Translate Evidentiality & Information Source

To translate evidentiality accurately, a translator must preserve more than the event described in a sentence. The translation must also preserve how the speaker knows about that event: whether it was seen directly, heard, reported by somebody else, inferred from evidence, assumed from context, or presented without a stated source. This is one of the least obvious but most important problems in accurate translation because two sentences can describe the same event while making very different claims about the speaker’s access to the evidence.

People searching for how to translate evidentiality, reported information, hearsay, inferential meaning, information source, apparently, reportedly, seems, must have, I heard that and related expressions are usually facing the same underlying problem: languages do not package evidence in the same way. One language may require a grammatical marker showing whether the speaker witnessed an event, while another may leave that information optional. A good translation therefore cannot simply replace words. It must preserve the relationship between the proposition and the evidence behind it.

This guide explains evidentiality in translation as a practical meaning-control problem. We will separate information source from certainty, distinguish direct evidence from reported and inferred evidence, work through examples, diagnose common machine-translation and human-translation failures, and build a repeatable method for translating evidential meaning across languages. The central rule is simple: do not make a speaker know more, less, or differently than the source allows.

Translation Is Not Only About What Happened

Suppose a source text communicates, in effect, “The train has left.” At first glance the proposition seems straightforward. The train was at the station and is no longer there. Yet the sentence can enter a conversation through many different paths. A speaker may have watched the train depart. A colleague may have phoned to report it. The speaker may have arrived at an empty platform and inferred the departure. A timetable may show that the scheduled departure time has passed. Each route gives the same broad proposition but a different evidential basis.

If a target language requires the translator to choose among forms that distinguish direct observation, report and inference, the translator must decide which evidential value the source supports. If the target language does not grammatically require such marking, the opposite problem appears: a source-language evidential marker may disappear unless the translator expresses it through an adverb, verb, clause or other construction. Translation therefore involves an evidence ledger as well as a meaning ledger.

The distinction matters in ordinary conversation, news, academic writing, witness statements, historical sources, scientific reporting, medical communication and any setting where readers care not only about a claim but also about its basis. “The device failed” is different from “The device apparently failed.” “She resigned” is different from “She is said to have resigned.” “The road is closed” is different from “The road seems to be closed.” A translation that erases these differences may preserve vocabulary while changing accountability.

What Evidentiality Means

Evidentiality is the linguistic signalling of information source. In some languages it is expressed mainly through optional words and constructions: apparently, reportedly, I saw, they say, it seems, judging from the data, according to the witness. In other languages, information source can be built into grammar so that a speaker routinely marks whether knowledge is direct, reported, inferred or otherwise obtained.

For translation, the theoretical label matters less than the operational question: what evidence relationship is encoded here, and where will that relationship live in the target sentence? The answer might be a verb ending, particle, auxiliary, adverb, reporting clause, parenthetical phrase, source attribution, word order choice or even a carefully maintained omission. Good translators do not insist that the same grammatical device survive. They insist that the same communicative distinction survive.

This connects directly to the broader problem of equivalence. Languages rarely provide one-to-one containers for meaning. eduKateSG’s guide to translation equivalence explains why the target expression may need a different form to perform the same semantic job. Evidentiality is an excellent example: the translator may have to move information from morphology into vocabulary, or from a reporting phrase into grammar, while keeping the evidence relationship stable.

Five Evidence Questions to Ask Before Translating

Did the speaker perceive the event directly? Was it seen, heard or otherwise sensed?
Was the information reported by another person or source? Is the speaker relaying testimony rather than claiming firsthand access?
Was the conclusion inferred from evidence? Is the speaker reasoning from traces, results or consequences?
Was the proposition assumed or treated as general knowledge? Is the source presenting a background expectation rather than fresh evidence?
Does the source deliberately leave the evidence unspecified? If so, would the translation wrongly force a stronger commitment?

