The Word for Both

translation criticism  ·  one Hebrew word, a dozen English ones, and its own opposite

There is a word the Hebrew Bible turns on — the steadfast love God keeps because he has bound himself to keep it. It is in the book 251 times, and no English word holds it. So the translators broke it: the Greeks into one word, the English into a dozen. And in two verses the same word means disgrace.

Ask a Hebraist for the most important word in the Hebrew Bible and a great many will say this one. חֶסֶד ḥesed is the quality God claims for himself at the burning heart of the book — the loyal, unearned, covenant-keeping love that does not let go. It is the refrain of the longest litany in the Psalter and part of God's self-description in Exodus 34:6. And it has no English equivalent. Not one. The history of trying is the history below.

Two facts set the problem. First, the word is common and load-bearing: it occurs 251 times in the Masoretic text — counted, here, straight from the tagged Leningrad Codex, not taken on report. Second, it is untranslatable in the strict sense: it bundles together things English keeps in separate words — love and loyalty, kindness and obligation, the warm feeling and the binding promise — so that any single English word drops most of it. A translator must choose what to lose.

This piece watches that choice happen — twice over, in opposite directions. The King James, working from the Hebrew, found no word big enough and so shattered ḥesed into about a dozen English words. The Septuagint, begun with the Torah in the third century BCE, did the reverse: it flattened ḥesed into a single Greek word — ἔλεος, pity — and that flattening became the Latin misericordia and the English mercy we still read. And underneath both: a homonym so sharp it reads like a warning.

The word for both

Begin with the strangest fact, because it is the one most easily checked. The Masoretes supplied the vowel pointing. The modern OpenScriptures Hebrew Bible tags two different words spelled חסד. The lexicons call them ḥesed I and ḥesed II. The first is the covenant love above. The second means shame, disgrace, a reproach. They are not just the same three consonants. In their plain dictionary form they are the same pointed word — חֶסֶד, segol and segol, vowels and all.

It happens exactly twice, and you can read both. In Leviticus the law against marrying one's sister calls the act a ḥesed — “a wicked thing.” In Proverbs, “sin is a ḥesed to any people” — a reproach. The very letters that, in Exodus, name God's abounding love name, in Leviticus, an incest. Toggle between them:

Instrument · the cruxone pointed word · two senses
חֶסֶד
ḥesed I  ·  the same five marks

The lexica (Brown–Driver–Briggs; Koehler–Baumgartner) treat these as two roots that fell together — ḥesed II is usually linked to an Aramaic and Syriac verb meaning to shame, to put to reproach. James Strong's 1894 concordance filed both under one number (H2617) and let its gloss run, tellingly, all the way from “(loving-) kindness, merciful (kindness), mercy” to “reproach, wicked thing.” The split here follows the augmented tagging in the OpenScriptures Hebrew Bible, which marks the two senses 2617 a and 2617 b. The count is what it is: 249 of the love, 2 of the shame.

The word for the love that keeps its promise is spelled exactly like the word for the promise broken. Hebrew left them to be told apart by the sentence alone.

The collapse — one Greek word

In the third century BCE, in Alexandria, Jewish translators translated the Torah into Greek. Other books followed, forming the Septuagint, the Bible of the Greek-speaking world and, later, of the early Church. Faced with ḥesed, they reached, overwhelmingly, for a single word: ἔλεος eleos — mercy, pity, compassion. One lexical study counts 213 of ḥesed's 245 occurrences rendered by ἔλεος — about 87%, rising past 90% in the Psalms and the Prophets.

It is the wrong word, and the wrongness is precise. ἔλεος is the feeling of the fortunate toward the wretched — what you extend to a beggar, the emotion an audience owes a tragedy. It is a mood. ḥesed is a bond: not pity flowing downhill but the faithfulness owed inside a relationship that both sides have entered. The Greek kept the tenderness and lost the covenant. And because Jerome's Latin Vulgate carried ἔλεος over as misericordia — miser + cor, “a heart for the wretched,” pity again — the entire Latin West read God's defining attribute as mercy for a thousand years. Even the King James, translating freshly from the Hebrew, kept the inheritance: its single most common rendering of ḥesed is the Septuagint's choice, wearing English clothes — mercy.

Watch one verse make the whole journey. Psalm 136 is built from a single refrain, repeated 26 times — כִּי לְעוֹלָם חַסְדּוֹ, “for his ḥesed is forever.” Click each layer to see the word it carried, and the verse, verbatim:

Instrument · the transmission corePsalm 136 :1 — “for his ḥesed is forever”
The covenant word leaves Hebrew as ἔλεος (pity), travels through Latin misericordia into English “mercy,” and holds for fifteen centuries — until the modern versions reach back for the bond the Greek dropped. Public-domain layers quoted verbatim; the RSV refrain is briefly quoted (in copyright).

