The Kaufmann Manuscript A15
For most of the twentieth century the received Hebrew Bible was described, with a confidence that by then felt like a fact of nature, as the recension of a single man. Aaron ben Moses ben Asher of Tiberias, the last and greatest of the Tiberian masoretes, was said to have gathered the text into final form in the early tenth century, and everything after that was copying. The Masoretic Text, on this account, was not a tradition but a decision. Once Ben Asher had settled the consonants, the vowels and the accents, the book was closed, and the copyist’s proper office was fidelity rather than judgement. It is a story about authority rather than about ink, and it flatters everyone who came after him, because it hands the whole burden of the text to a man working in one town a thousand years ago.
The story has a rival, and the rival had a name. In the other great Masoretic centre of the early middle ages, in Babylonia and in the regions of Galilee and Byzantium from which Babylonian practice drew, the school associated with Ben Naphtali maintained a different standard codex, with different readings and different habits of accentuation. Between the two authorities, scholars can enumerate roughly eight hundred and seventy-five known differences. The great majority of them concern where a stress accent falls. The remainder concern vowels and the spelling of consonants. That is an enormous amount of disagreement about a book whose defining claim is that it cannot disagree.
This essay is about one manuscript that makes the disagreement physical. Kaufmann A15, held in Budapest in the library of the Hungarian Academy of Sciences under the shelfmark Kaufmann A15, is a complete Hebrew Bible codex: a large parchment book set in three columns to the page, generally dated to the early tenth century, and generally traced to the Karaite communities of Crimea. In older literature the same manuscript circulates under other names, among them the Sumy Codex. Its significance is not that it is rare or handsome, though it is both. Its significance is that it is complete, that it belongs to the same generation as the codices that won, and that it was written in the other tradition.
Set A15 beside the standard account and the account stops working. No printed Bible, and so far as anyone can tell no surviving manuscript, follows Ben Asher uniformly. The received text is an eclectic result, assembled out of competing schools rather than copied down from one authorised book. A15 is not a damaged fragment or a curiosity from the fringes of the trade. It is a whole Bible, in a single binding, in the rival system, surviving from the very decade in which that system was supposed to have lost. The tradition did not close when the story says it closed. It closed later, by degrees, and the closing left fingerprints.
From Chufut-Kale to Budapest, how a Crimean codex reached Europe
Begin where the manuscript begins, which is not in Palestine but on the Black Sea. Crimea in the early middle ages was not a Jewish periphery. It held communities of Karaite Jews with a reputation for scriptural learning, a reputation that was uneven and often disputed by their Rabbanite neighbours but was never trivial. To the east of Bakhchysarai stood Chufut-Kale, a cave town and fortress whose name is Turkic for Jewish Fortress, and which the Karaite inhabitants themselves were content to call simply the Fortress. It had been a place of Karaite settlement for centuries, at various times a place of confinement and at others a place of refuge.
The town’s history is a reminder that communities are not permanent. After the Russian annexation of the Crimean Khanate in 1783, the restrictions that had kept the Jews of the region within certain settlements were relaxed, and the population of Chufut-Kale dispersed. By the middle of the nineteenth century the place had, in the ordinary sense, ceased to exist as a town, leaving a ruin and a name. The Karaite communities of the peninsula continued elsewhere, and the memory of Chufut-Kale remained a point of pride and a subject of antiquarian writing for generations afterwards. The physical settlement faded; the tradition it had housed did not.
What made the Crimean Karaite communities interesting to the history of the Hebrew Bible is not simply that they read scripture, but how they read it. The Karaite case against Rabbanite practice turned on a claim to return to the biblical text without the mediating apparatus of rabbinic interpretation. In practice this produced communities for whom the correct and defensible transmission of scripture was a practical, everyday, intensely serious concern. Codices with full vowel points, full accentuation and a complete masorah were not luxuries in such a setting. They were the working equipment of a minority that had staked its identity on the book being right.
This is the intellectual world that produced manuscripts of the type of A15. That the specific codex came from a Crimean Karaite holding is, as far as the surviving documentation allows one to say, a reasonable and generally accepted inference rather than a documented fact about a named scribe. The manuscript carries the marks of its class: the whole apparatus, the careful layout, the disciplined copying. It carries the marks of its period. But the route by which it left the Crimea, and the hands it passed through, are the sort of detail that the documentary record of the manuscript trade rarely preserves. We know the type of community. We infer the community.
