Translation equivalence
A Socratic walk-through of translation equivalence — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can a faithful translation differ greatly from the wording of its source?
We picture translation as substitution: find the other language's word for each word, keep the order, and the meaning arrives intact. Yet the translations we most admire often stray furthest from the wording, while the closest ones read as stiff or plainly wrong. That is an odd result if faithfulness means matching words. So what is a translator being faithful to?
Reasoning it through
REASONING #Start with the smallest unit and ask whether it even holds. Does every word have a counterpart? Take one as ordinary as "blue". Russian has no single everyday word covering it — siniy and goluboy split that stretch of colour in two, and a Russian speaker naming a sky must pick one. Going the other way, whichever the translator picks adds or loses something the original never decided. If words carve the world at different joints, word-for-word matching is not neutral; it is a run of small forced choices.
Now go up a level, to what grammar obliges you to say. Roman Jakobson put it sharply: languages differ less in what they may express than in what they must. English marks tense on nearly every verb; other languages do not, and some oblige you instead to mark how you came by the information — whether you saw it, or were told. A sentence that left something politely unspecified in the source may be untellable in the target without specifying it. The translator can only decide, and the decision is theirs, not the author's.
Then the layer above meaning: what the words do. Take a pun, a rhyme, a name that is also a joke. Reproduce the sense and the effect dies; reproduce the effect and the words must change. Which would you call faithful? Most of us, reading a comic novel, prefer the translator who found a new joke to the one who footnoted the old one — and notice what we just admitted: we were never measuring fidelity in words at all. The same holds for what a reader already knows; a reference that landed lightly at home may land abroad as a puzzle, and keeping it light means replacing it.
So the translator arbitrates among claims that cannot all be met at once: sense, form, effect, and the job the text was written to do. Texts rank them differently. A drug label puts unambiguous sense above everything; a poem may put sound and shape above literal reference; a slogan is judged only by whether it persuades, which is why slogans are re-written rather than translated.
The analogy
THE ANALOGY #Think of arranging a violin piece for the flute. A flute cannot sound two notes at once, so the arranger breaks a chord into a run, moves a line an octave, changes what is written — and a listener who knows the original says: yes, that is the piece. A note-for-note copy would be unplayable or thin. Faithfulness plainly lives in the effect rather than the marks, and nobody calls the arranger unfaithful for changing notes.
a melody exists independently of the instrument, so an arranger has something fixed to be faithful to. A text's meaning is not detachable that way — it is partly made by the words, which is why translators disagree about what the invariant even is. And a flute's limits are physical, whereas a language's are conventions a bold translator can sometimes stretch, coining a phrase the target language did not have.
Clarifying the model
THE MODEL #Two corrections. First, "free" is not the opposite of "accurate": literal and free renderings can both be inaccurate, and the disciplined question is which claim this text most needs kept, and whether the translator spent the others knowingly or by accident. Translation studies names the poles roughly as formal versus dynamic equivalence, and adds a strand that starts from the text's purpose; the durable insight is that fidelity is relative to a chosen aim, not that any school settled it.
Second, "untranslatable" is usually overstated. It rarely means the thought cannot be conveyed, only that no equally short, equally ordinary expression exists — so conveying it costs a clause, a footnote, or a gain made elsewhere. Translators call that compensation, and the honest ledger is drawn across the whole work rather than line by line.
A picture of it
THE PICTURE #How to readStart at the centre and read outward: each of the four branches is a different thing the word "faithful" can mean, and the leaves are the concrete claims a translator weighs. The picture is not a checklist to satisfy — it is a map of competing demands, and the point is that pulling hard on one branch usually slackens another. To judge a translation, first decide which branch this text most needed kept, then ask whether the translator spent the others deliberately.
What became clearer
WHAT CLEARED #A translation is not a copy in other words; it is a decision about which of several kinds of sameness to preserve when they cannot all survive. Wording is the most visible of them and often the least important, which is why a faithful translation can look, line by line, remarkably unlike its source.
Where to go next
ONWARD #- How translators of poetry decide between keeping the metre and keeping the images.
- Why machine translation is strongest where sense dominates and weakest where effect does.
- What retranslating a classic every generation says about the target language rather than the source.
Key terms
TERMS #| Term | What it means |
|---|---|
| Formal equivalence | a translation aimed at matching the source's wording and structure as closely as the target language allows. |
| Dynamic equivalence | a translation aimed instead at producing on its reader an effect comparable to the original's. |
| Compensation | deliberately recreating elsewhere in the text an effect that could not be preserved at the point where it occurred. |
Every term the collection defines is gathered in the glossary.