These questions prevent a common translation mistake: deciding too early which target word “means” the source marker. Evidential markers often interact with tense, aspect, modality, discourse stance and politeness. A translator who begins with the evidence relation has a better chance of choosing a target form that fits the whole sentence rather than mechanically mapping one marker to one gloss.

Direct Evidence: Preserve the Speaker’s Access

Direct evidence usually means the speaker presents the information as grounded in personal perception or immediate access. The simplest example is visual witnessing: “I saw the door open.” But direct evidence can also be auditory or otherwise sensory: “I heard the alarm stop,” “I felt the floor shake,” or a grammatical form that conventionally indicates firsthand knowledge without explicitly naming the sense involved.

When translating into a language that does not require a direct-evidence marker, the translator should not automatically expand every marker into “I saw” or “I personally witnessed.” That may overstate the source. A grammatical evidential may be routine and less emphatic than an explicit English clause. The job is to preserve communicative force, not to magnify a small grammatical signal into a dramatic claim of eyewitness authority.

Conversely, when translating into a language that requires evidential marking, the translator must avoid inventing firsthand access. If the source merely states an event and the surrounding context provides no evidence source, selecting a strongly direct form can create information that was never licensed. Context may resolve the choice, but if it does not, the translator may need a neutral or contextually conventional construction rather than pretending certainty about the speaker’s evidence.

Reported Evidence: Keep the Chain of Attribution Intact

Reported evidence tells the reader that the current speaker is not the original source. English can express this in many ways: “they say,” “I was told,” “according to the report,” “reportedly,” “the witness stated,” “the company says,” or “it is said that.” Other languages may encode reported information through particles or verb forms. Whatever the device, the translation should preserve the fact that the proposition arrived through another source.

This becomes especially important when a sentence carries legal, journalistic or reputational weight. Compare “The minister approved the plan” with “The newspaper reports that the minister approved the plan.” The first sentence directly asserts the approval. The second asserts that a newspaper made that report. A careless translation can collapse the attribution boundary and turn a report about a claim into the translator’s own unqualified assertion.

Think of attribution as a chain: event → original observer or source → reporter → current speaker → reader. Translation should not silently remove a link. This principle also connects with the `Translate | Reported Speech and Attribution` article already in the series. That article focuses on ownership of claims and speech; the present article focuses more specifically on how the source of knowledge itself is encoded and preserved.

Inference: Do Not Convert Evidence into Observation

Inferential evidentiality appears when the speaker reaches a conclusion from available evidence rather than witnessing the event itself. Imagine entering a kitchen and finding a wet umbrella, muddy shoes and a raincoat dripping by the door. You might say, “It must be raining,” even though you have not looked outside. The statement is evidence-based, but the evidence is indirect.

A translation failure occurs when “must have,” “apparently,” “it seems,” an inferential particle or similar device is reduced to a plain factual assertion. “He must have left” is not identical to “He left.” The first sentence contains reasoning: perhaps his coat is gone and his desk is empty. The second simply asserts departure. Conversely, adding “apparently” to a source that presents direct knowledge can weaken the speaker’s commitment and make reliable information sound speculative.

Inference often overlaps with modality, but the two are not identical. Modality concerns possibility, necessity, obligation, permission and related forms of stance; evidentiality concerns information source. An inferential form can sometimes carry both. The translator should therefore ask two separate questions: How does the speaker know? and How strongly does the speaker commit? Keeping those questions separate prevents many subtle errors.

Evidentiality Is Not the Same as Certainty

A person can be highly certain about reported information and uncertain about something personally witnessed. You might trust a laboratory report completely, yet be unsure whether the distant shape you saw was a person. Evidence source and confidence interact, but they are different dimensions. This distinction is crucial when translating words such as apparently, reportedly, seemingly, allegedly, evidently, must and seems.

