I read a novel in Finnish translation last year that stopped me cold in the middle of a sentence. The original was Japanese, a mid-century work I’d read in English years before and found competent but unremarkable. The Finnish translator had done something I still can’t fully explain. The rhythm of the prose had changed. Not just the words, but the spaces between them, the way one sentence leaned into the next, the way paragraphs breathed. I set the book down and thought: this is better than the original. And then I thought: what does “better than the original” even mean when we’re talking about translation?

That question has haunted me for months. It gets at the heart of what literary translation is and what we should expect from it. Most readers, myself included for many years, think of translation as a kind of sophisticated code-switching. You take the meaning in one language and reproduce it in another. Accuracy is the goal. Fidelity is the standard. A great translation is one where you forget you’re reading a translation at all.

I’ve come to believe that framework is almost entirely wrong.

A literary translation isn’t a reproduction. It’s a performance. Think of it like music. When a pianist plays a Chopin nocturne, nobody expects them to play it exactly the way Chopin would have played it. We expect them to interpret it. To bring their own understanding, their own sensibility, their own technical gifts to the material. Two pianists can play the same piece and produce utterly different experiences, and both can be magnificent. The score is the starting point, not the destination. Great literary translation works the same way. The source text is the score. The translator is the performer.

This analogy isn’t original to me. Gregory Rabassa, who translated Gabriel Garcia Marquez and Julio Cortazar into English, talked about translation as performance decades ago. But I think the publishing industry has been slow to internalize it. We still evaluate translations primarily on accuracy. Reviews of translated works almost always include some version of “the translation reads smoothly” or “the translator captures the author’s voice.” These are fine observations as far as they go, but they treat the translator as invisible, as a pane of glass through which you view the original. The best translators I’ve worked with are nothing like glass. They’re more like stained glass: they transform the light that passes through them.

Let me get specific. When we were working on the English edition of a Central European novel a few years ago (I won’t name it because the rights situation is complicated), we had two sample translations to choose from. Both were technically accurate. Both came from translators with excellent credentials and long track records. The difference between them was in the sentences. Translator A had produced prose that was correct, fluent, and somewhat flat. You could read it without stumbling, but it didn’t make you feel anything. Translator B had produced prose that was stranger, less immediately smooth, but alive in a way that Translator A’s version wasn’t. Translator B had made choices. Where the original used a common metaphor, Translator B had found an unexpected English equivalent that wasn’t quite the same thing but hit harder. Where the original had long, subordinate-clause-heavy sentences, Translator B hadn’t simply reproduced that structure in English (which would have read as clumsy) but had found a different way to create the same sense of accumulation and density.

We went with Translator B. It was the right choice. The reviews praised the novel’s prose style, which was really the translator’s prose style, shaped by and in conversation with the original author’s intentions.

This raises an uncomfortable question. If the translator is making creative choices, if they’re interpreting rather than reproducing, how do we judge their work? If accuracy isn’t the primary standard, what is? I’ve thought about this a lot, and I think there are several qualities that separate great literary translation from competent literary translation.

The first is voice. Every author has a voice, a characteristic way of putting sentences together that’s as distinctive as a fingerprint. When you translate that voice into another language, you can’t simply mirror it. English doesn’t work like Japanese. French doesn’t work like Arabic. The syntactic structures, the available vocabulary, the cultural connotations of individual words are all different. A great translator creates a voice in the target language that feels like it could belong to the original author, even though the original author would never have written those exact sentences. It’s a kind of controlled ventriloquism. The translator speaks in someone else’s character, using their own language and their own craft.

I think of Ann Goldstein’s translations of Elena Ferrante. Goldstein writes English prose that is clear, direct, and emotionally precise. It reads nothing like Italian. It’s not trying to. But it captures something about the way Ferrante thinks, the way her sentences build emotional pressure through accumulation and repetition. You feel Ferrante’s intelligence and intensity in Goldstein’s English, even though the two languages create that feeling through completely different means. That’s voice translation at its best.