Geography did the rest. Crimea sits on a narrow projection between two seas and at the hinge of the steppe, crossed by trade routes that carried salt, cloth, grain and people between the Ottoman interior, the Crimean Khanate, the Russian Empire and the markets beyond. Books moved inside those caravans, inside the baggage of people already travelling. The Crimean Karaite communities, tied into Ottoman and Russian commerce alike, sat where a Hebrew manuscript could change hands and change owner several times over without ever leaving the orbit of the Black Sea trade.
What altered the equation was the nineteenth century, and specifically the long shadow of the Russian Empire moving into the region. As Karaite learning declined in the later decades of the century, as communities thinned, dispersed and lost the institutions that had once sustained scholarship, the old codices changed from working instruments into property. Manuscripts that had been copied, corrected, collated and used for generations were offered for sale. They were sold, in the ordinary way that antiquities are sold, to dealers, to travellers, to scholars, and at the fairs where such things could be seen. The Crimea, and the smaller settlements around it, became a supplier of old Hebrew books to collectors who had come looking for exactly that.
The collector in this case was David Kaufmann, the Hungarian rabbi and scholar. He assembled his manuscript collection over years rather than months, and the Crimean material entered it in the course of that work, apparently during the 1870s and the decades around them. The details of individual purchases are, for the most part, recorded only approximately. It is generally accepted that much of what he gathered came from Karaite holdings in the Black Sea region, whether bought directly, bought through intermediaries, or encountered at fairs. The collection he built is remarkable for range as well as size, covering biblical codices, liturgical and paraliterary works, and Jewish books of several centuries and several provenances.
It is worth saying plainly that Kaufmann was not buying curiosities. He was a working scholar with a specific project, and the manuscripts were evidence for it. A collection assembled on that basis has a different character from one assembled as a trophy case, and the difference shows in the codices: they were acquired because somebody wanted to read them, and many carry the marks of having been read.
The collection outlived him. After his death in 1899 the manuscripts passed, by the terms of his bequest, to the Hungarian Academy of Sciences, which received them into its library in Budapest, generally recorded as an acquisition of the early 1890s. The Academy’s manuscript holdings retained the Kaufmann prefix, and the codices still answer to shelfmarks in the form Kaufmann A1, Kaufmann A15 and so on. The arrangement has the great advantage for a historian of provenance that the collection has kept its integrity: a manuscript that carries a Kaufmann shelfmark is a manuscript that can be traced, as a rule, back through a catalogue to a collector, a place and a purchase.
That traceability depends on the catalogue, and the catalogue for this collection is Beate Koester’s three-volume description, published in 1987. It sets out the manuscripts one by one, with their codicological descriptions, their scribal features, their colophons and their provenances. It is one of those works that readers consult rarely and then rely on absolutely. It is also the place to look when a claim about a Kaufmann manuscript needs to be checked rather than repeated.
One last detail of the collection’s geography, because it corrects a common confusion. The best known manuscript in the Kaufmann Collection is not A15. It is A50, a complete manuscript of the Mishnah, generally dated to the eleventh or twelfth century, much studied for the language of the Sages and for the separation between the hands that wrote the letters and the hand that added the vowel points. A15 entered an old and miscellaneous collection rather than one assembled to serve a single purpose, which is worth remembering when we come to a complete rival codex sitting quietly in Budapest since the nineteenth century, in full view of the scholars building the critical editions, and consulted by them without ever displacing anything.
What the codex is, a Masoretic Bible as a physical object
A15 is a book. That sounds trivial, and it is not. Before the codex, the Hebrew Bible existed in a scroll form, and the scroll carries a different set of constraints: a fixed length, a fixed sequence, and no room for anything that is not the biblical text itself. The Masoretic codex, which emerges in the tenth century, relaxes those constraints in exchange for adding a great deal. It is a bound collection of parchment leaves of large format, sewn so that they open flat, and it carries an apparatus whose only purpose is to make the text verifiable.
