How do you translate “some” without silently changing it to “not all”? Why can “or” imply a choice without logically forbidding both possibilities? What happens to the meaning of “even”, “only”, “may” and “few” when a sentence crosses languages? These are practical questions about scalar implicature in translation, pragmatic inference, quantifier translation, focus-sensitive particles and the difference between literal meaning and what a reader is likely to infer. Accurate human, machine and AI translation must preserve the source speaker’s commitment while leaving space for context-sensitive interpretations that the source itself leaves open.
The hidden difficulty is that a translation can preserve every visible noun and verb yet alter an invisible logical boundary. “Some participants agreed” can be true if all agreed, although ordinary conversation often suggests a smaller group. “You may submit a revision” normally expresses permission, not a duty. “Only the updated record was approved” excludes other relevant records; “even the updated record was approved” flags it as a surprising inclusion. A bilingual dictionary may offer plausible equivalents while a machine translation system or a human editor strengthens, weakens or relocates an implication. The remedy is a repeatable meaning-preservation workflow: separate entailment from conversational inference, identify the relevant alternatives, translate the assertion and its scope, then test how a target reader will interpret it.
Quick answer: preserve the asserted boundary, then audit the inference
- Identify exactly what the original sentence asserts, and do not promote a defeasible implication into an unconditional claim.
- Ask which stronger or weaker alternatives the reader would naturally compare: some/all, or/and, sometimes/always, possible/certain, a few/few.
- Treat only, even, at least, exactly and negation according to their actual scope and meaning; do not call every difference an implicature.
- Preserve uncertainty, permission, quantity, exclusion, surprise and contrast separately rather than collapsing them into one vague sense of emphasis.
- Use source-to-target and target-only checks, worked counterexamples, back-translation of meaning and an explicit risk register for high-stakes wording.
- Return to the Master Art of Translation root for the whole translation workflow, to the Vocabulary Learning Hub for acquiring the underlying words, and to How English Works for English-only structure and meaning.
This chapter is EDKSG-TRANS-MASTER-WORLD-1620. Its single ownership job is cross-language transfer of scalar alternatives and their inferences. The preceding Negative-Question Answer Fidelity chapter handles polarity in replies and corrections; this chapter handles what stronger alternatives a speaker did not choose. The two problems meet when a translated answer makes “not all” sound like “none”, but they are not the same mechanism. The present guide is world-facing: its methods apply wherever two language systems must carry the same intended boundaries, independently of place, language pair, exam board or industry.
Part I — The hidden translation problem: what a sentence says and what it invites
1. An accurate word can carry an inaccurate inference
A translator may reproduce the dictionary meaning of a quantifier and still mislead the reader. In a progress report, the phrase ‘some items were checked’ could describe a verified subset without saying anything final about the rest. A target version meaning ‘only some items were checked’ introduces an exclusion that the source may not contain. Conversely, translating a deliberately cautious ‘some’ with an unqualified ‘all’ invents completion. The error is not fluency. It is an unjustified change to the set of worlds in which the sentence is true.
The first diagnostic separates three questions. What must be true for the source statement to be true? What might a normal reader additionally infer in this situation? What did the speaker actively exclude or leave unsettled? A strong translation tries to preserve all three relationships without treating them as interchangeable. A reader may infer ‘not all’ from ‘some’ under ordinary expectations of informativeness, but that inference is defeasible and may disappear in a special context.
2. Scalar inference is not a single universal formula
In pragmatics, an implicature is broadly an implication conveyed beyond the sentence’s conventional assertion. Scalar cases often arise when a speaker chooses a weaker item from a salient group of alternatives. Yet neither speaker intention nor listener inference can be reduced to ‘the weaker word always excludes the stronger word’. Alternative sets differ by discourse, prosody, speaker knowledge, goal and language-specific conventions. A translator must therefore test plausible interpretations rather than convert a useful heuristic into a rigid decoding law.
The foundational discussion of scalar and other conversational implicatures in the <a href=”https://plato.stanford.edu/entries/implicature/” rel=”noopener”>Stanford Encyclopedia of Philosophy</a> (substantively revised 10 January 2024) is careful about this point. It discusses both the familiar some/all analysis and the theoretical difficulties of predicting which alternatives count. The <a href=”https://doi.org/10.3389/fcomm.2021.682009″ rel=”noopener”>Frontiers review of alternatives in language</a> similarly explains why alternatives, focus and context matter and why pragmatic enrichment is not obligatory in every setting. These are research anchors, not automatic rules for individual language pairs.
3. Assertion, entailment, inference, presupposition and conventional contribution
An assertion is the content the speaker puts forward. An entailment is something that must follow if the assertion is true, given the relevant interpretation. A conversational implicature is an additional contextual inference that can often be cancelled without direct contradiction. A presupposition is background content presented as already in play, although the precise analysis varies with construction and theory. Focus-sensitive expressions such as ‘only’ and ‘even’ introduce their own conventional structure and alternatives. The translator’s task is to label the right kind of contribution before choosing a target expression.
Consider ‘Some files were found—and in fact every file was found.’ This is pragmatically unexpected in many conversations, but not a simple contradiction. ‘Only one file was found—and a second file was found’ is different if the same domain and time are intended: the exclusion is being contradicted. ‘Even the missing file was found’ depends on an unexpected member of a contextual set; translating it as ‘only the missing file’ reverses its direction. A single list of synonyms cannot substitute for these distinctions.
4. Why the target reader’s expectations matter
Languages differ not only in how they express quantities but also in where they mark topic, focus, evidential source, habitual aspect and social commitments. A target-language equivalent of ‘some’ might carry a strongly restricted everyday reading in one construction, while another form is more neutral. Permission may be phrased by a modal, an impersonal construction or an explicit right. ‘Even’ may be marked by a particle, by word order, by intonation or by a paraphrase. It is unsafe to assume that similar-looking words carry identical contrast sets.
The solution is not to add more explanatory words automatically. Explanation can be essential when the brief demands explicitness, but it can also turn a cautiously open sentence into an overcommitted one. First model the source. Then compare natural target-language options with competent speakers and authentic contextual examples. Only after that should the translator decide how much additional wording the target requires.
5. A small set of logical tests saves large revisions
Write down a candidate source statement and a counterexample. Could the source sentence still be true if every member of the group satisfies the predicate? If so, a translation that states ‘not every’ may have overinterpreted ‘some’. Could both alternatives in an ‘or’ statement happen? Often the inclusive truth condition permits this, but contracts, menus and questions may impose an exclusive reading. Could an action fail to occur even though someone ‘may’ perform it? Yes: permission does not entail performance. These tests are compact, repeatable and unusually powerful.
The tests are not a substitute for expert judgment. Formal truth conditions do not settle every pragmatic expectation. Their role is to expose a sharp kind of error: an unintended logical commitment. Once the team detects that a target is stronger than the source, it can investigate whether the stronger commitment is really supported by the brief, the surrounding discourse or an authoritative definition.