For example, “reportedly” marks a reporting source. It does not automatically mean the current writer believes the report is unlikely. “Apparently” may indicate appearance, inference or a correction of prior expectation depending on context. “Allegedly” often signals attribution plus legal or rhetorical distance. Translating all three with one generic equivalent for “maybe” destroys distinctions the source is using deliberately.

A useful technique is to annotate two separate fields while analysing a difficult sentence: SOURCE OF KNOWLEDGE and DEGREE OF COMMITMENT. Only after those are clear should you choose the target phrasing. This two-axis method is especially helpful in news, research, compliance and historical texts where the difference between “someone said it” and “it is uncertain” can matter greatly.

Direct, Reported, Inferred and Assumed: A Working Translation Grid

Evidence relationship	Typical English signals	Main translation risk	Question to ask
Direct	I saw, I heard, direct grammatical marker	Over-expanding a routine marker into emphatic eyewitness language	How explicitly does the source claim firsthand access?
Reported	reportedly, according to, they say, I was told	Turning attributed information into the translator’s assertion	Who is responsible for the original claim?
Inferred	must have, apparently, seems, judging from	Turning a conclusion from evidence into a witnessed fact	What evidence supports the conclusion?
Assumed	presumably, of course, expected to, understood to	Confusing shared expectation with observation	Is this known, inferred or merely presupposed?
Unspecified	plain assertion or neutral form	Inventing a source because the target grammar encourages one	Does context license a more specific evidential choice?

Worked Example 1: “Apparently, the Server Restarted”

Start with the sentence: “Apparently, the server restarted during the night.” The event is a restart. But apparently tells us the writer does not simply present the restart as directly witnessed fact. Perhaps logs show a fresh boot time; perhaps somebody reported it; perhaps symptoms imply it. The immediate task is to discover which interpretation the document supports.

If surrounding sentences say, “The logs show a boot event at 02:13,” the evidential basis is inferential or documentary. A target form equivalent to “judging from the evidence” may fit. If the next sentence says, “The night operator told us it restarted,” the basis is reportative. If there is no further evidence, the translator may need a general marker of appearance or non-firsthand knowledge. The same English adverb can therefore map to different target choices.

The lesson is larger than this example: translate the evidential relationship, not the dictionary headword. A bilingual dictionary can offer candidate equivalents for apparently, but only context tells you which evidence relation is active. That is why translation quality depends on a context stack rather than isolated lexical replacement. See The Context Stack for the broader method.

Worked Example 2: “She Must Have Forgotten”

“She must have forgotten the meeting” contains a conclusion. The speaker does not necessarily possess direct access to the person’s memory. The evidence might be an empty chair, no reply to messages and a missed start time. In English, must have can signal a strong inference about a past event or state.

A word-for-word translation of must as obligation can fail badly. The sentence is not saying that she was required to forget. It is saying that forgetting is the speaker’s best explanation. The translator must therefore recognise the grammatical form as inferential rather than deontic. This is a classic case where surface vocabulary hides the true semantic job.

To test your translation, paraphrase the source before translating: “Given the evidence, I conclude that she forgot the meeting.” If your target version preserves that reasoning relation, it is probably on the right track. If it sounds like a command, duty or direct observation, the evidential value has been lost.

Worked Example 3: “They Say the Bridge Is Closed”

“They say the bridge is closed” has an intentionally vague source. The speaker is relaying circulating information but does not identify who originally said it. A translator may be tempted to tidy the sentence by replacing they with “officials,” “people,” “the authorities” or another plausible source. That would be an invention.

The correct translation preserves the level of attribution provided by the source. If the source is vague, the target should remain appropriately vague unless the broader document clearly identifies the referent. This is a fundamental translation discipline: ambiguity and underspecification are sometimes part of the meaning. Clarity does not mean adding unsupported facts.

This example also shows why translators should distinguish translation clarity from information completion. You can make syntax natural and still preserve an unspecified source. The target should be easy to understand, but it should not answer questions the original leaves unanswered.

