For Elizabeth Bishop, a successful translation was, in large part, a matter of luck. As someone who published translations of poetry and supervised the translations of her own poetry, she had definite ideas of what made for a good one. “You really should repeat a line exactly if the original repeats it exactly,” she told her Portuguese translator, urging her to avoid taking liberties. “You shouldn’t put in words that aren’t there…. You should pay attention to repeated words and phrases—etc.” Her punctiliousness about formal and semantic equivalence made luck loom all the larger. Yes, you needed skill and sensibility, but they weren’t enough. Translation was worthwhile when “a poem just happens to go into English without losing too much of the original.”*

Her friend Robert Lowell had created a stir in 1961 with Imitations, a collection that offered his free and impressionistic versions of poems written in other languages, among them Russian, which he didn’t know at all; for those, he relied on prose cribs. But what he had produced, Bishop thought, were Lowell poems, not translations. She was alert to the serendipities that poets seize on even when they’re writing in their own tongue—perhaps hitting on the arresting metaphor or the unforced rhyme that allows the memorable expression of an undistorted thought, the cliché that, given an elegant twist, yields an insight. Replicating such happenstance in another language could be a long shot. “Translating poetry is like trying to put your feet into gloves,” Bishop maintained. Sometimes, depending on the foot and the glove, the fit was quite acceptable. More often some jamming was needed, chafing the foot and stretching the glove.

That could be great for the glove. W.H. Auden believed that whenever English poetry got into a rut, translation jolted it out. Chaucer’s translation of Le Roman de la rose, he thought, heralded the arrival of a new sensibility and a new poetic form. When Dryden translated the ancients, he came to grasp fully the possibilities of the heroic couplet. For that matter, few poems had more influence on nineteenth-century French literary culture than Edgar Allan Poe’s—as rendered by Baudelaire. They’re quite “faithful,” in the usual ways, yet only Baudelaire could have written them.

Translations that show the jamming—that are self-evidently translations—may be especially generative. In seeking to be both accurate and majestic, the creators of the King James Bible ended up conjuring all sorts of strange idioms that are now utterly familiar (a thorn in the flesh, know for a certainty, the root of the matter, turn the world upside down). Some theorists of translation even think that there’s an ethical imperative to signal the strain.

The German theologian and philosopher Friedrich Schleiermacher, in a lecture published in 1816, distinguished between two strategies: “Either the translator leaves the author as undisturbed as possible and moves the reader toward him, or he leaves the reader as undisturbed as possible and moves the author toward him.” The French philosopher Antoine Berman, drawing attention to the lecture in the 1980s, decided that the second approach was ethnocentric and a betrayal of the original text, negating its strangeness under the “guise of transmissibility.” Lawrence Venuti, perhaps the most influential American theorist of translation, agreed: he defended “foreignizing” translations and denounced “domesticating” ones. Readers should stretch toward the author’s world and encounter unfamiliar syntactic patterns, cultural references, and linguistic peculiarities. They should be able to feel the foot within the glove.

An extreme example of this was Vladimir Nabokov’s notorious Englishing of Pushkin’s Eugene Onegin, which was referentially punctilious and unreadable as literature. Lowell’s venture was at the other extreme. “I have been reckless with literal meaning, and labored hard to get the tone,” he wrote in his preface to Imitations. “I have tried to write alive English and to do what my authors might have done if they were writing their poems now and in America.”

We’re still bound to ask, with Bishop, how far a rendering can drift from the referential and count as a translation. When Lowell imagines what Heinrich Heine would have written had the German poet been an American contemporary of his—a pack of Lucky Strikes in his jeans next to the keys to his Mustang—is he giving us Heine? It’s an enduring question, not one to hurry past.

