Translation Principles

0.1 For whom we translate

This project translates Theravāda commentarial literature — principally the Visuddhimagga of Buddhaghosa and the Chinese Vimuttimagga — into English and Chinese for contemporary non-specialist readers, not primarily for academic Buddhologists.

The reader we write for:

• English version: a thoughtful Western adult with no Pāli, no assumed Buddhist background, but curiosity and patience. Could be a meditator, a graduate student outside religious studies, a philosophically literate general reader.
• Chinese version: a college-educated reader from anywhere in the global Chinese-speaking world (世界華語讀者) — Mainland China, Taiwan, Hong Kong, Macau, Singapore, Malaysia, or diaspora communities — comfortable with 白話文, without assumed 文言文 reading fluency, without assumed 佛教學 training. Could be a practitioner, a student, a curious intellectual.

Neither reader should need to put the book down to consult a dictionary of Pāli or a concordance of Taishō numbers to follow the main thread. Technical apparatus supports the reading; it does not gatekeep it.

0.2 "A hundred years from now, we are the scholar."

The phrase from Max that founds this project. It carries two commitments:

First, we are not slavishly reproducing the conventions of the prior standard translations of the past century. Each of those translators was of their time and made defensible choices for their readers. Our reader is the reader of 2030 and beyond. Where existing conventions obscure rather than illuminate, we translate afresh.

Second, this is not an excuse for carelessness. The translation must be rigorous enough that when it becomes the default reference a generation from now, it will hold up to scholarly cross-examination. Precision is preserved; obscurity is not.

0.3 Relationship to existing translations

We consult but do not derive from:

• The standard English scholarly translation (1956) — in copyright until ~2030.
• The standard Chinese scholarly translation (1987) — in copyright.
• The standard English Vimuttimagga translation (1961) — in copyright.

We use these for terminology cross-reference and doctrinal sanity check. We do not copy phrasing, structure, or section breaks from them. Our translation is an independent creative work licensed under CC BY-NC-SA 4.0.

---

1.1 English register

Target: modern academic English, clear and direct, at roughly the register of The New York Review of Books or Charles Taylor's philosophical prose. Precise without being ponderous. Formal without being archaic.

Rejected registers:

• ❌ Victorian / 1920s PTS style: "whosoever desireth to cultivate the meditation subject..."
• ❌ Dense Pāli-laden academicese: most sentences containing 3+ Pāli terms in parentheses
• ❌ "Dharma-speak" informality: "the Buddha was all like..."
• ❌ New Age devotional: "Beloved Awakened One, whose radiance..."

Accepted models for calibration:

• Contemporary clarity-first translations (definite article use, modernist tone)
• Recent scholarly Nikāya translations (precision of technical distinctions)
• Contemporary philosophically-literate prose (minus the secularizing content): unafraid of long sentences

Concrete rules:

• Short to medium sentences preferred. Buddhaghosa's Pāli often runs to 80-word periods; break these into 2–3 English sentences rather than preserving the original syntax.
• Use contractions sparingly (acceptable in prefaces/introductions, avoided in main translation body).
• Avoid "one" as generic pronoun; use "you" for instructional passages, "the practitioner" or specific subject for descriptive passages.
• Latin abbreviations (e.g., i.e., viz.) allowed in footnotes, avoided in main text.

1.2 Chinese register

Target: 現代書面中文 (modern standard written Chinese) with technical Buddhist vocabulary, calibrated for the global Chinese-language readership (世界華語讀者) — Mainland China, Taiwan, Hong Kong, Macau, Singapore, Malaysia, and diaspora communities. Region-neutral; clear, precise, not archaic; not flavored by any one regional house style or sectarian lineage.