Part II — The mechanism map: twelve distinctions the translator must protect
1. A. Some versus all
A subset expression asserts a positive lower bound, while an all-expression asserts universal inclusion within the relevant domain. The common ‘not all’ inference is defeasible; in statistics or policy it can be safer to retain an open subset expression. The diagnostic is deliberately concrete: Check whether an all-true scenario would make the proposed target contradictory. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
2. B. Or versus and
Disjunction presents alternatives; conjunction requires both, although the exact relationship depends on syntax and scope. A conversational choice may invite an exclusive reading, but the wording does not necessarily assert exclusivity. The diagnostic is deliberately concrete: Test a scenario where both options are available; record whether the source accepts it. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
3. C. Sometimes versus always
Habitual frequency expressions locate events on different points of a scale. Often and frequently are not stable percentages, and sometimes should not be rewritten as rarely without evidence. The diagnostic is deliberately concrete: Ask what frequency the statement rules out rather than guessing a numerical rate. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
4. D. May versus must
Permission or possibility is weaker than necessity or obligation, but modal meaning depends on context and syntax. One modal can express epistemic possibility, another deontic permission; false equivalence changes the reader’s duties. The diagnostic is deliberately concrete: Write an allowed-but-unperformed counterexample to distinguish permission from requirement. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
5. E. Possible versus certain
An epistemic qualifier marks limits of speaker confidence or available evidence. A target that reads as ‘probably’ may be stronger than ‘possibly’; a target that reads as ‘certainly’ changes the claim radically. The diagnostic is deliberately concrete: Check whether the evidence described by the speaker justifies the target’s confidence. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
6. F. Not all versus none
Negated universal and universal negation are different logical statements. A translated ‘not all’ can be true when most members satisfy a property; ‘none’ denies every instance. The diagnostic is deliberately concrete: Construct an example with nine successes and one failure; test both readings. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
7. G. Few versus a few
Both refer to relatively small quantities, but English ‘few’ often frames the amount as disappointingly small while ‘a few’ can frame it as a meaningful positive set. Other languages may express the distinction through articles, particles or context rather than a direct lexical pair. The diagnostic is deliberately concrete: Ask whether the sentence is reporting scarcity or merely a small positive number. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
8. H. Only and exhaustivity
Only typically excludes relevant alternatives outside the focused expression. Changing the position of the focus can transform who is excluded and why. The diagnostic is deliberately concrete: Identify the focused constituent with brackets before translating. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
9. I. Even and unexpected inclusion
Even conventionally invites an ordering of alternatives in which the focused case is notable or unexpected. The surprise may concern a person, time, action or degree; ‘even’ is not a general synonym for ‘only’. The diagnostic is deliberately concrete: Ask which member of the alternative set was considered least likely. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
10. J. At least versus exactly
At least asserts a lower bound and leaves room for a higher count; exactly fixes the count under the relevant measurement. Replacing one with the other can cause compliance and reporting errors. The diagnostic is deliberately concrete: Try a value above the stated number and see whether the sentence remains true. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
11. K. Most versus all
Most typically expresses a large majority without universal quantification; its precise threshold depends on language and context. Translating most as all falsely certifies completion; as many can understate the source. The diagnostic is deliberately concrete: Ask whether one exception remains compatible with the source. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
12. L. Negation scope and alternatives
Not sometimes, sometimes not, not necessarily and necessarily not create different patterns of possibility. Particles that move across a verb or phrase can quietly reverse what remains possible. The diagnostic is deliberately concrete: Paraphrase the positive cases that remain permitted before selecting a target form. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
13. M. Focus accent
Stress may change which alternative a listener considers: the new report, the report, or the person submitting it. Written translations may require punctuation, particle choice or context because source audio carries the contrast. The diagnostic is deliberately concrete: Move the accent conceptually and inspect whether the explanation changes. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
14. N. Conditional antecedents
Under a conditional, a scale expression can behave differently from the same phrase asserted alone. Not every occurrence of ‘some’ under ‘if’ warrants ‘some but not all’. The diagnostic is deliberately concrete: Check whether the strengthened reading is licensed in the local clause, not merely at sentence level. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
15. O. Questions and alternatives
Questions can present possible answers without asserting either answer as true. A translation that changes ‘some?’ to ‘all?’ may alter the range of acceptable replies. The diagnostic is deliberately concrete: Generate two truthful responses to the original and test the target against both. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
16. P. Embedded speech
A reporter may attribute a cautious phrase to another speaker, preserving the speaker’s commitment rather than editorialising. The surrounding narrator’s beliefs do not authorise changing quoted possibilities into facts. The diagnostic is deliberately concrete: Keep speaker layers distinct in a commitment ledger. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
17. Q. Known versus unknown upper bounds
A writer can report a verified minimum without knowing the eventual maximum. Adding ‘only’ because a draft looks incomplete mistakes a measurement status for a fixed result. The diagnostic is deliberately concrete: Compare the counted observations to the universe of possible observations. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
18. R. Domain restriction
All, some, few and none quantify over a contextually relevant set. A target that expands ‘all inspected units’ to ‘all units’ changes the domain, even if it preserves ‘all’. The diagnostic is deliberately concrete: Underline the noun phrase and any restricting modifier before translating the quantifier. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
19. S. Alternative availability
A scalar inference relies on the idea that a stronger expression was relevant and available to the speaker. If the speaker did not know the stronger claim, silence may signal ignorance rather than negation. The diagnostic is deliberately concrete: Ask which facts the speaker possessed when choosing the weaker form. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
20. T. Literal contradiction versus cancelled inference
An implication is not an entailment merely because it is commonly heard. The naturalness of ‘some—indeed all’ helps expose a cancellable inference, although context still matters. The diagnostic is deliberately concrete: Test whether the continuation is contradictory or simply unexpectedly corrective. When the target fails the diagnostic, record the kind of drift, identify the source words that license or prohibit the stronger reading, and revise for commitment rather than for elegance. The principle extends beyond English: a translator may need a different target construction, but the underlying speaker commitment should remain recognisable.
Part III — A repeatable twenty-four-gate translation method
Gate 1 — Recover the whole utterance
Read the sentence with the preceding and following turns; identify whether the writer is correcting, hedging, granting permission or asking a question. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 2 — Name the decision the reader must make
An inventory note, licence, interface label and research abstract can carry the same quantifier but lead to different actions. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 3 — Bracket the quantified domain
Mark the full noun phrase, its modifiers and any explicit denominator. Do not silently broaden a local group into the entire population. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 4 — Separate assertion and inference
Write a literal paraphrase without ‘only’ or ‘not all’ unless those words are asserted; record likely pragmatic effects separately. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 5 — List candidate alternatives
Identify stronger relevant alternatives available to the speaker: all for some, and for or, must for may when the modal reading is comparable. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 6 — Check relevance of each alternative
Not every more precise sentence is a socially relevant alternative. Context determines which comparisons a listener is expected to make. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 7 — Check the speaker’s knowledge
A weaker statement can reflect uncertainty rather than a belief that the stronger claim is false. Keep this distinction explicit. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 8 — Identify polarity
Mark which constituent negation takes scope over and what positive situations the sentence still permits. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 9 — Identify modality
Classify whether may marks permission, possibility or a special institutional convention before choosing the target. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 10 — Locate focus
Underline the expression associated with only or even; ensure the exclusion or surprise concerns that expression, not a neighbouring one. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 11 — Map cardinal bounds
For at least, at most, exactly, more than and fewer than, write the accepted numerical set when useful. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 12 — Protect speaker attribution
If one participant says ‘perhaps’ and another says ‘definitely’, do not merge the commitments into an omniscient narration. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 13 — Draft two target candidates
Create the natural, low-commitment version and the more explicit version; ask which reproduces the licensed source meaning. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 14 — Run the all-true test
For ‘some’, imagine every relevant member satisfies the statement. If the target forbids this, ask whether the context genuinely does. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 15 — Run the both-true test
For ‘or’, imagine both alternatives are available. Determine whether the source prohibits, invites or ignores that combination. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 16 — Run the possible-not-actual test
For may and possible, imagine the event never occurs. Permission and possibility can remain true without actual performance. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 17 — Run the focal-substitution test
For only and even, substitute a different person or object into the focused slot and inspect what changes. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 18 — Run the exception test
For all and most, introduce one verified exception and see whether the source and target remain aligned. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 19 — Check natural target conventions
Consult parallel materials or competent target readers; avoid mechanically copying source particles when the target language packages meaning differently. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 20 — Check automatic translation output independently
Treat machine and AI drafts as hypotheses, not as proof that the stronger reading is valid. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 21 — Back-paraphrase commitments
Have a reviewer state in plain language what must be true, what is merely likely and what the target reader is invited to infer. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 22 — Classify errors by consequence
Distinguish a light stylistic effect from a change to permission, eligibility, total coverage, exclusions or evidence certainty. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 23 — Resolve with the brief
If the intended meaning is genuinely ambiguous in consequential content, query the owner rather than creating a confident but unsupported resolution. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Gate 24 — Re-test after revision
Read the final target alone, then compare with the source; sign off only when the relevant possible worlds, commitments and reader actions match. In practice, keep a one-line audit record containing the disputed source phrase, the intended alternative set, the target wording, and one concrete scenario that would reveal a mismatch. This makes the decision reusable in a translation memory or style guide. It also prevents a reviewer from ‘correcting’ a cautious translation into a confident error merely because the stronger version sounds more fluent. Never claim that a source unambiguously conveys a pragmatic inference when the surrounding discourse does not support that level of certainty.
