Ambiguity in Translation: Decisions, Author Queries and Multiple Readings

How to translate ambiguity accurately without making up the author’s meaning

How do you translate an ambiguous sentence without choosing a meaning that the original never established? A reliable translation method starts by listing the genuinely plausible readings, testing them against context, and deciding whether to clarify with the author, preserve the uncertainty, or translate an interpretation that is demonstrably supported. This guide explains ambiguity in translation, translating ambiguous words and sentences, lexical ambiguity, syntactic attachment, vague language, author queries and quality review for world-facing communication.

The hidden problem is that a fluent translation can resolve uncertainty without announcing that it has done so. One language may force the translator to choose an aspect, a pronoun relation, a quantity distinction, a level of formality or a word sense that the source leaves open. Other ambiguities are deliberate: a writer may sustain two interpretations for humour or narrative suspense. In either case, choosing a plausible reading is not the same thing as recovering an intended reading, and an unmarked choice can change instructions, commitments or a reader’s understanding.

Readers looking for how to translate unclear text, when to ask an author for clarification, how to preserve ambiguity, or how to avoid mistranslation caused by multiple meanings need more than a dictionary. The repeatable method in this chapter is to identify the ambiguity type, enumerate interpretations, determine which are supported, assess the cost of choosing incorrectly, issue a neutral clarification query when necessary, and preserve a decision trace for reviewers and future versions. Extensive worked examples show the method in technical, creative and everyday communication without assuming any particular country or language pair.

This is a cross-document decision and accountability chapter, not a replacement for eduKateSG’s existing pages about vocabulary meanings, English grammar, quantifier scope, reference or information structure. Those resources explain the linguistic building blocks. The distinct problem owned here is what a translator does when several readings remain viable, the intended reading is not recoverable, and a target-language output nonetheless must be prepared and reviewed. The safest result may be an author-approved choice, a controlled note, or a target phrase that carefully remains open.

The distinction between ambiguity, vagueness, underspecification and context sensitivity matters. Multiple discrete readings are not the same as a fuzzy boundary or a missing detail. The Stanford Encyclopedia of Philosophy discusses those distinctions explicitly; a translation workflow must therefore diagnose the uncertainty before selecting its cure. Guidance in ISO 11669:2024 concerns translation project specifications, communication, needs and risk, while ISO 5060:2024 describes analytic evaluation of translation output. Neither standard licenses translators to invent facts or author intentions.

The cases and sample author queries below are fictional teaching material. They are not established translations for every language, nor are they legal, clinical or operational instructions. When real decisions depend on an ambiguous source, qualified language and subject specialists should review the source, any clarification and the final target. An unresolved source may make a confident target impossible; documenting that limit is a successful professional decision rather than a failure of fluency.

Why deciding for the author is often the wrong translation

Translators regularly encounter a source phrase that supports more than one reading. Context can sometimes narrow that field to one. Sometimes the author can answer a neutral question. Sometimes an institution must specify the decision because the text governs a process. But in literature, deliberate indeterminacy may be the communicative point: removing it would damage the artistic effect. The translation challenge is to distinguish these cases rather than apply a blanket command to make every sentence maximally explicit.

There are two different questions in a good review. First, does the target sentence preserve the meaning supported by the source? Second, has the translator gone beyond the source in deciding something unestablished? A target might be plausible and idiomatic yet fail the second test. For a public-facing claim, the cost may be a wrong reader action. For a fictional narrative, the cost may be destruction of humour, surprise or doubt. The correct approach starts from the reader’s job and the limits of the available evidence.

Linguistic ambiguity often has two or more distinguishable interpretations. Vagueness instead concerns imprecise boundaries; underspecification concerns a dimension left unset without necessarily producing distinct lexical readings. For instance, a short request to bring a container may leave its size unstated, but does not automatically mean that the word container has exactly two competing senses. The distinction matters: the proper action may be to ask for a missing requirement rather than search for a new synonym.

The ambiguity triage ledger

A useful ledger assigns an ID to each ambiguous segment and stores: the exact source span, at least two plausible readings, evidence favouring or ruling out each, the kind of uncertainty, intended readership, decision consequence, clarification owner, query and answer, authorised wording, reviewer identity, and source version. Record a genuinely unresolved state rather than filling every cell with a comforting guess. This permits revision when new context appears without hiding the reasoning by which the original target was selected.

An author query should be neutral. Instead of asking “You mean that the technician inspected the valve, correct?”, a careful query offers the alternatives and identifies the affected sentence: “In the sentence about the technician and the valve, does the word ‘after’ modify the inspection or the repair? This changes the order we would communicate.” It is specific enough to answer and does not pressure the author toward the translator’s preferred interpretation. If the source owner is unreachable, the team must follow project rules for unresolved ambiguity.

Every source ambiguity is not equally urgent. A double reading in a marketing tagline might be intentional. An uncertain condition in a safety instruction can change whether a device may be operated. A vague adjective in a literary description may be acceptable. Risk triage asks whether the alternative readings change permissions, numbers, participants, timing, obligations, privacy, factual claims or safety. The scale of consequence should guide the level of clarification and specialist review.

Mechanism 1: Lexical homonymy

One written or spoken form can name two unrelated categories, as with an object name and an action. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A dictionary may return several glosses, but only the surrounding task reveals the relevant sense. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Build two complete interpretations, then test what the neighbouring verbs and objects permit. For a concrete diagnostic, consider this original training scenario: When a container label has the word seal, ask whether it names an animal, a closure or the act of closing; the domain will usually settle it. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Final acceptance question: which interpretation would an independent target reader infer if they saw only this line and no footnotes? If that reading differs from the author-verified or deliberately unresolved position, the translation needs revision. For reusable systems, store the selected meaning with a source identifier and rationale, not just the translated surface phrase; a later context change may invalidate the original choice.

Mechanism 2: Related word senses

A word may have connected senses that are hard to divide into dictionary entries. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The same term can describe a physical object, a process or an abstract state without a clean homonym boundary. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Choose a context-supported meaning instead of assuming a one-to-one source–target pair. For a concrete diagnostic, consider this original training scenario: The word channel can mean a physical passage, a signal route or a communications medium; retain the intended technical frame. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 3: Part-of-speech uncertainty

A form may function as a noun or verb, or a modifier may be read as an action. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Target grammar can force a category that the source leaves unstated. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Test grammatical role with an expanded paraphrase before selecting morphology. For a concrete diagnostic, consider this original training scenario: A short fragment such as record update can indicate an action to update the record or information about the record’s last update. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 4: Modifier attachment

A prepositional or adjectival phrase can modify different parts of a sentence. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Two parse structures can support different participants, locations or restrictions. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Draw brackets around each possible constituent and paraphrase both readings. For a concrete diagnostic, consider this original training scenario: Inspect the box with the camera might mean the box contains a camera or an inspection made using one; operational instructions cannot leave this accidental. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 5: Coordination attachment

A modifier or exception may apply to one item or several in a list. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The ambiguity may alter who is included and what requirements apply. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Rewrite alternative bracketings and ask which set the author intends. For a concrete diagnostic, consider this original training scenario: For storage and transport of hazardous materials, a single adjective may modify both processes or only one; do not decide by punctuation instinct. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 6: Quantifier scope