Damion Searls’s The Philosophy of Translation loiters productively around a number of such questions. It also reminds you that a gifted translator must be a gifted writer as well: to translate is, first and foremost, to write. His voice on the page is sometimes pettish and sometimes exuberant, but it’s never less than engaging. A literary translator of astonishing range—he has, among a great many others, taken on Patrick Modiano, Robert Walser, Jon Fosse, and Nescio, to single out four writers in four languages—he regularly wrestles with the angelic quirks of his authors, aiming to be responsive to sound, association, register, and movement. His pages are enlivened by his love of wordplay (he’ll chide a translation for its “word orders unidiomatic”) and a flair for the epigrammatic. “There are no rules, only decisions,” he says at one point about his craft.

Searls is especially riveting when he walks us through the choices he has made as a translator: this is OK, this is better, this kind of nails it. We learn how he worried that readers would think “Brage,” the two-syllable name of the narrator’s dog in Fosse’s Septology, rhymed with “page,” and how, when Fosse reminded him that Brage is the Norse god of poetry, he realized he could go with “Bragi,” as the god is often rendered in books about Norse myths.

He passionately rails against a starchy translation of Walter Benjamin. (Why, he asks, would the translator inflict such misery on his readers? “Does he hate them for not knowing German?”) He stoutly defends a maligned rendering of Rilke. He delves into the complications raised by culturally specific terms—do you call a dish “eba” (but what’s that?) or “vegetable paste” (yuck)? And he’s refreshingly attentive to how passages sound:

The acoustics of the text is so much of what matters in a translation and so little of what people usually talk about: translations are always reviewed and praised or criticized in terms of their accuracy, or perhaps faithfulness, when the sound of the thing is what makes or breaks it, the same as with anything else one reads.

So what of that contrast between the foreignizing and the domesticating? Searls has serious reservations. Maybe all “foreignizing” means is being respectful of the original? He’s impatient, too, with the distinction between a “free” and a “literal” translation: “No translation is either free or faithful.” When we’re assessing the fidelity of decent translators—translators who properly grasp the original text and who have some ability to recreate lyricism or humor, say—“we are really judging what the translator sees in the author or text and chooses to be faithful to.”

As Searls insists, a translation is an interpretation: what’s translated isn’t so much the text as an interpretation of the text. When you hear Glenn Gould playing Bach, he observes, you’re hearing Bach, but you’re also hearing Glenn Gould: “We read the book itself, but we read it, from a particular perspective, moving through it in a particular way. And a translation both is and isn’t the same as the original—it’s the translator’s path through it.”

Certainly a translation that makes a priority of replicating a rhyme scheme or even an emotional atmosphere isn’t exactly “free.” It’s just that one set of constraints has taken precedence over another. In the linguist Roman Jakobson’s terms, a rendition could foreground the “poetic function” of language or the “referential function” of language. Admittedly, translations we admire generally have some success at both, but most literary translators are continually calibrating the trade-offs between them.

To show how far you can depart from a line in order to get close to it, Searls turns to a couplet with which Walser concludes a poem (“There Are Thousands of People, Young and Old,” in Searls’s rendering of its title):

Nun sitzt die Zage genau wie solch ein Mann da, nichtsdestoweniger grüße ich dich herzlich, kleine liebe Wanda.

Or, in a graceless crib: “Now the timid one sits there exactly like such a man,/nevertheless, I greet you warmly, dear little Wanda.” Searls, judging that “the sound outranks the meaning,” says that “we are willing to lose the name of the addressee, ‘dear little Wanda,’ and even the fact that she is who the speaker is writing to, in order to keep the silly rhyme,” and he boldly offers this as a translation:

And now the shy lady sits there like such a man does, nonetheless I send you my warmest greetings, along with Amanda’s.

It’s a rhyme that would have pleased Cole Porter and perplexed Bishop. (Who’s this Amanda?) To understand why Searls views this as a model translation—exemplifying something like fidelity through freedom—it will help to turn to the philosophy of translation his title promises.