Part IV — Failure modes: how apparently fluent translation changes commitment
Failure 1: The automatic ‘not all’ insertion
Failure mechanism: Treating every some as an asserted exclusion. Repair: Leave the upper bound open unless the text explicitly excludes all. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 2: The universal completion upgrade
Failure mechanism: Turning most or some into all to make a summary decisive. Repair: Retain the source’s coverage level and any stated denominator. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 3: The exclusion-to-surprise swap
Failure mechanism: Treating only and even as interchangeable emphasis particles. Repair: Identify which alternatives are excluded versus unexpectedly included. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 4: The certainty inflation
Failure mechanism: Replacing might or possible with will. Repair: Restore the epistemic strength the speaker actually used. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 5: The obligation inflation
Failure mechanism: Rendering permission as duty. Repair: Ask whether non-performance would violate the source statement. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 6: The obligation dilution
Failure mechanism: Rendering must as may. Repair: Verify whether a non-performed action is allowed or prohibited. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 7: The exclusive-or assumption
Failure mechanism: Imposing ‘either but not both’ on every disjunction. Repair: Determine exclusivity from context, legal definitions or explicit wording. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 8: The and-to-or weakening
Failure mechanism: Replacing a combined requirement with alternatives. Repair: Test whether satisfying one alone meets the source condition. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 9: The none/not-all collapse
Failure mechanism: Reading partial failure as complete failure. Repair: Use one-success and one-failure counterexamples. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 10: The few/a-few inversion
Failure mechanism: Losing the evaluative framing of a small positive set. Repair: Preserve scarcity versus encouraging availability where relevant. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 11: The denominator disappearance
Failure mechanism: Dropping ‘of the inspected sample’. Repair: Protect the set over which the quantity is measured. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 12: The exact-count hallucination
Failure mechanism: Turning ‘at least ten’ into ‘ten’. Repair: Maintain the open upper range. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 13: The numerical rounding leak
Failure mechanism: Rewriting ‘more than 60%’ as ‘60%’. Repair: Keep inequality signs and strictness distinctions. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 14: The speaker-knowledge invention
Failure mechanism: Inferring negative evidence from a cautious statement. Repair: Record whether the speaker knew the stronger fact. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 15: The misplaced only
Failure mechanism: Excluding the wrong noun phrase. Repair: Bracket the focus and test alternative placements. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 16: The misplaced even
Failure mechanism: Attaching surprise to the action instead of the participant. Repair: Ask which element was least expected. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 17: The lost conditional
Failure mechanism: Strengthening a possibility inside an if-clause. Repair: Inspect the whole conditional rather than its embedded token alone. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 18: The quotation override
Failure mechanism: Applying an editor’s certainty to a quoted speaker. Repair: Keep each agent’s commitment separate. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 19: The scripted answer flip
Failure mechanism: Translating a yes/no reply without its question. Repair: Rebuild the dialogue polarity before transfer. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 20: The summary compression error
Failure mechanism: Removing ‘some’, ‘often’, or ‘usually’ from a short summary. Repair: Re-read summary claims against the fuller source. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 21: The interface-choice misfire
Failure mechanism: Showing one option in UI copy where two remain available. Repair: Test real user paths for both and neither where appropriate. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 22: The plural-rule shortcut
Failure mechanism: Assuming grammatical plural categories are semantic thresholds. Repair: Apply the locale’s grammar rules without inventing a quantity claim. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 23: The automated post-edit overcorrection
Failure mechanism: Preferring a smooth target over a cautious source. Repair: Require reviewers to justify every strengthened commitment. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 24: The genre mismatch
Failure mechanism: Translating informal hedges as scientific probability statements. Repair: Match genre and evidence standards, not merely dictionary frequency. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Failure 25: The false back-translation reassurance
Failure mechanism: Assuming reverse translation proves equivalent pragmatics. Repair: Use counterfactual truth tests and target-reader judgment as well. This problem becomes visible when the evaluator stops checking only lexical similarity and asks which events the target allows, forbids or implies. A useful audit note should state the disputed alternative and the consequence of a misreading. If the distinction bears on compliance, patient-facing communication, access, compensation or safety, escalate the ambiguity to a qualified human reviewer instead of resolving it by stylistic preference.
Part V — Worked diagnostic atlas: 160 contextual tests
The following mini-laboratories are not translations into any one target language. They are original source utterances and language-neutral target-meaning specifications. This is deliberate: a genuine bilingual solution must be checked against the relevant target language, grammar, audience and brief. Each laboratory separates the semantic commitment from a tempting but unsupported stronger reading, then sets a domain-specific test. A proposed target-language sentence succeeds only when it passes those tests.
Laboratory 1 — Archive accession and provenance
Scene: The archivist must release the digital catalogue. The relevant material is documented through signed comparison sheets and a dated accession log. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a late-discovered handwritten correction. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 1: Archive accession and provenance — Some, all, and an open upper bound
Some accession records were reconciled with the original register.
Tempting but unsafe target meaning: Only some accession records were reconciled with the original register; others definitely failed.
Commitment-preserving target specification: A verified positive subset of accession records were reconciled with the original register; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in signed comparison sheets and a dated accession log may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the archivist intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a late-discovered handwritten correction. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to release the digital catalogue and create mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 2: Archive accession and provenance — Or, and, and an unwarranted exclusive choice
The archivist may authorize a page-by-page audit or a selected-page audit.
Tempting but unsafe target meaning: The archivist must select exactly one of a page-by-page audit and a selected-page audit, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a page-by-page audit might precede a selected-page audit, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a page-by-page audit followed by a selected-page audit. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to release the digital catalogue, with the risk of mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 3: Archive accession and provenance — May, permission, and unasserted performance
The archivist may request another provenance check.
Tempting but unsafe target meaning: The archivist is required to request another provenance check now.
Commitment-preserving target specification: The archivist has permission to request another provenance check; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the archivist from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the archivist decides not to request another provenance check because the evidence in signed comparison sheets and a dated accession log is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for release the digital catalogue; a false permission can equally compromise control and cause mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 4: Archive accession and provenance — Sometimes, often, and frequency without an invented rate
The archivist sometimes compares accession numbers against the physical labels.
Tempting but unsafe target meaning: The archivist always compares accession numbers against the physical labels.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the archivist compares accession numbers against the physical labels and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind release the digital catalogue and overlook mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 5: Archive accession and provenance — Not all, none, and scope of negation
Not all accession records were reconciled with the original register.
Tempting but unsafe target meaning: No accession records were reconciled with the original register.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 14 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to release the digital catalogue and either conceal useful progress or cause mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 6: Archive accession and provenance — Only, focus, and exclusion of alternatives
Only the handwritten entry was reconciled with the original register.
Tempting but unsafe target meaning: Even the handwritten entry was reconciled with the original register; perhaps many others did too.
Commitment-preserving target specification: The handwritten entry satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the handwritten entry, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the archivist believe other items are eligible when deciding to release the digital catalogue, creating mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 7: Archive accession and provenance — Even, expectation, and surprising inclusion
Even the damaged ledger was reconciled with the original register.
Tempting but unsafe target meaning: Only the damaged ledger was reconciled with the original register; every alternative was rejected.
Commitment-preserving target specification: The damaged ledger satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the handwritten entry also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to release the digital catalogue and may cause mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 8: Archive accession and provenance — Few, a few, and evaluative scarcity
Few accession records were reconciled with the original register.
Tempting but unsafe target meaning: A comfortably sufficient number of accession records were reconciled with the original register.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 14 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the archivist expected before release the digital catalogue. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 9: Archive accession and provenance — At least, exactly, and an open numerical ceiling
At least 14 accession records were reconciled with the original register.
Tempting but unsafe target meaning: Exactly 14 accession records were reconciled with the original register, no more.
Commitment-preserving target specification: The verified count is 14 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a late-discovered handwritten correction. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in signed comparison sheets and a dated accession log increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of release the digital catalogue and produce mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 10: Archive accession and provenance — Most, all, and majority without universal completion
Most accession records were reconciled with the original register.