The shatter — a dozen English words

The King James translators were better Hebraists than the Greek of Alexandria gave them credit for, and they could feel that mercy would not cover every ḥesed. So where the Septuagint had collapsed the word into one, they did the opposite — they let it fracture, choosing a different English word almost sentence by sentence: mercy here, kindness there, lovingkindness, goodness, favour, pity, kindly. True to each verse, and false to the fact that the Hebrew had used one word the whole time. Click any rendering to see the King James verse it sits on — every one a place where the underlying Hebrew is the single word ḥesed:

Instrument · the shatterone Hebrew word → the KJV's many
חֶסֶד
one word, in the Hebrew, every time

Tap a rendering above to see its King James verse, verbatim.

The last two chips are the homonym: the King James, following the dominant reading, read those two verses as ḥesed II and rendered them reproach and a wicked thing. They belong in the spray because they are spelled the same; they belong apart because they mean the reverse. The instrument shows both at once: a word so capacious it took a dozen English words to carry, and so double it carries its own negation.

The repair

The long climb back began early. In 1535, Miles Coverdale, making the first complete printed English Bible, hit ḥesed in the Psalms and found — as everyone did — that no English word fit. So he built one: he set loving beside kindness and wrote louinge kyndnesse. The Oxford English Dictionary's earliest record of lovingkindness is that 1535 Psalter; the King James kept it; it is in the language now because a translator needed it and English did not have it. A word coined to patch a hole left by a word that would not translate.

The deeper repair waited four centuries. The Revised Standard Version (Old Testament, 1952) made a decision the Septuagint had made impossible to inherit: it rendered ḥesed, as its standard, “steadfast love” — two words, holding both halves, the warmth and the bond. Most modern translations followed. The covenant sense the Greek filtered out in Alexandria was, at last, back in the English.

What the word meant all along is its own scholarly story, and an honest one to tell because it is unsettled. Nelson Glueck's 1927 dissertation argued that ḥesed was essentially obligation — the conduct owed within a covenant, almost a legal term. Katharine Doob Sakenfeld's 1978 study pushed back: ḥesed is help given freely, to someone in real need, by the one positioned to give it — not contractual, but not unbound either; loyalty that chooses to act. The truth the translators kept stumbling on is that the word lives exactly where English splits the world in two — between the feeling and the duty, the gift and the promise — and refuses to land on either side.

The Greek made it a feeling. The Latin made it a feeling. The English spent four hundred years climbing back toward the promise.

So there is no villain here and no error, only the cost of carrying a word across a border it does not fit through. The Septuagint chose the tenderness and lost the bond; the King James kept the bond by giving up on one word; Coverdale minted a word; the moderns reached back. Read in the original, ḥesed occurs 251 times across divine and human contexts, including the two shame uses. Read in translation, it is the sound of every language that received it deciding, again and again, which half of the love to keep.

The apparatus

Verifier — research/chesed/verify.mjs · 19/19 PASS, offline.
The Hebrew counts are recomputed, not cited: 251 occurrences of the lemma ḥesed (Strong's H2617) in the tagged Leningrad Codex (OpenScriptures Hebrew Bible), split 249 (ḥesed I) / 2 (ḥesed II) by the source's augmented tags. The centrepiece — that Exodus 34:6 (love) and Leviticus 20:17 (disgrace) are the same pointed word חֶסֶד — is checked codepoint for codepoint. The 26-fold Psalm 136 refrain is counted. Every King James rendering in the shatter sits on a verse independently confirmed to carry H2617, with Psalm superscriptions reconciled (Hebrew vs. English versification).
Honest edges

What is recomputed vs. cited. Recomputed from primary tagged text: every Hebrew count, the two-sense split, the pointed-form identity, and that each quoted verse carries ḥesed. Cited (marked as such in the prose): the Septuagint's ~87% rendering of ḥesed by ἔλεος (a published lexical count, on a corpus of 245 — slightly fewer than the Masoretic 251, the usual edition-and-homonym variation); the Vulgate's misericordia; the OED's dating of lovingkindness to Coverdale 1535; and the Glueck / Sakenfeld debate.

On the count itself. Published totals for ḥesed range from roughly 245 to 251 depending on the text edition and whether the homonym and a few textually disputed forms are counted. The figure here, 251, is exactly what the OSHB-tagged Leningrad Codex yields for lemma 2617 — stated as a reproducible count of one specific text, not as the single “correct” number.

On the two senses. ḥesed II is a genuine, if minority, lexical entry, not a scribal quirk; BDB and HALOT both list it. A handful of scholars have proposed reading even Leviticus 20:17 as a faded sense I (“an act between kin”); the dominant reading, and the OSHB lemma tag followed here, is the disgrace sense. The crux is real, not manufactured — but it rests on two verses, and the piece says so.