The second quality is attention to register. Register is the level of formality or informality in language, and it shifts constantly within a text. A character might speak in slang, think in elevated prose, and write letters in bureaucratic jargon. Each of those registers carries meaning. In the original language, those shifts are obvious to a native reader. In translation, they can easily get flattened into a single, undifferentiated “translated” voice. Great translators preserve the shifts. They find English (or French, or German, or whatever the target language is) equivalents that carry the same social and emotional signals as the original registers. This is extraordinarily difficult work. It requires deep knowledge of both languages at a level that goes way beyond vocabulary and grammar, into the sociology of how people actually speak and write.

The third quality is rhythm. Prose has rhythm the same way poetry does, just less obviously. The length of sentences, the placement of stressed syllables, the way commas and periods create pauses. In the original language, these rhythms are part of the reading experience. They speed you up or slow you down. They create tension or release. When you translate, you can’t preserve the original rhythms because different languages have different rhythmic structures. English tends toward shorter sentences and front-loaded emphasis. German allows for longer sentences with the verb kicked to the end. Japanese piles up subordinate phrases before arriving at the main clause. A translator who simply reproduces the original sentence structures in English will produce prose that sounds wrong, even if every word is correct. A great translator finds new rhythms in the target language that produce the same emotional effect.

This is something I didn’t appreciate until I started working with translators professionally. I used to think that the hardest part of translation was handling wordplay, idioms, and culture-specific references. And those things are hard, certainly. How do you translate a pun that only works in French? How do you handle a reference to a TV show that nobody in the target culture has seen? But these are surface-level problems with surface-level solutions (footnotes, creative substitution, strategic omission). The deeper challenge is capturing the music of the prose. I’ve seen translators agonize for hours over a single paragraph, not because they couldn’t understand it, but because they couldn’t make it sing in English.

The fourth quality, and this one is controversial, is willingness to depart from the source. Strict fidelity to the original is not always a virtue. Sometimes the original text uses a technique that doesn’t work in the target language. Sometimes a metaphor that’s fresh and surprising in Spanish is a dead cliche in English. Sometimes the original has weaknesses that a good translator can quietly fix. I know that last point will make some purists angry, but I’ve seen it done, and done well. A translator who spots a clunky transition in the original and smooths it out in translation isn’t betraying the author. They’re serving the author’s intentions better than the author served them in that particular passage.

The Italian phrase “traduttore, traditore” (translator, traitor) gets quoted in every conversation about translation, usually to warn against taking too many liberties. I think the phrase is more interesting than its usual usage suggests. All translation is betrayal, because all translation involves loss. The question isn’t whether the translator betrays the original, but which betrayals are productive and which are destructive. A translator who flattens a complex voice into bland fluency is betraying the original just as much as one who takes wild creative liberties. Maybe more so, because the bland version doesn’t even acknowledge that something has been lost.

At ScrollWorks, we publish a modest number of translated works each year, and the translation process is the most labor-intensive part of our editorial workflow. We start by having two people read the source text independently: one who reads in the original language and one who reads an existing translation (if available) or a rough literal translation. Both write reports on what makes the book work, what its voice sounds like, what its rhythmic patterns are, what might be difficult to render in English. Then we commission sample translations from two or three translators, usually 15 to 20 pages each. We read those samples not for accuracy (we assume competent translators will be accurate) but for life. Does the English version feel alive? Does it have its own energy? Does it make you want to keep reading?

The selection process is subjective, and I’m comfortable with that. If translation is performance, then choosing a translator is like casting a role. You’re not looking for the most technically skilled performer. You’re looking for the performer whose sensibility matches the material. I’ve turned down translators with decades more experience than the person we ultimately hired because the less experienced translator simply heard the original text more clearly. Experience matters, but connection to the material matters more.