When English Uses “Apparently”

Apparently is deceptively flexible. It can indicate that something appears true from available evidence, that the speaker learned it indirectly, that a surprising fact has become known, or that the speaker is distancing themselves slightly from the proposition. Because of that range, mechanically translating it with one fixed equivalent is risky.

Compare: “Apparently the shop closes at six,” “Apparently he has three passports,” and “The machine is apparently functioning normally.” The first may be based on information someone supplied; the second may signal surprising reported information; the third may indicate an assessment from observed behaviour or tests. Good translation begins by asking what justifies the statement in its local context.

A practical habit is to replace apparently temporarily with a fuller paraphrase: “according to what I have learned,” “judging from what we can observe,” “it now appears that,” or “the available evidence suggests.” The paraphrase is not necessarily the final translation. It is a diagnostic tool that reveals which meaning must survive.

When English Uses “Reportedly”

Reportedly normally signals that the proposition comes from reports rather than from the writer’s direct assertion. “The company reportedly plans to sell the division” means that reports exist claiming such a plan. It does not necessarily say who produced those reports, how reliable they are, or whether the writer endorses them fully.

In translation, the most important feature is attribution distance. If the target equivalent sounds like “the company announced,” the translation has strengthened the source. If it sounds like “perhaps,” it may have converted report source into uncertainty. If it drops the marker entirely, the translated writer may appear to assert as fact what the source only attributes to reports.

Journalistic translation benefits from a source audit: underline every phrase that answers “who says?”, “according to what?”, “based on which document?” or “from what evidence?”. Then verify that each source relationship remains visible after translation. This simple pass catches errors that ordinary grammar checking misses.

Evidentiality in Academic and Scientific Writing

Academic writing constantly manages evidence source, even when it does not use grammatical evidentials. Phrases such as “the data indicate,” “previous studies suggest,” “participants reported,” “the authors observed,” “the model predicts,” and “the results appear to show” distinguish different routes from evidence to claim. Translation must preserve those routes.

Consider the difference between “participants experienced fatigue” and “participants reported fatigue.” The second statement tells us that fatigue was self-reported. A translation that removes reported can accidentally present subjective reports as independently measured outcomes. Likewise, “the analysis suggests an association” should not become “the analysis proves an association.” Evidence verbs are part of research logic.

When translating research, create a small verb map for evidential functions: observe, measure, report, indicate, suggest, infer, estimate, predict, demonstrate, establish. These are not stylistic synonyms. They encode different relationships between data and claim. The eduKateSG Vocabulary Learning Hub is useful here because strong translation depends on depth of word knowledge, collocation and semantic precision rather than a large list of approximate substitutes.

Evidentiality in News and Public Information

News writing frequently separates confirmed facts, official statements, eyewitness accounts, anonymous reports, preliminary assessments and interpretation. Translation should maintain those categories. “Police confirmed,” “witnesses said,” “local media reported,” “video appears to show,” and “investigators believe” are not interchangeable openings.

A dangerous error is source promotion: moving a claim upward in certainty or authority. An eyewitness statement becomes an official confirmation; a report becomes a fact; an analyst’s interpretation becomes the institution’s position. The reverse error is source demotion: confirmed information becomes merely alleged or speculative. Both change the informational architecture of the text.

The translator should therefore preserve not only names and quotations but the verbs that connect sources to claims. In high-stakes documents, it can help to mark each proposition with a compact label during revision: DIRECT, OFFICIAL, REPORTED, WITNESSED, INFERRED, ESTIMATED, UNKNOWN. The final reader never sees the labels; they are a quality-control layer for the translator.

Evidentiality in Historical Sources

Historical texts often contain nested evidence. A chronicler may report what a traveller claimed to have heard from local residents about an event that occurred decades earlier. Modern translation can easily flatten this chain because contemporary prose often prefers shorter sentences and more direct statements. Yet the chain is part of the historical evidence.