In the most innovative section of his book, Searls draws on accounts of perception that he sees as having been jointly advanced by the French phenomenologist Maurice Merleau-Ponty and the American psychologist James J. Gibson. Both thinkers opposed a Cartesian model (one that had been thoroughly critiqued by William James, John Dewey, and Martin Heidegger, among others) in which perception is conceived as a disembodied mind processing sensory stimuli. Rather, the world discloses its reality to us in an immediate, prereflective way. To see a chair is to see something that can be sat on.

The term Searls favors for pinpointing our relation to that chair is a coinage of Gibson’s: “affordances.” A chair, Gibson wrote, is something that “affords” sitting, as air affords breathing, a hole affords escape from predators, and a branch, if you’re a bird, affords perching. For animals, including humans, the features of our landscape are first and foremost defined by what they afford us. What’s more, these affordances are intrinsic, invariant, and objective properties—they’re simply out there, and we’re bound to perceive them. This model, Searls thinks, catapults us beyond those old debates about liberties and literalness:

The point is to get around the false dilemma of whether a translator (or translation) is or should be “free.” We should stop thinking of translators as “free to choose” their translation: we are actually guided by the original in the same way that a chair “makes us” see it as a chair…. Texts “ask to be translated” in a certain way, and when translators “respond to” a text, that is actually what we’re doing.

What are we to make of this effort to translate an account of perception into the realm of translation? Is the act of literary translation really like seeing a chair? At moments, Searls insists that it is:

The question of how I “choose” which words to “use” in a translation feels to me a bit like asking what pictures I “choose” to see in my mind when I’m reading an Austen or Dickens novel. Or what I “choose” to see and hear in a given environment, when my eyes are pointed at a sunset or my ears are near a barking dog—this is the analogy with perception. I’m not choosing.

He’s not choosing? For page after page he beguilingly explains his choices as a translator, often reached through considerable struggle and second-guessing (and sometimes in consultation with the author he’s translating). Hence his dictum “There are no rules, only decisions.” Of course he’s choosing. Amanda didn’t show up without a chaperone.

Searls takes it to be “one of the bizarre facts of twentieth-century intellectual history” that the French philosopher immersed in Heidegger and Edmund Husserl and the American psychologist who’d studied US pilots “ended up with the exact same model of perception, even though they didn’t know each other’s work at all.” (Maybe not so bizarre. Both were attentive readers of the same Gestalt psychologists: Kurt Lewin and Kurt Koffka, whose approach they adopted and modified, purging its residual Cartesianism.) But it’s relevant that their accounts aren’t exactly the same.

For Gibson, perception was, in a phrase Searls relishes, direct “information pickup.” For Merleau-Ponty, however, perception was not only an embodied engagement with the world but something connected to “the cultural tools that my education, my previous efforts, my history have prepared for me.” His notion of “horizon” is crucial here: every perceived object exists within a larger field of potential meanings, and our conscious experience always extends beyond what’s immediately given. When you visit someone’s house and see a very pricey piece of French furniture, what catches your eye may be less what the chair can afford than what the host can afford. But this richer account of perception—as culturally embedded, historically informed, laden with social meaning—is hard to reconcile with the Gibsonian model Searls gravitates toward, where objective properties of the landscape are directly apprehended.

In a later chapter, Searls enlists “affordances” to designate what writers in particular languages can do: what effects can be easily achieved given their very different tropisms, idioms, registers, and more. But the model of affordances, invariants of our environment, doesn’t readily afford this application. For Gibson, to be sure, affordances could be multiple: that lovely round rock could be a projectile or a hammer or a seat. But wait—one of my fellow cave dwellers has just figured out that it can be a wheel! And my wheel inventor didn’t simply detect an affordance; she created a new way of relating to the world. What counts as a meaningful use of a rock shifts because a whole pattern of practical engagement shifts. Similarly, once we’re trying to see the world through other eyes—like translators trying to provide monolingual readers an experience otherwise unavailable to them—perception requires reflection: for me this chair affords sitting, but maybe for that toddler it affords climbing.

