Is Translation Carrying Meaning Across, or Building a World Again?

Geoffrey Chen

How a Person Builds a World · Season One: “Reading, Memory and Judgment” · Article Seven

When a sentence passes from one language into another, every dictionary item may appear to have found a counterpart and the sentence may still be wrong. Its facts are intact and its grammar acceptable, but the speaker’s distance, politeness, irony or hesitation has disappeared. Another version may depart from word-for-word correspondence while giving its readers an understanding closer to the original.

Translation has never dealt with words alone. Meaning in the source depends upon syntax, tone, genre, historical setting, shared knowledge and relations with other sentences. If that meaning is to continue working in another language, the translation must find new supports for those relations.

Both familiar metaphors—transport and reconstruction—are therefore only partly right. Translation does not pour a sealed quantity of meaning from container A into container B. But neither is it free construction independent of the source. It rebuilds a set of workable relations under the constraint of an existing text.

No word carries a whole world by itself

Dictionary correspondence is indispensable. Without lexical constraint, translation becomes impossible to verify. Yet the function of a word in a particular work is usually larger than a listed definition.

The English word “home” may refer to a residence, a place of belonging, or a relation to departure, return and loss. The Chinese 家 connects with kinship, ethics, property and feeling. The two can translate one another in many sentences, but their relational fields do not coincide in every case. The translator must decide which part of the structure the current sentence activates.

Philosophical terms make the issue especially visible. Intuition in a Kantian context cannot be understood merely as an everyday feeling or hunch. Its established Chinese rendering, 直观, needs explanation precisely because contemporary ordinary usage can pull a reader elsewhere. There is no self-executing transfer available from a dictionary. The term takes meaning from its position in an argument, and the translation must reconstruct that position as far as possible.

This is also why a repeated source word may sometimes need a stable translation and sometimes need contextual variation. Consistency can preserve a conceptual thread while producing rigid prose. Variation can produce natural language while concealing a deliberate pattern in the author. Translational judgement occurs between local readability and whole-text relation.

A target language may require the implicit to become explicit

Source texts omit information that their readers can supply through grammar, convention or shared culture. In another language, the same implication may fail. A translator may need to add a subject, specify a relation or make a logical connection visible. Translation studies often discuss such changes under the heading of explicitation.

Corpus research has compared source texts and translations to ask whether translated language tends to spell out what the original left implicit. One example is this corpus-based study of text length and explicitation. Claims of a universal tendency must remain cautious, however. Language pairs, genres and editorial processes differ, and research outside commonly studied European languages has not always reproduced proposed “translation universals”. See, for example, a Persian corpus study asking whether translation universals are really universal.

The important point is not a mechanical law. It is that translation necessarily makes decisions about visibility. Making the implicit explicit may help a target reader understand, but it can also remove a deliberate ambiguity. In a legal text, an open formulation may preserve interpretive space. In literature, an absent subject may leave responsibility uncertain. Greater clarity is not automatically greater fidelity.

Every explicitation should therefore answer two questions. What enabled source readers to understand? Can the target text rely upon an equivalent relation? If not, does the added material restore a condition of understanding, or does it decide something the source withheld?

Fidelity is not the discovery of one substitute sentence

We often imagine good translation as a problem with one correct answer. There is a single target sentence corresponding to the source; alternatives are deviations. In practice, multiple versions may preserve different relations. One retains rhythm and tone, another terminological consistency, and a third the accessibility of a historical allusion.

This does not mean that translation has no errors. A fact can be reversed, a negation dropped, a relationship misunderstood, or a technical term rendered inconsistently. Source, context and target language impose real constraints. After those constraints are met, however, more than one defensible solution may remain.

Quine’s philosophical discussion of the indeterminacy of translation pushes this possibility to a radical conclusion: distinct translation schemes might be equally compatible with observable linguistic behaviour. The Stanford Encyclopedia of Philosophy entry on Quine gives an accessible account of the argument and its controversies. One need not accept the full thesis to draw a moderate lesson. Meaning is not a single object concealed behind a sentence, waiting to be extracted.

Fidelity is better understood as sustaining significant relations. A translator should be able to explain what was prioritised, where a sacrifice occurred and which meanings could not be carried together in the main text. Prefaces, notes and glossaries matter because one translated sentence may not hold every relation simultaneously.

A translator is neither a transparent pipe nor a replacement author

The principle that translators should remain invisible expresses a legitimate ethic: they should not use translation as permission to rewrite an author. Complete disappearance, however, is impossible. Choosing one word, dividing a sentence, or explaining an allusion affects the work that target readers encounter.

What must be resisted is not the translator’s existence but judgement without a trace. A responsible edition identifies the source version, terminological principles and major disputed choices. It allows readers to know that the new voice did not emerge automatically from the original.

Conversely, recognising creativity in translation cannot turn fidelity into an optional style. Translational creativity operates under constraint. The translator does not create a new world that the author never wrote. The translator creates conditions under which the work can appear in another linguistic world. The more consequential the alteration, the stronger the reason it requires.

Machine translation and generative AI sharpen this distinction. They can produce fluent candidate translations quickly. Fluency shows that a target sentence resembles acceptable language; it does not prove that crucial relations have survived. Without context, a terminology plan, version control and human review, a system can be locally plausible in every sentence while allowing the conceptual structure of a work to drift. Technology can expand the field of candidates. Somebody must still decide which relations may not be sacrificed.

Translation makes a language encounter its own limits

Translation gives a work new readers, but it also changes the target language. To receive new thought, languages borrow, coin terms, alter existing usage and sometimes develop new scholarly syntax. Many philosophical and scientific expressions that now appear native once entered through disputed translational decisions.

The source language also becomes strange to itself. When a native speaker is asked why an expression cannot be directly translated, previously invisible assumptions emerge. Translation does not merely measure the distance between two languages. It makes hidden structures within each language visible.

This is its distinctive value for a knowledge life. Within one language, concepts and formulations can feel natural. Moving between versions reveals that every apparently obvious expression involves decisions. Translation does not train indiscriminate scepticism. It trains sensitivity to relation.

From a Sustenesis perspective, meaning is a structure maintained through difference and constraint. It is neither an isolated word nor the private possession of an author’s intention. Text, linguistic convention, historical situation and readerly interpretation enable it to operate. Translation must give that structure relative continuity under changed conditions.

Translation does build a world again, but “again” is not freedom from obligation. The new world must still answer to the constraints of the old text. It should let readers enter and also ask how the route was made.

The best translation does not make difference disappear. It lets meaning arrive while retaining enough of a seam for the reader to know that what arrived is not an unchanged duplicate. It is a bridge towards the source, one that still permits return and correction.

Primary sources and further reading

Continue reading: Explore the How a Person Builds a World series.

Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.