A history of the Masoretic Text, the Hebrew Bible that every English Old Testament is made from

Written by families who spent three centuries counting letters, pointed with vowels younger than Islam, and kept so exactly that the mistakes are still in it.

A history in 12 parts · about 43 minutes · c. 600 – today

30 November 1947 · Aleppo, the Central Synagogue

A Book That Opens in the Middle of a Curse

The manuscript that a thousand years of Jewish scholarship treated as the standard against which other Bibles were corrected begins, as you can see it today, halfway through a curse. Its first surviving words are the end of Deuteronomy 28:17 — cursed shall be thy basket and thy store — and everything before them is gone. Genesis, Exodus, Leviticus, Numbers and nine tenths of Deuteronomy: absent. So are 2 Kings 14:21 to 18:13, most of the second book of Chronicles, the first twenty-five psalms, the Song of Songs from 3:11 onward, and the whole of Ecclesiastes, Lamentations, Esther, Daniel and Ezra–Nehemiah.

The book is the Aleppo Codex, written at Tiberias on the Sea of Galilee around 930. It is a stout thing, thirty-three centimeters high and twenty-six and a half wide, gathered in quires of ten leaves, three columns to a page and twenty-eight lines to a column. Yosef Ofer, who has spent decades on it, measured what those pages hold: an average of 1,774 characters each, 591 to the column, 21.1 to the line. Ofer’s is also the only reconstruction of the whole book computed from the manuscript itself rather than by comparison with a printed Bible, and it makes the complete codex 481 leaves. Two hundred and ninety-four of them are left.

Those two numbers are worth pausing on, because the form in which you will meet them is wrong. Almost every account says that 294 of an original 487 pages survive. They are not pages, they are leaves, and the word has been carried through the popular literature for decades. The arithmetic gives it away: a book of 487 pages holding 1,774 characters each could not contain a Hebrew Bible at all, while 294 leaves out of 481 is sixty-one percent, which is exactly the “around forty percent missing” that the scholarship reports. Even that figure is not agreed. The Ben-Zvi Institute, which has custody of the book, says in its own UNESCO paperwork that one third is missing; Ofer, Matti Friedman, Ronen Bergman and the Israel Museum all say about two fifths. Both are estimates against a reconstruction that carries an error bar of its own.

What happened to the rest of it has been argued over for seventy years and is not settled. The United Nations voted to partition Palestine on 29 November 1947, and the mob came to the Jewish quarter of Aleppo the following morning. The codex was kept in a basement chapel of the Central Synagogue, said to be the cave where Elijah hid, in an iron safe with two locks whose keys were held by two different men. Sam Sabbagh, describing the day to Michael Glatzer forty-one years afterward, said the crowd poured petrol at the foot of the metal door, burned the wooden core out of it and let the plates collapse inward; they turned the safe face-down, cut through three layers of tin at its back expecting money, and found a book. Sabbagh could have carried the whole surviving codex away. His hands were trembling, he said, from the horrors he saw; he picked up one loose scrap and left the rest. He kept it in his wallet in a plastic sheath for sixty years, through open-heart surgery, as a charm. It is a piece of Exodus 8, the plague of frogs on one side and the swarms of flies on the other, and when the Israel Museum laboratory finally flattened it out it came apart into four pieces and a scatter of shreds, every one of which fitted back together.

The story everybody knows is that the codex burned. It did not, and the rumor was manufactured on purpose: the Aleppo elders put it about to conceal that the book had been saved, in case somebody came looking for it. During the restoration at the Israel Museum between 1986 and 1993, Michael Maggen established that the purple staining along the edges of the surviving leaves, read for decades as charring, was mold; the finding went into Nature in 1988 under the title “Fungi not fire damaged Aleppo Codex.” The claim is usually stated more broadly than the evidence supports, though, and it is worth narrowing. It is the bound leaves that show no burning. Ofer writes that certain marks on Sabbagh’s fragment cannot be seen in ordinary light because the parchment there was fire-damaged, and Sabbagh’s own testimony says the same twice. The volume did not burn. A piece already loose on the floor did.

Where the missing leaves went is the part nobody can close. The codex reached Israel ten years later, carried out of Syria by a cheese merchant named Murad Faham whose wife wrapped the sack in cheesecloth and hid it in a washing machine; the memo recording its arrival in Jerusalem gave the name of the book and nothing else, no condition report and no count of leaves. Matti Friedman’s investigation of 2012 concluded that most of the loss happened around the time it left Syria or after it arrived, and named five men who might have known more without naming a culprit. Yifat Erlich, working independently in 2014, found no evidence of significant loss before it reached Jerusalem. Paul Sanders, reviewing Friedman, entered the fair objection that he could have been more critical of the witnesses who supported his own reconstruction. There are at least ten accounts of how the book was rescued, and they contradict one another.

That is where this page begins, and it begins there because of what the loss was a loss of. The men who made this book were the end of a tradition that had spent something like three hundred years counting the letters of the Hebrew Bible. That is not a figure of speech. They counted them, wrote the totals in the margins, flagged every form that occurred only once, and recorded how many times every other form occurred and where. They built an apparatus whose whole purpose was that nothing could go missing without somebody noticing. Their masterpiece then lost two fifths of itself, and no one can say for certain in which decade.

What they made is called the Masoretic Text, and the name describes a discipline rather than a book. Every Old Testament in English is translated from it. Very little of what is commonly said about it is true — not that its vowels are ancient, not that the men who made it corrected the text, not that they were working against Christianity, and not, as the most-quoted proof of their exactness concedes in its own next sentence, that the counting worked.

c. 600–1000 · Tiberias, on the Sea of Galilee

The Family Business

Tiberias had been the Jewish capital of Palestine for six hundred years by the time any of this happened. Herod Antipas built it around 20 AD on a graveyard, which made it unclean and kept observant Jews out of it for a century; then Rabbi Johanan founded an academy there, the patriarch moved down from Sepphoris, the Palestinian Talmud was redacted in the city, and under Islam it became the capital of Jund al-Urdunn from 636. It had hot springs, black basalt, and the Palestinian Yeshiva, which is the part that matters: the prestige of the Tiberian reading came from the academy sitting in the town, not from any special purity in the local air. The medieval explanation — that the Tiberians never left Palestine and so preserved the original pronunciation — does not survive examination. Geoffrey Khan notes that the Tiberian reading did change, and was in several respects less conservative than the Babylonian tradition it beat.

