How an Ancient Text Reached Us: Work, Manuscript, Fragment, Translation and Edition

A work is not one physical object. It can be copied, acquire variants, move into another language and be quoted within another work. A manuscript is a particular material carrier: a codex, scroll, leaf or fragment produced at a particular time. It preserves not the work “in general”, but one of the states of its text available to us.

Scholars therefore speak of a manuscript witness. The term does not mean an infallible eyewitness; it means a source of particular readings. A witness may be a complete manuscript, a small fragment, an ancient translation or a quotation by another author. Each can answer only those questions for which it actually preserves evidence.

H54 is dated not through a guess about the age of the Didache but through its own colophon: the codex was copied in 1056, and the scribe is named as Leo the notary. This tells us when the codex was produced. It does not directly tell us when a work contained within it was first composed. That requires other evidence. For the Didache, this evidence does not produce one undisputed narrow date.

No single document can fill the chronological gap. H54 tells us reliably that we have a copy from 1056. An earlier fragment can prove that particular lines already existed at the end of the fourth century. It does not date the rest of the text or reveal the whole preceding history. The conclusion arises not from one discovery but from comparing witnesses of different dates and different scope.

The question of a “thousand years” does not require us to imagine an unbroken chain in which every copyist is known. Intermediate carriers may have disappeared. The researcher works with what actually survives and separates direct observation from inference: here is the date of the object, here are the legible lines, here are the relations between readings, and here is the boundary beyond which the evidence does not yet allow us to proceed.

H54 has an earlier interlocutor: P.Oxy. XV 1782. Despite the designation P.Oxy., it is not papyrus but two small parchment leaves from a miniature codex, each less than six centimetres high. They are dated to the end of the fourth century. Only two passages of the Didache survive on them: Did. 1.3c–1.4a and Did. 2.7b–3.2a.

The word “only” matters. The fragment directly confirms particular lines, not the whole book. It shows that these portions existed long before H54 was copied and allows their readings to be compared with the later manuscript. Two leaves, however, do not become an “early complete Didache”. Everything beyond their physical limits requires other evidence.

This gives us a first picture of transmission. A work does not travel through the centuries inside one miraculously surviving volume. We see a late continuous codex and earlier remnants, each preserving its own part of the history. What matters is not only the age of an object but also which text it carries, how much of it and in what condition.

The fragment’s small size does not make it insignificant. It fixes several things at once: the material of the carrier, the approximate date of production, the language and the readings of the surviving lines. Yet its strength is inseparable from its boundary. The more precisely we state what survives, the more reliably the fragment can participate in the larger picture — and the less temptation there is to make it testify to what it does not contain.

Copying is not instantaneous cloning. Spelling changes; a scribe may omit a line, repeat a word, confuse similar letters or correct what appears to be an error. Deliberate editing is sometimes possible. But the mere presence of a difference does not prove intentional corruption. Ordinary processes of writing and copying must be examined first.

A difference between two forms of one passage is called a variant reading. This is a neutral description, not a verdict. A variant may be an error, linguistic normalisation, an editorial alteration or a trace of another line of transmission. Its cause has to be argued separately. The age of a manuscript matters, but cannot be the only criterion: an early carrier may contain a later reading, while a later carrier may preserve an earlier one.

First Clement offers a particularly clear contrast. Codex Alexandrinus, a fifth-century manuscript, lost one leaf, and with it the passage 1 Clem. 57.7–63.4 is absent from that codex. H54 was produced about six centuries later but preserves the passage. The later manuscript does not thereby become “better in every respect”. It simply contains what was physically lost from the earlier one.

Manuscripts cannot therefore be placed in an automatic ranking by date. A researcher evaluates the preservation of a particular passage, the distribution of readings and which change best explains the rise of the alternatives. The same witness may be especially valuable in one place and less reliable in another.

Comparison always takes place at particular points. The forms of the text are first recorded without premature judgement: one witness reads this way, another differently, while a third has not preserved the passage at all. Only then do we ask which reading best explains the others and how it is distributed among the sources. This order protects against the convenient but false scheme in which one preferred codex is granted all the correct answers in advance.

Sometimes the principal extensive witness to a work survives not in the language of its early form. An ancient translation then becomes an important part of its textual history. It may transmit a large body of text absent from surviving manuscripts in the source language, while also reflecting the translator’s choices, the possibilities of another language and its own history of copying.