If a source says, in effect, “It is related by merchants that the city once stood farther east,” translating it as “The city once stood farther east” removes the historian’s distance from the claim. The original may be valuable precisely because it records a tradition, not because it independently verifies geography. A responsible translation lets readers see the evidence status the historical writer provided.

When translating old texts, resist the urge to modernise evidential structure simply because it feels repetitive. Repeated “he says,” “it is told,” “they report” or “I learned from” phrases may be the author’s method of documenting knowledge pathways. Style can be modernised carefully, but evidence provenance should remain recoverable.

Machine Translation and Evidential Errors

Machine translation systems can produce fluent output while mishandling evidentiality because the information may be distributed across morphology, particles, syntax and discourse context. A system may choose a natural target sentence that communicates the event but drops how the speaker knows. Fluency can hide the error because the result reads smoothly.

Another failure occurs when the system regularises unusual evidence marking into common target phrasing. A reportative marker may become a generic past tense; an inferential form may become a direct assertion; a direct evidential may be expanded unnecessarily. Automated output therefore needs semantic review, not only grammar review.

A good post-editing question is: Would a reader of the translation know the claim came from the same kind of evidence as a reader of the source? If not, the translation needs repair even if every noun and verb appears plausible. Translation quality is not the absence of awkward sentences. It is the preservation of consequential distinctions.

A Step-by-Step Method for Translating Evidentiality

Identify the proposition. State plainly what event, state or claim the sentence is about.
Identify the evidence source. Direct perception, report, inference, assumption, document, measurement or unspecified source?
Separate evidence source from certainty. Do not treat reported as uncertain or inferred as weak by default.
Locate the source-language signal. Is evidential meaning carried by morphology, a particle, adverb, verb, clause or context?
Choose the target-language carrier. Preserve the function even if the grammatical form changes.
Check attribution. Make sure the translated speaker does not inherit somebody else’s claim.
Check strength. Have you made the evidence more direct, more certain or more authoritative than the source?
Read the paragraph, not only the sentence. Evidence source may be established earlier and maintained across several sentences.
Run a reverse paraphrase. Explain what your translation says about how the speaker knows. Compare that explanation with the source.

Why Paragraph Context Matters

Evidentiality often cannot be translated sentence by sentence in isolation. A paragraph may begin with “According to the inspection report” and then contain five unmarked sentences describing defects. Those sentences inherit the report as their evidence source. Repeating “the report says” in every target sentence may be clumsy, but deleting the opening attribution may wrongly make the translator the source.

The opposite pattern also occurs. A paragraph may shift from reported evidence to the writer’s own observation: “Witnesses said the lights failed at midnight. Photographs taken the next morning show damage to the control panel.” A translation must preserve the shift because the second sentence has a different evidence base.

This is why the translation unit is sometimes larger than a sentence. The relevant meaning may live across a paragraph or document. eduKateSG’s Translation Unit article develops this principle in detail: translate the smallest unit that contains enough information to make the correct decision, not the smallest unit that happens to be separated by punctuation.

Common Failure 1: Deleting the Evidence Marker

This happens when the translator treats an evidential marker as optional decoration. “The road appears to be blocked” becomes “The road is blocked.” “The witness reportedly left” becomes “The witness left.” The target sentence may sound cleaner, but it makes a stronger claim than the source.

To catch the error, underline every source element that modifies the writer’s relationship to the proposition. Then ask where each function appears in the target. It does not need a one-to-one lexical match, but it needs a functional home. If the function has nowhere to live, it has probably been lost.

Common Failure 2: Adding Unsupported Certainty

Translators sometimes strengthen sentences because direct statements sound more natural. A source equivalent to “It seems the shipment was delayed” becomes “The shipment was delayed.” This may feel harmless, especially if delay is likely, but translation is not an invitation to settle uncertainty on the author’s behalf.