Tempting but unsafe target meaning: All accession records were reconciled with the original register, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a late-discovered handwritten correction. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger release the digital catalogue and cause mistakenly treating unverified objects as cleared for exhibition. Compare the target against signed comparison sheets and a dated accession log. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 2 — Software security release
Scene: The release manager must authorize the production rollout. The relevant material is documented through build logs, test hashes and release notes. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a dependency updated after the test window. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 11: Software security release — Some, all, and an open upper bound
Some security patches passed the regression suite.
Tempting but unsafe target meaning: Only some security patches passed the regression suite; others definitely failed.
Commitment-preserving target specification: A verified positive subset of security patches passed the regression suite; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in build logs, test hashes and release notes may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the release manager intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a dependency updated after the test window. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to authorize the production rollout and create shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 12: Software security release — Or, and, and an unwarranted exclusive choice
The release manager may authorize a complete retest or a targeted retest.
Tempting but unsafe target meaning: The release manager must select exactly one of a complete retest and a targeted retest, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a complete retest might precede a targeted retest, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a complete retest followed by a targeted retest. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to authorize the production rollout, with the risk of shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 13: Software security release — May, permission, and unasserted performance
The release manager may postpone the deployment.
Tempting but unsafe target meaning: The release manager is required to postpone the deployment now.
Commitment-preserving target specification: The release manager has permission to postpone the deployment; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the release manager from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the release manager decides not to postpone the deployment because the evidence in build logs, test hashes and release notes is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for authorize the production rollout; a false permission can equally compromise control and cause shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 14: Software security release — Sometimes, often, and frequency without an invented rate
The release manager sometimes reviews the dependency report before deployment.
Tempting but unsafe target meaning: The release manager always reviews the dependency report before deployment.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the release manager reviews the dependency report before deployment and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind authorize the production rollout and overlook shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 15: Software security release — Not all, none, and scope of negation
Not all security patches passed the regression suite.
Tempting but unsafe target meaning: No security patches passed the regression suite.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 7 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to authorize the production rollout and either conceal useful progress or cause shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 16: Software security release — Only, focus, and exclusion of alternatives
Only the authentication patch passed the regression suite.
Tempting but unsafe target meaning: Even the authentication patch passed the regression suite; perhaps many others did too.
Commitment-preserving target specification: The authentication patch satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the authentication patch, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the release manager believe other items are eligible when deciding to authorize the production rollout, creating shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 17: Software security release — Even, expectation, and surprising inclusion
Even the legacy integration patch passed the regression suite.
Tempting but unsafe target meaning: Only the legacy integration patch passed the regression suite; every alternative was rejected.
Commitment-preserving target specification: The legacy integration patch satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the authentication patch also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to authorize the production rollout and may cause shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 18: Software security release — Few, a few, and evaluative scarcity
Few security patches passed the regression suite.
Tempting but unsafe target meaning: A comfortably sufficient number of security patches passed the regression suite.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 7 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the release manager expected before authorize the production rollout. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 19: Software security release — At least, exactly, and an open numerical ceiling
At least 7 security patches passed the regression suite.
Tempting but unsafe target meaning: Exactly 7 security patches passed the regression suite, no more.
Commitment-preserving target specification: The verified count is 7 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a dependency updated after the test window. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in build logs, test hashes and release notes increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of authorize the production rollout and produce shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 20: Software security release — Most, all, and majority without universal completion
Most security patches passed the regression suite.
Tempting but unsafe target meaning: All security patches passed the regression suite, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a dependency updated after the test window. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger authorize the production rollout and cause shipping a component with an untested vulnerability. Compare the target against build logs, test hashes and release notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 3 — Crop seed quality
Scene: The quality lead must label the supply batch. The relevant material is documented through sample identification and signed germination counts. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: an incomplete sample from a storage bin. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 21: Crop seed quality — Some, all, and an open upper bound
Some seed lots met the germination criterion.
Tempting but unsafe target meaning: Only some seed lots met the germination criterion; others definitely failed.
Commitment-preserving target specification: A verified positive subset of seed lots met the germination criterion; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in sample identification and signed germination counts may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the quality lead intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including an incomplete sample from a storage bin. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to label the supply batch and create mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 22: Crop seed quality — Or, and, and an unwarranted exclusive choice
The quality lead may authorize a laboratory germination test or a controlled field test.
Tempting but unsafe target meaning: The quality lead must select exactly one of a laboratory germination test and a controlled field test, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a laboratory germination test might precede a controlled field test, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a laboratory germination test followed by a controlled field test. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to label the supply batch, with the risk of mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 23: Crop seed quality — May, permission, and unasserted performance
The quality lead may repeat the sampling procedure.
Tempting but unsafe target meaning: The quality lead is required to repeat the sampling procedure now.
Commitment-preserving target specification: The quality lead has permission to repeat the sampling procedure; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the quality lead from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the quality lead decides not to repeat the sampling procedure because the evidence in sample identification and signed germination counts is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for label the supply batch; a false permission can equally compromise control and cause mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 24: Crop seed quality — Sometimes, often, and frequency without an invented rate
The quality lead sometimes checks moisture records before packaging.
Tempting but unsafe target meaning: The quality lead always checks moisture records before packaging.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the quality lead checks moisture records before packaging and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind label the supply batch and overlook mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 25: Crop seed quality — Not all, none, and scope of negation
Not all seed lots met the germination criterion.
Tempting but unsafe target meaning: No seed lots met the germination criterion.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 11 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to label the supply batch and either conceal useful progress or cause mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 26: Crop seed quality — Only, focus, and exclusion of alternatives
Only the stored seed lot met the germination criterion.
Tempting but unsafe target meaning: Even the stored seed lot met the germination criterion; perhaps many others did too.
Commitment-preserving target specification: The stored seed lot satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the stored seed lot, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the quality lead believe other items are eligible when deciding to label the supply batch, creating mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 27: Crop seed quality — Even, expectation, and surprising inclusion
Even the oldest seed lot met the germination criterion.
Tempting but unsafe target meaning: Only the oldest seed lot met the germination criterion; every alternative was rejected.
Commitment-preserving target specification: The oldest seed lot satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the stored seed lot also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to label the supply batch and may cause mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 28: Crop seed quality — Few, a few, and evaluative scarcity
Few seed lots met the germination criterion.
Tempting but unsafe target meaning: A comfortably sufficient number of seed lots met the germination criterion.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 11 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the quality lead expected before label the supply batch. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 29: Crop seed quality — At least, exactly, and an open numerical ceiling
At least 11 seed lots met the germination criterion.
Tempting but unsafe target meaning: Exactly 11 seed lots met the germination criterion, no more.
Commitment-preserving target specification: The verified count is 11 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with an incomplete sample from a storage bin. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in sample identification and signed germination counts increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of label the supply batch and produce mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 30: Crop seed quality — Most, all, and majority without universal completion
Most seed lots met the germination criterion.
Tempting but unsafe target meaning: All seed lots met the germination criterion, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include an incomplete sample from a storage bin. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger label the supply batch and cause mixing acceptable and unverified lots into one shipment. Compare the target against sample identification and signed germination counts. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 4 — Research ethics documentation
Scene: The project coordinator must resume participant enrolment. The relevant material is documented through version history and committee approval correspondence. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: an older form still circulating with the team. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 31: Research ethics documentation — Some, all, and an open upper bound
Some consent forms contained the revised disclosure.
Tempting but unsafe target meaning: Only some consent forms contained the revised disclosure; others definitely failed.
Commitment-preserving target specification: A verified positive subset of consent forms contained the revised disclosure; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in version history and committee approval correspondence may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the project coordinator intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including an older form still circulating with the team. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to resume participant enrolment and create using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 32: Research ethics documentation — Or, and, and an unwarranted exclusive choice
The project coordinator may authorize a participant recontact or a documentary amendment.
Tempting but unsafe target meaning: The project coordinator must select exactly one of a participant recontact and a documentary amendment, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a participant recontact might precede a documentary amendment, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a participant recontact followed by a documentary amendment. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to resume participant enrolment, with the risk of using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 33: Research ethics documentation — May, permission, and unasserted performance
The project coordinator may consult the ethics officer.
