What Survives When a Poem Is Translated — and What That Proves

Vladimir Nabokov, finishing his English Eugene Onegin in 1964, did what a novelist of his gifts had no business doing: he threw away the music. He rendered Pushkin’s diamond-cut stanzas into flat, literal, footnote-shackled prose, and let a commentary swell across four volumes until it dwarfed the poem it served. He called every rhyming version a fraud. Edmund Wilson, his friend of two decades, attacked the result in print; the friendship did not survive it. The episode gets filed under literary feud. It reads better as a forced confession. Nabokov had stared straight at the question every translator eventually faces and refused the comfortable lie. What, exactly, is the thing a poem is, once you take its own words away?

That sounds like a parlor puzzle and is in fact the sharpest instrument we have for testing a far larger claim: that literature carries something beneath its sentences at all. We talk loosely about a novel’s vision, a poem’s soul, the spirit of a line. These are flattering nouns, and flattering nouns are cheap. Translation is where the bill comes due. To move a poem into another language is to strip it to the studs and ask which timbers were load-bearing. The words go; they must, since the new language has different ones. If anything stands after that demolition, we have evidence — not proof, but real evidence — that the poem was never only its words.

The two easy answers fail

Two tempting answers present themselves, and both buckle under load. The first says what survives is the meaning. But meaning, in poetry, is welded to sound and shape in ways no paraphrase can pry apart. Take Hopkins: ‘The world is charged with the grandeur of God.’ Render the proposition faithfully — the world is full of divine glory — and you have kept the sense and lost the poem. The charge, the metallic shock of that verb, both electric current and moral summons, is the meaning, inseparable from the syllable that delivers it. The second answer says what survives is the form: keep the sonnet a sonnet, the iamb an iamb. But forms are local. A French alexandrine and an English pentameter do not measure the same breath, and a Japanese count of morae is not a count of stresses at all.

A poem is instructions for an experience, written in disappearing ink.

So neither the words, nor the meaning, nor the form crosses intact. This looks like a verdict against translation, and many honest people have reached exactly that verdict. Frost’s much-repeated quip, that poetry is the thing lost in translation, gets quoted as a closed door. Dante, in the Convivio, warned that nothing harmonized by the Muses can pass from its own tongue into another without breaking its sweetness, and offered the Psalter as his proof. These are not careless men. If poetry is purely an effect of one language’s particular sounds, then a translated poem is a taxidermied bird: the shape of flight and none of it. The case against the whole enterprise is strong, and any argument worth making has to walk straight through it rather than around.

What actually crosses

Walk through it, then, and look at what does arrive on the far shore. The plainest case is narrative and image. When Seamus Heaney brought Beowulf into modern English in 1999, the Old English alliterative line and its dense kennings — the sea as the whale-road — could not survive whole. But the shape of the hero’s three fights, the cold ceremony of the funeral pyre, the particular ache of a civilization watching its own dusk: these arrive undiminished, sometimes sharpened. Heaney opened with a single blunt syllable, ‘So.’, the voice of a man clearing his throat to tell you something true. That word was not in the original. It was a discovery the original made possible — a structure of feeling the words had been carrying, and that a different set of words could carry again.

Push to the harder case: the lyric, where image thins and sound takes over. Here the evidence is subtler and, I think, more decisive. Take the ghazal, a Persian and Urdu form built on a repeated end-word, a chiming rhyme just before it, and the poet’s own name folded into the final couplet. Agha Shahid Ali spent his life smuggling the form into American English, insisting the refrain and rhyme be honored, and he showed that a structure of return and recognition — the same word arriving again and again, each time changed by what preceded it — lands in English much as it lands in Urdu. The specific sounds are gone. The architecture those sounds were building is portable. That architecture is the thing beneath the sentences.

The turn

Here is the turn, and it inverts the whole complaint. Translation seems to destroy poetry because we have been measuring at the wrong altitude. We compare word to word, sound to sound, and naturally find loss everywhere, since words and sounds are precisely what cannot cross. Run the comparison one level up — at the level of pattern, relation, gesture, the choreography of attention a poem performs on a reader’s mind — and the picture reverses. The losses are real but local. The survivals are structural and large. A poem is not a string of beautiful words; it is a set of instructions for an experience, encoded in words. The encoding is untranslatable. The instructions, often, are not.

This explains a strange fact: some poems survive translation better than they survive paraphrase in their own language. Rewrite Emily Dickinson into smoother English prose and you destroy her; carry her into French and a real Dickinson can appear, because the translator is forced to ask what the dashes and the hymn-meter were doing, not merely to copy them. Constance Garnett’s Russians all read a little like Garnett, and Joseph Brodsky scorned the smoothing — yet Dostoevsky’s moral vertigo reached millions in English who could not read a syllable of the original. The vertigo lives in the deep structure. It survived a translator who, by exacting lights, was flattening the prose as she went. That survival is the experiment returning a positive result.

What the failures prove

The cases that fail teach as much as the ones that succeed, because they show where the load was actually bearing. Pure sound-poetry does not cross: the nonsense lullaby, the patter where the syllables are the whole event, the pun that is two meanings at once. Finnegans Wake resists translation not because it runs too deep but because, in places, there is nothing under the wordplay to recover — the surface is the substance, which is a rare and genuine achievement. When a poem cannot be translated at all, we have learned something precise about it: that it staked everything on its own language’s particular body and kept no architecture in reserve. Untranslatability is not always a compliment. Sometimes it marks a poem that is all skin and no skeleton.

“Poetry is what is lost in translation.”— attributed to Robert Frost (Untermeyer, 1964)

Frost’s line, then, is not wrong; it is half a thought. Something is lost — the body, the breath, the irreplaceable grain of the first sound. But the loss is the test’s whole point. We pour the poem from one vessel into another, and whatever evaporates was bound to the first vessel’s shape. Whatever pours through was free of it. The Septuagint’s Greek is not the Hebrew of the Psalms, yet the Psalms have crossed three thousand years and a hundred tongues, and the man weeping over them in Spanish is weeping at something that was in the Hebrew and is now also in the Spanish. That persistence is not sentiment. It is data.

The cleanest test

This is why translation is the cleanest test we have — cleaner than criticism, cleaner than theory. Criticism can declare a poem deep and never be falsified; the claim costs nothing. Translation puts the claim on trial with a real penalty. Strip the words and either something stands or nothing does, and the result is public, repeatable, checkable against the original by anyone who knows both tongues. A poem that survives many good translations has shown, under maximal stress, that it was carrying cargo beneath its surface. A poem that survives none has shown the opposite — or shown that its surface was its cargo, which is a different and narrower kind of greatness.

Nabokov, in the end, proved the case against his own intention. He withheld the music to protect the meaning, certain the two could be cleanly divorced. But readers went on loving Pushkin in other men’s rhyming English — in Charles Johnston’s deft 1977 version above all, which kept the impossible Onegin stanza intact — loving a structure of irony and tenderness and lethal lightness that Nabokov’s literalism had preserved as fact and killed as experience. The thing beneath the sentences turned out to be transmissible after all, just not by the method of refusing to transmit it. What survives a translation is never the words and never quite the meaning. It is the shape the words were making in the dark — and that shape, the evidence keeps insisting, was the poem all along.