The scholars are called Masoretes, from masorah, and even that word is disputed: whether it comes from masar, to hand on, or from a term for a bond in Ezekiel. The English spelling Massorah, with its doubled s, is worse than disputed. It is a nineteenth-century grammarian’s conjecture with nothing in any manuscript behind it.

They worked in families. The house of ben Asher runs through five generations, or six depending which reference work you open, ending with Aaron ben Moses ben Asher in the tenth century, who wrote a treatise called the fine details of the accents and whose name the Aleppo Codex carries on a dedication written about a century after the book. The other name is ben Naphtali. They are always described as rivals, and there is no documented rivalry at all. For roughly a century afterward the two traditions were held to be equally authoritative: Mishael ben Uzziel, who compiled the Book of Differences between them, gives priority to neither, and the only rule anyone laid down was that a reader must take one of them whole and not mix them, since a man who picks the best of each reads by no rule at all. Maimonides ranked ben Asher first. David Qimhi, who died in 1235, seems to be the first to decide between them on the evidence of the differences themselves, three hundred years after both men were dead.

And the differences are the second thing worth knowing. Mishael’s book records 867 places where the two disagree, and 406 more where they agree against a third authority he usually does not name. You will very often see the figure given instead as about 875, nine tenths of them concerning the accents; that formulation comes from the Jewish Encyclopedia of 1901 and is still circulating. The modern count is Lipschütz’s, from his critical edition of 1965. And what the disagreements are about is narrower than accents in general: a secondary mark of stress and length. A handful touch spelling or word division. None touches a consonant. After five or six generations of two houses doing nothing else, the sum of the argument between them is where to lengthen a syllable.

Even that overstates how firm the two names are. The Aleppo Codex — pointed, on the traditional account, by Aaron ben Asher himself — agrees with the readings Mishael attributes to ben Asher in only ninety-four percent of the recorded cases. Leningrad reaches ninety, the older leaves of a British Library manuscript eighty, and the Cairo Codex of the Prophets thirty-three. And behind both famous names stands the third school Mishael recorded four hundred times and never identified.

Whether the ben Ashers were Karaites has been argued for a hundred and sixty years, and the argument has a discoverable starting point. No Rabbanite source accuses Aaron ben Asher of anything of the kind before 1860, when Simhah Pinsker printed the charge in Vienna on the principle that a Masorete with no documented engagement in the Talmud ought to be suspected of it. Graetz, Klar and Paul Kahle took it up, Kahle writing that we know it with certainty; Zucker and Dotan argued the other way; Yeivin and Raphael Zer revived the case from the codex’s own pointing. Khan’s conclusion, after setting out both sides, is that there is no incontrovertible evidence either way. Two facts sit against the confident versions. There is no evidence of a Karaite community in Tiberias during the masoretic period at all — Karaite immigration into Palestine began in the later ninth century and went to Jerusalem. And the codex’s Karaite dedication was written around the middle of the eleventh century, a hundred years after the book, while its first folio carried a separate notice dedicating it to the Rabbanite community. Karaite ownership is not evidence of Karaite authorship.

Then the school stops, and not for the reason usually given. It did not end because Tiberias was destroyed: the masoretic work ceased in the tenth century while the city was still prospering, and the earthquakes that ruined it came in 1033 and 1068. What happened in between is that the Yeshiva moved to Jerusalem and took the authority with it. The Tiberian signs went on to conquer the Jewish world; the Tiberian reading they had been invented to record fell out of use within a generation or two, so completely that no later pronunciation of Hebrew descends from it directly. Most medieval communities went on writing the signs while saying something else entirely — reading the notation in an accent it was never designed for. A thousand years later Khan reconstructed the sound, and the recordings made for his volumes are described as the first time the Hebrew Bible has been read aloud that way since. As for when the last of these men died, there is no document at all. The only evidence for Aaron ben Asher’s death is a colophon of 989 calling him may he rest in the Garden of Eden, which fixes nothing except a date he was already dead by.

The margins · What the apparatus is and what it counts

An Accountancy of Consonants

The masorah is not a commentary. Nothing in it explains a verse, and nothing in it argues about what a verse means. It is an audit, written in the margins of the page in an Aramaic shorthand so compressed that most of it is single letters with dots over them, and its purpose is to make a copy checkable.

It is conventionally divided into three. The masorah parva sits in the side margins next to the columns and is almost entirely counts. The masorah magna runs along the top and bottom of the page and expands those counts into full lists, giving the place of every occurrence. The masorah finalis is collected at the end of a book or of the whole codex, and consolidates the totals: verses, words, letters, the middle verse, the longest and shortest sections. A small circle written above a word, the circellus, is the footnote marker; it means nothing on its own except look in the margin.

The commonest note in the whole apparatus is a single lamed with a dot over it, short for the Aramaic word for there is none, meaning that the form beside it occurs nowhere else in Scripture. Letters double as numerals for the rest: two, three, four. There are marks for a defective spelling and for a full one, for a word that is read differently from the way it is written, and for a form a reader is likely to correct and must not. When the magna lists the places a form occurs it identifies each one by quoting a snippet of the verse, because it could not do anything else — chapter and verse numbers are a Christian invention of the thirteenth century, three hundred years after these men had finished. By the tenth century the lists were also circulating on their own, pulled out of the margins and bound as a book, the Sefer Okhlah we-Okhlah, named after the first pair of words in its opening list.

Now the part that is usually left out. The most-quoted evidence for masoretic precision is a passage in the Babylonian Talmud, at Kiddushin 30a, which explains the title soferim, scribes, by deriving it from a word meaning to count: the early sages counted all the letters in the Torah. The same passage supplies the famous midpoints — the middle letter of the Torah is the vav of gachon at Leviticus 11:42, the middle words are darosh darash at Leviticus 10:16, the middle verse begins at Leviticus 13:33 — and the tradition thought them important enough to write into the scroll itself, since both the vav at 11:42 and the gimel at 13:33 are written oversized in every manuscript tradition anyone has surveyed.

They are all wrong, and the passage says so itself two lines later. Rav Yosef is made to concede that the count cannot be repeated: they were expert in the full and defective spellings and could count precisely, and we are not expert in this. Nor, he adds, are we expert about verses — followed by a report that a verse Babylonia counted as one was counted in the land of Israel as three. The figures bear him out. Kiddushin 30a says a Torah scroll has 5,888 verses; the received Torah has 5,845, which is forty-three short. It says Psalms has eight more verses than that and Chronicles eight fewer, and those are not near misses but impossibilities: the Masoretic Psalter has 2,527 verses and Chronicles under 1,800, so neither can be within eight of 5,888 on any counting anyone recognizes. Rav Hai Gaon suggested the Talmud was describing one particular scroll found in Jerusalem whose verse divisions differed. Nobody has done better since.