The Gospel of Thomas provides such a case. Its continuous Coptic text occupies pages 32–51 of Nag Hammadi Codex II, dated to the fourth century. The Greek witnesses P.Oxy. 1, 654 and 655 survive only in fragments. The Coptic version therefore supplies extent, while the Greek fragments offer direct access to particular passages in Greek.

But the Coptic codex does not become a Greek manuscript. A translation cannot mechanically yield one unique source word at every point: one Coptic expression may correspond to several Greek formulations, and the translator may have reshaped the syntax. The translation expands our knowledge, but its evidence has its own limits.

The Coptic text and Greek fragments are not two halves of a torn leaf. They pose questions to one another. A Greek passage helps to test a particular point in the translation; the continuous Coptic text shows the context of sections not preserved in Greek. The comparison becomes stronger precisely when the difference between the languages is not erased.

Quotation is another path of preservation. One verifiable route to Papias of Hierapolis runs through Eusebius: in *Ecclesiastical History* 3.39.4–6, Eusebius quotes Papias on the gathering of traditions. What survives is not a codex of Papias but a fragment of his text embedded in another author’s work.

Such a witness is both valuable and complex. The quoting author selects a passage, frames it and may abbreviate it. The locator must therefore lead not to an imaginary page of the lost work, but to the actual passage in Eusebius that preserves the quotation. Without further evidence, one extract cannot define the extent or structure of Papias’s whole work.

Once witnesses have been assembled, their readings are compared. The editor records what each source reads, evaluates their relations and explains the choice. The existence of a variant and the decision to prefer one reading belong to different levels: the first concerns the evidence; the second concerns the editor’s argument.

A critical edition is therefore not a photograph of one manuscript. It may present a text established through the comparison of several witnesses and show alternatives in an apparatus. One decision may rest on an early fragment, another on a late complete codex, a translation or indirect evidence. The method and aim of the edition must be named: different projects may attempt to reconstruct an early recoverable state, follow a principal manuscript or display several states in parallel.

This does not mean that editors have “restored the original” word for word. They do not possess one authorial page but a set of witnesses from different periods. Textual criticism proposes the best-supported reconstruction of a stated early textual state and indicates where confidence is limited. A strong edition does not hide the distance between document and conclusion; it makes that distance visible.

Three editorial layers should also be distinguished. A facsimile answers: what does the document look like? A transcription asks: which signs can be read on this page? A critical text asks: which form does the editor propose after comparing the sources? All three may appear in one edition, but the movement from image to reading, and from reading to decision, must not remain invisible.

A critical apparatus is not a storehouse of “errors beneath the line”. In compressed form, it displays alternative readings, their witnesses and sometimes the degree of confidence and the editor’s responsibility. The reader can examine not only the final sentence but the route that led to it. This capacity for verification distinguishes an argued edition from a text printed without an explanation of its origin.

On a screen these levels are easily confused. A digital image shows a page. A transcription renders legible signs in searchable type. A digital scholarly edition connects images, transcriptions, variants, identifiers and editorial decisions, explains its method and makes precise citation possible.

A scan can be invaluable: it opens a document to readers anywhere. But the presence of a scan or a website does not by itself make a scholarly edition. The decisive distinction is transparency. Can the reader understand which sources were used, how the text is linked to the image, who made the editorial decision and where the document ends and interpretation begins?

In a strong digital edition, the reader can move in both directions: from a convenient text to a particular witness, and from a page image to a transcription, variant and explanation of the choice. Stable identifiers make exact location possible, while accessible data allow the work to be checked again. The value of the digital environment lies not in the shine of the interface but in how clearly it reveals the structure of the evidence.

An ancient text does not reach the reader as one magically preserved “original”. What survives are objects and traces of different kinds: a complete late codex, an early fragment, a translation, a quotation, a variant, an old printed facsimile, a modern critical edition or a digital edition. They are not interchangeable, but together they make more precise questions possible.

The history of the Didache makes this work visible. H54 provides a continuous Greek text; two parchment leaves from Oxyrhynchus carry particular lines back to the end of the fourth century; comparison shows where the evidence is direct, where it is local and where editorial reconstruction begins. The central question is therefore not “Do we possess the original?” but: which witnesses survive, what exactly does each preserve and how far do they allow our conclusions to go?