The page of A15 is generally described as set in three columns. This is a practical arrangement before it is a scholarly one. Parchment is expensive, and a scribe who can fit three columns to the page gets more text onto a given quantity of skin, while the aspect ratio of the column stays in the range the eye can track without effort. Three columns also suit the masorah, which wants a stable inner margin in which to write short notes beside the text they qualify. The consequences run both ways. The reader gets a large book that holds an entire Bible in one volume, and the page gets a density of signs that is legible only to somebody trained from childhood.
The script is generally described as square. In masoretic codicology, square does not mean geometric perfection. It means a ductus whose proportions, spacing and stroke shapes conform to a recognised type, and it is a classification, not a compliment. Script type is one of the primary handles by which a scribe’s school is identified, because the training that produces a hand is exactly the training that produces a text. A15’s letter forms are classed as Galilean rather than Tiberian. That classification is doing real work in the argument of this essay, so it is worth being precise about its status: it is the general description given by specialists in the codicology of the manuscript, not a claim that can be made from a printed photograph, and it is a classification of the hand rather than of the text it carries.
The written text itself has three layers. The first is the consonantal skeleton, the ancient text, the letters without which nothing else means anything. The second is the vowel system, the points that indicate the vowels of the language. The third is the accentuation, the marks that indicate stress, melody and the syntactic shape of the phrase, and which exist to be sung. A fourth layer sits outside the text altogether. Between them these layers convert a row of letters into a recited text, and they do it by carrying information that the letters themselves cannot carry.
That fourth layer is the masorah, and it comes in more than one form. The short version, the masorah parva, runs in the margins beside the text, a compressed notation keyed to the words near it. The long version, the masorah magna, runs in the upper and lower margins in longer alphabetical sequences, and a further body of long notes is collected at the end of each book, where it is conventionally called the masorah finalis. The language is largely Aramaic with Hebrew elements, and it is written in a script of abbreviations that assumes a reader who already knows what every abbreviation means.
What the masorah actually says is worth stating plainly, because the modern reader often imagines it as commentary and it is closer to bookkeeping. It records how many times a particular word or form occurs in scripture. It records whether a given form is written full or defective, and it registers exceptions to the general rule. It records words that occur only once. It records the cases where the written form and the reading form differ, the famous pairs where the consonantal text says one thing and the tradition says it should be read as another. And it cross-references, constantly, so that any given point in the text can be checked from several directions at once.
This is where the numerical discipline enters, and it is what makes the Masoretic codex a genuinely technical object. The masoretes counted letters, words, verses and occurrences, and treated those counts as a checksum on the text. A scribe who copied a whole book and then checked the counts against a reliable standard could be reasonably confident that he had introduced no error in the middle, because a single wrong letter changes a count and the change propagates. The system is not foolproof; no medieval copying system was. But it means that the Masoretic codex is not a religious document that happens to be written carefully. It is a document with error control built into its architecture.
It also means the masorah is evidence rather than ornament. When a codex carries a full masorah, it is declaring which textual tradition it belongs to, in a vocabulary specific enough that a trained reader can tell. That is why the apparatus on A15 matters so much for what follows. The notes are not only checking the text against a count. They are placing the text inside a school, and the school is not the one that won.
The manuscript carries a colophon, and the colophon is generally read as dating the copy to 982 CE. The reading should be held with the ordinary caution of a document whose dating has been argued over, and specialists continue to discuss the details. What is not in dispute is the general period: A15 is a tenth century codex, which places it in the same generation as the Aleppo Codex and only a few decades before the Leningrad Codex. A date in that range puts the manuscript inside the decisive decades, when the rival traditions were both still being copied by professionals who knew exactly what the disagreement was.
Layered onto that, A15 carries a vocalisation that is generally classed as Babylonian in type rather than Tiberian, combined with the Galilean letter forms described above. The combination is what gives the manuscript its peculiar standing. A scribal system is a package: the points, the accents, the ductus and the masorah move together, and they were learned together. A codex that mixes the systems is not necessarily careless, and it is not necessarily a copy of one system made sloppily by a scribe of the other. It is more likely to be a witness to how these traditions actually travelled, which was as neighbouring practices that borrowed from one another rather than as sealed systems that never met.