All, some, each, two or at least one can interact with other operators. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Different logical readings may create different obligations or distribution. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Represent the alternatives using natural-language test scenarios, not only symbols. For a concrete diagnostic, consider this original training scenario: Each team may assign a reviewer can mean one shared reviewer is allowed or each team may choose its own; the intended allocation matters. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 7: Negation scope

Not may target an action, an obligation, a whole condition or one participant. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Misplacing the negative can turn an optional step into a prohibition. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Ask precisely which proposition is denied and test against a simple counterexample. For a concrete diagnostic, consider this original training scenario: Not required to approve is different from required not to approve; a short target must maintain that distinction. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 8: Temporal reference

Soon, next, before, after and by can depend on speaking time or context. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A translator may add a date or sequence not established by the source. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Record event time, reference time and delivery time separately. For a concrete diagnostic, consider this original training scenario: The next inspection may refer to a scheduled future inspection or the inspection immediately following another event; the timeline must be checked. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 9: Duration versus deadline

Within a duration can mean completion by a deadline or continued occurrence over a period. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The target may impose a measurable requirement the source never intended. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Identify whether the specified time is a point, interval or maximum elapsed period. For a concrete diagnostic, consider this original training scenario: Finish within two hours is an elapsed-time limit, not necessarily a clock-time appointment; timestamps cannot be guessed. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 10: Reference ambiguity

A pronoun, demonstrative or repeated noun phrase can point to different people or objects. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A target language may force gender, number or grammatical class and thereby choose a referent. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Trace entity chains and ask the source author when the referent changes the action. For a concrete diagnostic, consider this original training scenario: When Ana spoke to Bea after she arrived, she may refer to either person, depending on context. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 11: Zero arguments

Some languages omit subjects or objects that another language requires. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Inserting a named actor from context can be wrong when multiple actors are salient. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Distinguish recoverable grammatical omission from genuinely unresolved reference. For a concrete diagnostic, consider this original training scenario: A report may simply say was approved; if the approving authority is not stated, a translation must not invent one. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 12: Collective versus individual

A group noun or plural verb can describe collective action or separate actions. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Target agreement may make the division explicit. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Test whether the action happened once for the group or once per member. For a concrete diagnostic, consider this original training scenario: The reviewers signed can mean a single shared document or separate forms; ask before translating into a system requiring individual records. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 13: Modal force

Can, may, must, should and would can convey ability, permission, possibility or obligation. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Apparent synonymy disguises changes in authorisation. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Compare permitted actions under every reading before choosing a target modal. For a concrete diagnostic, consider this original training scenario: You may open the cover can describe permission, while the cover may open describes possibility; a safety notice cannot confuse them. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 14: Conditional reading

If, unless, only if and provided that can express different logical conditions. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Target restructuring may reverse the direction of an implication. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Write test cases where the condition is true and false and check the authorised action. For a concrete diagnostic, consider this original training scenario: Access is allowed if approved does not itself imply that approval is the only possible condition or that approval has been granted. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 15: Comparative baseline

Better, more, less or higher can omit the group or period used for comparison. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The target may wrongly supply a numerical baseline or universal claim. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Identify the compared items and leave the comparator unspecified if the source does. For a concrete diagnostic, consider this original training scenario: A report saying response improved could compare with yesterday, a control group or a target; the translator must not choose silently. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 16: Approximation

About, roughly and nearly express imprecision whose bounds depend on context. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The target may make a rounded statement look exact. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Preserve stated precision and avoid converting a vague estimate into a fixed threshold. For a concrete diagnostic, consider this original training scenario: About ten hours does not automatically mean exactly ten hours; in a contract the range may need clarification. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 17: Vagueness

Terms such as adequate, soon or substantial may have borderline cases rather than two fixed senses. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A translation can create a false numerical standard. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Treat vagueness as a request for criteria where a decision depends on it. For a concrete diagnostic, consider this original training scenario: A sufficiently strong signal needs an operational test if engineers must accept or reject it, but fiction may use the phrase without thresholds. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 18: Underspecified actor

An instruction may omit who must perform an action. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Adding a subject changes responsibility. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Identify whether an imperative or passive structure already establishes an implicit actor. For a concrete diagnostic, consider this original training scenario: Check before release might leave it unclear whether a designer, operator or reviewer must check the item. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 19: Underspecified object

An instruction may name an action without specifying its affected item. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A translator may select a nearby object that is not uniquely intended. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Use source-context traceability and escalate if several objects fit. For a concrete diagnostic, consider this original training scenario: Replace after inspection is incomplete when the equipment has several replaceable components. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 20: Presupposition

A question or construction can treat a proposition as already true. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Translating background assumptions as new facts or vice versa changes argumentative force. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Separate what is asserted from what the wording assumes. For a concrete diagnostic, consider this original training scenario: Why did the supplier stop sending reports presupposes that reports were sent before; it is unsuitable if that history is contested. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 21: Implicature

Speakers may imply more than the explicit proposition through context and convention. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Literal target wording can communicate a different indirect request or judgement. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Test what a competent reader would infer and whether that inference was intended. For a concrete diagnostic, consider this original training scenario: That window is open could be an observation or a request to close it; do not automatically convert every report to an order. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 22: Speech-act ambiguity

A question may function as a request, warning, suggestion or factual inquiry. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: An over-literal or over-direct target can change politeness and obligation. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Distinguish the sentence’s grammatical form from the social act it performs. For a concrete diagnostic, consider this original training scenario: Could you review this might be a polite request, but could ask about ability in a training interview; context matters. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 23: Quotation boundary

A text may not clearly mark whether a phrase is quoted or editorial paraphrase. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The target may invent verbatim speech or change attribution. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Check punctuation conventions and source-document structure. For a concrete diagnostic, consider this original training scenario: A report using he said the room is safe may allow a direct quoted sentence or an indirect description, but quotation marks should not be fabricated. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 24: Anaphoric ellipsis

A later clause may omit a repeated verb phrase whose antecedent is disputed. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: The translator can fill in the wrong action. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: List eligible antecedents and evaluate structural and discourse cues. For a concrete diagnostic, consider this original training scenario: Sam prepared the kit and Jo did too may share the preparation action, but an intervening clause can complicate which action is repeated. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 25: Document heading inheritance

A heading may govern several bullets or only the first group. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A translator can attach a condition to the wrong table rows. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Inspect the source layout and document hierarchy before linearising. For a concrete diagnostic, consider this original training scenario: A warning heading above two sections may be intended to apply to both; check nesting and approved formatting. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 26: Table column ambiguity

A label may refer to a count, total, change or status whose unit is hidden. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A localised table can look precise yet change the meaning of values. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Recover field definitions and units from the data dictionary. For a concrete diagnostic, consider this original training scenario: A column labelled net might mean net amount, remaining quantity or network status in different applications. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 27: Multimodal cue

Image, timing, gesture or audio stress may disambiguate written words. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A standalone transcript may appear ambiguous when video resolves the intended interpretation. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Inspect authorised media and preserve meaningful cues in captions or notes. For a concrete diagnostic, consider this original training scenario: A speaker points at the left item while saying that one; the transcript alone may lose that reference. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 28: Prosodic contrast