The midpoints miss too. On the standard modern count of 304,805 letters in the Torah the midpoint is letter 152,403, which falls in Leviticus 8; the vav of gachon sits about five thousand letters past it, a little under two percent of the scroll. The middle word misses by more than nine hundred words in the same direction. And the letter total itself is not stable: 304,805 is the figure in Even-Shoshan’s concordance and on every scribe’s wall chart, while the Keter edition yields 304,801, four fewer. It is a modern count either way. No masoretic manuscript gives it.

None of this makes the Masoretes careless, and it is worth being clear about why. The counts are not measurements of the text they had; they are an inherited tradition about a text somebody else had, written down by men who knew it no longer matched and wrote it down anyway. That is the habit this whole page is about. When the tradition and the page disagreed, they did not correct the tradition and they did not correct the page. They recorded both and left the disagreement standing where a reader could see it. The Leningrad Codex, which is the manuscript most modern Bibles are made from, disagrees with its own masoretic apparatus in several hundred places. The book argues with itself across the gutter, and nobody reconciled it.

Ketiv and qere · The word on the page and the word in the mouth

What Is Written and What Is Read

The clearest case of that habit has a name, and it is the single most characteristic thing the Masoretes did.

In something between eight hundred and sixteen hundred places, depending on which manuscript and which edition you count, the tradition held that the word on the page was not the word to be said aloud. The written form is the ketiv, the read form the qere. What the Masoretes did about it is the interesting part. They did not replace the letters. They left the written consonants standing exactly where they were, pointed them with the vowels belonging to the word that was to be spoken, and put the spoken word’s consonants out in the margin under a small qof. The form left standing in the line is therefore a chimera — one word’s consonants wearing another word’s vowels, a thing that was never a Hebrew word and could not be pronounced by anybody. It is not a correction. It is a correction refused, in writing, with both readings preserved and neither allowed to displace the other.

Nobody can tell you how many there are, and the reason is worth having. The figures in circulation run from about 848 to 1,566, and the spread is not carelessness: the qere notes are a medieval marginal apparatus that grew and was copied unevenly, so different manuscripts genuinely carry different numbers of them. Emanuel Tov sharpens it further — the ketiv and qere variations are essentially absent from the scrolls found in the Judean desert, so the system as a written apparatus belongs to the middle ages rather than to antiquity. Biblion serves the Leningrad text, and on this site the number is 1,108: that many verses carry a written form beside the spoken one.

Some substitutions are so constant that the note stopped being written at all. These are the qere perpetuum, and the largest is the divine name, which stands in the text as four letters pointed with the vowels of the word said in its place. The others are quieter. The third-person feminine pronoun is written as the masculine and read as the feminine a hundred and twenty times in the Torah, with eleven exceptions; the word for a girl is written as the word for a boy twenty-two times; Jerusalem is written throughout without the letter the spoken name requires. Once you know they are there you are reading, in every printed Hebrew Bible, a text nobody has ever pronounced as it stands.

Where the system came from has never been settled, and printing the positions is more useful than picking one. Christian Ginsburg held that the Masoretes treated the qere as authoritative but thought the ketiv usually the better reading. Robert Gordis argued in 1937 that it began as a direction to the synagogue reader, above all to keep him from pronouncing the divine name, and became a way of recording optional variants that hardened into obligatory ones. Harry Orlinsky proposed in 1959 that the Masoretes were neither correctors nor selectors: faced with three manuscripts they vocalized the majority reading and left the minority one standing unvocalized, two votes to one, mechanically. His evidence is the awkward kind — in a great many cases the ketiv is plainly better and the qere plainly worse, and the very first ketiv and qere in the Bible, at Genesis 8:17, has the correct form in the line and an anomalous one in the margin. Israel Yeivin refused to choose, observing that every suggestion about the origin is speculative, and noting that the notation cannot predate the eighth century because it presupposes the vowel signs.

What that discipline cost them is visible wherever the text they inherited was plainly broken, because they left it broken. The first verse of 1 Samuel 13 gives Saul’s age at his accession and the number of years he reigned, and as it stands in the Hebrew it is unusable: the figures are missing. Open an RSV, an NRSV or a NASB at that verse and you will find a gap on the page. Three of the great English Bibles chose to print a hole rather than a number, because the manuscript prints a hole.

At Judges 18:30 the priest of the shrine at Dan is named as the son of Gershom son of Manasseh, and the nun that makes the name Manasseh rather than Moses is not written on the line at all. It floats above it. Four letters in the whole Hebrew Bible are written suspended like this, and this is the only one anybody can explain: the traditional account is that it was put there to spare Moses the disgrace of a grandson keeping an idol’s shrine, and the rabbinic chronicle Seder Olam, quoting the verse, reads Moses. Whether the letter records a deliberate act or an older variant is not settled. What is not in doubt is the manner of it. Ellicott’s description has never been bettered: the letter is not boldly inserted, but is timidly and furtively suspended. Even in the act of changing a name they declined to put the change in the line.

There are fifteen passages, ten of them in the Torah, where dots are written over the letters — the puncta extraordinaria, fifty-three dots above letters and three below. The medieval explanation is that they mark words of doubtful authenticity, and it is probably wrong. Tov argues that they began as deletion marks, the Alexandrian scribal convention for canceling letters that were technically hard to erase from a leather scroll, and were never meant to be copied forward. Proto-Masoretic scribes copied them forward anyway. The parallel is on the ground: the great Isaiah scroll from Qumran dots two letters at Isaiah 35:10, and at Isaiah 44:9 the very words that the Masoretic Text dots appear in that scroll as a squeezed-in addition above the line. A cancellation mark somebody refused to act on, preserved for a thousand years and then reinterpreted as a mark of doubt, is as good a summary of the whole enterprise as anything in it.

That verse is the plainest example there is. The Hebrew says that Cain said to Abel his brother, and then says nothing that Cain said. Every ancient translator who met it put words in his mouth; among the Targums only Onkelos let the silence stand, and so did the Hebrew scribes, and so, following them, did Tyndale and the Geneva and the King James. The Masoretes had an apparatus capable of noting anything they liked about that verse. They did not use it to supply the missing sentence, and it is still missing.