Rejected registers:

• ❌ 文言-inflected scholarly-standard style: "彼於爾時作如是思惟..."
• ❌ Japanese-syntactic-traced translations: unnatural Chinese phrasing
• ❌ 漢傳傳統 commentary style: "今者此句正明..."
• ❌ Popular Buddhist publishing style: emotionalized or devotional tone
• ❌ Region-coded vocabulary or idiom (e.g., 台灣 corporate-Buddhist phrasing, 大陸 simplified-only colloquialisms, 港式 Cantonese-influenced syntax)

Concrete rules:

• 白話 syntax. Short-to-medium sentences. Punctuation per PAPER_STYLE.md全形 rules.
• Retain technical Buddhist vocabulary where 漢傳 tradition has precise terms: 戒、定、慧、五蘊、十二因緣、八正道, etc. — these are pan-regional standard.
• Avoid 文言 grammatical constructions (之、乎、者、也 patterns). 「的、是、了」 in modern usage.
• Primary script: 繁體 (Traditional Chinese) as source of truth — read natively in Taiwan/HK/Macau and remains accessible to 簡體 readers with mild adjustment. Auto-conversion to 簡體 for distribution channels that require it; never maintain divergent content between scripts.
• When a term has different conventional renderings across regions, choose the form most widely intelligible across all regions, with footnote noting alternates.
• When a traditional 漢傳 translation is imprecise, use direct Pāli transliteration with gloss on first occurrence, then continue with the transliteration.
• Pronoun rule: 祂 for 佛陀, 阿賴耶識, 如來藏 (per book-phase rule). Otherwise 他 / 她 / 它 per standard usage.

---

2.1 Three tiers

Tier A: Loanwords that have entered English — keep transliterated, no italics.

• buddha, dharma/dhamma, karma/kamma, nirvana/nibbāna, saṅgha, bodhi, samsara/saṃsāra, sutra/sutta
• These function as English words now. Italicizing them makes the prose jumpy. Gloss on first occurrence in each chapter.

Tier B: Technical terms where translation serves the reader better than transliteration.

• sīla → virtue / ethical conduct (context-dependent)
• samādhi → concentration (consistently; one footnote per chapter on the upacāra/appanā distinction when relevant)
• paññā → wisdom
• sati → mindfulness (with footnote on how this differs from contemporary secular mindfulness)
• anussati → recollection (never "mindfulness-of"; the directionality matters)
• jhāna → absorption (for Vism work; "jhāna" as transliteration acceptable in popular work)
• bhāvanā → cultivation / meditative development
• khandha → aggregate
• nāmarūpa → name-and-form
• paṭicca-samuppāda → dependent origination
• saṅkhāra → formation (contextual: mental formation, volitional formation, conditioned thing)

Tier C: Keep transliterated with gloss; these lose too much in translation.

• metta, karuṇā, muditā, upekkhā (goodwill/compassion/sympathetic-joy/equanimity — loanwords plus English gloss on first occurrence)
• kammaṭṭhāna (meditation subject; English gloss first occurrence)
• upacāra-samādhi / appanā-samādhi (access / absorption concentration — first occurrence in each chapter defines)
• vitakka / vicāra (applied thought / sustained thought)
• pīti / sukha (rapture / pleasure — but see GLOSSARY.md for the case against "rapture")
• dukkha (traditionally "suffering"; keep as dukkha because "suffering" systematically misleads Western readers about the technical scope)

2.2 First-occurrence gloss protocol

On first occurrence of a Tier B or Tier C term in each chapter, render as:

vitakka (applied thought) ... [thereafter: "applied thought" or "vitakka" as flow demands]

In Chinese:

vitakka（尋，applied thought）... [thereafter: 尋 or vitakka as flow demands]

This is lighter than full parenthetical apparatus on every occurrence but heavier than zero support. The GLOSSARY.md is the normative reference.

2.3 Chinese technical term strategy

For Chinese, the 漢傳 tradition already has precise equivalents for most terms. Defaults:

• Use 漢傳 term where it is both traditional and accurate: 戒 = sīla, 定 = samādhi, 慧 = paññā, 五蘊 = khandha, 念 = sati, 隨念 = anussati.
• Use direct Pāli transliteration where 漢傳 term distorts: macchera → 「macchera（慳吝）」rather than just 慳 (which in modern Chinese lost its technical sting).
• Use modern Chinese gloss where both 漢傳 and transliteration are opaque: upacāra-samādhi → 近行定 (with explanation in first-occurrence footnote).

2.4 Enumerated-formula first-occurrence policy (Pāli-italic + lemma form)

Locked 2026-05-28 (v1.0 polish pass; tasks/PRE_V1.0_POLISH_PASS.md Item 2 — Max-ratified Option D + devas-form name-register ruling). Sets convention for any chapter containing an enumerated formula of proper Pāli names: deva-realm taxonomies (VII.6, Ch IX Brahmavihāra, Ch XI deva-references), attribute lists, and the like.

