04 · Five Kinds That Do Not Translate

In 645 Xuanzang came back to Chang'an after sixteen years travelling overland to India, and what he left behind as a translator was not a theory of how to translate but a list of five kinds of word that must not be translated. The list is called 五种不翻 — "five kinds that do not translate" — five categories of term to be transliterated rather than rendered by meaning.

Roughly two and a half centuries before that, another man did the same job by the opposite method. Kumārajīva (344–413), of Kuchean and Indian parentage, arrived in Chang'an in 401 and led the translation bureau there. Over about eleven years his team produced a run of sutras, among them the Lotus Sutra, finished in 406 in seven fascicles, which became the standard East Asian version from then on, and the Diamond Sutra, translated between 402 and 412.

Kumārajīva put meaning ahead of wording. One exchange with his disciple Sengrui is on record. Sengrui had rendered a line literally as 天見人，人見天, "heaven sees man, and man sees heaven." Kumārajīva would not have it, and had it changed to 人天交接，兩得相見, "man and heaven connect, and the two are able to see each other." Elsewhere he compared rendering Sanskrit incantations literally to 嚼飯與人, chewing up rice and feeding it to someone else: the food is pre-digested before it reaches the mouth, so none of the taste survives.

Xuanzang (602–664) worked the other way. He set out overland for India, stayed there sixteen years, returned to Chang'an in 645, and founded the Yogācāra school with his disciple Kuiji — the same school China calls Faxiang, Weishi and Ci'en — and the Daci'en Temple, established in 648, became its ancestral seat.

The difference between the two methods can be measured on a single line of the same sutra. In the Heart Sutra, Kumārajīva's version (Taishō T223) reads 非色異空，非空異色, a doubled negation: "it is not that form differs from emptiness, it is not that emptiness differs from form." Xuanzang's later version (T251) reads 色不異空，空不異色 — the same meaning, differently built, saying directly that form is not different from emptiness instead of going around through the double negative. The chiastic core line, 色即是空，空即是色, is identical in both versions and does not move at all.

Beyond that line, what separates the two men is terminological consistency. Xuanzang set a standing Chinese equivalent for each technical Sanskrit term and pinned it there, while Kumārajīva used a more fluid idiom and followed the meaning. The wording the source uses is that Xuanzang is "generally described as" combining literal with free method rather than being purely literal, and this book does not tighten that phrase beyond what the source wrote.

The list of five is not advice. It is a sorting rule, decidable one word at a time: this word falls under one of the categories, or it falls under none of them. The five, as recorded:

1. Secrets — dhāraṇī (陀羅尼), Sanskrit ritual speech and mantras.
2. One word, many meanings — Bhagavant (薄伽梵) carries six meanings at once: sovereignty, ablaze, solemnity, fame, auspiciousness and esteem. No single Chinese word covers all six.
3. Nothing of the kind in China — the jambu tree (閻浮樹) does not grow on Chinese soil, so Chinese has no name of its own for it.
4. Follow what is already there — terms earlier translators had transliterated until the transliteration stuck, such as Anouputi (阿耨菩提) for anuttara-samyak-saṃbodhi.
5. Preserve the weight of the word — keep Prajñā (般若) rather than translating it as 智慧, the Chinese word for wisdom, which sits lighter.

Only one of the five is about a missing word. Category three alone concerns a thing China does not have; the other four are about something else entirely. Category one says the word works by sound and not by sense — translate it and the sound is gone. Category two says the source language lets one word hold six meanings at the same time, and the target language forces the choice of one. Category four says the term already has a history in the target language — render it afresh and the reader can no longer trace it back. Category five says the register of a word is part of what that word says, and lowering the register says something else.

A rule for what not to translate is stronger than any theory of translation, for two reasons. First, it can be wrong in a way anyone can see: take a word in category two, render it in one Chinese word carrying all six meanings, and category two falls on the spot. A theory holding that translation is a negotiation between source and target can never be knocked over that way, because nothing sticks out to knock. Second, it is operational. A translator stuck on a single word reads this list and knows what to do with that word on the next line.

What this book's dossier does not have is the text that records the five categories, or that text's date. The dossier gives all five categories and an example for each, but does not say who wrote them down, where, or when. What can be held is Xuanzang's own lifespan, 602 to 664, so the rule belongs to the seventh century as far as the evidence in hand certifies, and no further.

This book's contract follows Xuanzang's fourth category exactly, without having set out to. The requirement is that technical terms and proper names stay in their original form in both the Thai and the English — evidentiality, parthenos, kawanatanga, BLEU, and 五种不翻 itself. The reason is not laziness, and it is not taste. These terms already have a history where they came from. A reader who meets evidentiality on this page can type it out and walk from here to the source work; a reader handed a freshly coined Thai word instead cannot walk anywhere, holding a word nobody outside this page uses.

The seventh century answered one of this book's questions already. The answer was not that everything translates, and not that nothing does. It was to sit down and write out what cannot cross, then lift those things out and set them aside in their original form. Whoever does that takes one measured loss in front of him in order to keep it from spreading through the whole book.

What could not cross: The sound of a dhāraṇī, and the six meanings Bhagavant holds at once. Sanskrit lets a syllable work as sound without having to mean anything, and lets one word hold several meanings simultaneously. Chinese obliges the translator to pick one meaning, and obliges him to write it out in characters that always mean something.

Overturned if: Someone renders Bhagavant in a single Chinese word that a Chinese reader takes in all six meanings without a footnote — category two falls. And if someone points to the text that records 五种不翻 and that text is dated outside Xuanzang's lifetime, this entry's attachment of the rule to the seventh century falls with it.

Sources:

• Wikipedia, "Chinese translation theory" — https://en.wikipedia.org/wiki/Chinese_translation_theory
• Wikipedia, "Kumārajīva" — https://en.wikipedia.org/wiki/Kum%C4%81raj%C4%ABva
• Wikipedia, "Lotus Sutra" — https://en.wikipedia.org/wiki/Lotus_Sutra
• Xue, "Study on Xuanzang's Translation of Heart Sutra" — https://www.semanticscholar.org/paper/Study-on-Xuanzang's-Translation-of-Heart-Sutra-%E2%80%94-of-Xue/04b07be9bdd8bcfe889969b5dae027ba8404c9b5