Stress, pauses and intonation can distinguish alternatives invisible in text. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Subtitles or transcripts may lose emphasis that selects a contrast. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Listen to source audio when available and annotate only supported features. For a concrete diagnostic, consider this original training scenario: I said the red folder can contrast colour or object depending on spoken stress; translator should not fabricate prosody. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 29: Strategic ambiguity

A speaker may intentionally avoid committing to a specific reading. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: An overhelpful translation can turn careful diplomatic, creative or commercial language into a commitment. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Preserve the degree of openness unless the authorised source owner instructs otherwise. For a concrete diagnostic, consider this original training scenario: We are considering options may deliberately avoid naming a chosen option; the target should not imply a settled decision. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 30: Literary indeterminacy

An author may sustain two interpretations for aesthetic effect. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Clarifying every allusion can remove mystery or humour. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Use candidate expressions that permit comparable reader discoveries. For a concrete diagnostic, consider this original training scenario: An ambiguous ending may intentionally leave the narrator’s knowledge unresolved; an explanatory translation would spoil the story. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 31: Humour and wordplay

A pun may exploit sound, spelling or two meanings simultaneously. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Literal substitution can preserve words while destroying the dual reading. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Identify the two meanings, their timing and their effect before adapting. For a concrete diagnostic, consider this original training scenario: A headline can use a term that means both a tool and an organisational role; a target may need a different pun with the same double function. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 32: Cultural frame

A phrase may evoke a culturally shared practice that changes interpretation. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A bare literal wording may not expose the alternatives to unfamiliar readers. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Check the source culture and choose a functional explanation or unobtrusive note where justified. For a concrete diagnostic, consider this original training scenario: An invitation to bring a plate can mean different things in different social settings; assuming physical crockery may misstate the event. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 33: Technical term collision

The same token can be a common word, a formal parameter or a product feature. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Translating a defined identifier as prose can break user action or software. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Use the governing schema and preserve defined tokens when required. For a concrete diagnostic, consider this original training scenario: A string such as title in markup is a formal attribute; a nearby ordinary use of title is translatable as natural language. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 34: Machine-generated confidence

A tool may select one interpretation because it produces one fluent output. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A single candidate is not proof that competing readings were excluded. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Require explicit candidate enumeration and human evidence before approving meaning. For a concrete diagnostic, consider this original training scenario: An automated sentence about a vessel might choose ship over container because of corpus frequency; context must establish it. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 35: Author correction

A source owner may answer a query with wording that changes the original text. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: A translation may reflect the answer while other language versions remain tied to the old source. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Record source revision, approved reading and propagation to all target locales. For a concrete diagnostic, consider this original training scenario: If the author clarifies that all reviewers need separate signatures, update both source instructions and translations if permitted. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

Mechanism 36: High-stakes uncertainty

An unresolved source can support conflicting permissions, duties or safety actions. This is a specific interpretive problem, not a general failure to read carefully. An analyst should show what the source form itself licenses, what the surrounding passage supports, and what information is still missing. If a word is merely underspecified, the analyst should not invent an arbitrary menu of senses; if the construction genuinely has two readings, both should be stated explicitly before any target is selected.

Translation risk: Publishing one confident reading can cause harmful behaviour. The review can test this by describing one situation in which a proposed target is true and a second situation in which the source remains acceptable but the target is not. That contrast reveals precisely what meaning the translation silently selected or erased. Where the two readings would change a person’s action, the item should be escalated before publication rather than hidden behind fluent language.

Recommended method: Escalate, hold or explicitly identify limits according to project authority. For a concrete diagnostic, consider this original training scenario: A machine instruction saying remove the guard before testing needs clarification if the temporal sequence changes exposure to hazards. Write two source-language paraphrases that make the competing interpretations explicit. Next ask which facts, document attachments or authority decisions could select one. If that evidence does not exist, record an unresolved decision instead of trusting the first plausible output.

The 20-gate ambiguity decision workflow

Gate 1: Establish the reader’s task

Identify what a reader needs to decide or understand from the target text. A sentence in fiction has different tolerances from an instruction that authorises access.

Output required for this gate: a short, inspectable entry in the ambiguity register that cites the evidence used or states that no decisive evidence was available. A high-quality review does not say only “checked.” It identifies what competing meaning was considered, what reader inference would result from a wrong choice, and how the target preserves the authorised interpretation or an intentional openness. If this cannot be demonstrated, the line remains a query or review item.

Gate 2: Freeze the actual source version

Store the precise source sentence, document revision and adjoining material. A clarification against the wrong version can create an invisible disagreement.

Gate 3: Read a full context window

Inspect several neighbouring paragraphs, headings, tables, illustrations and referenced documents before deciding that a short phrase is ambiguous.

Gate 4: Identify ambiguity versus missing detail

Decide whether the source licenses multiple readings, has a fuzzy boundary, depends on an unstated referent or simply leaves a feature irrelevant.

Gate 5: Enumerate plausible readings

Write each interpretation as a complete, testable proposition. Do not inflate the list with implausible grammatical curiosities.

Gate 6: Find discriminating evidence

Look for source cues that would be true under one candidate and false under another, such as event sequence, participants or a data definition.

Gate 7: Check domain conventions

Review approved terminology, specification language or established genre conventions without treating custom as a substitute for the document’s actual intent.

Gate 8: Determine intentionality

Ask whether ambiguity contributes to humour, persuasion, politeness, narrative suspense or a deliberately open position.

Gate 9: Assess consequence and severity

Measure how different readings change permissions, financial figures, identity, timing, obligations or factual claims.

Gate 10: Select clarification owner

Identify the author, technical lead, policy owner, researcher or other person authorised to answer the disputed question.

Gate 11: Write a neutral query

Quote the source and present at least two plausible interpretations without leading the owner toward the translator’s preferred answer.

Gate 12: Record the response literally

Store the answer, responder role, timestamp and whether the response clarifies one sentence or changes the source document.

Gate 13: Decide one of three outcomes

Choose an evidenced interpretation, deliberate preservation of multiple readings, or a documented hold or limitation when a consequential ambiguity remains unresolved.

Gate 14: Draft target-language alternatives

Write idiomatic candidate renderings for the chosen outcome. Target grammatical constraints may require a different sentence structure to keep both readings available.

Gate 15: Run a back-reading test

Give the target wording to a competent reviewer without the source and ask what it permits, forbids and leaves uncertain.

Gate 16: Compare unintended commitments

Check whether the target invented a particular actor, number, date, certainty, legal duty, causal relation or status that the source did not establish.

Gate 17: Check short formats

Review headings, captions, labels and one-line messages where a note or contextual cue may be lost.

Gate 18: Apply qualified domain review

For high-impact texts, have appropriate practitioners examine the intended consequence and interpretation alongside a language specialist.

Gate 19: Propagate approved source changes

Update dependent translated strings and documentation when the source owner revises the original, while retaining earlier versions for audit.

Gate 20: Release with an accountable record

Attach the target to a source version, decision trace, reviewer approval and known unresolved questions so later teams can reconstruct the choice.

