The most important word in the book has no word in English. One translator gave it six. One gave it the same word every time. One reached for “virtue,” and said so — adding, on the record, that he meant it only for want of a better term.
Lionel Giles, translating the Analects in 1907, stopped to make a confession most translators only think. “The most important word in the Confucian vocabulary,” he wrote, “is jên, which in the following extracts is translated ‘virtue’ only for want of a better term.” The English word, he went on, “has so many different shades of meaning and is withal so vague” that it loses the one thing 仁 insists on.
He was not the first to find the word too big for its English. 仁 — rén — is the axis the whole Analects turns on; it appears 109 times across 59 of its passages. It is usually Englished as benevolence, virtue, goodness, humaneness, or love — and the reason there are so many candidates is that no one of them is right. The character is built from 人 (rén, “person”) and 二 (èr, “two”): the oldest dictionary glosses it simply as closeness, the good that arises between people. English has the abstract nouns or the warm verb; it does not have the single word that is both.
This page lays selected passages from three public-domain translations over the Chinese and colours their renderings of 仁, so the disagreement is visible at a glance. Giles’s version is an abridgement. Then it does the thing the venue keeps finding most damning: it shows that the disagreement is not only between translators. It runs inside a single one.
James Legge’s 1893 Analects is the scholarly standard, the version every later translator argued with. Reading straight through it, you meet 仁 rendered as a different English word almost every time the context shifts — not from carelessness, but because each English word fits the local sentence and none fits the whole. Here is the same character, in one book, by one hand:
The choice was never “which word is correct.” It was “which loss can I bear in this sentence” — and Legge bore a different one each time.
Now set Legge beside two others. Leonard Lyall (1909) made the opposite decision: he picked one English word — love — and held to it everywhere, letting the sentences strain rather than the word. Lionel Giles (1907) reached for “virtue,” “goodness,” “charity,” by turns — and confessed the reach. Tap through the passages; the colour marks what each did with 仁.
▷ Choose a passage. The Chinese 仁 is marked in red; each translator’s rendering of it is highlighted in its family colour.
The sharpest moment is XII.22. A disciple asks what 仁 is, and Confucius gives a two-word answer: 愛人 — ài rén, “love people.” The Master’s own gloss on the word is love. And yet the translators who refuse “love” as their headword for 仁 — who call it virtue, goodness, benevolence — are forced to write “love” here, in the answer, where the text gives them no room to avoid it. The word they could not settle on, the text settles in passing.
Why is the word so hard to carry? Because in Chinese it is not one idea but a knot of them, held together by a single graph. The oldest character dictionary, the Shuōwén Jiězì (許慎, c. 100 CE), gives the whole etymology in four characters: 仁，親也。从人从二。 — “仁: closeness; composed of 人 ‘person’ and 二 ‘two’.” Tap the parts, then open the four senses the one word carries — no two of which English keeps in the same word.
▷ The four lives of one word — each attested in the dictionaries, none of them kept together by a single English word:
The venue’s rule is to let the translators testify in their own words rather than be caught out. Two of them, generations apart, said the quiet part aloud.
“The most important word in the Confucian vocabulary is jên, which in the following extracts is translated “virtue” only for want of a better term. Our English word “virtue” has so many different shades of meaning and is withal so vague, that in using it, the idea of altruism is often hardly present to our mind.”
A Chinese translator, writing for Victorian readers, rendered 仁 throughout as “the moral life” and “the moral character” — deliberately naturalising it into the ethical vocabulary of Matthew Arnold’s England. In a footnote he conceded the trade: the word, he wrote, “means literally ‘humanity.’”
Set the three strategies side by side and you have the whole predicament. Legge let the word fracture into its English pieces — true to each sentence, false to the unity. Lyall welded it to one English word and made the sentences carry the strain. Giles split the difference and told you he was doing it. None was wrong. Each took a different loss — which is the venue’s recurring finding, here in its starkest form: a word can be too important to translate consistently, because consistency would betray it as surely as variety does.