Seventh to tenth centuries · The signs, and what they are for

Vowels Younger Than Islam

Hebrew is written with consonants. The vowels that appear in a printed Hebrew Bible as dots and short strokes above and below the letters were not part of the text as it was written down; they were devised in Tiberias in the early Islamic period, along with the accents, to record a way of reading that was already old. The dates given for the three systems vary and all of them are estimates rather than documents. The Jewish Encyclopedia put the Babylonian system at somewhere between 500 and 700, the Palestinian around 700 and the Tiberian around 750, and noted that the earliest reference to any of it in a datable source comes from Saadia Gaon and his contemporaries, around 900. Modern surveys are content to say the second half of the first millennium. Whichever figure you take, the marks are younger than the Qur’an.

The clearest way to see that they are a commentary on the text rather than part of it is to look at a Torah scroll. It carries no vowels and no accents, and it never has. This is not merely custom: writing them in is held to render the scroll invalid for public reading. The rule is not in the Gemara and appears only as a hint in the earlier literature, which is itself a small piece of evidence about how late the signs are. Two quite different reasons are given for it in the tradition. The Rashba, in the thirteenth century, argued that only what was given at Sinai belongs in the scroll. The Radbaz, in the sixteenth, argued the opposite way round — that unpointed letters hold many possible readings at once, and the scroll is left bare so that the seventy faces of the Torah stay available. So the same text exists in a synagogue in two forms at the same time. On the shelf, or in the study, the codex: three columns, the vowels in, the accents in, the margins black with notes. In the ark, the scroll: bare consonants on skin, sewn with sinew, unreadable without the tradition already in the reader’s head. And it is the bare one that gets read aloud, cued by a man moving his hand.

The accents do more work than the vowels. They mark which syllable is stressed; they divide the verse into units of sense, functioning as a system of punctuation with a strict hierarchy in which a disjunctive always outranks a conjunctive; and they carry the melody the text is chanted to. Gesenius treats the music as the origin and the other two as the grammatical use that survives, which is a better way round than the modern shorthand. Even counting them is unstable: Gesenius lists twenty accents for the twenty-one prose books, thirteen disjunctive and seven conjunctive, while a modern table shows about twenty-eight, the difference being whether positional variants count separately. Psalms, Proverbs and Job have a system of their own that reuses many of the same signs for different purposes.

One thing about the accents ought to be dropped and never is. The scheme that ranks the disjunctives as emperors, kings, dukes and counts — still taught in almost every popular account of cantillation — is not a Jewish scheme and is not masoretic. It was invented by a Christian Hebraist, Samuel Bohl, at Rostock in 1636, and Gesenius, having named the book it came from, adds that as a source of manifold confusion it had better be given up. It has not been given up. As for the claim that the original melody has been recovered, Suzanne Haïk-Vantoura’s reconstruction treats each accent as a single note of a scale, an assumption at odds with every surviving tradition, and specialists reject the method. The recordings circulate anyway.

1538–1675 · What the points cost the men who studied them

The War Over the Points

That the vowel signs were added late looks now like a technical observation. For about a hundred and fifty years it was a question men lost careers over, and the reason is worth stating plainly before the story, because otherwise the story reads as pedantry. The Protestant case rested on Scripture alone. Catholic scholars — Jean Morin above all, working from the Samaritan Pentateuch and the Septuagint — were arguing that the Hebrew text was too unstable to carry that weight, and that the Church and its Latin Bible were therefore indispensable. If the vowels of the Hebrew Bible turned out to be a medieval addition, then the meaning of the Hebrew Bible was, at the joints, a medieval decision. The vowel points were the narrowest point of that front.

Elijah Levita opened it, and he was not a Protestant or a Catholic but a Jewish grammarian who had spent thirteen years in Rome in the household of Cardinal Egidio da Viterbo, teaching him Hebrew, learning Greek in return, keeping his family on the arrangement, and being attacked by his own community for it. His Masoret ha-Masoret was printed at Venice in 1538 in three parts, titled First Tables, Second Tables and Broken Tables, which is a joke about Moses with the broken ones at the end. In the third introduction he argued that the vowel points were invented by the Masoretes after the close of the Talmud, on two grounds: the Talmud and the Midrashim never mention the signs or their names, and the names themselves are Aramaic rather than Hebrew. He was right about the direction and several centuries too early about the date.

Johannes Buxtorf the Elder, who held the Hebrew chair at Basel for thirty-eight years, set himself against that dating in a commentary of 1620 named, of all things, Tiberias. Then, in 1622, a Huguenot professor at Saumur named Louis Cappel finished a manuscript demonstrating Levita’s case in detail and did the courteous thing: he sent it to Buxtorf for an opinion. Buxtorf read it, agreed with it, and wrote back advising him not to publish — pointing out the injury it would do the Protestant cause and the advantage it would give Rome on the question of the infallible accuracy of the text of Scripture. Thomas Erpenius at Leiden then had it printed anonymously in 1624 on his own responsibility, under the title The Secret of the Pointing Revealed.

The story is usually told with a detail attached to it that belongs to a different book: that no Protestant printer would touch it. The 1624 anonymity was Erpenius’s decision and Buxtorf’s advice, not a boycott. The boycott is real, but it belongs to Cappel’s Critica Sacra, which he finished in 1634 and could not print because of the opposition of the Protestants at Geneva, Leiden and Sedan. That book went further and that is why: it denied that even the consonantal text preserved the original wording, and proposed emending the Hebrew from the ancient versions and by conjecture. Conceding that the vowels were late cost little. Conceding that the consonants were corrupt cost everything. It sat unprinted for sixteen years and appeared at Paris in 1650 only because Cappel’s son Jean, who had gone over to Rome, obtained the king’s leave to print it. The son’s name is on the title page as editor.

Buxtorf’s son answered for the family in 1648, nineteen years after his father’s death, with a treatise whose title names Cappel’s book as its target, and again in 1653 against the Critica Sacra. Cappel’s reply to the first appeared in Amsterdam in 1689, thirty-one years after he died, edited by another son. The English theater of the same war ran on well past all of them: Brian Walton’s London Polyglot took Levita’s side in the 1650s, John Owen answered in 1659 holding that the points had been completed by Ezra and his companions under divine direction, and John Gill was still arguing for their antiquity in 1767.

