Try translating a family nickname. A dictionary may supply the meaning, but explaining why one person is allowed to use it can take considerably longer. An outsider might hear an insult where the family hears affection. A relative might remember a childhood incident that nobody else knows. The explanation grows until the nickname begins to seem like the shortest available version of a rather long story.
Translation encounters this problem wherever words carry a history between people. A polite request can sound cold when moved too carefully into another language. An apology can become more intimate than the speaker intended. Even when the information survives, the listener may have difficulty recognizing the person offering it.
For someone who moves between languages, the difficulty can begin before there is a listener. Which language offers the words for this experience? Which one carries the embarrassment? In a hypothetical argument, a person might find it easier to begin in the language learned at university, then return to a childhood language for the part that hurts. Another speaker might move in the opposite direction. Neither owes us a tidy story about having two personalities.
What a language remembers
Research on bilingual emotion offers reasons to attend to that variation. Karina Febre, Nafiseh Faghihi and Jyotsna Vaid studied 110 Spanish-English bilingual adults living in the United States. They asked participants to judge short situations presented in each language and compared people reporting more or less experience translating informally for others, including family members. The article appeared online in 2024 and in a 2025 issue of the International Journal of Bilingualism. The study
Among participants with more language-brokering experience, positive situations received more pleasant ratings in Spanish than in English. The study found no corresponding group or language differences in ratings of emotional intensity, or in the valence ratings of negative and neutral situations. The result is narrower than a claim that a heritage language is always more emotional. It suggests that what people have done in a language belongs in the explanation of how that language feels.
The history matters. A language used to help a parent understand a document has been part of a relationship involving responsibility. The same vocabulary learned for an examination may have quite different associations. Proficiency alone cannot tell us what the speaker remembers when a phrase arrives.
It also cannot tell us how much work a fluent speaker is doing. Imagine a family member asked to interpret a difficult conversation. The task includes choosing an equivalent word, deciding when to interrupt and noticing that someone has stopped following. If a phrase sounds cruel in direct translation, the interpreter has to decide whether softening it would misrepresent the speaker. Being able to speak both languages makes these decisions possible. It does not settle them.
When translation gets easier
A machine can remove some of this work. That is a substantial benefit for someone trying to understand a notice or begin a conversation without an available interpreter. Research published in Nature in 2024 described a neural machine-translation system covering 200 languages, with particular attention to languages receiving relatively little digital research and training data. The project also made the unequal availability of parallel text central to its account of translation quality. The NLLB Team's paper
The achievement concerns a technical system and its evaluation. It does not establish that all languages receive equivalent performance, or that a translated sentence carries everything its speaker intended. Yet there is no need to dismiss the practical gain in order to ask what remains. A readable first translation can give someone a way into a conversation that would otherwise be closed.
One useful test is whether the tool leaves a person able to correct it. Suppose a translated message is grammatically clear but makes its author sound unusually formal. The author should be able to change the tone, supply a missing relationship or keep a word untranslated. A recipient should be able to ask what was meant. Fluency becomes less helpful when it persuades everyone that clarification is no longer necessary.
This is especially important when the speaker's style is itself information. Hesitation may express care. An indirect phrase may give another person a graceful way to refuse. Repetition can be affectionate, impatient or simply habitual. A system asked to make everything concise and professional may produce an excellent message from someone rather unlike the person who sent it.
Room to speak
The same pressure exists without a machine. A newcomer who speaks carefully can be treated as though careful speech exhausts their personality. In an imagined staff meeting, the person with the best joke may still be assembling its final clause when the subject changes. Offering a little time would change who gets to be funny in that room. Nobody needs to discover a new theory of bilingualism before allowing it.
There are choices here for institutions that describe themselves as multilingual. Translating the application form helps at one stage. Providing a way to ask about an unusual circumstance helps at another. A translated performance gives a work a new audience; retaining access to the original voice gives that audience another way to listen. Each arrangement answers a more specific question than whether translation is available.
At home, the choices can be less formal and just as consequential. A speaker may want help finding the exact word. On another occasion they may want to finish an imperfect sentence without being corrected. Someone listening has to discover which kind of help is welcome. Automatic improvement is a poor substitute for that small piece of attention.
An emotionally complete life does not require every feeling to pass smoothly between languages. Some expressions will remain awkward. A joke may need a footnote and suffer for it. People can still learn how a word is used, who says it and why its literal meaning is insufficient. The effort can become part of the relationship.
Return to the nickname. Keeping it untranslated may be the most accurate decision available, provided the people hearing it can ask about it. An explanation can then begin with the relative who first said it, or the incident everyone remembers differently. The conversation has taken longer than a translation tool promised. By then, it may also have given the listener a reason to use the word with care.