A useful discipline is to imagine that the translation will be used as evidence in a dispute. Would the translated wording make the author responsible for a stronger claim than they actually made? If yes, restore the evidential or modal boundary. The same principle applies in everyday prose even when stakes are lower.

Common Failure 3: Treating Reported Evidence as Doubt

Reported information is information obtained from another source. It is not automatically unreliable. A translator who routinely renders reportative forms with words equivalent to “maybe” or “supposedly” can introduce scepticism absent from the original. That changes stance.

Ask whether the source marks origin or evaluation. “According to the laboratory” identifies origin. “The laboratory’s questionable claim” evaluates reliability. Do not merge these. A careful translator lets the author decide when to express doubt.

Common Failure 4: Confusing Inference with Obligation

English must provides a familiar example. “You must leave now” expresses obligation. “He must be home already” often expresses inference. A target language may use entirely different constructions for these functions. Translating the surface modal rather than the function can generate absurd or misleading results.

The diagnostic is to paraphrase. If the meaning is “the evidence strongly supports this conclusion,” choose an inferential solution. If the meaning is “a rule or necessity requires this action,” choose an obligation solution. The word is the same; the semantic architecture is not.

Common Failure 5: Over-Explaining a Grammatical Evidential

The opposite of deletion is inflation. A small grammatical marker can be translated into a heavy phrase such as “I personally witnessed with my own eyes that…” when the source merely uses an ordinary direct evidential. This makes the target sound emphatic, defensive or theatrical.

Functional equivalence requires attention to prominence. Ask how noticeable the evidential is to an ordinary reader of the source. If it is routine grammar, the target expression should usually feel routine too. Meaning includes information structure and rhetorical weight, not only dictionary definition.

Build an Evidence Ledger

For difficult texts, create a compact evidence ledger before final revision. Each important proposition receives four fields: claim, source, access, commitment. For example: “Factory closed — local authority notice — documentary/reportative — high commitment.” Or: “Cause was electrical — investigator inference — indirect — provisional.” The ledger can be kept mentally for short texts or written in a table for complex work.

The ledger helps with long documents because the same fact may recur in several forms. An early paragraph may say “witnesses reported smoke,” while a later summary says “smoke was observed.” Is the later sentence still based on witnesses, or does the document now introduce photographic evidence? Without tracking evidence provenance, translators can accidentally merge sources.

This method also supports revision by another person. A reviewer can challenge a choice with a concrete question: “Why is this direct in the target when the source is reportative?” That is far more useful than a vague comment such as “this sounds too strong.” Good translation workflows make reasoning inspectable.

Vocabulary Depth Matters

Evidential translation depends on recognising fine differences among verbs and adverbs. See, observe, notice, witness, hear, learn, report, allege, claim, state, indicate, suggest, imply, infer, deduce, assume, presume, estimate, appear, seem all participate in evidence and stance, but not in the same way.

A learner who knows only rough dictionary meanings may treat these as interchangeable. A translator needs deeper lexical knowledge: typical subjects, collocations, register, strength, grammatical patterns and discourse effects. That is why translation and vocabulary learning support each other. Translation exposes where “knowing a word” must become “knowing what the word does in a sentence.”

For English learners, How English Works provides the broader language-system connection, while the Vocabulary Learning Hub develops word knowledge, retrieval and precision. The translation lane should use those owners rather than duplicate them: translation is where language knowledge is stress-tested by movement between systems.

Practice Set: Identify the Evidence Relationship

“The lights appear to have failed after the storm.”
“According to the technician, the fuse had already blown.”
“I heard the generator start at 3 a.m.”
“The footprints suggest that two people entered.”
“The company is reportedly considering a sale.”
“The figures seem inconsistent with last month’s report.”
“Witnesses say the vehicle turned left.”
“The sensor recorded a sudden temperature increase.”
“She must have taken the earlier train.”
“It is generally understood that the gate remains locked overnight.”

