Roadmap

The goal is simple to say and long to do: get the untranslated Galen into readable English, in public, with the source text alongside every claim.

How we pick targets

• No existing English. Not the Loeb, not CMG, and not an unpublished thesis either: a translation nobody indexed is still a translation. The atlas exists partly so effort goes where nothing exists.
• A digitized source text. Machine-readable Greek (or a published Arabic edition) so work can start immediately, with the original published next to the translation.
• Bounded and self-contained. A book or treatise of finishable length, with a definite end in view.
• Human interest. Texts that show ancient life — recipes, patients, prices, complaints — over technical scaffolding, at least at first.

Phase 0 — Map the corpus live now

Publish the atlas, with translation status flagged by confidence, and open it to corrections. The unglamorous foundation: you can't plan a rescue without a map.

Phase 1 — Pilot: the hair & scalp book one chapter, then paused

Book 1 of On the Composition of Drugs According to Places (Kühn XII 378–497): Galen's remedies for baldness, falling hair, hair dye, depilatories, dandruff, lice, and scalp sores. No published English rendering of these chapters is known. The preface is drafted and unreviewed; the remaining nine chapters are untouched. Work moved to the strand below before they were begun, and saying otherwise would misdescribe the state of the project.

Phase 1b — Collation on a text with two traditions where the work actually went

The commentary on the Aphorisms survives in Greek and in Ḥunayn's Arabic, both anchored to the same Kühn pagination and to numbered Hippocratic lemmata, so the two collate mechanically. One verse is published with both traditions set out. This strand, not the drug book, is what tested the method — and what drew the correction that rewrote the witness protocol, including an admission that our own showcase example had presented a construal as though it were a variant. It also means the Arabic path has already been exercised end to end, which Phase 3 below no longer waits on.

Phase 2 — First full campaign shortlist checked twice

The first sustained translation effort will be chosen publicly. The shortlist has been checked twice. On 12 August 2026 the four candidates were checked against the BBAW translation catalogue, which lists each work with an empty English column. On 17 August they were checked again, this time outside the catalogue — and one candidate came off. On the Differences of Fevers has been translated in full already: John Ware rendered both books, and De temperamentis after them, in a University of Edinburgh thesis of 1928 that was never published and can be downloaded today. On Antidotes turns out to be partly done by the same route: Book 1, chapters 1–5 and 7 appear in Mucci's 2023 Warwick thesis, leaving chapter 6 and the whole of Book 2. Both theses were read, not merely cited; the atlas records which pages. The pulse treatises remain withdrawn because Johnston and Papavramidou's 2024 translation covers them.

The lesson is worth stating plainly, since it nearly cost a campaign. BBAW indexes published translations, so an unpublished one is invisible to it: the catalogue shows De differentiis febrium as untranslated into any language, and a century-old thesis says otherwise. Checking the catalogue is necessary and it is not sufficient. Institutional repositories are where the second look belongs, and corrections from anyone who knows of a translation we have missed remain the most useful thing a reader can send.

• On the Composition of Drugs According to Places — continue past Book 1 (eyes, ears, teeth, chest… ten books of ancient pharmacy).
• On Antidotes — theriac, mithridatium, and the emperor's daily dose; the founding text of the antidote tradition. Book 1 is partly translated already, so a campaign here would begin at chapter 6 and Book 2.
• On Crises — how a Roman doctor predicted the turning point of a disease.

Vote on the ballot →

Phase 3 — The Arabic frontier later

Works lost in Greek that survive in the Arabic of Ḥunayn ibn Isḥāq's ninth-century Baghdad workshop — some translated by scholars already, some (like On Medical Names, which has only a German version of 1931) still without English. The tooling is no longer the obstacle: Phase 1b already reads Arabic alongside Greek. What these need is competence we do not have in-house, which is why the arrangement described under review below matters more here than anywhere else.

How a campaign runs

1. The source text is split into chunks — passages anchored to Kühn page numbers.
2. Each chunk gets an AI-assisted draft that compares accessible witnesses, asking of any divergence which underlying Greek reading would account for it, and distinguishing a variant in the text from a construal of it.
3. Draft, source text, citations, concise translation rationale, and notes are published side by side, flagged as a draft until a qualified human reviewer signs off.
4. Anyone can challenge a rendering by opening an issue pointing at the Greek — the fastest way to improve the text.

Review, and what is honestly in place partly arranged

Every draft on this site carries a flag saying it awaits a qualified reviewer. For a long time that was an aspiration rather than an arrangement, and the flag was doing more work than the reality behind it. The position now: one substantive correction has been taken from a specialist in the Greek-into-Arabic tradition and applied across the site, and paid occasional advice from a doctoral researcher with both Greek and Arabic is being arranged. Nothing here has yet been signed off by a reviewer, and every draft still says so.

Infrastructure: recording divergences as data designed, not built

Divergences between witnesses are currently prose. The intention is to record each as structured data — what every witness reads, which underlying Greek would account for it, and whether the claim is a variant in the text or a construal of it — so that the distinction is enforced by the format rather than by the drafter's memory, and so the corpus can answer questions like “where does the Arabic imply a variant?” at scale. This is the part of the work an amateur project is actually well placed to do: not philological judgement, but systematic and auditable bookkeeping.

Contributing via pull request

The whole site is a public repository of plain HTML, JavaScript, and versioned JSON — no framework, no build step. Fixing the catalogue means editing works.json with a dated source for any claim marked checked; adding a translation means filling a source-pinned packet from the harness. Preview locally with python3 -m http.server 8000; one work or one chunk per pull request, sources cited; merged means live.

Details for contributors: CONTRIBUTING.md.