Tempting but unsafe target meaning: The project coordinator is required to consult the ethics officer now.
Commitment-preserving target specification: The project coordinator has permission to consult the ethics officer; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the project coordinator from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the project coordinator decides not to consult the ethics officer because the evidence in version history and committee approval correspondence is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for resume participant enrolment; a false permission can equally compromise control and cause using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 34: Research ethics documentation — Sometimes, often, and frequency without an invented rate
The project coordinator sometimes records the approved version before recruitment.
Tempting but unsafe target meaning: The project coordinator always records the approved version before recruitment.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the project coordinator records the approved version before recruitment and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind resume participant enrolment and overlook using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 35: Research ethics documentation — Not all, none, and scope of negation
Not all consent forms contained the revised disclosure.
Tempting but unsafe target meaning: No consent forms contained the revised disclosure.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 18 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to resume participant enrolment and either conceal useful progress or cause using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 36: Research ethics documentation — Only, focus, and exclusion of alternatives
Only the translated consent form contained the revised disclosure.
Tempting but unsafe target meaning: Even the translated consent form contained the revised disclosure; perhaps many others did too.
Commitment-preserving target specification: The translated consent form satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the translated consent form, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the project coordinator believe other items are eligible when deciding to resume participant enrolment, creating using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 37: Research ethics documentation — Even, expectation, and surprising inclusion
Even the archived consent form contained the revised disclosure.
Tempting but unsafe target meaning: Only the archived consent form contained the revised disclosure; every alternative was rejected.
Commitment-preserving target specification: The archived consent form satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the translated consent form also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to resume participant enrolment and may cause using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 38: Research ethics documentation — Few, a few, and evaluative scarcity
Few consent forms contained the revised disclosure.
Tempting but unsafe target meaning: A comfortably sufficient number of consent forms contained the revised disclosure.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 18 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the project coordinator expected before resume participant enrolment. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 39: Research ethics documentation — At least, exactly, and an open numerical ceiling
At least 18 consent forms contained the revised disclosure.
Tempting but unsafe target meaning: Exactly 18 consent forms contained the revised disclosure, no more.
Commitment-preserving target specification: The verified count is 18 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with an older form still circulating with the team. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in version history and committee approval correspondence increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of resume participant enrolment and produce using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 40: Research ethics documentation — Most, all, and majority without universal completion
Most consent forms contained the revised disclosure.
Tempting but unsafe target meaning: All consent forms contained the revised disclosure, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include an older form still circulating with the team. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger resume participant enrolment and cause using language that participants were never shown. Compare the target against version history and committee approval correspondence. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 5 — Transport service dispatch
Scene: The dispatcher must publish the service summary. The relevant material is documented through timestamped operating records and verified dispatch updates. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a vehicle arriving just after the permitted window. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 41: Transport service dispatch — Some, all, and an open upper bound
Some scheduled services departed within the agreed interval.
Tempting but unsafe target meaning: Only some scheduled services departed within the agreed interval; others definitely failed.
Commitment-preserving target specification: A verified positive subset of scheduled services departed within the agreed interval; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in timestamped operating records and verified dispatch updates may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the dispatcher intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a vehicle arriving just after the permitted window. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to publish the service summary and create announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 42: Transport service dispatch — Or, and, and an unwarranted exclusive choice
The dispatcher may authorize an additional shuttle or a revised dispatch sequence.
Tempting but unsafe target meaning: The dispatcher must select exactly one of an additional shuttle and a revised dispatch sequence, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, an additional shuttle might precede a revised dispatch sequence, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes an additional shuttle followed by a revised dispatch sequence. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to publish the service summary, with the risk of announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 43: Transport service dispatch — May, permission, and unasserted performance
The dispatcher may activate a standby vehicle.
Tempting but unsafe target meaning: The dispatcher is required to activate a standby vehicle now.
Commitment-preserving target specification: The dispatcher has permission to activate a standby vehicle; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the dispatcher from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the dispatcher decides not to activate a standby vehicle because the evidence in timestamped operating records and verified dispatch updates is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for publish the service summary; a false permission can equally compromise control and cause announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 44: Transport service dispatch — Sometimes, often, and frequency without an invented rate
The dispatcher sometimes compares actual departures with the control sheet.
Tempting but unsafe target meaning: The dispatcher always compares actual departures with the control sheet.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the dispatcher compares actual departures with the control sheet and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind publish the service summary and overlook announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 45: Transport service dispatch — Not all, none, and scope of negation
Not all scheduled services departed within the agreed interval.
Tempting but unsafe target meaning: No scheduled services departed within the agreed interval.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 9 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to publish the service summary and either conceal useful progress or cause announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 46: Transport service dispatch — Only, focus, and exclusion of alternatives
Only the evening service departed within the agreed interval.
Tempting but unsafe target meaning: Even the evening service departed within the agreed interval; perhaps many others did too.
Commitment-preserving target specification: The evening service satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the evening service, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the dispatcher believe other items are eligible when deciding to publish the service summary, creating announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 47: Transport service dispatch — Even, expectation, and surprising inclusion
Even the first replacement service departed within the agreed interval.
Tempting but unsafe target meaning: Only the first replacement service departed within the agreed interval; every alternative was rejected.
Commitment-preserving target specification: The first replacement service satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the evening service also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to publish the service summary and may cause announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 48: Transport service dispatch — Few, a few, and evaluative scarcity
Few scheduled services departed within the agreed interval.
Tempting but unsafe target meaning: A comfortably sufficient number of scheduled services departed within the agreed interval.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 9 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the dispatcher expected before publish the service summary. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 49: Transport service dispatch — At least, exactly, and an open numerical ceiling
At least 9 scheduled services departed within the agreed interval.
Tempting but unsafe target meaning: Exactly 9 scheduled services departed within the agreed interval, no more.
Commitment-preserving target specification: The verified count is 9 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a vehicle arriving just after the permitted window. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in timestamped operating records and verified dispatch updates increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of publish the service summary and produce announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 50: Transport service dispatch — Most, all, and majority without universal completion
Most scheduled services departed within the agreed interval.
Tempting but unsafe target meaning: All scheduled services departed within the agreed interval, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a vehicle arriving just after the permitted window. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger publish the service summary and cause announcing total reliability after only partial confirmation. Compare the target against timestamped operating records and verified dispatch updates. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 6 — Library preservation programme
Scene: The digital librarian must open the digital collection. The relevant material is documented through image checksums, page maps and review signatures. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a folded page not visible in the first scan. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 51: Library preservation programme — Some, all, and an open upper bound
Some scanned manuscripts passed the legibility review.
Tempting but unsafe target meaning: Only some scanned manuscripts passed the legibility review; others definitely failed.
Commitment-preserving target specification: A verified positive subset of scanned manuscripts passed the legibility review; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in image checksums, page maps and review signatures may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the digital librarian intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a folded page not visible in the first scan. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to open the digital collection and create making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 52: Library preservation programme — Or, and, and an unwarranted exclusive choice
The digital librarian may authorize higher-resolution capture or manual transcription review.
Tempting but unsafe target meaning: The digital librarian must select exactly one of higher-resolution capture and manual transcription review, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, higher-resolution capture might precede manual transcription review, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes higher-resolution capture followed by manual transcription review. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to open the digital collection, with the risk of making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 53: Library preservation programme — May, permission, and unasserted performance
The digital librarian may schedule another imaging session.
Tempting but unsafe target meaning: The digital librarian is required to schedule another imaging session now.
Commitment-preserving target specification: The digital librarian has permission to schedule another imaging session; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the digital librarian from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the digital librarian decides not to schedule another imaging session because the evidence in image checksums, page maps and review signatures is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for open the digital collection; a false permission can equally compromise control and cause making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 54: Library preservation programme — Sometimes, often, and frequency without an invented rate
The digital librarian sometimes checks the page sequence before archiving.
Tempting but unsafe target meaning: The digital librarian always checks the page sequence before archiving.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the digital librarian checks the page sequence before archiving and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind open the digital collection and overlook making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 55: Library preservation programme — Not all, none, and scope of negation
Not all scanned manuscripts passed the legibility review.