Two schools and one text in the Ben Asher problem
The standard narrative begins with Ben Asher. He was the last of a distinguished family of Tiberian masoretes extending back into the later eighth century, and by the early tenth century he was the leading figure of the school. He produced a standard codex embodying his own opinions, and that codex is generally identified with the Aleppo Codex, the volume that was kept for centuries in a synagogue in Aleppo and is now in Jerusalem. His standing was such that when the century turned, his text was effectively the text. The rival was Ben Naphtali, associated with Babylonia, whose authority competed with his, and it is generally held that the great Babylonian scholar Saadia Gaon opposed Ben Asher.
And yet the rival did not lose cleanly, because it did not lose. The differences between the two authorities number approximately eight hundred and seventy-five. Nine-tenths of them, as the standard accounts record, pertain to the placement of accents. The remainder concern vowels and consonantal spelling. That distribution is itself informative. Two authorities can disagree about a letter and the disagreement is a small thing, correctable in principle by anyone who checks. But two authorities can disagree about where the stress falls in a word, and the disagreement propagates through the whole melody of a verse, through its chanting in a room full of people, and through the shape of how the sentence is heard.
Accent is where a Masoretic school is most socially embedded, because accents are sung. A textual variant nobody ever hears aloud can persist for centuries without consequence. A variant in the position of a stress accent is a variant in the tune of a verse, and a tune is learned, argued over and defended. The Ben Naphtali disagreements are, in a real sense, the sound of a rival congregation.
Here the standard narrative runs into its first serious obstacle. The record of the disagreements survives not as a codex but as lists and quotations: masoretic compilations, and later writers such as David Kimhi and Jedidiah Norzi who transmitted what their predecessors had transmitted. And those records do not always agree with one another about the precise nature of any given difference. Where the lists are fuzzy and the disagreements are numerous, the received text cannot be expected to follow either system with perfect consistency. The most straightforward explanation of the evidence is also the least flattering to the standard narrative: the received text does not follow uniformly the system of Ben Asher, nor does it follow uniformly the system of Ben Naphtali, and this is probably because it never followed either of them.
This is the point at which the received text stops being a recension and becomes a compilation. Ben Asher’s reputation eventually won the argument. His codex became the standard text of the Hebrew Bible, and the standing of Tiberias in later halakhic practice reinforced that outcome. But winning the argument is not the same as supplying every leaf that was thereafter copied. For reasons that nobody has fully reconstructed, neither the printed text nor any manuscript which has been preserved is based entirely on Ben Asher. They are all, in the phrase that recurs in the literature, eclectic.
The other great axis of disagreement is the vowel system, and here the two traditions differ in kind rather than in degree. The Tiberian vocalisation, the one the received text uses, is a carefully specified system in which the quality and the length of a vowel are distinguished with a precision that took generations of grammarians to develop and that the accent signs cooperate with. The Babylonian system, simpler in some respects and different in others, marks vowels according to a different logic, with its own conventions for the treatment of a shewa and for the lengthened reading indicated by the ga’ya. Neither system is a corruption of the other. They are two answers to the same question, produced by people who took the question seriously.
Twentieth century scholarship did much to make sense of this. Saul Lieberman, working on the masorah and on the masoretes themselves, argued that the Tiberian reading was a genuine, coherent and teachable tradition, transmitted by people who knew what they were doing and who could say what they were doing. Geoffrey Khan, working on the reading tradition and on the Tiberian pronunciation tradition, took the argument further and reconstructed, from the masoretic literature and the evidence of reading, a system of Tiberian pronunciation that is coherent at a level of detail that had never before been demonstrated. The two bodies of work are not identical in every emphasis, and specialists continue to disagree about particular conclusions, but they share a central claim that matters here.
The shared claim is that the Tiberian tradition was a system, and that a system is something people could learn, teach, copy and enforce. That is precisely what makes A15 significant. If the Tiberian reading was a system, then an alternative system is not noise. It is not a set of random errors accumulated by careless provincial scribes. It is a rival answer produced by people with comparable training, comparable care and, evidently, comparable confidence. The existence of A15 forces the question of which system was right to be asked as a real question about rival scholarship, rather than as a formality about a text already known to be settled.