Once we’ve selected a translator, the editorial process is collaborative in a way that monolingual editing isn’t. Our editor works with the translator through multiple drafts, reading the translation alongside the original (or alongside a literal crib, for languages our editors don’t read). The conversations are about feel as much as meaning. “This paragraph moves too fast in the English.” “This character’s dialogue sounds too formal; in the original, she speaks like a teenager.” “This image is beautiful in the source, but in English it sounds like a cough medicine commercial.” These are real notes from real editorial sessions. They’re the kind of notes that only make sense when you treat translation as a creative act rather than a mechanical one.

I want to mention compensation because it’s part of the picture. Literary translators are chronically underpaid. The standard rate for literary translation in the English-speaking world is somewhere between eight and twelve cents per word, depending on the language and the publisher. For a 300-page novel, that works out to roughly $6,000 to $9,000. A good literary translator might spend six months to a year on a book. Do the math. These are highly skilled professionals, usually with graduate degrees and deep expertise in multiple languages and literatures, earning less than minimum wage for their work. And yet the quality of literary translation in English is remarkably high. That’s a tribute to the translators who do this work because they love it, often subsidizing their translation income with teaching, freelance editing, or other work.

We’ve tried to do better at ScrollWorks. We pay above the standard rates, and we offer royalty participation so that translators benefit from a book’s success. We also credit translators prominently on covers and in marketing materials. “Translated by” should appear on the front cover, in my opinion. The translator is a co-creator of the English edition, and readers deserve to know who they are. I’ve bought books specifically because of the translator. When I see that Jhumpa Lahiri has translated something from Italian, or that Margaret Jull Costa has translated something from Portuguese, that name on the cover tells me something meaningful about what the reading experience will be like.

The technology question is impossible to ignore. Machine translation has improved dramatically. Google Translate in 2022 produces output that would have been unimaginable ten years ago. DeepL is even better for many language pairs. And there are more specialized AI translation tools emerging every year. Can machines produce literary translation? Not yet. Not even close. Machine translation can handle the code-switching part, the conversion of meaning from one language to another. What it cannot do is perform. It cannot hear the music of the source text and compose new music in the target language. It cannot make the creative choices that give translated prose its life. I’m sure machine translation will continue to improve, and I’m sure it will become useful as a first-draft tool for translators, the way a pianist might use a recording as a starting point for their own interpretation. But I don’t think it will replace human literary translation in my lifetime, for the same reason that synthesizers haven’t replaced orchestras.

One thing I wish more readers understood is that a book in translation is, in a very real sense, two books. It’s the book the original author wrote, and it’s the book the translator wrote. They overlap but they’re not identical. When you read Dostoevsky in the Pevear and Volokhonsky translation, you’re reading a different book than if you read Dostoevsky in the Constance Garnett translation. Not a little different. Substantially different. The same events happen. The same characters appear. But the texture of the experience, the way the prose feels in your mind as you read it, changes radically depending on the translator. This is not a flaw in the system. It’s the system working as it should. Multiple translations of the same work give readers multiple paths into that work, and each path reveals different things.

I sometimes wonder if the ideal reader of translated literature is someone who reads the same book in two or three different translations. That’s impractical, obviously. Most people don’t have time to read a book once, let alone three times. But for the books that matter most to you, it’s an exercise worth trying. Read a chapter of Proust in the older Scott Moncrieff translation and then in the newer Lydia Davis translation. The differences are illuminating, not because one is right and the other wrong, but because they show you how much of what you experience as “the book” is actually the translator’s interpretation of the book.

This is what I keep coming back to. A great literary translation is not the one that’s most accurate. It’s not the one that tracks most closely to the source. It’s the one that creates a living, breathing work of literature in the target language, one that honors the original by being something worthy in its own right. It’s the translation that stops you cold in the middle of a sentence, that makes you set the book down and think. Not “that was well translated” but “that was well written.” Because in the end, that’s what a great translator is: a writer. A writer working within constraints that most writers never face, making art from someone else’s blueprint, finding their own voice inside another person’s vision. When it works, when the translator and the author and the target language all come together, the result is something close to magic. Two minds, two languages, one experience that belongs fully to both.