Tempting but unsafe target meaning: No scanned manuscripts passed the legibility review.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 23 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to open the digital collection and either conceal useful progress or cause making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 56: Library preservation programme — Only, focus, and exclusion of alternatives
Only the fragile manuscript passed the legibility review.
Tempting but unsafe target meaning: Even the fragile manuscript passed the legibility review; perhaps many others did too.
Commitment-preserving target specification: The fragile manuscript satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the fragile manuscript, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the digital librarian believe other items are eligible when deciding to open the digital collection, creating making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 57: Library preservation programme — Even, expectation, and surprising inclusion
Even the faded manuscript passed the legibility review.
Tempting but unsafe target meaning: Only the faded manuscript passed the legibility review; every alternative was rejected.
Commitment-preserving target specification: The faded manuscript satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the fragile manuscript also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to open the digital collection and may cause making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 58: Library preservation programme — Few, a few, and evaluative scarcity
Few scanned manuscripts passed the legibility review.
Tempting but unsafe target meaning: A comfortably sufficient number of scanned manuscripts passed the legibility review.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 23 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the digital librarian expected before open the digital collection. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 59: Library preservation programme — At least, exactly, and an open numerical ceiling
At least 23 scanned manuscripts passed the legibility review.
Tempting but unsafe target meaning: Exactly 23 scanned manuscripts passed the legibility review, no more.
Commitment-preserving target specification: The verified count is 23 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a folded page not visible in the first scan. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in image checksums, page maps and review signatures increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of open the digital collection and produce making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 60: Library preservation programme — Most, all, and majority without universal completion
Most scanned manuscripts passed the legibility review.
Tempting but unsafe target meaning: All scanned manuscripts passed the legibility review, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a folded page not visible in the first scan. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger open the digital collection and cause making unreadable pages appear complete in the catalogue. Compare the target against image checksums, page maps and review signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 7 — Laboratory instrument verification
Scene: The metrology coordinator must release the measurement report. The relevant material is documented through calibration certificates and dated instrument logs. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a calibration shift between two verification runs. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 61: Laboratory instrument verification — Some, all, and an open upper bound
Some measurement channels remained within tolerance.
Tempting but unsafe target meaning: Only some measurement channels remained within tolerance; others definitely failed.
Commitment-preserving target specification: A verified positive subset of measurement channels remained within tolerance; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in calibration certificates and dated instrument logs may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the metrology coordinator intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a calibration shift between two verification runs. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to release the measurement report and create publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 62: Laboratory instrument verification — Or, and, and an unwarranted exclusive choice
The metrology coordinator may authorize a full recalibration or a limited verification run.
Tempting but unsafe target meaning: The metrology coordinator must select exactly one of a full recalibration and a limited verification run, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a full recalibration might precede a limited verification run, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a full recalibration followed by a limited verification run. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to release the measurement report, with the risk of publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 63: Laboratory instrument verification — May, permission, and unasserted performance
The metrology coordinator may quarantine the measuring instrument.
Tempting but unsafe target meaning: The metrology coordinator is required to quarantine the measuring instrument now.
Commitment-preserving target specification: The metrology coordinator has permission to quarantine the measuring instrument; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the metrology coordinator from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the metrology coordinator decides not to quarantine the measuring instrument because the evidence in calibration certificates and dated instrument logs is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for release the measurement report; a false permission can equally compromise control and cause publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 64: Laboratory instrument verification — Sometimes, often, and frequency without an invented rate
The metrology coordinator sometimes checks reference standards before issuing results.
Tempting but unsafe target meaning: The metrology coordinator always checks reference standards before issuing results.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the metrology coordinator checks reference standards before issuing results and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind release the measurement report and overlook publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 65: Laboratory instrument verification — Not all, none, and scope of negation
Not all measurement channels remained within tolerance.
Tempting but unsafe target meaning: No measurement channels remained within tolerance.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 6 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to release the measurement report and either conceal useful progress or cause publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 66: Laboratory instrument verification — Only, focus, and exclusion of alternatives
Only the reference channel remained within tolerance.
Tempting but unsafe target meaning: Even the reference channel remained within tolerance; perhaps many others did too.
Commitment-preserving target specification: The reference channel satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the reference channel, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the metrology coordinator believe other items are eligible when deciding to release the measurement report, creating publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 67: Laboratory instrument verification — Even, expectation, and surprising inclusion
Even the temperature-sensitive channel remained within tolerance.
Tempting but unsafe target meaning: Only the temperature-sensitive channel remained within tolerance; every alternative was rejected.
Commitment-preserving target specification: The temperature-sensitive channel satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the reference channel also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to release the measurement report and may cause publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 68: Laboratory instrument verification — Few, a few, and evaluative scarcity
Few measurement channels remained within tolerance.
Tempting but unsafe target meaning: A comfortably sufficient number of measurement channels remained within tolerance.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 6 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the metrology coordinator expected before release the measurement report. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 69: Laboratory instrument verification — At least, exactly, and an open numerical ceiling
At least 6 measurement channels remained within tolerance.
Tempting but unsafe target meaning: Exactly 6 measurement channels remained within tolerance, no more.
Commitment-preserving target specification: The verified count is 6 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a calibration shift between two verification runs. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in calibration certificates and dated instrument logs increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of release the measurement report and produce publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 70: Laboratory instrument verification — Most, all, and majority without universal completion
Most measurement channels remained within tolerance.
Tempting but unsafe target meaning: All measurement channels remained within tolerance, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a calibration shift between two verification runs. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger release the measurement report and cause publishing values from an out-of-tolerance channel. Compare the target against calibration certificates and dated instrument logs. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 8 — Documentary subtitle production
Scene: The subtitle editor must approve the final subtitle track. The relevant material is documented through time-coded audio and signed editorial notes. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: an off-screen speaker overlapping the narrator. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 71: Documentary subtitle production — Some, all, and an open upper bound
Some subtitle segments matched the approved dialogue.
Tempting but unsafe target meaning: Only some subtitle segments matched the approved dialogue; others definitely failed.
Commitment-preserving target specification: A verified positive subset of subtitle segments matched the approved dialogue; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in time-coded audio and signed editorial notes may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the subtitle editor intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including an off-screen speaker overlapping the narrator. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to approve the final subtitle track and create attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 72: Documentary subtitle production — Or, and, and an unwarranted exclusive choice
The subtitle editor may authorize a transcript comparison or a speaker-attribution check.
Tempting but unsafe target meaning: The subtitle editor must select exactly one of a transcript comparison and a speaker-attribution check, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a transcript comparison might precede a speaker-attribution check, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a transcript comparison followed by a speaker-attribution check. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to approve the final subtitle track, with the risk of attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 73: Documentary subtitle production — May, permission, and unasserted performance
The subtitle editor may request a new listening pass.
Tempting but unsafe target meaning: The subtitle editor is required to request a new listening pass now.
Commitment-preserving target specification: The subtitle editor has permission to request a new listening pass; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the subtitle editor from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the subtitle editor decides not to request a new listening pass because the evidence in time-coded audio and signed editorial notes is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for approve the final subtitle track; a false permission can equally compromise control and cause attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 74: Documentary subtitle production — Sometimes, often, and frequency without an invented rate
The subtitle editor sometimes replays difficult exchanges before delivery.
Tempting but unsafe target meaning: The subtitle editor always replays difficult exchanges before delivery.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the subtitle editor replays difficult exchanges before delivery and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind approve the final subtitle track and overlook attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 75: Documentary subtitle production — Not all, none, and scope of negation
Not all subtitle segments matched the approved dialogue.
Tempting but unsafe target meaning: No subtitle segments matched the approved dialogue.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 31 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to approve the final subtitle track and either conceal useful progress or cause attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 76: Documentary subtitle production — Only, focus, and exclusion of alternatives
Only the closing subtitle matched the approved dialogue.
Tempting but unsafe target meaning: Even the closing subtitle matched the approved dialogue; perhaps many others did too.
Commitment-preserving target specification: The closing subtitle satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the closing subtitle, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the subtitle editor believe other items are eligible when deciding to approve the final subtitle track, creating attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 77: Documentary subtitle production — Even, expectation, and surprising inclusion
Even the overlapping-speaker subtitle matched the approved dialogue.