It must be said, however, that A15 is not Ben Naphtali’s codex. The codex that Ben Naphtali is generally supposed to have produced has not survived. A15 is a relative, not the thing itself, and the degree to which it faithfully represents him is a matter of judgement. Some specialists treat its readings as substantially Ben Naphtalian; others regard it as a codex of the Ben Asher sphere with a Babylonian-style vocalisation laid over a Galilean hand, a different and more interesting claim. Scholarship is divided, and the honest position is to report the division rather than to resolve it.
There is a further complication that resists tidying. The school of Ben Naphtali is traditionally associated with Babylonia, but the scribal habits that A15 actually carries are classed as Galilean. Whether the Ben Naphtali tradition should be located in Babylonia proper, or whether the label describes a school defined by its opposition to Tiberias rather than by its own geography, is not settled. The relationship between Tiberian and Galilean practice, and the degree to which the two were in dialogue rather than opposed, is an active question in the field. A15 sits directly in the middle of it.
What can be said without hedging is this. Before A15, the rival tradition existed in the standard account chiefly as a list of eight hundred and seventy-five differences, a discrepancy to be explained away. After A15, it exists as a book. There is a particular kind of evidence in a physical object that cannot be reproduced in an apparatus entry, however carefully the entry is compiled. An apparatus tells you that two authorities differed at a point. A codex tells you that somebody sat down with a whole Bible, in the rival system, and did the work properly, with the full apparatus and the checking and the centuries behind the practice. That is a different kind of claim, and it lands differently.
The rival witness, what survives, what is lost, and what must be kept
Begin with the awkward fact. No complete codex of Ben Naphtali has survived. The standard accounts are explicit about it: Ben Asher wrote a standard codex, Ben Naphtali likely did as well, and Ben Naphtali’s has not survived. What survives instead is a set of lists and quotations, uneven, sometimes contradictory, and stripped of the physical context that would let a reader see how the differences actually functioned on the page.
The reason no Ben Naphtali codex survives is not mysterious once the standard narrative is taken seriously, and that is exactly what makes it troubling. Ben Asher’s text became the standard. Once it was the standard, rival codices stopped being worth copying. And once they stopped being worth copying, the generations that came after had no reason to preserve them, because to them they were not rivals but superseded. The better the decisive victory, the less of the defeated text remains. Survival in the manuscript tradition is not random, and it is not a fair sample. It is a sample of what people thought would matter, filtered through what they could afford to keep.
A15 survived for reasons that have nothing to do with merit. It survived because it lay in a Crimean Karaite holding, in a community that was shrinking rather than growing, in a region whose libraries were being dispersed by economic and political change rather than by any decision about the Hebrew Bible. It survived because a collector with a scholarly project and the resources to travel was buying material of this kind. It survived because a bequest carried it into an institutional collection in Budapest that still exists and is still properly catalogued. Any one of those links could have failed. The codex we have is the product of a chain of accidents, and the accident happened to land on the one manuscript of its class that anyone has found.
It is important to be exact about what A15 did and did not do in the history of editions. It was not the basis of the printed Hebrew Bible. That honour belongs elsewhere, and the printed text descends through a different line altogether. What A15 appears to have been, in one decisive episode, is one of the principal variant witnesses consulted for the third edition of Biblia Hebraica, published in 1937. The third edition took the Leningrad Codex as its base text, and A15 was collated against it as a witness from outside the line of that base.
The circumstances of that edition make the point sharply. When Paul Kahle was revising Biblia Hebraica in the 1920s, he tried and failed to obtain a photographic copy of the Aleppo Codex, and the refusal of access forced him to use the Leningrad Codex for the third edition that appeared in 1937. The most authoritative Ben Asher manuscript in existence was unavailable to the very editor who most wanted it. In that situation, a complete Bible of the tenth century written outside the Ben Asher line was not a curiosity but a working instrument, and A15 is generally understood to have filled exactly that role in the collation. The irony is exact. The Ben Asher case was made in the base text; the rival case was made in the apparatus.