How ambiguous translations fail: thirty diagnostic patterns

When a source supports multiple meanings, a polished target can appear to settle a question that the writer never answered. The patterns below give reviewers a vocabulary for these failures. Each is an operational distinction, not a reason to avoid natural phrasing. A good correction names the discarded source reading and supplies evidence for the accepted one—or leaves the issue genuinely unresolved.

Failure mode 1: Dictionary-first guessing

Observable error: The most familiar sense is selected without context. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Compare usage in the current subject area and write two full readings. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 2: Unapproved author intention

Observable error: An editor explains what the writer probably wanted to say. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Query the actual source owner and distinguish explanation from fact. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 3: Accidental legal commitment

Observable error: May is translated as shall or a confirmed permission. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Compare the actions permitted under each candidate reading. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 4: Overclarifying fiction

Observable error: A deliberately unresolved scene is given a definitive explanation. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Preserve the intended narrative openness and ask about artistic effect. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 5: Vagueness forced into a threshold

Observable error: A general term becomes an exact number without authority. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Request an operative criterion when decisions depend on a boundary. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 6: Underspecification disguised as error

Observable error: An irrelevant missing detail is unnecessarily inserted. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Ask whether the text needs the distinction at all. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 7: Syntactic attachment guess

Observable error: A modifier is linked to an intuitive but unverified noun. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Bracket the competing structures and test them against context. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 8: Wrong quantifier distribution

Observable error: A shared requirement becomes individual or vice versa. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Use a concrete example to test group versus individual allocation. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 9: Negation scope reversal

Observable error: Not obliged is translated as forbidden. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Write a two-state action test and preserve the logical relation. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 10: Ambiguous deadline invented

Observable error: A relative time becomes a specific calendar date. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Identify the reference date before computing or inserting a deadline. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 11: Pronoun arbitrarily resolved

Observable error: A target gender or number identifies an unconfirmed referent. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Find documentary evidence or use a legitimate neutral recast. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 12: Passive agent invented

Observable error: The translator names the actor who performed an unnamed action. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Preserve the unknown actor or query the author. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 13: Collective entity misread

Observable error: A team-level act becomes an act by every member. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Confirm whether participation is collective or distributive. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 14: Question turned into command

Observable error: A polite inquiry becomes an order. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Use discourse context to identify intended speech act. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 15: Implicature presented as literal

Observable error: A suggested action is translated as an explicit decision. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Distinguish what was stated from what was inferred. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 16: Acronym expanded wrongly

Observable error: A context-dependent abbreviation is given one plausible full form. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Check the domain glossary and source definition. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 17: Formal keyword translated

Observable error: A code identifier is treated as everyday vocabulary. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Check the schema and preserve nontranslatable token spelling. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 18: Quote boundary invented

Observable error: Paraphrase is set in quotation marks as verbatim speech. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Verify exact words before presenting a direct quote. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 19: Headline strips qualifier

Observable error: A careful passage becomes an unqualified claim when shortened. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Put the critical distinction near the claim. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 20: Figure cue lost

Observable error: The words this one are detached from their image. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Keep a verified reference or accessible disambiguating description. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 21: Audio stress erased

Observable error: A transcript loses the contrast carried by stress. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Consult authorised audio when it exists. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 22: Source answer not propagated

Observable error: One translation reflects an answer while other locales retain guesses. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Record the decision and update every affected target. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 23: Source version mixed

Observable error: A clarification for a later source revision is applied to an older text. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Bind answers to exact source versions. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 24: Intermediary treated as author

Observable error: A reviewer paraphrase becomes the author’s position. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Keep separate records for source author and translator. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 25: Machine fluency mistaken for evidence

Observable error: One model output is accepted as the only possible reading. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Require context and explicit alternative analyses. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 26: Glossary applied outside scope

Observable error: A preferred term is used for the wrong semantic category. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Check definition and genre before substitution. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 27: Unresolved query hidden

Observable error: The item is marked approved although the author never responded. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Maintain an unresolved state or release hold. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 28: Cultural assumption imported

Observable error: A familiar local convention is applied to another audience. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Ask which underlying social practice the text refers to. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 29: Purposefully open statement narrowed

Observable error: A policy statement is translated as a definite commitment. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Preserve authorised degree of commitment. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Failure mode 30: No independent reader test

Observable error: Reviewers assume the target communicates the intended uncertainty. The important problem is that a possible interpretation is turned into an asserted interpretation without a documented reason. The review should compare what the target allows a reader to believe with what the source can actually support. A sentence can sound idiomatic, precise and confident while still incorrectly assigning responsibility, timing, scope or intended force.

Corrective action: Ask a target-only reader what is permitted and left open. Record the competing reading, the evidence used and the specific target wording that survives the check. If no evidence is decisive, keep the source question open and choose an authorised interim treatment instead of writing a fabricated resolution. Test the same line in a headline or compact interface element, because context lost through compression can reintroduce the ambiguity.

Worked diagnostic cases: what to ask before a translation chooses meaning

The following records are fictional. Their paired readings are English semantic paraphrases to clarify a problem; they are not promises of how any particular target language grammatically expresses the distinction. Each case includes an evidence source that could settle the issue and a consequence of selecting wrongly. The exercise is to learn how to formulate a useful author query and to preserve unresolved meaning rather than replacing it with a guess.

Worked diagnostic 1: Equipment — “Inspect the cover with the camera”

Reading one: Use a camera as the inspection tool. Reading two: Inspect a cover containing a camera. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Approved engineering drawing and inspection protocol. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: An operator could inspect the wrong component. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Drafting strategy: prepare a target version for each supported meaning and, where possible, a third option that retains the source’s indeterminacy without creating a misleading new one. Compare those versions using a target-only reader. Ask the reader what is permitted, required, known and still open. The reader’s answer must match the author-approved interpretation or the deliberately preserved set of readings, not a guess introduced by target grammar.

Quality record: attach the exact source sentence, the two candidate paraphrases, the clarification question, any answer and the revision number to the approved segment. If the source owner changes the original after answering, update other translated versions that depend on the old reading. An answer applied to one locale but omitted in others is a new consistency defect, not proof that the original ambiguity was harmless.

Worked diagnostic 2: Equipment — “Replace the filter after testing the valve”

Reading one: The valve test must precede filter replacement. Reading two: The phrase after testing refers to a different preceding step while replacement time is unresolved. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Numbered maintenance steps and task sequence. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Maintenance work could occur in an unsafe order. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 3: Equipment — “Close the vent and stop the motor if smoke appears”

Reading one: Both actions depend on smoke detection. Reading two: Only stopping the motor depends on smoke detection. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Safety procedure specifying the scope of if. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The source may change when each action is required. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 4: Workplace — “The supervisor told the assistant that they were selected”

Reading one: The assistant was selected. Reading two: The supervisor was selected. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Personnel selection record or author response. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The wrong employee could be told they have a role. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 5: Workplace — “Each group may nominate a reviewer”

Reading one: Every group may nominate its own reviewer. Reading two: Groups may combine to nominate one common reviewer. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Approved nomination policy and eligibility matrix. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Selection rights and numbers can change. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 6: Workplace — “Managers and trainees with certificates can apply”

Reading one: Both managers and trainees must hold certificates. Reading two: Only trainees require certificates. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Eligibility table for each role. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The target could exclude eligible applicants. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 7: Scheduling — “Submit the file by the next review”