Modern proposals such as “humaneness” and “co-humanity” try to keep the 人, the person, audible inside the virtue. Individual words and short phrases are not protected by U.S. copyright; the gallery’s restriction to public-domain translations is an editorial choice. These, too, are attempts to say in English what the dictionary said in four characters two thousand years ago: 仁 is what passes, for good, between two people. Giles reached for “virtue” for want of a better term. There is, still, no better single term.
research/analects-ren/verify.mjs re-extracts every Chinese passage from the ctext source files and every English quotation from the named Gutenberg / scan texts, and checks each one byte-for-byte against this page; it also recomputes the 109/59 count and the Shuōwén etymology, and asserts the honest-edge tags are present. Run: node research/analects-ren/verify.mjs. The Han glyphs are set in a self-hosted subset of Noto Serif CJK SC (SIL Open Font License) covering exactly the characters quoted. Font delivery makes no third-party request.
Corrected 2026-10-14: The VI.30 gallery cell said Giles’s rendering was “not isolable in this version”; his Individual Virtue section includes the passage and renders the moral term as “moral virtue” (found by the assay line, 2026-10-14). Giles.
Corrected 2026-10-14: The opening quotation printed “is 仁”; Giles wrote “is jên” (found by the assay line, 2026-10-14). Giles.
Corrected 2026-10-14: The apparatus said Legge, Lyall and Giles each number passages differently; Legge and Lyall use chapter numbers, while Giles arranges unnumbered passages by theme (found by the assay line, 2026-10-14). Giles.
Corrected 2026-10-14: The page described “three complete public-domain translations”; the gallery shows six selected passages from three translations, including Giles’s abridgement (found by the assay line, 2026-10-14). Giles.
Corrected 2026-10-14: The page said “No third-party request”; the fonts are self-hosted, but the live page loads Cloudflare’s analytics beacon (found by the assay line, 2026-10-14). Live page.
Corrected 2026-10-14: The page said copyright prevented quotation of modern rendering terms; individual words and short phrases are not protected by U.S. copyright (found by the assay line, 2026-10-14). U.S. Copyright Office, Circular 33.
Corrected 2026-10-14: The related-layer note attributed the horns on Michelangelo’s Moses to a missing vowel; the horned image belongs to the Vulgate translation tradition, and the verb qāran itself admits a horned reading (found by the assay line, 2026-10-14). Moses’ Shining or Horned Face?.
Corrected 2026-10-14: The markdown said righteousness and ritual were the two terms 仁 is “always paired against”; Analects IX.29 groups 仁 with wisdom and courage (found by the assay line, 2026-10-14). Analects IX.29.
Written 2026-06-17. Claims re-read against their sources on 2026-10-14: 132 checked, 112 confirmed, 20 wrong, 0 unverifiable. By the assay line (research/assay-line/queue-1014c/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex; repair review by Codex.
This was the directed queue-1014c repair pass for for-want-of-a-better-term, read at 85f512d191. Codex re-read the evidence for the triage items before editing, recorded all 21 decisions, and retained the 111 CONFIRMED claims that the N2 Codex checker re-read against the N1 Codex source record. The retained claims below explicitly attribute their evidence to N2’s pinned f5e5f84f34 pass; carrying them forward does not claim a new independent source visit for every one. This pass corrected this layer’s text for nine WRONG reports covering eight distinct errors and one MINOR item, left ten listed MINOR items unchanged, and rejected X8 on the source evidence. Repair review confirmed unannotated copies of the Moses and copyright/consensus claims in another layer’s preserved corpus. MINOR lines are included in wrong: 9 WRONG + 11 MINOR = 20. The rejected X8 is CONFIRMED, giving 112 CONFIRMED lines. X17/X21, X7/X20 and X13/X14 retain their shared error labels. One dated correction answers X17/X21. The N2 OBSERVED and UNVERIFIABLE entries outside the triage are not silently recast as confirmed claims in this record. The HTML is authored directly; there is no layer page generator. Only the layer’s own verifier mirror was refreshed from its source.
This record is long (132 claim lines and 20 findings), so it is not part of the page until you open it. It loads here now if your browser runs scripts; otherwise, or to keep it, read the whole record as plain text.
The assay office: what a record is, and every layer re-read so far.