Today’s critical editions lean elsewhere. The Aleppo Codex and the Leningrad Codex are the primary witnesses for the received text, and the reasons are understandable. They are the two witnesses that agree with the mainstream tradition most consistently, and the Leningrad is complete and, for many generations, reasonably accessible. But they are not the same kind of evidence, and neither is whole. The Aleppo is a Ben Asher witness damaged in the twentieth century, with a substantial portion of its leaves lost and its contents reconstructed. The Leningrad is complete but is of a later vintage and a Ben Asher line. A15 is complete, and different.
Three witnesses, three shapes of evidence, three different sets of blind spots. A modern editor who wants to know what the tenth century looked like has no single codex that will answer the question. The Aleppo answers for one school and is incomplete. The Leningrad answers for another and is later. A15 answers for a third and is a relative rather than a specimen. That is not a failure of the sources. It is the actual shape of the evidence, and it has to be edited around rather than wished away.
The archivist’s reflex, when confronted with a case like this, is simple and slightly cold. Preserve the rival witness. Not because you expect to need it, and not because you have a taste for heterodoxy, but because the alternative is to make your own conclusion unfalsifiable. The moment a discipline decides that one school was correct and the other was wrong, the evidence for having been wrong ceases to be available. It is not that the losing manuscripts are suppressed. It is that they stop being interesting. Once a reading is a mistake, its witnesses are no longer witnesses to anything except their own errors, and the discipline has quietly lost the ability to check itself against the possibility that it was mistaken.
I have found this to be the most durable lesson of working with manuscripts, and it is a lesson about method rather than about any particular text. A codex in the wrong system, kept carefully, collated honestly, described accurately, and never quietly explained away, is a permanent check on the certainty of the winning system. It costs a great deal to keep. It costs shelf space in Budapest, a place in a three-volume catalogue, a line in every apparatus, and a small persistent irritation in the footnotes of editions that would prefer not to be reminded. It is worth every one of those things.
So return to the object. A15 is a complete Hebrew Bible, copied in the early tenth century, in the tradition that lost, with a vocalisation and a letter form that say so on the page before a word is read, and with the checking apparatus of the masoretes running through the margins to prove that its copyist intended it to be right. It reached Europe from a shrinking Karaite community on the Black Sea, through a nineteenth century market in old books, into the collection of a Hungarian scholar who knew what he was buying, and from there into an Academy library that catalogued it properly and kept it.
The Masoretic Text, we are told, is a single thing, fixed by a single man, and the only question worth asking about it is how faithfully later copies have preserved it. A15 is the physical proof that this is not quite true. Not because it corrects the received text, which it does not, and not because it should have been printed, which it should not. But because it is a complete book, of the right age, from the other side of the argument, and because as long as it sits on the shelf in Budapest with its margins intact, the claim that there was ever only one Masoretic Text has to be reargued rather than assumed. The tradition had a rival at the moment it supposedly closed. The rival had a codex. The codex is still there.
References
- “Kaufmann Manuscript” - article on the Kaufmann Collection, the Hungarian Academy of Sciences Library in Budapest, shelfmark MS A50, and the collection’s provenance; consulted for the collector’s identity, the Academy’s holding, and the 1987 Van Gorcum and Fortress Press source volume
- “Masoretic Text” - article on the Masoretic Text, the age of the masoretes, Ben Asher and Ben Naphtali, the approximately 875 differences between them, the Tiberian and Babylonian centres, and the forms of the masorah parva, masorah magna and masorah finalis
- “Aleppo Codex” - article on the Ben Asher standard codex, its keeping in the Aleppo synagogue, the loss of leaves in 1947, and Paul Kahle’s failed attempt to obtain a photographic copy for Biblia Hebraica
- “Leningrad Codex” - article on the codex used as the base text for Biblia Hebraica from 1937, and on the origin of its modern name
- “Chufut-Kale” - article on the Crimean Karaite fortress town east of Bakhchysarai, its meaning as Jewish Fortress, and its decline after the Russian annexation of 1783
- Die Handschriften der Sammlung Kaufmann - Koester, Beate, Die Handschriften der Sammlung Kaufmann, 3 vols, Assen/Maastricht/Philadelphia, Van Gorcum and Fortress Press, 1987; the codicological catalogue of the Kaufmann Collection, Hungarian Academy of Sciences Library, Budapest