When the source presents an enumerated list of proper Pāli names — e.g. the eightfold deva-class enumeration Cātumahārājikā · Tāvatiṃsā · Yāmā · Tusitā · Nimmānaratī · Paranimmita-vasavattī · Brahmakāyikā (+ tat'uttariṃ, "higher than that"):

1. Uniform name-register, both languages. Render each name in one uniform register across BOTH the English and Chinese translation lines. The ruling register is devas-form: a-stem names take the nominative-plural long ending (Yāmā, Tusitā, Cātumahārājikā); consonant -in-stem names — which have no -ā plural — take the -ī citation form (Nimmānaratī, Paranimmita-vasavattī), never the source genitive-plural -ino. A verbatim Pāli source-quotation (blockquote) preserves the original case-forms (Nimmānaratino / Paranimmita-vasavattino); the register normalization applies only to the rendered translation lines.
2. Uniform Pāli-italic on first occurrence. Every member carries its Pāli in italic on first occurrence, uniformly across the whole list — no selective italic-some / bare-roman-others. Bare thereafter. Where a descriptor accompanies the name, the Pāli sits in italic parentheses ("the Thirty-Three (Tāvatiṃsā)" / 「三十三天（Tāvatiṃsā）」); where the name is used directly, the name itself is italic ("the Yāmā" / 「夜摩天（Yāmā）」).
3. Structural framing is not forced uniform. The choice between descriptor-plus-Pāli and direct-Pāli-name follows per-member readability; only the name-register and italicization are required uniform. (This is Option D's deliberate scope over Options A/B, which would also have forced a single structural frame.)
4. Naturalized loanwords stay roman. Standalone deva / devas / brahmā used as English loanwords remain roman (non-italic), consistent with project-wide usage; italic is reserved for Pāli technical compounds (devatā, devatānussati) and for enumerated proper class-names on first occurrence per (2).

Cross-ref GLOSSARY §XXVII (VII.6 deva-class enumeration). Closes the deferred tasks/PRE_V1.0_POLISH_PASS.md Item 2.

---

3.1 English

• "The Buddha" always capitalized as title; never "The Buddha."
• "he" lowercase for the Buddha when pronoun-required. This matches modernist practice and avoids Christian-divine overtones that "He" would import.
• Monks: "the venerable Sāriputta," "Sāriputta" after introduction; not "Ven. Sāriputta" in running prose.
• Laywomen/men: "the layman Citta," "Citta" after introduction.
• Devas/brahmas: capitalize proper names (Sakka, Brahmā), lowercase class terms (devas, brahmas).
• No Roman italics for Pāli proper names: Sāriputta, Mahākassapa, Buddhaghosa — these are names, not terms.

3.2 Chinese

• 祂 for 佛陀, 阿賴耶識, 如來藏 (book-phase rule, extended here).
• 他 for 論師 (Buddhaghosa, 龍樹, 玄奘 etc.), 比丘, 居士, 天人.
• 她 for 比丘尼, 女居士, 女性天人.
• 它 for non-sentient references (器世間, 法, 境 when treated as objects).
• Titles: 佛陀 / 世尊 / 如來 (context-appropriate), 阿羅漢 (not 阿羅漢尊者 in running prose).
• Monk names: 舍利弗, 目犍連 — use standard 漢傳 forms.

---

4.1 Primary source: PTS Rhys Davids 1920–21 for Pāli, CBETA for Chinese

Pāli Visuddhimagga: Pali Text Society edition by C. A. F. Rhys Davids (London: PTS, 1920–21), in 2 volumes. Public domain (>100 years post-publication). PTS pagination anchors form the citation backbone of this project (see §5.1).

Digital text acquisition: We do not have access to a faithful pre-existing digital transcription of the PTS edition (SLTP digitizes only the Tipiṭaka proper; BuddhaDust digitizes only the Nikāyas and Vinaya; GRETIL is license-incompatible). We therefore generate our own digital text via OCR of the Archive.org scans of PTS 1920–21, refined by multi-engine OCR cross-checking and LLM-assisted diacritic correction against the source images. The resulting pali.md files are derivative of a public-domain source through a mechanical transcription process; per Bridgeman v. Corel (US, 1999) and parallel principles in EU jurisdictions, no new copyright attaches to faithful transcription of public-domain text. Our cleaned text is released CC BY-NC-SA 4.0 alongside the translations.