Reading one: The file is due before the meeting begins. Reading two: The file is due by the conclusion of the review session. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Published milestone definition and clock time. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A submission could wrongly appear late. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 8: Scheduling — “The team will meet after the client call on Monday”

Reading one: Both events occur Monday with the meeting second. Reading two: The call occurs Monday but the meeting occurs later. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Calendar invitation and timeline. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: An incorrect meeting day could be published. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 9: Scheduling — “Send the summary within two hours”

Reading one: Send within two hours of receiving the source message. Reading two: Send within two hours of completion of a separate task. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Trigger event in process requirements. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The elapsed-time obligation may be misstated. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 10: Customer service — “We can send a replacement next week”

Reading one: The provider has the capability to send next week. Reading two: The provider has authorised a replacement and intends to ship next week. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Stock and decision records. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A possibility can become an unsupported promise. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 11: Customer service — “The adviser saw the case with the red label”

Reading one: The case carried a red status label. Reading two: The adviser used a red-labeled device to view the case. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Interface screenshot or internal glossary. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A support status could be incorrectly assigned. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 12: Customer service — “The form is available after registration”

Reading one: Each user must register before accessing the form. Reading two: The form becomes available after the registration period ends. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Access rules and product specification. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Readers may face a false access requirement. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 13: Research — “The findings improved in two treatment groups”

Reading one: Both treatment groups showed improvement. Reading two: Only two of several groups improved. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Statistical tables and group count. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The scope of an observed effect can be exaggerated. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 14: Research — “The technician stored the sample with the control in the cabinet”

Reading one: Both sample and control were stored together. Reading two: The cabinet contains a control apparatus used in storage. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Lab storage log and item identifiers. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Chain-of-custody relations might be invented. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 15: Research — “The evidence is consistent with a change in exposure”

Reading one: The evidence permits exposure change among several explanations. Reading two: The evidence establishes exposure change as the only explanation. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Methods and authors’ interpretation. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A tentative result could be presented as proof. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 16: Policy — “No employee is required to sign”

Reading one: Signing is optional. Reading two: Signing is prohibited. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Policy owner and full operative wording. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Employees’ rights can be reversed. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 17: Policy — “Access is available if authorisation is granted”

Reading one: Authorisation alone is sufficient. Reading two: Authorisation is necessary but not necessarily sufficient. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Access-control matrix. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The translation could confer a right not actually granted. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 18: Policy — “The revised clause applies to new and existing accounts with premium status”

Reading one: Premium applies to both new and existing accounts. Reading two: Premium qualifies only existing accounts. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Scope table and legal review. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Account coverage could be misrepresented. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 19: Health text — “The clinician spoke to the caregiver after discharge”

Reading one: The discussion took place after discharge. Reading two: The caregiver became involved after discharge, with discussion timing unclear. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Clinical notes and event timestamps. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A medical chronology could be altered. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 20: Health text — “The reader should take the product with food in the evening”

Reading one: The product is taken with an evening meal. Reading two: In the evening modifies a different regimen step while food is a general condition. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Authorised instruction layout and qualified review. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A reader could follow the wrong regimen. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 21: Health text — “The results indicate possible inflammation”

Reading one: A provisional observation is noted. Reading two: A specific inflammatory diagnosis is confirmed. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Responsible clinician’s interpretation. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The target may turn a tentative observation into a diagnosis. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 22: Data — “Show the last three approved versions”

Reading one: Show the most recent three among versions already approved. Reading two: Show the last three versions only if each has approval. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Version history filtering definition. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: An unapproved version could be displayed. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 23: Data — “New records are reviewed weekly”

Reading one: Each week’s newly added records get reviewed. Reading two: A new review process checks all records every week. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Pipeline population and cadence definitions. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Audit coverage may be overstated. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 24: Data — “The rate is five percent higher”

Reading one: The rate increased relatively by five percent. Reading two: The rate increased by five percentage points. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Underlying values and measure definition. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The size of the change can be materially wrong. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 25: Journalism — “Officials said the route could reopen”

Reading one: Reopening is considered possible. Reading two: Reopening is authorised and scheduled. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Original public statement and closure notice. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Readers may infer access that does not exist. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 26: Journalism — “The witness saw the driver with binoculars”

Reading one: The witness used binoculars. Reading two: The driver possessed binoculars. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Original interview or follow-up question. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Evidence about possession could be invented. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 27: Journalism — “The organisation plans to expand services soon”

Reading one: A plan exists but its date is unsettled. Reading two: A firm launch schedule has already been approved. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Published plan and decision status. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: An intended expansion may be reported as available. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 28: Education — “Students discussed the instructor’s answer in pairs”

Reading one: Students formed pairs for discussion. Reading two: The answer itself concerned pairs. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Lesson plan and assignment prompt. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The learning activity can be described incorrectly. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 29: Education — “The rubric values clear and concise writing”

Reading one: Clarity and concision are separate graded dimensions. Reading two: One integrated criterion judges both qualities jointly. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Rubric criteria and weighting. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Students may optimise the wrong scoring element. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 30: Education — “The task is optional after the first session”

Reading one: The task becomes optional only for later sessions. Reading two: The task is optional but only becomes available after the first session. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Syllabus instructions and event sequence. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Learner workload and access may change. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 31: History — “A manuscript places the envoy at the bridge at dawn”

Reading one: The reported arrival happened at dawn. Reading two: The manuscript was written at dawn while event timing is unresolved. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Manuscript dating and historical context. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The translation might invent an event timestamp. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 32: History — “The writer described the captain near the hill”

Reading one: The captain stood near the hill. Reading two: The writer was near the hill while describing the captain. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Wider narrative location and syntax. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A historical location may be attributed to the wrong person. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 33: History — “They never returned to the village”

Reading one: They did not return at any later time. Reading two: They had not returned before the narrator’s present reference point. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Later narrative and tense sequence. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The translator could make a temporary absence permanent. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 34: Transport — “The bus stops by the station with the green roof”

Reading one: The station has a green roof. Reading two: The bus has a green roof. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Route illustration and vehicle data. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A passenger could identify the wrong landmark. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 35: Transport — “The operator will confirm after check-in”

Reading one: A booking confirmation occurs only after check-in is completed. Reading two: An existing booking confirmation is followed by another check-in message. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Booking status flowchart. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Travellers may assume a reservation is already secure. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 36: Transport — “Arrive before the doors open”

Reading one: Arrival is a strict access prerequisite. Reading two: Arrival is advice to avoid queues rather than a required condition. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Venue admission policy. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Users may plan around a false restriction. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 37: Emergency text — “The shelter stays open until the warning ends”

Reading one: Closing occurs when the warning ends. Reading two: Operating hours are guaranteed through that point but closure may be later. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Shelter operations decision. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: People may believe a safe facility closes automatically. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 38: Emergency text — “Call the coordinator using the radio in the vehicle”

Reading one: The radio is kept inside the vehicle. Reading two: The coordinator is located inside the vehicle. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Communications plan and equipment map. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Readers may go to an unsafe or incorrect location. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 39: Emergency text — “Inspect the passage and close the hatch if water is present”