On copyright. The Hebrew (Leningrad Codex), the Septuagint (Swete), the Clementine Vulgate, and the King James (1611/1769) are public domain and quoted verbatim. The RSV (1952) and other modern versions are in copyright; their renderings are named and described, never quoted at length.

Sources

Hebrew text + morphology — OpenScriptures Hebrew Bible (OSHB / morphhb), the Westminster Leningrad Codex with Strong's tagging. github.com/openscriptures/morphhb
Strong's H2617 gloss — J. Strong, A Concise Dictionary of the Words in the Hebrew Bible (1894), Open Scriptures JSON. github.com/openscriptures/strongs
Septuagint — H. B. Swete, The Old Testament in Greek (Ps 135:1–2, verbatim). The 213/245 equivalence figure: lexical studies of ἔλεος / ḥesed in the LXX.
Latin — Clementine Vulgate (Ps 135:1), public domain.
English — King James Version (1611/1769), public domain; lovingkindness dated to M. Coverdale's Bible (1535) per the OED.
Lexicography — Brown–Driver–Briggs; Koehler–Baumgartner (HALOT). Studies — N. Glueck, Das Wort ḥesed (1927); K. D. Sakenfeld, The Meaning of Ḥesed in the Hebrew Bible (1978).
Type — Noto Serif Hebrew (SIL OFL), GFS Didot (SIL OFL), self-hosted subsets.

Corrections

Corrected 2026-10-14: The page called ḥesed the first thing God says about his character. Exodus 20:5 and 22:27 already describe his character; ḥesed belongs to the self-description in Exodus 34:6 (found by the assay line, 2026-10-14). Exodus in the KJV.

Corrected 2026-10-14: The page placed the Septuagint a thousand years before the King James. The Greek Torah dates to the third century BCE, and other books followed; the King James appeared in 1611 (found by the assay line, 2026-10-14). Dorival on the Septuagint.

Corrected 2026-10-14: The page attributed the homonym tags to the Masoretes. The lemma tags used here are modern OSHB data; the Masoretes supplied the vowel pointing (found by the assay line, 2026-10-14). OSHB documentation.

Corrected 2026-10-14: The Strong's gloss was quoted as “(loving-)kindness, mercy”. It reads “(loving-) kindness, merciful (kindness), mercy” (found by the assay line, 2026-10-14). Strong's H2617.

Corrected 2026-10-14: The page said the King James read the two shame verses “like everyone”. Some interpreters read Leviticus 20:17 as loyalty or kindness (found by the assay line, 2026-10-14). AJS 2006, page 277.

Corrected 2026-10-14: The page described 251 occurrences as repetitions of God's claim about himself. The total counts ḥesed across divine and human contexts, including the two shame uses (found by the assay line, 2026-10-14). Ruth 3:10 and Leviticus 20:17 in the KJV.

Corrected 2026-10-14: The Greek Psalms row was dated “3rd–2nd c. BCE”. The cited study dates the Greek Psalms to the second or first century BCE (found by the assay line, 2026-10-14). Olofsson on the Psalter.

Corrected 2026-10-14: The page said the RSV rendering was named but not quoted. The transmission core briefly quotes the refrain from Psalm 136:1 (found by the assay line, 2026-10-14). Psalm 136 in the RSV.

Corrected 2026-10-14: The layer's markdown said “No third-party request.” The served page includes a Cloudflare beacon script from another domain; the fonts are self-hosted (found by the assay line, 2026-10-14).

The claims record 135 claims re-read against their sources, 14 October 2026

Written 2026-06-18. Claims re-read against their sources on 2026-10-14: 135 checked, 115 confirmed, 20 wrong, 0 unverifiable. By the assay line (research/assay-line/cited-1014b/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex.

This was the directed corrective pass over all 21 items in research/assay-line/cited-1014b/the-word-for-both/triage.json, against this layer's authored HTML, inline JavaScript and markdown at f5e5f84f34. The fixer re-read the evidence for every edit, applied 12 WRONG fixes and one MINOR fix, left seven listed MINOR findings unchanged, and rejected X11 with evidence. This record also carries forward all 114 CONFIRMED lines from the earlier Codex N2 check, explicitly attributed to that check rather than presented as 114 fresh source re-reads by this fixer. The 20 in `wrong` comprises 12 WRONG and eight MINOR lines. Every WRONG or MINOR has its own finding. X11 is CONFIRMED. The inherited interpretive confirmations retain N2's qualifications. Ellipses in copied claims or assessments mark omitted wording or punctuation. The 44 OBSERVED and three UNVERIFIABLE entries from N2 remain in that source record and are outside this directed triage; no new disposition is claimed for them. See the layer's fix-report.md for the edit ledger, exact correction lines, checks and limitations.

This record is long (135 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.