In literary language, the creative dimension is especially obvious: the features that Searls wants to call affordances aren’t just perceived but enacted. When Woolf and Faulkner devised their ways of capturing consciousness on the page, they were creating new capabilities, expanding what novels could do. Idioms, too, are created and become capabilities; a poem doesn’t merely follow but revamps rules and conventions. Literary innovation has always taken place in a dynamic space between discovery and invention.

Consider that text of Walser’s. It’s from a poem that belongs to his “microscripts,” works penciled in a minuscule hand, mainly in the mid-1920s. Across forty-nine lines of irregular rhyme and meter—the faltering rhythm contributing to an air of disorientation and introspection—we learn that the poet, having seen a woman who had meant something to him, is reminded of a sort of beggar he was once transfixed by, and he feels a surge of tenderness. This man was down and out, but he had the air of a defeated general, dirty and disheveled and yet capable of a steely calm. While others might be impressed by cheap triumphs, it’s defeat that stirs our poet, because all splendors ultimately come crashing down. And all this is prelude to his addressing “dear little Wanda” (in the intimate du register), who sits timidly like this man and to whom the poet still says a heartfelt hello. The sense is that he’s thinking: My dear little Wanda, I see your timidity, and I see through it.

Readers of Walser will find it impossible not to recall the character of Wanda from The Robber, a novel he composed around the same time. She’s a girl from a conservative family who is taken up and then cruelly cast aside by the title character. When we catch up with her later, we’re informed, “Poor little Wanda was living in utter seclusion. She’d become the talk of the town and felt she could not show her face.” Yet in both works she’s not simply a victim or an outcast; there may be a steely resistance in her reclusion.

If this were some other poem, Searls’s jaunty, whimsical version (“I send you my warmest greetings, along with Amanda’s”) might be a marvelous translation. But as the conclusion of Walser’s haunted meditation on shame, defiance, and dignity? The irony is that Searls’s own practice—his generally meticulous attention to context, his sophisticated interpretive moves, his sense of translating as finding a path through a text—demonstrates forcefully why translation can’t be reduced to unmediated information pickup. Captivated by Gibson’s stringent anti-mentalism, someone who is usually a penetrating and deeply reflective reader seems to have theorized away his own greatest strengths.

What happens when we consider translation not as a solitary act of perception but as a negotiation—one shaped by institutional constraints imposed from without or even by unequal relationships of global power? Speaking in Tongues, a dialogue between J.M. Coetzee and Mariana Dimópulos, repeatedly shifts the focus from individual choices to the structural forces that shape translation. He is, of course, the South African novelist and Nobel laureate long resident in Australia; she is an Argentine author of four novels and specializes in translating German works into Spanish.

Both worry about the global hegemony of English. That’s what led Coetzee to have his most recent novel, The Pole (2023), appear first in Spanish, in Dimópulos’s rendering, as El polaco. When something he wrote posed a difficulty for her, it was the English text, still in manuscript, that gave way. Intent on inverting the usual relationship between a work and its translation, he wanted the Spanish to “carry no marks of a secondary status,” he explains in Speaking in Tongues. He even tried, without success, to persuade foreign publishers to translate from the Spanish version.

That’s a fascinatingly provocative position to adopt. Julian Barnes, for one, has worried that when writers get too chummy with their translators, the result can be “the sort of international prose that is like an airline meal: it feeds all, doesn’t actually poison anyone, but isn’t noticeably nutritious.” Coetzee has a radically different perspective. Writing The Pole, he recounts, he “deliberately starved it of what I think of as native nutrients.” In recent years, in fact, “I had lost interest in sounding like a native—that is, like a native speaker…. I told myself I was feeling my way towards a rootless language,” one with little sociocultural specificity. He began to work more closely with his translators, wondering why translators had to be faithful to his English. In his judgment, the Coetzee novels produced by his German translator Reinhild Böhnke are “in no respect inferior” to the originals. Had the German versions appeared first and his English versions been presented as their translations, “I would be praised for the accomplishments translators are usually praised for: fidelity to the original, mastery of idiom, et cetera.”