Reading one: Both actions depend on water being present. Reading two: Inspection is unconditional while hatch closure is conditional. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Emergency sequence and qualified procedure owner. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The wrong condition can cause a missed prerequisite. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 40: Creative narrative — “She watched the lights disappear as the vehicle left”

Reading one: The vehicle’s departure caused the visual disappearance. Reading two: The lights went out independently while the vehicle left. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Authorial context and intended suspense. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A narrative ambiguity about cause may be destroyed. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 41: Creative narrative — “He said he would return in the morning”

Reading one: He means the next morning. Reading two: He means some later morning within the story’s open time frame. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Narrative viewpoint and planned scene. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A precise time may be invented. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 42: Creative narrative — “The note said I know what happened”

Reading one: The writer has actual knowledge. Reading two: The writer bluffs to unsettle the addressee. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Subsequent plot and author intent. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Explaining the threat can spoil the story. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 43: Security systems — “Review the logs with the administrator account”

Reading one: Use an administrator account to review logs. Reading two: Review logs that belong to the administrator account. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Audit guide and interface permission model. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Security investigators may check the wrong records. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 44: Security systems — “The credential can be revoked after expiry”

Reading one: Revocation is allowed only after expiry. Reading two: Revocation remains possible after expiry without prohibiting earlier revocation. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Protocol and state machine. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: A false rule can prevent necessary revocation. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 45: Security systems — “Verified devices and accounts with enhanced access may connect”

Reading one: Verification applies to devices only. Reading two: Both devices and high-access accounts need independent verification. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Consequence of choosing incorrectly: An incorrect permission could grant access improperly. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 46: Cultural communication — “Bring a plate to the gathering”

Reading one: Bring crockery for serving food. Reading two: Bring a prepared dish to share. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Event invitation convention and organiser response. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: Attendees may arrive with the wrong contribution. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 47: Cultural communication — “We will consider the proposal carefully”

Reading one: A decision is pending and no acceptance is implied. Reading two: The proposal is being implicitly accepted under a local politeness convention. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Speaker intention and relationship context. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The target may create a commitment that does not exist. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Worked diagnostic 48: Cultural communication — “The invitation is open to the family”

Reading one: All relatives are invited. Reading two: Only members of the immediate household are invited. These are candidate interpretations to assess, not two equally probable facts. The source expression has to be understood in a full document, with genre, speaker and task specified. A translator should write out both propositions and ask which grammatical attachment, contextual referent, professional convention or missing definition would distinguish them. A single attractive translation is not a substitute for that analysis.

Discriminating evidence: Host’s stated invitation scope. This evidence would let a reviewer select the meaning on a traceable basis rather than because one wording feels more natural. If the proposed evidence does not exist, the analyst should ask a neutral question showing the two readings, or follow the project’s approved process for preserving an unresolved point. Avoid a leading query that effectively answers itself.

Consequence of choosing incorrectly: The target can change who may attend. State the consequence in the decision ledger so the source owner understands why the question matters. Different interpretations might change the actor, action, object, deadline, threshold, permission or the audience’s background assumptions. A low-risk artistic double reading may be retained intentionally, whereas a high-risk operational uncertainty can require a publication hold.

Transfer exercises: become reliable when the source has more than one reading

These workshops move the method beyond individual sentences. They are deliberately varied: some require a query, some require preservation, some require a hold and others require a reviewable choice. Their common deliverable is a decision trace. A translation is accepted when the target reader receives the same justified meaning or the same intentional openness that the original author supplied, not merely when it sounds natural.

Transfer workshop 1: Write a neutral author query

Situation: A sentence names two possible approvers with the pronoun they. Task: Draft a query quoting the whole sentence and offering the two possible referents. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: A good question does not imply that the nearest noun must be the answer; it specifies what changes in the approval record if one referent is selected. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 2: Preserve a narrative question

Situation: A fictional note hints that a character knows a secret but never specifies which event. Task: Write an English semantic guide for a target-language passage that leaves the same uncertainty. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: The translation should not reveal a secret or name an action merely because the target grammar prefers explicit subjects. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 3: Distinguish optional from forbidden

Situation: A policy states that employees are not required to attend a meeting. Task: Show two test cases, one where attendance is allowed and one where it is declined. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Both actions may be permissible; translating the source as a prohibition would be a logical error. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 4: Ask for a date anchor

Situation: A notice says the form is due next Friday but gives no publication date. Task: Identify the evidence needed to convert the relative expression to a calendar date. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Do not calculate a deadline from the translator’s clock; use the authoritative source date or retain the relative term with an approved note. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 5: Resolve a measured unit

Situation: A result is described as five percent higher with the baseline omitted. Task: Propose two numerical examples distinguishing percentage points and relative change. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: The analyst must seek underlying rates, not substitute a mathematically convenient interpretation. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 6: Classify uncertainty correctly

Situation: The word adequate is used in a test procedure with no threshold. Task: State whether this is a fuzzy standard or a discrete dictionary ambiguity. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Request acceptance criteria, because a new synonym cannot supply the missing measurement rule. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 7: Identify modification scope

Situation: A numbered list refers to trained staff and contractors with current certificates. Task: Write the two eligible-set interpretations with their own examples. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Do not guess whether trained applies to contractors as well; an eligibility decision must follow the authorised policy. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 8: Separate image dependence

Situation: A screen instruction says click the button beside this icon. Task: Write the accessibility information needed for a screen-reader user. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Identify a stable control name, label and relationship instead of assuming readers share the same visual layout. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 9: Preserve source intentionality

Situation: A promotional sentence uses an ambiguous expression for playful effect. Task: Design a brief that tells the translator which two meanings make the wordplay work. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: A creative adaptation can change surface words while preserving the dual reading and reader effect, subject to the brief. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 10: Test a machine suggestion

Situation: A model outputs one fluent translation of a sentence with two possible actors. Task: List the two source readings and compare each against the model output. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: A model’s single answer is a hypothesis; source evidence and language review decide whether the selected actor is justified. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 11: Review an audio transcript

Situation: A speaker stresses one item in a sentence containing several alternatives. Task: Explain how listening could narrow the reading and how to record that evidence. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Audio-based disambiguation must be supported by real prosody, not by punctuation invented after transcription. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 12: Update source and target

Situation: An author clarifies a modifier and rewrites the source sentence. Task: Specify which translation memory segments, captions and help pages need retesting. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Propagate the clarification to all dependent targets and record the new source version before treating the issue as closed. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 13: Check scope across paragraphs

Situation: A heading says exceptions apply while subsequent bullets span two sections. Task: Use document hierarchy to decide what text the exception governs. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: The translator should not infer section scope solely from visual proximity in a flattened text export. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 14: Rank risk without distorting meaning

Situation: Compare a pun in a novel with a conditional clause in an emergency procedure. Task: Explain why both are ambiguity problems but need different release paths. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: The artistic ambiguity may be preserved; the hazardous instruction may require a qualified reviewer and a publication hold. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 15: Use target-only review

Situation: Give a reviewer a translated notice without its source and ask who can act. Task: Write the reviewer questionnaire about actor, prerequisite, deadline and unresolved facts. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: The answers show what the target actually communicates, which may differ from the translator’s own intended reading. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 16: Handle an unresponsive owner

