How to transcribe a manuscript: the conventions

A transcription records what is written on the leaf, in type. The craft is in the honesty: every expansion, every guess, every hole in the parchment is marked, so a reader always knows what the scribe wrote and what you supplied.

A diplomatic transcription aims to represent selected features of a document closely, but every transcription applies an editorial model. Projects differ on letterforms, abbreviations, capitalization, punctuation, word division, lineation, damage, additions, and deletions. The conventions below form one workable house style rather than a universal scholarly core.

The first rule: type what you see

• Keep the original spelling, capitalization, and word division. Wrong spellings are evidence of the writer's dialect and ear.
• Decide whether u/v and i/j represent visual forms, normalized letters, or both through encoded alternatives.
• State whether capitalization, punctuation, and word division follow the document or an editorial normalization.
• State how line, column, page, and folio boundaries are represented. A plain-text pipe is one local option; TEI uses an <lb> milestone for a line boundary.

Abbreviations: the main event for older hands

Medieval and early modern hands use strokes, hooks, superscript letters, and special signs as abbreviations. An edition may preserve the abbreviation, expand it, or encode both readings. Italics and parentheses are local display conventions. TEI <choice> can pair an abbreviation with its expansion.

• Italic style: dominus. The scribe wrote dns with a stroke above; the italic letters are yours.
• Parenthesis style: d(omi)n(u)s. Same information, plain-text safe.

Both are respectable; pick by output format (italics don't survive every database field) and never mix them. If an abbreviation is ambiguous and the expansion is a judgment call, expand it and flag the doubt: p(er) and p(ro) hang on the direction of a single stroke.

Damage, gaps, and guesses

The following plain-text symbols are this guide's chosen policy. They overlap with conventions used in several editorial fields, but their meanings are not universal. Projects based on Leiden, EpiDoc, TEI, documentary editing, or repository transcription rules may encode the same evidence differently:

• [text]: letters lost to damage but restored by you: [con]uentus where a wormhole ate the start of the word.
• [...]: letters lost and not restorable; use one dot per letter if you can count the space, […] if you can't.
• [word?]: legible enough to guess, not enough to assert.
• [illegible]: you can see writing but can't read it. Say how much: [illegible, one word].
• ⟨text⟩: letters the scribe omitted by mistake that you supply: tem⟨p⟩us. Angle brackets keep your corrections distinct from damage repairs.

The distinctions look fussy and are the entire point: plain conuentus, [con]uentus, ⟨con⟩uentus, and [conuentus?] make four different claims about the leaf.

The scribe's own corrections

Manuscripts correct themselves: words struck through, letters dotted for deletion (expunction), insertions floated above the line with a caret, passages scraped away and rewritten. Record the event as well as the result:

• Deleted text: keep it, marked as ⟦propria⟧ or an explicit [struck through: propria].
• Inserted text: mark where it entered, as \su(per)script/ or [added above the line: nostri].
• Erased text: an erasure is [erasure], with the underlying reading supplied when it can still be made out.
• A later hand is worth naming when you can tell: [added in the margin, s. xv hand]. Corrections stratify a manuscript's history the way numbering systems do.

A worked line

Modern documents: the same honesty, a lighter kit

For nineteenth- and twentieth-century material (letters, diaries, ledgers, fieldbooks) the abbreviation problem largely disappears, and large transcription programs deliberately simplify the kit so that thousands of contributors produce consistent, searchable text. The Smithsonian's crowdsourced projects, for instance, use double brackets for every editorial voice, as in [[strikethrough]]hate[[/strikethrough]], [[image]], or [[Abigail Adams]] for an optional corrected spelling, and drop the fussier distinctions entirely: insertions are typed where you would read them aloud, and words hyphenated across a line break are joined.

The lighter kit is a policy tuned to its purpose rather than a lesser standard. Searchability rewards whole words and minimal clutter, and a diary legible to any modern reader doesn't need per-letter damage accounting. The scholarly kit above and the crowdsourcing kit make the same promise in different dialects: the reader can always tell scribe from editor. If you're contributing to an institutional project, its rules win; read the project's instructions before typing. For your own research transcriptions, borrow from either kit, and say which.

Cross-written letters, personal shorthand, faded pencil, and business cursive require project-specific choices. For cross-writing, state whether layers follow writing order, documentary position, or a reconstructed reading order. Do not infer an oral order without evidence.

State your policy, once

Because practice varies, every serious transcription travels with a short note: "Abbreviations expanded in italics; original punctuation and u/v retained; line breaks marked | ; damage in square brackets." Four clauses, written once, and every bracket in your file becomes unambiguous. A wrong reading can be corrected by the next scholar; a transcription whose rules nobody can reconstruct cannot.

For explicit encoding systems, consult the EpiDoc guidelines and the TEI chapter on primary sources. The UK National Archives' palaeography tutorial provides practice with English documents, while the Smithsonian transcription instructions document a different policy for collaborative modern transcription.