The high-water mark is the Formula Consensus Helvetica, drafted at Zurich in 1675 by Johann Heinrich Heidegger with Lucas Gernler of Basel and Francis Turretin of Geneva, which made the inspiration of the Hebrew vowel points a term of Swiss Reformed ministry. Its second canon holds the Hebrew original inspired not only in its consonants but in its vowels, and its third forbids emending the Hebrew from the Septuagint, the Samaritan Pentateuch or the Targums — Cappel by description, without the name. Two things soften that and both are usually left out. The canon does not say the marks are ancient: it says either the points themselves, or at least the power of the points, six words drafted to let a man subscribe without lying about the date. And Turretin, who helped write it, allowed in his own Institutio that the points might have been added lately, and called the question rather grammatical than theological.

It did not last. Basel stopped requiring signature in 1686, Geneva’s obligation lapsed by 1706 and was renounced outright in 1725 by Jean-Alphonse Turrettini, undoing his father’s work, and the whole thing was disused by about 1740. It was first printed in 1714, so for its first four decades it bound men who could not buy a copy of it. Philip Schaff’s verdict is the one to keep: it was allowed to die and be buried without mourners.

Cappel won the date and lost nothing else, and the modern position is not quite his either. The signs are medieval and nobody defends Mosaic vowel points. But specialists hold that the Tiberian pointing records an oral reading tradition that is genuinely ancient and corresponds closely to the Hebrew of the Second Temple period, which makes it evidence about old Hebrew rather than a medieval invention, while conceding that particular vocalizations here and there are late. Cappel had said as much on his own first pages in 1624: late marks, authentic tradition. The seventeenth century fought for thirty years over a thesis whose loser had conceded the point that mattered before the argument began.

1477–1525 · Venice, and the edition that lasted four hundred years

The Book a Proofreader Made

The first printers found the masorah unprintable. A Psalter came off a press at Bologna in 1477 with vowels and accents in the first four psalms and then nothing at all: the compositor started, saw what he had taken on, and stopped. The first complete Hebrew Bible was printed at Soncino, a small town in Lombardy, in 1488 by Joshua Solomon Soncino, four years before the expulsion from Spain scattered Hebrew books and Hebrew printers across the Mediterranean. A later Soncino edition from Brescia is the one Luther used; his own copy is an octavo he could hold in one hand, bought second-hand from a south German seller before 1519, and about half his annotations in it are in the Pentateuch.

The book that mattered came out of Venice, and the man who paid for it was a Christian. Daniel Bomberg, from Antwerp, got his Venetian privilege to print Hebrew in December 1515 at the fifth attempt and at a cost of some five hundred ducats. His first Rabbinic Bible appeared in 1516 and 1517, edited by Felix Pratensis, and it did not sell to the people it was made for: Pratensis was a Jew who had become a Christian and an Augustinian friar, and the title page carried a papal privilege. A Jewish household was not going to buy its Bible on those terms.

So Bomberg commissioned a second one, and the editor he found for it is the reason this section exists. Jacob ben Hayyim ibn Adonijah had fled Tunis in his forties, passed through Rome and Florence, and was working as a proofreader in a Christian printer’s shop in Venice. The Second Rabbinic Bible of 1524 and 1525 is his, and it became the received text of the Hebrew Bible for four hundred years. It was the first edition to print the masorah in a usable arrangement beside the text it belonged to, and it fixed the ketiv and qere apparatus, the arrangement of the rabbinic commentaries around the page, and the verse divisions that Hebrew Bibles have used since. His introduction is a working document rather than a preface. He describes the manuscripts Bomberg’s agents brought him as being in the utmost disorder and confusion, so much so that there was not a sentence to be found without a blunder. He complains that people ask what profit can be derived from the masorah, and that for this reason it has almost been forgotten and lost. He states his own doctrine flatly — that the qere and the ketiv are both a law of Moses from Sinai — and then, having to, admits that the Talmud is sometimes at variance with the masorah.

You will read everywhere that ben Hayyim ended his life a Christian, and the irony is so good that it is worth saying clearly: there is no evidence for it. The Jewish Encyclopedia says late in life he embraced Christianity and offers nothing; the modern reference works cite the Encyclopaedia Judaica, which is one encyclopedia citing another; the frequent gloss sometime after 1527 looks like an inference from the last dated work with his name on it. Ginsburg, who translated ben Hayyim’s introduction and annotated it closely, says nothing about a conversion at all. It may be true. Nobody has produced a document.

What is certainly true about his Bible is less satisfying and more important. He was working from late manuscripts of no great quality, because those were what a Venetian printer’s agents could buy, and he corrected them by the masorah rather than by better copies. A Catholic reference work of 1909 put it that he paid too much attention to the masorah and too little to reliable old manuscripts, and Frants Buhl, writing a year or so later, delivered the general verdict: none of these editions presents the Masoretic text in its original form. The received Hebrew text of the Bible — the one the Geneva and the King James Old Testaments were translated from, the one printed and reprinted until the twentieth century — is a proofreader’s reconstruction from second-rate copies, made in a hurry, in a Christian’s print shop, in 1525.

One decision he took that morning is still on the page in front of you. Open a Hebrew Bible and an English one at Exodus 21. The Hebrew has thirty-seven verses and the English has thirty-six, because ben Hayyim followed a division that Isaac Nathan of Arles had adopted in the 1440s for his Hebrew concordance — and Isaac Nathan had taken his chapters from Christian Bibles, having found, as he explained, the same division into chapters in the Christian version and decided to use it. The chapter numbers in every Hebrew Bible printed since are a thirteenth-century Christian invention, borrowed by a Jewish lexicographer in the fifteenth century and fixed by a Venetian print shop in the sixteenth. Printed verse numbers took longer still: the first Hebrew Bible a Jewish reader could use with them was Joseph Athias’s, at Amsterdam in 1659, a full century after English readers had them in the Geneva Bible.

1929–1977 · The year the base text changed

Available

For the first two editions of Rudolf Kittel’s Biblia Hebraica, in 1906 and 1913, the printed base was still ben Hayyim’s. Then Paul Kahle, preparing the third, tried to get a photograph of the Aleppo Codex, and the Aleppo elders refused. They had let Ben-Zvi look at the case in 1935 without opening it; the one scholar ever allowed to collate the book against a printed Bible was Umberto Cassuto, in 1943, and photography was forbidden outright. So Kahle used the other one. The third edition of Kittel, published in fascicles between 1929 and 1937, was the first printed Bible built on the Leningrad Codex, and it was chosen in the editors’ own words as the best available representative at this time of the ben Asher text.