Situation: A contract clause is consequentially ambiguous and the author cannot be reached. Task: Describe the escalation and status in the release ledger. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Do not fabricate a resolution or announce that a legal meaning is settled; follow the organisation’s qualified-review and hold rules. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 17: Avoid overexplaining a cultural expression

Situation: An invitation contains an expression with two culturally plausible interpretations. Task: Write a query that asks about the expected practical contribution. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Preserve the social intent in a clear target once the organiser confirms what participants should bring or do. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 18: Audit a previous localisation

Situation: An old translation uses an exact noun where the source named no actor. Task: Identify the historical source version, the risk and the minimum necessary correction. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Add a review record for the target, avoid silently rewriting unrelated existing material, and apply the approved meaning only to authorised new versions. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 19: Translate multiple readings deliberately

Situation: A narrator uses a word with two meanings that recur in later paragraphs. Task: Prepare a mapping plan that tracks where each meaning becomes apparent. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: A one-line solution is insufficient when a later reveal depends on keeping both meanings alive in the reader’s mind. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

Transfer workshop 20: Write acceptance tests

Situation: A brief says no sentence may imply authorisation until a separate approval appears. Task: Define positive and negative reader scenarios for the translated interface. Begin by recording the original statement, its plausible interpretations and the evidence available. If the expression is merely under-specified, say which detail was left unexpressed rather than forcing an arbitrary two-way ambiguity. Next produce an idiomatic target or a neutral clarification query that fits the reader’s actual task.

Acceptance criterion: Test both approved and pending states; the target must not turn a preliminary request into an authorised action. A second reviewer should reconstruct what the target says and what it does not say. Put an unresolved item on hold where competing readings change a consequential reader action. Keep intentional ambiguity when it serves the source’s communicative purpose and a natural target counterpart is available. In both cases, retain the reasoning with the source version so a future edit cannot silently undo the decision.

A portable ambiguity register and quality scorecard

An ambiguity register may be a table, database or editorial note, but it needs identifiers and a disciplined state machine. Suggested fields are item ID, source version, exact source string, surrounding context, ambiguity class, alternative interpretations, decision consequence, evidence available, author-query owner, question text, answer text, response time, adopted reading, target strings affected, reviewer, acceptance status and follow-up version. Not all fields are required for every low-stakes phrase, but an auditable record is essential when a wrong reading can change an obligation or a technical action.

Use statuses such as identified, context-resolved, query-sent, author-confirmed, intentionally-preserved, qualified-review-needed, held and released. A status of resolved must have evidence. A status of preserved must document why keeping alternatives is faithful. A held item should state which authority may release it. This avoids the common mistake of treating silence from the source author as agreement with a translator’s first interpretation.

A severity model can distinguish critical, major, moderate and minor errors according to consequence and recoverability. Critical cases affect safety, permissions, rights or significant factual claims. Major cases change scope, participants, quantities or status in important content. Moderate cases change emphasis or user effort without destroying a central decision. Minor cases are local clarity problems with no material semantic impact. The exact definitions belong to project specifications; these are an illustrative training taxonomy, not a universal certified scoring scale.

For a reproducible target-only test, give two independent reviewers the target without the original. Ask each to write the actor, action, scope, condition, date, authority and unresolved aspect. Then compare those entries with the source-author decision. A disagreement should be investigated by locating the specific target wording responsible. Counting that discrepancy is more informative than asking whether the sentence sounds good. If the wording needs contextual media, test it with and without the media under the actual publication conditions.

A four-week learning and implementation plan

Week 1: see alternatives before choosing

Collect thirty short passages from unrelated genres, mark the ambiguous spans and paraphrase each plausible reading. Classify the item as lexical ambiguity, structural ambiguity, vagueness, underspecification or context dependence. Discuss cases where a reader assumes a missing fact but the sentence itself never states it. The exercise should train the habit of producing more than one hypothesis before a preferred translation becomes psychologically fixed.

Deliverable: one documented ambiguity ledger, source-to-target comparison and a short explanation of how the chosen sentence protects the reader’s interpretation. Judge progress by reduction of unsupported meaning choices and the clarity of author queries, not the number of extra words added to a translation.

Week 2: obtain and document clarification

Create an author-query template with exact source quotation, two alternatives, consequence and proposed neutral question. Practise on fictional records, exchange queries with a peer and answer them without seeing the drafter’s preferred translation. Record each response as either a clarified source, a revised source or an unresolved item. Compare how much unnecessary back-and-forth is avoided by a focused question.

Week 3: translate into actual reader contexts

Use qualified reviewers to translate short passages into a chosen target language and adapt them into labels, notifications, instructions, narratives and accessible captions. Compare target-only reader interpretations across formats. Note where target grammar requires choosing a person, number or evidential form not specified in the original, and document the authorised resolution or workaround.

Week 4: maintain the decision through revision

Revise several original sentences, including one that changes the previously accepted reading. Check whether all affected translations and metadata are identified, updated and revalidated. Create release criteria that distinguish intentional openness from dangerous unresolved instructions. Audit an earlier sample after a delay to see whether reviewers still preserve the meaning when they no longer remember the initial context.

Glossary of ambiguity and decision terminology

Lexical ambiguity

A form can legitimately express different word senses. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Structural ambiguity

More than one syntactic analysis may correspond to a sentence. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Attachment ambiguity

A modifier can belong to more than one phrase. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Scope ambiguity

An operator can govern different parts of a proposition. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Referential ambiguity

An expression can identify more than one participant or object. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Anaphora

A later expression refers back to an antecedent. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Coreference

Two expressions designate the same entity. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Vagueness

A category lacks a sharp boundary in relevant cases. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Underspecification

The source leaves a feature unset without necessarily offering discrete senses. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Context sensitivity

An expression’s interpretation depends on context of use. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Polysemy

Related meanings may be associated with one form. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Homonymy

Distinct meanings share a pronunciation or spelling. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Pragmatic inference

A reader derives meaning from discourse and communicative context. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Implicature

An inferred meaning not identical to the explicitly asserted proposition. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Presupposition

Information taken as background by a construction. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Speech act

The action accomplished through an utterance, such as requesting or warning. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Ellipsis

Omission of material understood from linguistic or contextual context. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Contrastive stress

Prosodic emphasis that distinguishes alternatives. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Neutral author query

A clarification request presenting possibilities without steering the answer. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Source-owner response

An answer from someone authorised to clarify the content. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Candidate reading

A plausible interpretation under evaluation, not an established intention. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Decision trace

A recoverable record of the interpretation, evidence and approval. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Source version

A specific revision of the original text and context. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Target version

A specific authorised revision of translated text. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Back-reading test

An independent reader reconstructs meaning from target wording alone. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Target-only review

Review performed without showing the source to measure reader inference. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Translation specification

A project-level description of requirements, readers, risk and acceptance. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Semantic invariant

A meaning relation that must survive the language change. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Disambiguation

Selection or identification of a particular reading. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Ambiguity preservation

Retaining more than one intended interpretation in the target. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Intentional indeterminacy

An author deliberately refrains from fixing one reading. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Release hold

A state preventing publication until a required clarification or approval. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Quality severity

The consequence level of a translation discrepancy. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Reviewer authority

The scope of decisions a reviewer is qualified and permitted to make. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Evidence sufficiency

Whether available material justifies selecting one interpretation. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Quotation boundary

The distinction between source words and a reported paraphrase. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Language-specific constraint

A grammatical requirement that may force distinctions not in the source. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Terminology entry

A controlled relation between a concept, designation and context. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Revision propagation

Updating dependent translations after an approved source change. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Reader consequence

What a reader may incorrectly believe or do because of the target. In a real project, record how this concept appears in the precise source sentence and which source evidence justifies its handling. This term is an analytic aid, not an automatic one-word translation. A reviewer should test whether a target expression creates an extra interpretation or destroys a legitimate source reading.