Tempting but unsafe target meaning: Only the overlapping-speaker subtitle matched the approved dialogue; every alternative was rejected.
Commitment-preserving target specification: The overlapping-speaker subtitle satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the closing subtitle also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to approve the final subtitle track and may cause attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 78: Documentary subtitle production — Few, a few, and evaluative scarcity
Few subtitle segments matched the approved dialogue.
Tempting but unsafe target meaning: A comfortably sufficient number of subtitle segments matched the approved dialogue.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 31 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the subtitle editor expected before approve the final subtitle track. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 79: Documentary subtitle production — At least, exactly, and an open numerical ceiling
At least 31 subtitle segments matched the approved dialogue.
Tempting but unsafe target meaning: Exactly 31 subtitle segments matched the approved dialogue, no more.
Commitment-preserving target specification: The verified count is 31 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with an off-screen speaker overlapping the narrator. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in time-coded audio and signed editorial notes increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of approve the final subtitle track and produce attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 80: Documentary subtitle production — Most, all, and majority without universal completion
Most subtitle segments matched the approved dialogue.
Tempting but unsafe target meaning: All subtitle segments matched the approved dialogue, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include an off-screen speaker overlapping the narrator. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger approve the final subtitle track and cause attributing a cautious statement to the wrong speaker. Compare the target against time-coded audio and signed editorial notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 9 — Supplier inspection
Scene: The inspection supervisor must sign the incoming-goods certificate. The relevant material is documented through batch trace codes and calibrated measurements. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a returned component mixed into an inspected batch. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 81: Supplier inspection — Some, all, and an open upper bound
Some delivered components met the dimensional specification.
Tempting but unsafe target meaning: Only some delivered components met the dimensional specification; others definitely failed.
Commitment-preserving target specification: A verified positive subset of delivered components met the dimensional specification; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in batch trace codes and calibrated measurements may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the inspection supervisor intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a returned component mixed into an inspected batch. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to sign the incoming-goods certificate and create claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 82: Supplier inspection — Or, and, and an unwarranted exclusive choice
The inspection supervisor may authorize a full measurement campaign or a random acceptance sample.
Tempting but unsafe target meaning: The inspection supervisor must select exactly one of a full measurement campaign and a random acceptance sample, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a full measurement campaign might precede a random acceptance sample, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a full measurement campaign followed by a random acceptance sample. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to sign the incoming-goods certificate, with the risk of claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 83: Supplier inspection — May, permission, and unasserted performance
The inspection supervisor may hold the incoming batch.
Tempting but unsafe target meaning: The inspection supervisor is required to hold the incoming batch now.
Commitment-preserving target specification: The inspection supervisor has permission to hold the incoming batch; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the inspection supervisor from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the inspection supervisor decides not to hold the incoming batch because the evidence in batch trace codes and calibrated measurements is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for sign the incoming-goods certificate; a false permission can equally compromise control and cause claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 84: Supplier inspection — Sometimes, often, and frequency without an invented rate
The inspection supervisor sometimes checks lot traceability before issuing approval.
Tempting but unsafe target meaning: The inspection supervisor always checks lot traceability before issuing approval.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the inspection supervisor checks lot traceability before issuing approval and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind sign the incoming-goods certificate and overlook claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 85: Supplier inspection — Not all, none, and scope of negation
Not all delivered components met the dimensional specification.
Tempting but unsafe target meaning: No delivered components met the dimensional specification.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 27 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to sign the incoming-goods certificate and either conceal useful progress or cause claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 86: Supplier inspection — Only, focus, and exclusion of alternatives
Only the replacement component met the dimensional specification.
Tempting but unsafe target meaning: Even the replacement component met the dimensional specification; perhaps many others did too.
Commitment-preserving target specification: The replacement component satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the replacement component, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the inspection supervisor believe other items are eligible when deciding to sign the incoming-goods certificate, creating claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 87: Supplier inspection — Even, expectation, and surprising inclusion
Even the hand-finished component met the dimensional specification.
Tempting but unsafe target meaning: Only the hand-finished component met the dimensional specification; every alternative was rejected.
Commitment-preserving target specification: The hand-finished component satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the replacement component also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to sign the incoming-goods certificate and may cause claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 88: Supplier inspection — Few, a few, and evaluative scarcity
Few delivered components met the dimensional specification.
Tempting but unsafe target meaning: A comfortably sufficient number of delivered components met the dimensional specification.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 27 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the inspection supervisor expected before sign the incoming-goods certificate. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 89: Supplier inspection — At least, exactly, and an open numerical ceiling
At least 27 delivered components met the dimensional specification.
Tempting but unsafe target meaning: Exactly 27 delivered components met the dimensional specification, no more.
Commitment-preserving target specification: The verified count is 27 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a returned component mixed into an inspected batch. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in batch trace codes and calibrated measurements increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of sign the incoming-goods certificate and produce claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 90: Supplier inspection — Most, all, and majority without universal completion
Most delivered components met the dimensional specification.
Tempting but unsafe target meaning: All delivered components met the dimensional specification, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a returned component mixed into an inspected batch. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger sign the incoming-goods certificate and cause claiming every component conforms without exhaustive evidence. Compare the target against batch trace codes and calibrated measurements. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 10 — Power-system monitoring
Scene: The control engineer must publish the reliability statement. The relevant material is documented through event logs and independently checked telemetry. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: an alert that cleared temporarily but returned. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 91: Power-system monitoring — Some, all, and an open upper bound
Some monitoring alerts were resolved before the reporting cut-off.
Tempting but unsafe target meaning: Only some monitoring alerts were resolved before the reporting cut-off; others definitely failed.
Commitment-preserving target specification: A verified positive subset of monitoring alerts were resolved before the reporting cut-off; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in event logs and independently checked telemetry may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the control engineer intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including an alert that cleared temporarily but returned. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to publish the reliability statement and create closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 92: Power-system monitoring — Or, and, and an unwarranted exclusive choice
The control engineer may authorize an on-site inspection or a remote telemetry check.
Tempting but unsafe target meaning: The control engineer must select exactly one of an on-site inspection and a remote telemetry check, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, an on-site inspection might precede a remote telemetry check, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes an on-site inspection followed by a remote telemetry check. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to publish the reliability statement, with the risk of closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 93: Power-system monitoring — May, permission, and unasserted performance
The control engineer may trigger a diagnostic review.
Tempting but unsafe target meaning: The control engineer is required to trigger a diagnostic review now.
Commitment-preserving target specification: The control engineer has permission to trigger a diagnostic review; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the control engineer from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the control engineer decides not to trigger a diagnostic review because the evidence in event logs and independently checked telemetry is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for publish the reliability statement; a false permission can equally compromise control and cause closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 94: Power-system monitoring — Sometimes, often, and frequency without an invented rate
The control engineer sometimes reconciles event timestamps before closing incidents.
Tempting but unsafe target meaning: The control engineer always reconciles event timestamps before closing incidents.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the control engineer reconciles event timestamps before closing incidents and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind publish the reliability statement and overlook closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 95: Power-system monitoring — Not all, none, and scope of negation
Not all monitoring alerts were resolved before the reporting cut-off.
Tempting but unsafe target meaning: No monitoring alerts were resolved before the reporting cut-off.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 13 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to publish the reliability statement and either conceal useful progress or cause closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 96: Power-system monitoring — Only, focus, and exclusion of alternatives
Only the voltage alert was resolved before the reporting cut-off.
Tempting but unsafe target meaning: Even the voltage alert was resolved before the reporting cut-off; perhaps many others did too.
Commitment-preserving target specification: The voltage alert satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the voltage alert, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the control engineer believe other items are eligible when deciding to publish the reliability statement, creating closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 97: Power-system monitoring — Even, expectation, and surprising inclusion
Even the intermittent alarm was resolved before the reporting cut-off.
Tempting but unsafe target meaning: Only the intermittent alarm was resolved before the reporting cut-off; every alternative was rejected.
