Antiquity to Today (2006) — The growth of the seventy-two translators traced author by author across two thousand years, in both the Jewish and the Christian traditions. This is the book behind the section on the cells.
• Moses Hadas, Aristeas to Philocrates (1951) — The forgery itself, in Greek and English, with an introduction that lays out the anachronisms. Read it and see how good it is; it is not a clumsy fake.
• Sylvie Honigman, The Septuagint and Homeric Scholarship in Alexandria (2003) — The argument that the Letter is a charter myth, told on the model of the Homeric editing being done at the Museum down the road, and that its curse is aimed at scholars.
• Emanuel Tov, The Text-Critical Use of the Septuagint in Biblical Research (2015) — How the Greek is actually used as evidence about the Hebrew, by the man who edited the Dead Sea Scrolls. Technical, and the place where the Jeremiah and Samuel arguments are set out properly.
• Tessa Rajak, Translation and Survival: The Greek Bible of the Ancient Jewish Diaspora (2009) — Reads the Greek Bible as an instrument of Jewish survival rather than assimilation, and makes the case against the idea that Jews simply abandoned it.
• Albert Pietersma and Benjamin G. Wright, editors, A New English Translation of the Septuagint (2007) — The Greek Old Testament in English, free online, with an introduction to every book by its editor. Those introductions are the quickest way to see how unlike each other the translators were.
• Henry Barclay Swete, An Introduction to the Old Testament in Greek (1900) — Old, free, and still the place where the nineteenth-century figures are printed — which is what makes it possible to catch several of them, since a good many of the numbers on the modern internet were first written down here.
• Frederick Field, Origenis Hexaplorum quae supersunt (1875) — Where the surviving fragments of the Hexapla are actually collected. Field spent his life on it and declined, at the end, to say what year it was destroyed.
• Martin Hengel, The Septuagint as Christian Scripture: Its Prehistory and the Problem of Its Canon (2002) — The canon question at book length, and the plainest statement that no delimited pre-Christian Jewish collection larger than the Hebrew Bible can be shown to have existed.