The AI Journal

In the second of Seneca’s letters to Lucilius there is a sentence of four words: Nusquam est qui ubique est. Read plainly it says: he is nowhere who is everywhere. The line is a Stoic paradox, and in Latin it is also a small piece of carpentry, the same two adverbs turned around, nusquam and ubique, so that the sentence folds shut on itself. In English the folding mostly survives. What does not survive is the shrug. Latin can say this and let it stand as a proverb. English wants to explain it, or to make it pretty, or to smooth the corners until it sounds like something an English essayist would have written on purpose.

The temptation has a name. Lawrence Venuti named the two poles of the translator’s practice domestication and foreignization, borrowing an older opposition from German thought, and argued in his history of the field that Anglo-American publishing had quietly settled on the first: it rewards translations that read as though they had never been translated at all. On that account fluency is the reigning criterion and the translator’s reward is invisibility, since the better the work, the less of it the reader is allowed to notice (Venuti 2017).

I want to argue that this is one method wearing two faces. The reason is not subtle. A translation is written in the target language. It is not a border crossing that occasionally forgets to land; it lands, necessarily, in English. Every word it can use is an English word, every syntax an English syntax, every connotation one that English speakers already carry. There is no third language in which a translation could be done, and no neutral idiom waiting outside both, from which the source could be addressed without distortion. Domestication and foreignization therefore name not two routes but two manners of the same unavoidable act. The foreignizing translator does not escape domestication; she domesticates visibly.

That the target language is not a vessel but a set of obligations is clearest in grammar. Roman Jakobson observed that languages differ less in what they can express than in what they oblige a speaker to specify (Jakobson 1959). English forces a commitment on every noun phrase, a or the; on every verb, tense and aspect; on every predicate, number. The translator does not choose whether to make these commitments, only which. The Latin of Seneca’s line leaves the subject implicit and the paradox open; English must decide who is nowhere and when. Something is settled in that decision that the original never settled, and it is settled by the grammar of the receiving tongue rather than by any taste of the translator’s.

The same is true higher up, at the level of technique. Antoine Berman catalogued the deforming tendencies that a translation imposes on a foreign text: rationalization, clarification, expansion, ennoblement, the leveling of rhythm and of the networks of meaning that run beneath a sentence (Berman 2021). What matters for my argument is not whether those deformations are losses; Berman thought several of them were. It is that they are operations of the target language. Even a translator determined to let the foreign show through performs that showing with target-language tools, an archaic word here, an unfamiliar syntax there, a calque lifted on purpose. Foreignness, in a translation, is manufactured at home.

The opposite camp is no different in kind. Eugene Nida made the reader the measure of the work: a translation succeeds when the receptor responds to it as the original receptor is imagined to have responded (Nida 1964). That is the domesticating program stated openly, and it has been the practical orthodoxy of scripture translation and much else since. Itamar Even-Zohar pushed the point further still, showing that translated literature is not a neutral window onto other cultures but a position inside the receiving culture’s own literary system, central or peripheral, and that what gets translated, and how, is shaped by that position (Even-Zohar 2021). Note what the fluent program and its critics share. Both describe the translator’s task entirely in terms of what the receiving language and reader can be brought to accept. Neither camp ever names a translation that was not made out of the target language, because there is none.

If the two methods are one, what choice actually faces a translator? I think it is narrower and more modest than the grand debate suggests: whether the finished text shows the join where the original was prised apart, or grinds it smooth. When I put Seneca’s line into English and leave the paradox slightly odd, he is nowhere who is everywhere, I am not being more faithful to the Latin in any measurable sense. I am declining to hide that an operation has been performed. A fluent rendering, you are never really anywhere if you are everywhere, hides the same operation by making it sound like ordinary English wisdom. Both are English. Only one of them tells you it is a translation.

The strongest objection to this is that it is a distinction without a difference, dressed up as ethics. The objector will say that I have conceded the only thing ever in dispute, that both versions are target-language products, and then smuggled back a preference by calling one honest and the other dishonest. Difficulty is not virtue, the objection runs. A visible seam can be its own kind of lie, flattering the reader into believing she has touched the original when all she has met is a translator’s idea of what foreignness ought to feel like. And the fluent version, whatever its conceits, has actually carried Seneca to readers who would never have climbed over an awkward sentence to reach him. On this account foreignizing is a taste, and a slightly smug one, and my claim that it is more honest is merely that taste asking for a moral certificate.

I accept most of this, and I want to narrow my claim rather than defend the grand one. I am not arguing that difficulty is a virtue, that a foreignizing translation stands closer to its source, or that readers are owed an obstacle. I am arguing something smaller: disclosure is not a method but a threshold, and the threshold is whether the text tells the reader what it is. The objection treats honesty as an achievement, one more quality on the same scale as fluency, and then complains that I have favored it. But honesty is not a style that competes with fluency; it is the floor both styles stand on. A translation that conceals its seams does not merely fail to be difficult. It makes a claim it cannot support, that you are reading Seneca, when what you are reading is English. My narrower point is only that when the claim is false, the craftsmanship that made it persuasive stops being a virtue and becomes the mechanism of the deception. The distinction survives for that reason and no larger one: not because the visible join carries you closer to Rome, but because it stops you short of pretending that you have arrived.

So when I come to Nusquam est qui ubique est, I do not ask whether I have been faithful. The question has no clean answer, because there is no English sentence that is the Latin sentence. I ask whether my English admits what has happened to it, and I leave the join where a reader can see it.