Verification protocol: Each ingested chapter is sample-verified against the original PTS 1920–21 scans on Archive.org. Sampling: 5 paragraphs per chapter, distributed across the chapter's range. Pass threshold: ≥98% character-level match per paragraph AND chapter average ≥98%. Diacritic differences are not normalized in the comparison — a missing ṃ or ñ is a real error. Failures trigger paragraph-level re-verification; chapters that cannot reach 98% are escalated rather than silently accepted. Verification logs live in the chapter's notes.md. The OCR pipeline itself is documented in tools/ingestion/.

CSCD variant readings: The Vipassana Research Institute Chaṭṭha Saṅgāyana (CSCD / Burmese Sixth Council) edition (tipitaka.org) is consulted as a secondary source for variant readings. CSCD represents a different recension lineage than PTS, and we do not use it as base text — doing so while citing PTS pagination would be a lineage incoherence. Where CSCD differs substantively from PTS in a way that affects translation, the variant is recorded in notes.md. CSCD readings do not enter pali.md.

Chinese Vimuttimagga (T1648): CBETA (Chinese Buddhist Electronic Text Association) edition of the Taishō Tripiṭaka. Public domain.

Forbidden sources for direct textual derivation:

• GRETIL Pāli edition (CC-BY-SA 4.0 — share-alike conflicts with our CC-BY-NC-SA). Usable as reading aid only; no text enters our files.
• Any source whose license is non-permissive or ambiguous.

Revision note (v0.2, 2026-04-15): The v0.1 document named SLTP as the primary Pāli source; this was factually erroneous (SLTP does not include commentarial literature). An interim revision named BuddhaDust as the digital transcription source; this was also erroneous (BuddhaDust digitizes only canonical texts). The current §4.1 reflects the actual situation: no clean public-domain digital transcription of PTS Vism exists, so we generate one through OCR-of-scans with multi-stage validation. Source-provenance assertions in this project are now backed by direct verification, not assumption.

4.2 Parallel text structure

Each chapter / section has four files:

• pali.md — source Pāli with PTS anchors and paragraph numbering
• english.md — translation
• chinese.md — translation
• notes.md — translator's decisions, variant readings, cross-references

The english.md and chinese.md files are standalone readable — a reader should be able to follow the translation without the Pāli open beside them. Key Pāli terms appear inline on first occurrence (per §2.2); extensive Pāli parallels live in pali.md for those who want them.

4.3 Diacritics

Unicode IAST throughout. Full diacritics: ā ī ū ṃ ṅ ñ ṭ ḍ ṇ ḷ. No simplified transliteration, no ASCII-Velthuis (aa, .m, etc.) anywhere in published files.

---

5.1 PTS pagination inline

Primary citation: inline brackets with PTS reference.

...and so buddhānussati reaches only access concentration [V i 197].

For passages spanning pages: [V i 197–198].

5.2 CSCD / VRI paragraph numbers as secondary

In notes.md only, for cross-verification with VRI CSCD. Not in main translation body.

5.3 Sūtra / canonical references

Standard scholarly: AN 11.11, SN 11.3, MN 10, DN 22, T1509 (Taishō number for Chinese canonical references), preceded by full title on first occurrence in chapter.

5.4 Cross-reference to four-gates books

When a passage is specifically drawn upon by a 四行門 volume, notes.md logs: "Used in NIAN P03 §3.4 for X argument." Running index lives in top-level CHANGELOG.md.

---

7.1 Semantic versioning per chapter

Each chapter versions independently. Version string in chapter README:

• v0.x — draft, used internally for book writing; not publicly advertised
• v1.0 — first public release, aligned with corresponding 四行門 book publication
• v1.x — patches for corrections, no interpretive changes
• v2.0 — substantive revision (new edition, significant reinterpretation)

7.2 Changelog granularity

Top-level CHANGELOG.md logs chapter version bumps with date and brief description. Per-chapter changes (paragraph-level edits) tracked via git, not the changelog.

---