4. What translators actually do

The project’s founding charter named a question the genre never asks: how do professional translators actually handle these words when they meet them in a real text? A word that genuinely defeats translation should leave marks — translators keeping it in the original, footnoting it, apologizing for it. The evidence the project gathered is uneven; for more than half the words it found no usable data point at all, and it says so on every page where that is true. But where evidence exists, it points one way.

Saudade. Clarice Lispector’s 1964 novel The Passion According to G.H. uses the word three times, and a 2012 review set two published English translations side by side. Neither translator kept the Portuguese word or settled on one English equivalent. Ronald Sousa (1988) wrote “I shall have to bid a nostalgic good-bye to beauty,” “A suffering with a great sense of loss,” and “I long to go back to hell”; Idra Novey (2012) wrote “I’ll have to bid farewell with longing to beauty,” “A suffering of great longing,” and “I miss Hell.” A working Portuguese–English literary translator, Ilze Duarte, describing her own published choices, reports “longing” in one novel and “missing you” in a song lyric, and generalizes that nostalgia, longing, homesick, or the verb miss will all do depending on the sentence. The one purported counter-example — a radio piece said to have left the word untranslated — dissolved when checked: the piece actually in question translates the Portuguese statement into English first and only then introduces “saudade” as a separately glossed foreign term. Across every instance the project could find, a small and unsystematic sample, no published English prose leaves saudade bare (saudade).

Hygge. The Danish author Dorthe Nors wrote a short story titled, in Misha Hoekstra’s English translation, simply “Hygge,” and it has appeared under that untranslated title three times, in a magazine, a literary website, and a 2021 collection. The project obtained the full text of the magazine printing and found the words hygge and hyggelig nowhere in the story’s own prose. Where the narrative names the concept, it says “we’re going to have us a cozy time” and “We sat there and were cozy.” The loanword survives as a title, where it works as a hook (one reviewer read it as “a pretty deliberate repudiation of the warm, marketable coziness” the word had acquired), and is paraphrased everywhere it has descriptive work to do (hygge).

Lagom. The project’s best early evidence came not from a search of novels but from a 2013 University of Wisconsin dissertation on Swedish-to-English literary translation, whose appendix tracks how three published translators rendered lagom across three novels. Of the instances she tabulated, nine are paraphrased in ordinary English (“sufficiently,” “medium,” “perfect,” “just right,” “just the right level of”), one is omitted, and two are kept as “lagom” — both on one page of one novel, both by the dissertation’s own author, whose thesis is that internet-era translators are more willing to leave such words alone. The two earlier translators in her sample, working in 1973 and 1986, never once kept it. The project read this as only partial support for her thesis: the one translator most inclined to keep the word still paraphrased most of her own instances (lagom).

Abhiman. The most rigorous data point on the whole file is a 2018 peer-reviewed corpus study that coded all thirty-six occurrences of the Bengali word in a sample of twelve Rabindranath Tagore short stories, two novels, and two plays, as rendered by native-English and native-Bengali translators. For the older sense, translators converged: “pride” was used in 44 percent of English-translator instances and 27 percent of Bengali-translator ones, and when one group rendered a passage as “pride,” the other largely did too (one translator excepted). For the newer, relational sense, they scattered — hurt, sulking, resentment, self-respect, grievance, offense, wounded pride, petulance — and 60 percent of the English translators’ non-“pride” choices were used exactly once. The authors’ conclusion, that “no English word appears to have a script close to the current usage of obhiman,” is the closest thing on the project’s file to a measured untranslatability — and note what it measures: not that the word could not be translated, but that no two translators translated it the same way (abhiman).

Toska. Chekhov’s 1886 story is titled with the single word «Тоска». In English it has been “Heartache” (Payne), “Misery” (Constance Garnett in 1921, and a more recent translator), and “Anguish” (Pevear and Volokhonsky) — three different titles across a century; the critic Gary Saul Morson calls the two common ones, “Misery” and “Anguish,” “inadequate,” since neither carries the longing. Within the parallel Russian and English texts of Nabokov’s own Lolita, the linguist Apresjan reports the word rendered as anguish, heartache, yearning, longing, and weariness by context; the project did not verify this against the novel (toska).

The loanword kept. Two documented cases show the word left untranslated, and both are instructive about where that happens. South Africa’s 1993 Interim Constitution, an English legal text, reads “a need for ubuntu but not for victimisation” — the drafters italicized the word rather than substitute “humaneness”; a Constitutional Court judge, interpreting it two years later, wrote that “ubuntu translates as humaneness” and, more fundamentally, as “personhood and morality,” while arguing the fuller sense exceeds any one word (ubuntu). And Lu Xun’s 1934 essay on mianzi, in a 1959 published translation (as quoted at second hand), renders the word simply as “face” — because the calque had long since done the translating (mianzi). English business and journalistic prose likewise keeps jugaad and Schadenfreude as loanwords as a matter of course — a pattern the project observed across its own sources but did not count as a translator-practice finding, since no source discusses it as a translator’s choice. The word stays foreign, in other words, where it functions as a name, a term of art, or an already-naturalized borrowing — not where a translator is wrestling with a sentence.

What could not be checked. Orhan Pamuk’s translator Maureen Freely published a first-person essay on translating his hüzün passage — “by the time I had translated my way to the end of this sentence, the hüzün had so permeated my imagination that I could barely see through it” — but the project never read the translated passage itself, so how she rendered the word stays reported, not confirmed (hüzün). Kundera’s novel exists in two English translations, one from Czech in 1980 and one reportedly made from his own revised French text in 1996; the project could not legitimately obtain both passages to compare them (litost). For Tamil oodal the question did not quite arise: a classical poem given in four independent English translations dramatizes a sulking scene without ever using the word, because in the classical system the word is a critic’s label for a situation, not vocabulary inside the poem (oodal). And for fourteen words — komorebi, wabi-sabi, Schadenfreude, gezellig, meraki, jugaad, sisu, ya’aburnee, gigil, teranga, tizita, gemas, sankofa, and iktsuarpok — the project found no translator-practice data point at all (for Schadenfreude it checked two Nietzsche translations and found the word in neither; for sankofa and tizita it never searched); for mamihlapinatapai there is essentially no literary corpus to translate from.

Put together, the evidence is modest in quantity and consistent in direction. Where a working translator meets one of these words in running prose, the word gets translated — with an ordinary English word chosen for that sentence, a different one for the next sentence, and without ceremony. In the theorists’ terms, keeping “Hygge” as a title is what Lawrence Venuti calls a foreignizing move, and “cozy” inside the story is, by the same token, Eugene Nida’s dynamic equivalence (the cultural turn). The genre’s implicit picture — a translator stopped dead before an untranslatable word — appears nowhere in the record. What the record does show, in the abhiman study, is the honest form of the difficulty: many translations, none of them wrong, few of them the same.