Is a perfect translation possible, then? Coetzee is of two minds, because he’s equally interested in everything that impedes translation. The title of his latest novel, which centers on a Polish pianist, has meanings in English (a pole marking an axis of rotation or suggesting a relation of antithesis) that the Spanish demonym lacks. And the way Spanish tends to gender animate nouns (you’re not a poet; you’re un poeta or una poeta) creates a host of complications. Coetzee wonders whether

full translation is possible between a gendered language, whose speakers may still, in an obscure way, feel the presence of archaic forces in the world around them, and a genderless language, from which those forces are excluded.

What about languages with color terms that don’t map onto English ones? The Xhosa language has one word, luhlaza, for blue and green. Describing the color of the sky, a speaker will qualify the word for green-blue (appending a term that means “like the sky”), and Coetzee is convinced that to “gloss a word is not the same as translating it.” Plainly, I can explain a joke that depends on wordplay that’s specific to a language you don’t know, but in delivering understanding, I won’t be delivering the joke. It doesn’t seem as if I’ve translated it. Coetzee, for his part, wonders whether amanzi aluhlaza—it can refer to green water or blue water—simply defies translation.

Dimópulos doesn’t think so. Translation, she makes plain, involves countless difficulties like this, and the translator’s job is to come up with work-arounds. One set of challenges relates to “specification,” as Coetzee puts it. Sometimes the target tongue lacks a distinction found in the original: when you’re translating most European languages into English, you must figure out how to deal with their pronominal contrast between formal and intimate address.

Even more vexingly, a target language may tend to impose a qualification that isn’t in the source language. In Vietnamese, Coetzee says, there’s no word that indicates you’re someone’s brother without specifying whether you’re older or younger. “Roger and his brother caught a bus,” someone writes, and he imagines a Vietnamese translator being perplexed to learn that the author doesn’t have a view about their relative ages. Yes, you could laboriously state, in Vietnamese, that two boys of shared parentage caught the bus, but it would be at odds with the stylistic simplicity of the source text.

Our interlocutors also broach the difficulties posed by cultural distance. Coetzee cites the Movimento Antropofágico of the 1920s, which urged Brazilians not to imitate European models but to devour and transform them into something distinctively Brazilian—an act of incorporation in which one’s own flesh is fed by another’s. “The more distant two cultures are, the more likely it is that mixing becomes the only way to make the new content understandable,” Dimópulos says. From a global perspective, she knows, the cultural distance between major-market languages like English and Spanish tends to be relatively narrow, especially when contrasted with the distance that separates these from indigenous languages and communities. Readers might consider how The Pole assumes familiarity with the world of concert pianists, metropolitan cultural institutions, Dante’s Beatrice. The pianist plays Chopin in a style that reins in his romanticism and positions him as a descendant of Bach (think Maurizio Pollini). There’s actually a lot of sociocultural specificity here; it’s just shared across the major-market languages.

In general Coetzee focuses on the limits of translation, Dimópulos on its possibilities. “Each natural language has a special music, a particular taste for images, and an array of preferred structures that fixes words as if in a gold chain,” she says. But she takes the “principle of fidelity” seriously. As a literary translator, she thinks that honoring a text requires that she creatively render corresponding beauties in her own language, which rules out literal translation but still means being constrained by what she calls the “guide-text.”

Like Searls, Coetzee and Dimópulos reject the dictionary fallacy: they know that translation isn’t necessarily hindered because one language lacks a direct equivalent for a word in another. What’s true, though, is that words may carry a certain fragrance that’s hard to reproduce. Take the comic pedigree trailed by “blurb”—Speaking in Tongues bears one from me—which, coined by the author of “The Purple Cow,” still traipses around with a jester’s cap and bells. Perhaps there’s slightly greater dignity when a French book editor solicits a recommandation for the quatrième de couverture or a German one seeks a Lobzitat? (Though these days, both might use the English term.)