Available. The better manuscript was in Aleppo, and ten years after the last fascicle appeared it was lying in pieces on a synagogue floor. Every critical edition since, and behind them a great many of the translations on an ordinary bookshelf, descends from a decision forced by a community that would not let a German scholar take a picture.

The manuscript that got the job is worth knowing in its own right, because it is the one your Old Testament is made from. The Leningrad Codex was written at Cairo and dated by its colophon to 1008 or 1009, depending how the reckoning is resolved. Its scribe was Samuel ben Jacob, and the Genizah has preserved both ends of his career. In one letter, newly arrived in Fustat, he asks a charity administrator for even a coin and signs off that he has nothing and is helpless. Within twenty years the same hand is charging twenty-five gold dinars for a Bible, against an average monthly income of two dinars and a year’s house rent of five or six, and a contract of 1021 has him supplying not only the script, the vowels and the masorah but the binding and a rigid slipcase. A book like that took the skins of something like a hundred and twenty sheep. He signed his work with a private mark, a colon and a circle and a colon, by which unsigned manuscripts are still identified as his; on one folio he wrote his name inside a drawn star in letters small enough to need magnification. His colophon dates the book by five eras at once, and reckons the Muslim year from the kingdom of the little horn, which is Daniel 7:8 used as an insult.

How it left Cairo is less edifying. The manuscript is Firkovich B19a, from the collections of Abraham Firkovich, whose first collection the Imperial Library bought in 1862 for 125,000 rubles, ten times its entire annual budget for manuscripts, and who bought the contents of the Cairo Karaite genizah the following year for a thousand silver rubles at the age of seventy-six, traveling with his wife and his grandson. Firkovich is also known to have altered colophons and gravestone dates to make Karaite settlement look older than it was, and he explained his motive himself in a letter of February 1839: he conceded that the Karaites separated from the Rabbanites around 640, which would make them participants in the murder of Jesus, which is not good for us — whereas if a stone could be found from a sufficiently early year, then the Christians would accord us much honor and praise. That is a forger setting out his reasons in writing, and it is why every Firkovich colophon gets looked at twice. Nothing has been shown against B19a specifically, and nothing has cleared it either.

The rest of the line is quickly told. Biblia Hebraica Stuttgartensia, complete in 1977, is Leningrad again with a rewritten apparatus, and it has been on every scholar’s desk for fifty years. Biblia Hebraica Quinta has been appearing fascicle by fascicle since 2004. The Hebrew University Bible Project went the other way and based itself on the Aleppo Codex; founded in 1956, it has published four books of the thirty-nine in seventy years, and the name of Shemaryahu Talmon, who helped found it and died in 2010, stands on a volume that appeared in 2024. Mordechai Breuer, denied access to the codex, reconstructed its text from the masorah and other early manuscripts and found his result matched it almost exactly; the Jerusalem Crown of 2000 is built on his method, set in a face cut from the codex’s own lettering.

And one thing is worth knowing before anyone tells you they have read the masorah: the masorah you can buy is not the masorah in the manuscript. BHS prints a masorah parva in its margins, but it is Gérard Weil’s edited version rather than the codex’s — he harmonized notes the manuscript gives only sporadically, merged notes together, and supplied notes where the manuscript has none. One word carries a rubric of twelve occurrences; Samuel ben Jacob marked four of them and Weil marked all twelve, in the edition everyone calls diplomatic. The masorah magna was removed from the page altogether into a separate volume of which only the first part was ever finished. The apparatus even carries flags found nowhere in the manuscript, marking notes Weil could not resolve and notes that contradict the text they annotate.

The short version of all of it is a swap almost nobody outside the field noticed. For four hundred years the English Old Testament rested on a printed book of 1525. Since 1937 it has rested on a manuscript of 1008. No English translation committee decided that; a German print shop did, because a community in Syria said no to a camera.

1947–today · The Judean desert, and what it settled

What the Scrolls Actually Said

Until 1947 the oldest Hebrew biblical manuscript anybody had was a single sheet in a Cambridge drawer — the Nash Papyrus, four fragments carrying the Ten Commandments and the Shema, bought from a dealer in Egypt in 1898, probably about 150 BC. It is not a scroll but a prayer card, the daily liturgy of the Second Temple, and its text is not the Masoretic one. Behind it there was nothing at all until the ninth century. Between the Judean desert and Tiberias lie seven hundred and fifty years from which no Hebrew Bible survives, and the reason for the silence is piety: a worn scroll is not shelved, it is buried.

What came out of the caves cut in two directions at once, and both halves matter. From Masada, Wadi Murabba’at, Nahal Hever and three other wadis came about twenty-five biblical scrolls, copied between roughly 50 BC and AD 115 and hidden by men who lost two wars against Rome. They agree with the book finished in Tiberias eight hundred years later to within about two percent of their words. In 2016 a carbonized lump of leather from En-Gedi that could not physically be opened was read by X-ray tomography, and its text of Leviticus 1 and 2 turned out to agree with the medieval consonantal text letter for letter — the first ancient manuscript ever published of which that could be said. Whatever else is true, a recognizable Masoretic text existed a thousand years before the Masoretes, and it was copied with extraordinary care.

At Qumran itself the picture is completely different, and here the numbers are a lesson in themselves. Emanuel Tov’s classification of the Qumran biblical scrolls is what every popular account cites, and he has rewritten it in every edition of his textbook. The proto-Masoretic share stood at sixty percent in 1992, thirty-five in 2001, forty-five in 2012 and thirteen in 2019; the non-aligned share went from ten percent to forty-five. Nobody found new scrolls in between. What changed was how one scholar sorted them, and any page that quotes a single figure as the finding of the Dead Sea Scrolls is quoting a moving target. The other number worth holding onto is Tov’s estimate that about five percent of the ancient Hebrew biblical text has physically survived at all. The famous agreement statistics are measured across that five percent.