Frequently asked questions

1. What is ambiguity in translation?

It occurs when a source expression legitimately supports multiple interpretations and translation requires deciding whether to preserve, clarify or select among them.

For practice, take a contextual example and document the two candidate interpretations or the specific missing detail. Identify what evidence would answer the question. A correct decision is one the reader can reconstruct from target text and the reviewer can justify from the authorised source, not merely a fluent phrase.

2. Should a translator always resolve ambiguity?

No. Accidental ambiguity in consequential instructions may require clarification, but deliberate literary or rhetorical ambiguity may be part of the meaning to preserve.

3. How do you translate an ambiguous sentence?

List the readings, consult context, separate evidence from guesses, assess consequences, seek clarification when appropriate and back-test the target.

4. How is vagueness different from ambiguity?

Vagueness concerns imprecise boundaries, while ambiguity often involves distinguishable alternative meanings; some expressions have both properties.

5. What is underspecification?

A dimension such as size, identity or precise time can be left unstated without creating two separate word meanings.

6. When should I ask the author?

When plausible readings change material content and context cannot establish the intended one, especially for permissions, instructions, amounts, identities or dates.

7. What should an author query say?

Quote the exact span, present neutral alternative readings and explain the specific consequence without suggesting a preferred answer.

8. What if the author does not respond?

Follow project risk rules. Preserve legitimate uncertainty where suitable, escalate unresolved consequential content and never present a guess as confirmed author intention.

9. How can ambiguity be preserved across languages?

Use natural target constructions that allow the same relevant readings, or an authorised editorial solution when target grammar forces a distinction.

10. Can a dictionary decide the correct meaning?

A dictionary lists possible senses; it does not determine which one is intended in a specific passage.

11. Do machine translation systems detect ambiguity?

They may propose readings, but a single fluent output does not establish that all alternatives were considered or excluded.

12. Can source context settle ambiguity?

Often, yes. Neighbouring paragraphs, entity references, diagrams, timing and genre conventions can eliminate unsupported candidates.

13. Is an ambiguous word the same as an unclear sentence?

Not necessarily. A sentence may be unclear because it lacks needed detail rather than because a word has multiple senses.

14. What is modifier attachment ambiguity?

It arises when a phrase such as with the camera could describe the tool used or the object being described.

15. Why is negation scope important?

Because not required and required not to describe different permissions; confusing them can reverse the source’s meaning.

16. How should numbers be handled?

Preserve approximate versus exact status, reference group, units and the difference between percentage points and relative percentages.

17. How should pronouns be handled?

Track candidate referents and use contextual evidence; do not invent a gender, actor or entity when the original leaves it unresolved.

18. What about intentionally ambiguous fiction?

Identify the effect, later reveals and available target-language devices. Avoid explanatory wording that removes intended uncertainty.

19. What are examples of high-risk ambiguity?

Conditions for safety actions, access permissions, financial amounts, formal approvals, legal rights and clinical instructions.

20. How do I test a translated sentence?

Give the target alone to a competent reader and ask who acts, under what conditions, when, and what remains unresolved.

21. What is an ambiguity register?

A record of source spans, readings, evidence, author queries, decisions, versions and review status for traceable management.

22. Can translators invent missing dates?

No. A relative date requires an authoritative reference point; otherwise its unresolved status must be managed.

23. What does author-approved mean?

The responsible source owner has confirmed the intended reading or changed the original to express it clearly.

24. Does a native speaker automatically resolve every ambiguity?

No. Native competence supports language analysis but does not supply facts, author intention or missing technical specifications.

25. How do controlled terms relate to ambiguity?

A terminology record can fix domain meanings, but the term must be used in the relevant concept and context.

26. How does this chapter relate to Vocabulary?

Vocabulary resources teach word senses and usage. This chapter addresses the decision process when a translation must handle competing source interpretations.

27. How does it relate to How English Works?

English grammar and discourse resources explain structures such as reference and scope; the translation method manages interpretation choices across languages.

28. What is the final release rule?

The target must communicate only interpretations that are supported, intentionally preserved or authorised, with high-risk unresolved choices held for review.

Research and standards: where the terminology comes from

Stanford Encyclopedia of Philosophy — Ambiguity offers a rigorous discussion of lexical and structural ambiguity, context sensitivity and the distinction between multiple readings, generality, underspecification and vagueness. The taxonomy is used here to avoid treating every unclear phrase as the same problem.

Stanford Encyclopedia of Philosophy — Vagueness examines borderline cases and their relation to ambiguity. An uncertain threshold can demand a decision rule rather than a choice between two dictionary entries.

ISO 11669:2024 — Translation Projects: General Guidance provides guidance for planning translation projects, specifications, communication and risk assessment. It does not itself provide an output-quality scoring procedure or guarantee a particular reading.

ISO 5060:2024 — Translation Services: Evaluation of Translation Output describes guidance for human analytic evaluation of human, post-edited and unedited machine translation output, including error types, evaluator competence and sampling. It is not a substitute for an authorised source clarification.

W3C Style Guide — Write for a Global Audience gives useful internationalisation writing principles, including terminology choices and internationally comprehensible formats. These support clear source writing; they do not imply that all creative ambiguity should be eliminated.

Structural-Ambiguity-Aware Translation from Natural Language to Signal Temporal Logic (2026) is a narrow formal-language research example of retaining multiple plausible structural analyses rather than always selecting one early. Its application is not evidence that the same algorithm is established for ordinary human translation.

Next routes: return to translation foundations without taking over protected content

Return to the parent architecture: Master Art of Translation | The Complete System for Moving Meaning Between Languages. The preceding chapter on topic, focus, contrast and given–new information is Master Art of Translation | The Information Structure System. These chapters answer neighbouring questions, not the same editorial decision problem.

For word-sense explanations and learning strategies, use the protected Vocabulary Learning Hub. For English-specific grammar, reference, sentence structure and discourse, use the protected How English Works | Translation and Meaning Across Languages. Both remain independent subject owners, linked only where readers need their foundations.

The end-to-end standard is not to make every sentence unambiguous. It is to be exact about the uncertainty that exists, the evidence that can settle it, and the authority under which a translator may choose. Preserving meaning includes preserving an author’s unanswered questions where those questions are real. A successful target reader can tell what is known, what is deliberately open, and what should not be acted on until it is clarified.

Architecture reference: EDKSG-TRANS-MASTER-WORLD-1310. This chapter is a standalone method for ambiguity decisions and does not claim ownership of existing vocabulary, syntax, discourse or English-learning treatments.

Master Art of Translation | The Ambiguity Decision System — How Multiple Readings, Author Queries and Unresolved Meaning Stay Honest