Coetzee rather beautifully evokes the experience of scouring a thesaurus for the right word he thinks is there and then settling:

Every now and again, in the trail of writing we leave behind us, there is a word with a secret sign hovering over it, a sign of shame or defeat: the word on the page is not the right word but a pretender, the younger brother of the rightful word.

(Oh, younger: our Vietnamese translator is relieved.) This image of the secret sign gets at something significant about both writing and translation. We reach for the perfect word or phrase across languages just as we reach for it within our own, sensing, maybe only fleetingly, the gap between what we mean and what we manage to say. In that reaching, though, we just may create something new: a younger brother who grows to be his own man.

At one point in Speaking in Tongues, Dimópulos, prodded about a seemingly untranslatable phrase, offers a powerful rejoinder. Untranslatable? No, she says, these “extremely difficult cases of equivalence” are exactly what translators are for. “The rest is pure mechanics”: a task that in the near future machines will take on. She’s breathtakingly undefensive here: she evidently sees that a hybrid model—human-aided machine translation—will become the default.

Searls will have none of it. Toward the end of his book, he says that Google Translate, though useful for tourists, is basically just a souped-up bilingual dictionary, able to look up all the words in a passage at once. As for ChatGPT, it can’t really translate because it’s incapable of the “deeply human act of reading,” by which he means a process that’s not only objective (being responsive to a shared reality) but also subjective (being performed by an individual capable of intentionality). He apparently thinks that the enterprise requires an experiencing subject. Has he been caught with his hand in the Cartesian cookie jar?

To support his skepticism he shows how, for example, a machine translator, given a sentence from a French work of art history, renders essais as “essays” where the apposite rendering would be “efforts” or “attempts.” Such arguments inevitably have a god-of-the-gaps ring. Just as science’s explanatory lacunae were once treated as proof of God’s existence, today’s machine-translation glitches are taken to show what machines will never succeed at doing. Yet both Google Translate and the large-language-model chatbots rely on a neural network architecture, Transformer, that’s barely eight years old. Computer scientists are now devising architectures that could vastly expand a system’s “context window”—the amount of text or data that a language model can consider at once. Humility is in order. LLMs are ultimately crystallizations of human creativity. If they can be trained to translate inventively, it’s because human beings have translated inventively.

And, of course, defining “success” in translation is a slippery business; it all depends on what you’re trying to achieve. If you’re translating song lyrics, preserving emotional resonances might be your main goal, but you must also fit the metrics to the music. If it’s an industrial manufacturing guide, the vibes are immaterial—quantitative precision is paramount. A legal contract? Every semantic nuance matters. And the standards of success will depend not just on what it’s for but on who it’s for. Depending on the intended audience, you might have to explain a lot of background assumptions—or not.

“The translator is continually having to juggle the claims of the literal meaning of words against the values of those words in a landscape of sound and tone,” Coetzee observes of poetic language, and throughout these books you’re reminded that Dimópulos and Searls are old hands at such juggling, combining patience and prowess. Yes, poetry gets lost in translation, but, Searls acutely remarks, poetry also gets created in translation, “for the new words in the new language are in a new relationship with the Idea, and they convey whatever they convey in a new way, within a new linguistic system and for a new audience.” In this sense, he suggests, one could fairly say that “poetry is what gets gained in translation.” What you can’t know in advance, of course, is whether it will be good poetry.

You imagine Bishop in her estudio in Samambaia, the scent of coffee and orchids mingling in the air, as she weighs the prospects of such gains against her conviction that linguistic cunning won’t suffice for meaning and music to align. She squints at a much-loved poem, probing for parities, foraging for felicities, hoping for happenstance. Before setting her pen to paper, she pauses. Is she feeling lucky?