The disagreements, where they occur, are not trivial. The Masoretic book of Jeremiah is about an eighth longer than the Greek one and puts the oracles against the nations in a different place, and a Hebrew manuscript of the shorter form was found at Qumran — which means the two are not translator’s liberties but two editions of a book. In 1 Samuel the Masoretic Goliath is six cubits and a span, about nine foot nine; in the Qumran Samuel scroll, in both great Greek codices and in Josephus he is four cubits and a span, about six foot nine. The story does not change. The monster does. And the New Revised Standard Version prints a paragraph before 1 Samuel 11, about Nahash the Ammonite gouging out the right eye of every man of Gad and Reuben so that he could not sight along a shield, which is in no Masoretic manuscript at all; it comes from the Qumran Samuel scroll and from Josephus. The NIV and the ESV leave it in a footnote. Two Bibles on the same shelf disagree about whether a paragraph exists. Frank Cross held that it had dropped out of the Hebrew; Alexander Rofé held that it was never there.

The Samaritan Pentateuch differs from the Masoretic in something like six thousand places, of which perhaps nineteen hundred agree with the Greek. In 1815 a young Wilhelm Gesenius at Halle sorted those variants into eight classes and pronounced the Samaritan text secondary, and his verdict stood for a century until five manuscripts from a cave overturned it: the pre-Samaritan scrolls from Qumran carry the whole expanded, harmonized Samaritan Torah and say nothing whatever about Mount Gerizim. The Samaritans did not write their Bible. They inherited a text that was already circulating among Jews, and added a paragraph to it.

So what is the Masoretic Text? It is one edition of the Hebrew Bible, which existed in more than one, transmitted with a fidelity that the desert has now demonstrated rather than assumed. It is not the original and it does not claim to be. What it is, uniquely, is the edition that somebody kept an audit trail for.

The accusation · Psalm 22, Isaiah 7, and what the evidence will bear

Did They Change It?

There is a persistent claim that the Masoretes altered the Hebrew Bible to make it harder to read as pointing to Christ, and it deserves to be met on the evidence rather than waved off, because the men making it are usually pointing at real passages.

Take the strongest one. Psalm 22 has a line that the Greek and Latin traditions render as they pierced my hands and my feet, and the Masoretic Text reads three consonants that mean something like a lion. The dispute is genuinely open and the honest position is untidy. A fragment from Nahal Hever, copied between about AD 50 and 68, appears to read the form behind pierced — but the ink at the crucial letter is faint, its editor measured the stroke against its neighbors and pronounced it one letter, and other scholars looking at the same photograph say it cannot be told; three letters away the same scribe misspelled the word for my hands. The state of the scholarship is fairly captured by three articles in the leading journal of the field, published in 1997, 2000 and 2005, titled The Old Guess, More Guessing and A New Guess. And Jerome, who translated the psalm twice, is his own best witness against a simple story: from the Greek in about 387 he wrote foderunt, they dug through; from the Hebrew about four years later he wrote vinxerunt, they bound. Neither is pierced, and the psalter Latin Christendom actually sang for a thousand years is the one he made from the Greek.

Isaiah 7:14 is a translation dispute rather than a Masoretic one. The Hebrew word is almah, which means a young woman of marriageable age; the word that means virgin specifically is betulah; the Greek translators, centuries before any Christian controversy, rendered it parthenos. Jerome conceded the philology in as many words — that the Jews object that almah does not mean a virgin but a young woman, and that to speak truth a virgin is properly called betulah — and then put virgo in the Vulgate anyway. Luther offered to pay a hundred guilders to anyone who could show that almah meant young woman, and the commentator Ulrich Luz’s note on that offer is that he owes them. The heat has not gone out of it. In November 1952 a minister in Rocky Mount, North Carolina, climbed onto the bed of a truck with a Revised Standard Version, tore out the leaf carrying Isaiah 7:14 and burned that one page, keeping the rest of the Bible. And the reading virgin in the New American Bible was, in Raymond Brown’s footnote, imposed by a decision of the American bishops on translators who had wanted young woman in 1970. They got it in 2011.

Against the general charge, four things stand. The proto-Masoretic scrolls from Masada and Murabba’at were copied eight hundred years before the Masoretes and match their text; whatever the shape of the Hebrew Bible was, it was that shape long before there was a Christianity to argue with. The apparatus records rather than conceals: a tradition minded to remove an inconvenient reading would not have invented a system whose whole design is to leave the disputed word standing in the line. The Masoretes preserved readings that embarrassed their own tradition, from the suspended letter that admits it is hiding Moses’s name to the verse that leaves Saul’s age blank. And the same discipline is the reason the disputes are visible at all: it is only because they wrote down what they did not fix that anyone can now argue about it.

There are eighteen places where the tradition itself says the scribes changed the text out of reverence — the tiqqune sopherim — and even that number is a convention rather than a count. The early sources do not agree: one gives eleven, another seven, another thirteen, the masoretic lists fifteen or eighteen, and two Genizah fragments in Cambridge give twenty-two and eight. The terminology shifts too, from the early phrase that the verse used a euphemism — which says the author wrote it that way — to the later one that the scribes emended it. Carmel McCarthy’s study of 1981 concluded that only three of the eighteen are genuine scribal emendations. The Cambridge Genizah unit’s verdict is that they are not a masoretic phenomenon at all but a midrashic one. The popular version, that here are eighteen places where scribes changed the Bible, is a great deal stronger than the evidence for it.

The last word on the whole enterprise belongs to that dedication, because it is the only surviving statement of purpose any of these men left. It does not say the book is holy, or perfect, or beyond question. It says that if anyone of the seed of Israel, Karaite or Rabbanite, wishes on any day of the year to check a full or defective spelling, or a section break, or one of the accents, the book is to be brought out to him to see and check and so gain understanding. It was written for auditing. Everything in this page — the counts that do not add up and were recorded anyway, the readings marked wrong and left in the line, the blank where Saul’s age should be, the sentence Cain never got to say — follows from taking that sentence seriously.

The Hebrew on this site is that text: the Leningrad Codex, pointed, in 23,144 verses. In 1,108 of them the written word and the spoken one part company, and the line can be switched from the one to the other — which is as close as a screen gets to what the margin was for. You can go and look at the disagreements. That was the idea.