For each sentence, write two notes before translating: what is claimed and how the writer knows. Then translate. Finally, cover the source and inspect the target alone. Can you still recover the evidence relationship? If not, revise. This exercise trains the habit that matters more than memorising labels.

Practice Set: Repair the Over-Strong Translation

Imagine that each source below contains the evidential wording shown, but the draft target has been reduced to a plain assertion. Your job is to restore the missing relationship:

Source: “The medication appears to have caused the reaction.” Draft target meaning: “The medication caused the reaction.”
Source: “The committee reportedly rejected the proposal.” Draft target meaning: “The committee rejected the proposal.”
Source: “Judging from the damage, the impact came from the east.” Draft target meaning: “The impact came from the east.”
Source: “Residents say the water was discoloured.” Draft target meaning: “The water was discoloured.”
Source: “The results suggest a seasonal effect.” Draft target meaning: “The results prove a seasonal effect.”

Notice that some repairs need only one adverb; others need a reporting clause or evidence phrase. Do not force one grammatical pattern across all items. Translation competence grows when you learn to preserve a function through different surface forms.

A Review Checklist for Evidential Translation

Have I identified the source of knowledge for every important claim?
Did I preserve direct versus non-direct evidence where the distinction matters?
Did any report become an unqualified assertion?
Did any inference become direct observation?
Did I confuse information source with uncertainty?
Did I invent an authority that the source never named?
Did I weaken a confirmed fact into mere possibility?
Did I over-explain a routine grammatical marker?
Does the evidential relationship remain coherent across the paragraph?
Can a reader tell who owns each claim?

Where Evidentiality Sits in the Translation System

Evidentiality is not an isolated grammar curiosity. It sits at the intersection of context, modality, attribution, discourse and lexical choice. The translator first needs enough context to identify the evidence relationship, then enough grammatical flexibility to express it naturally, then enough vocabulary precision to avoid strengthening or weakening the claim.

Within eduKateSG’s wider translation architecture, this article belongs under the principle that translation moves meaning relations, not strings. The Context Stack helps determine the evidence source; the Translation Unit helps choose how much text must be considered together; the Equivalence article explains why the target form may differ; and the Terminology System protects specialised evidential vocabulary in technical domains.

The Deeper Principle: Do Not Change the Reader’s Right to Know

When a source tells readers how a claim was obtained, that information gives them a basis for interpretation. A witnessed event, a rumour, a laboratory measurement, a model prediction and an inference from circumstantial evidence are different kinds of knowledge. Readers may reasonably respond to them differently. Translation should preserve that opportunity.

This is why evidentiality belongs to translation ethics as well as grammar. A translator should not quietly convert testimony into fact, inference into observation or report into endorsement. Nor should the translator insert scepticism where the source expresses none. The target reader deserves access to the same evidence relationship the source reader received.

The practical rule is worth repeating: do not make the speaker know more, less, or differently than the source allows. Once that rule becomes habitual, evidential markers stop looking like small optional words. They become what they really are: instructions for how a claim is connected to knowledge.

Final Translation Drill

Take any paragraph from a news report, research paper, memoir or technical incident log. Before translating, circle every expression that signals information source. Include reporting verbs, evidence verbs, attribution phrases, inferential modals and appearance verbs. Then write a one-line evidence map for the paragraph. Only after that should you translate.

On revision, perform a source-blind test. Read only the translation and ask: Which claims did the writer witness? Which came from other people? Which are inferred? Which are measured? Which remain uncertain or merely assumed? If your answers differ from what the source supports, the translation is not finished.

That discipline scales from a single sentence to a long document. It makes translations more accurate, more accountable and more useful because it preserves not only information but the route by which information became knowable. That is the real work of translating evidentiality.

Translate | Evidentiality and Information Source — Preserve Seen, Heard, Reported and Inferred Meaning Across Languages