150 BC – today

The whole thing, in order

1. c. 150 BCThe Nash Papyrus: for forty-five years the oldest Hebrew Bible on earth, and not a Masoretic text.
2. 50 BC – AD 115The scrolls hidden in six wadis, which agree with Tiberias eight centuries early to within about two percent.
3. 636Tiberias, long the Jewish capital of Palestine, becomes capital of Jund al-Urdunn under Muslim rule.
4. 749An earthquake brings down its Roman theater, seating seven thousand, and much of the city with it.
5. c. 750–900The Tiberian vowel signs and accents are devised. The earliest datable reference is Saadia Gaon’s.
6. c. 930The Aleppo Codex. Shlomo ben Buya’a writes the consonants; the points, accents and masorah are Aaron ben Asher’s.
7. c. 950The Yeshiva leaves Tiberias for Jerusalem and the masoretic work stops, in a city still prospering.
8. 989A colophon calls Aaron ben Asher may he rest in the Garden of Eden — the only evidence of when he died.
9. 1008Samuel ben Jacob finishes the Leningrad Codex at Cairo, twenty years after begging a charity for a coin.
10. 1100The codex is ransomed from the Crusaders with 230 others, on money borrowed at Alexandria.
11. c. 1180Maimonides cites it for the section breaks and the layout of the songs, and for nothing else.
12. 1375Traditionally the year it reaches Aleppo, carried by a descendant of Maimonides. No document says so.
13. 1477A Bologna Psalter is pointed for four psalms and then gives up.
14. 1488Soncino prints the first complete Hebrew Bible, four years before the expulsion from Spain.
15. 1516–17Bomberg’s first Rabbinic Bible: edited by a convert, licensed by the Pope, unsaleable to Jews.
16. 1524–25Jacob ben Hayyim’s Second Rabbinic Bible becomes the received Hebrew text for four hundred years.
17. 1538Levita argues in his Broken Tables that the vowel points were invented after the Talmud closed.
18. 1624Cappel’s Arcanum is printed anonymously at Leiden, after Buxtorf read it, agreed, and told him to suppress it.
19. 1650Critica Sacra, blocked sixteen years, appears at Paris by the king’s leave, got by the author’s Catholic son.
20. 1675The Formula Consensus Helvetica makes the vowel points a term of ministry, with an escape clause in the canon.
21. 1725Geneva renounces the Formula, at the hand of a son of one of its drafters.
22. 1860Pinsker prints the charge that ben Asher was a Karaite. No Rabbanite source had said it before.
23. 1896Schechter reaches the genizah at Fustat and calls it a battlefield of books.
24. 1929–37Refused a photograph of the Aleppo Codex, Kittel’s third edition is built on Leningrad instead.
25. 30 November 1947The mob comes to the Central Synagogue at Aleppo. The safe is cut open from behind.
26. 1957–58The codex leaves Syria in a washing machine, and is receipted in Jerusalem by a memo naming nothing.
27. 1977Biblia Hebraica Stuttgartensia is complete, and has been on every scholar’s desk since.
28. 1988The purple staining on the surviving leaves is shown to be mold, not fire.
29. 2007Sabbagh’s page of Exodus 8 leaves his wallet after sixty years and falls into four pieces that fit.
30. 2016The En-Gedi scroll is read without being unrolled, and agrees letter for letter.
31. 2019Tov’s proto-Masoretic share of the Qumran scrolls reaches thirteen percent, from sixty in 1992.

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 choice. A surprising number of the figures everybody quotes turn out to be estimates, unit errors, or numbers with nothing behind them, and saying which is which is most of what a page like this can usefully do.

Three things here rest on one person each, and the sentences name them on purpose. Yosef Ofer’s is the only reconstruction of the Aleppo Codex computed from the manuscript rather than from a printed Bible, and every leaf count in the first section is his. Geoffrey Khan’s introduction to the Tiberian pronunciation is where the ben Asher and ben Naphtali numbers come from, by way of Lipschütz’s edition of Mishael ben Uzziel; the older figures still in circulation come from the Jewish Encyclopedia of 1901. And what the Masoretes were doing in the margins is set out best by Israel Yeivin, whose habit of refusing to choose between explanations has been imitated here more than once.

Several things above are contested and are marked as contested: whether the ben Ashers were Karaites, where the ketiv and qere system came from, whether the tiqqune sopherim record real changes, and what the faint letter at Nahal Hever is. Where a claim rests on a single late witness, or on none — the 1375 journey to Aleppo, ben Hayyim’s conversion — the page says so instead of repeating it. The count of ketiv and qere notes is given as a range because the manuscripts genuinely differ and no single total exists. The one figure here that is a measurement rather than a report is the 1,108, counted in Biblion’s own database.

Further reading

• Geoffrey Khan, The Tiberian Pronunciation Tradition of Biblical Hebrew (2020) — The current standard, in two volumes, and freely readable online. The long introduction is the best short history of the Masoretes there is, and it is scrupulous about what cannot be known.
• Israel Yeivin, Introduction to the Tiberian Masorah (1980) — The book to open if you want to know what the marks in the margin actually say. Translated by E. J. Revell, and still the working manual.
• Emanuel Tov, Textual Criticism of the Hebrew Bible (1992–2019) — Four editions, and the differences between them are part of the subject: the author revised his own classification of the Qumran scrolls so heavily that the headline figure fell from sixty percent to thirteen.
• Page H. Kelley, Daniel S. Mynatt and Timothy G. Crawford, The Masorah of Biblia Hebraica Stuttgartensia (1998) — How to read the apparatus in the edition most students own, including the notes in it that are not in the manuscript.
• Yosef Ofer, A Fragment of the Aleppo Codex (Exodus 8) that Reached Israel, in Textus 26 (2016) — The leaf-by-leaf reconstruction of what the codex held, and the account of Sam Sabbagh’s wallet. Freely available, and the source of most of the numbers in part one.
• Matti Friedman, The Aleppo Codex (2012) — A journalist’s investigation of the missing leaves, vigorously argued and vigorously disputed. Read it with Paul Sanders’s review, which is a scholar checking the working.
• Christian D. Ginsburg, Introduction to the Massoretico-Critical Edition of the Hebrew Bible (1897) — Old, enormous and still the place where the lists are actually printed — the tiqqune sopherim, the dotted letters, the suspended letters, the sevirin.
• Christian D. Ginsburg, The Massoreth ha-Massoreth of Elias Levita (1867) — Levita’s book in English with an eighty-four-page introduction, which is where the argument about the vowel points can be read in the words of the man who started it.
• Ernst Würthwein, The Text of the Old Testament (1979) — The standard short textbook, and the quickest route to the printed editions and what each of them is based on.
• Carmel McCarthy, The Tiqqune Sopherim (1981) — The study that went through the eighteen emendations of the scribes one at a time and concluded that three of them are emendations.
• Ronen Bergman, The Aleppo Codex: A High Holy Whodunit, in the New York Times Magazine (2012) — The long-form account of the trial, the sealed transcripts and the men who handled the book. The source of several details here that exist only as testimony.