Commitment-preserving target specification: The intermittent alarm satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the voltage alert also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to publish the reliability statement and may cause closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 98: Power-system monitoring — Few, a few, and evaluative scarcity
Few monitoring alerts were resolved before the reporting cut-off.
Tempting but unsafe target meaning: A comfortably sufficient number of monitoring alerts were resolved before the reporting cut-off.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 13 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the control engineer expected before publish the reliability statement. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 99: Power-system monitoring — At least, exactly, and an open numerical ceiling
At least 13 monitoring alerts were resolved before the reporting cut-off.
Tempting but unsafe target meaning: Exactly 13 monitoring alerts were resolved before the reporting cut-off, no more.
Commitment-preserving target specification: The verified count is 13 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with an alert that cleared temporarily but returned. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in event logs and independently checked telemetry increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of publish the reliability statement and produce closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 100: Power-system monitoring — Most, all, and majority without universal completion
Most monitoring alerts were resolved before the reporting cut-off.
Tempting but unsafe target meaning: All monitoring alerts were resolved before the reporting cut-off, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include an alert that cleared temporarily but returned. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger publish the reliability statement and cause closing an unresolved fault because the summary sounds complete. Compare the target against event logs and independently checked telemetry. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 11 — Educational assessment moderation
Scene: The moderation lead must release the assessment outcomes. The relevant material is documented through versioned marking records and discrepancy reports. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a late appeal that changes the relevant cohort. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 101: Educational assessment moderation — Some, all, and an open upper bound
Some assessed scripts received the second marking pass.
Tempting but unsafe target meaning: Only some assessed scripts received the second marking pass; others definitely failed.
Commitment-preserving target specification: A verified positive subset of assessed scripts received the second marking pass; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in versioned marking records and discrepancy reports may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the moderation lead intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a late appeal that changes the relevant cohort. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to release the assessment outcomes and create conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 102: Educational assessment moderation — Or, and, and an unwarranted exclusive choice
The moderation lead may authorize double-marking of all scripts or targeted moderation of borderline results.
Tempting but unsafe target meaning: The moderation lead must select exactly one of double-marking of all scripts and targeted moderation of borderline results, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, double-marking of all scripts might precede targeted moderation of borderline results, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes double-marking of all scripts followed by targeted moderation of borderline results. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to release the assessment outcomes, with the risk of conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 103: Educational assessment moderation — May, permission, and unasserted performance
The moderation lead may convene an additional moderation review.
Tempting but unsafe target meaning: The moderation lead is required to convene an additional moderation review now.
Commitment-preserving target specification: The moderation lead has permission to convene an additional moderation review; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the moderation lead from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the moderation lead decides not to convene an additional moderation review because the evidence in versioned marking records and discrepancy reports is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for release the assessment outcomes; a false permission can equally compromise control and cause conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 104: Educational assessment moderation — Sometimes, often, and frequency without an invented rate
The moderation lead sometimes compares marking comments before final sign-off.
Tempting but unsafe target meaning: The moderation lead always compares marking comments before final sign-off.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the moderation lead compares marking comments before final sign-off and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind release the assessment outcomes and overlook conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 105: Educational assessment moderation — Not all, none, and scope of negation
Not all assessed scripts received the second marking pass.
Tempting but unsafe target meaning: No assessed scripts received the second marking pass.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 42 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to release the assessment outcomes and either conceal useful progress or cause conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 106: Educational assessment moderation — Only, focus, and exclusion of alternatives
Only the borderline script received the second marking pass.
Tempting but unsafe target meaning: Even the borderline script received the second marking pass; perhaps many others did too.
Commitment-preserving target specification: The borderline script satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the borderline script, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the moderation lead believe other items are eligible when deciding to release the assessment outcomes, creating conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 107: Educational assessment moderation — Even, expectation, and surprising inclusion
Even the partially scanned script received the second marking pass.
Tempting but unsafe target meaning: Only the partially scanned script received the second marking pass; every alternative was rejected.
Commitment-preserving target specification: The partially scanned script satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the borderline script also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to release the assessment outcomes and may cause conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 108: Educational assessment moderation — Few, a few, and evaluative scarcity
Few assessed scripts received the second marking pass.
Tempting but unsafe target meaning: A comfortably sufficient number of assessed scripts received the second marking pass.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 42 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the moderation lead expected before release the assessment outcomes. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 109: Educational assessment moderation — At least, exactly, and an open numerical ceiling
At least 42 assessed scripts received the second marking pass.
Tempting but unsafe target meaning: Exactly 42 assessed scripts received the second marking pass, no more.
Commitment-preserving target specification: The verified count is 42 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a late appeal that changes the relevant cohort. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in versioned marking records and discrepancy reports increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of release the assessment outcomes and produce conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 110: Educational assessment moderation — Most, all, and majority without universal completion
Most assessed scripts received the second marking pass.
Tempting but unsafe target meaning: All assessed scripts received the second marking pass, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a late appeal that changes the relevant cohort. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger release the assessment outcomes and cause conflating a reviewed sample with the entire submission. Compare the target against versioned marking records and discrepancy reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 12 — Emergency supply logistics
Scene: The logistics coordinator must update the distribution bulletin. The relevant material is documented through signed receiving notes and transport status messages. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a vehicle held at a transfer point. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 111: Emergency supply logistics — Some, all, and an open upper bound
Some relief consignments arrived at the distribution depot.
Tempting but unsafe target meaning: Only some relief consignments arrived at the distribution depot; others definitely failed.
Commitment-preserving target specification: A verified positive subset of relief consignments arrived at the distribution depot; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in signed receiving notes and transport status messages may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the logistics coordinator intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a vehicle held at a transfer point. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to update the distribution bulletin and create telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 112: Emergency supply logistics — Or, and, and an unwarranted exclusive choice
The logistics coordinator may authorize an express delivery or a redistribution from available stock.
Tempting but unsafe target meaning: The logistics coordinator must select exactly one of an express delivery and a redistribution from available stock, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, an express delivery might precede a redistribution from available stock, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes an express delivery followed by a redistribution from available stock. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to update the distribution bulletin, with the risk of telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 113: Emergency supply logistics — May, permission, and unasserted performance
The logistics coordinator may order a second dispatch.
Tempting but unsafe target meaning: The logistics coordinator is required to order a second dispatch now.
Commitment-preserving target specification: The logistics coordinator has permission to order a second dispatch; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the logistics coordinator from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the logistics coordinator decides not to order a second dispatch because the evidence in signed receiving notes and transport status messages is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for update the distribution bulletin; a false permission can equally compromise control and cause telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 114: Emergency supply logistics — Sometimes, often, and frequency without an invented rate
The logistics coordinator sometimes checks arrival receipts against the dispatch register.
Tempting but unsafe target meaning: The logistics coordinator always checks arrival receipts against the dispatch register.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the logistics coordinator checks arrival receipts against the dispatch register and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind update the distribution bulletin and overlook telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 115: Emergency supply logistics — Not all, none, and scope of negation
Not all relief consignments arrived at the distribution depot.
Tempting but unsafe target meaning: No relief consignments arrived at the distribution depot.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 16 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to update the distribution bulletin and either conceal useful progress or cause telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 116: Emergency supply logistics — Only, focus, and exclusion of alternatives
Only the priority consignment arrived at the distribution depot.
Tempting but unsafe target meaning: Even the priority consignment arrived at the distribution depot; perhaps many others did too.
Commitment-preserving target specification: The priority consignment satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the priority consignment, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the logistics coordinator believe other items are eligible when deciding to update the distribution bulletin, creating telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 117: Emergency supply logistics — Even, expectation, and surprising inclusion
Even the re-routed consignment arrived at the distribution depot.
Tempting but unsafe target meaning: Only the re-routed consignment arrived at the distribution depot; every alternative was rejected.
Commitment-preserving target specification: The re-routed consignment satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the priority consignment also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to update the distribution bulletin and may cause telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 118: Emergency supply logistics — Few, a few, and evaluative scarcity
Few relief consignments arrived at the distribution depot.
Tempting but unsafe target meaning: A comfortably sufficient number of relief consignments arrived at the distribution depot.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 16 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the logistics coordinator expected before update the distribution bulletin. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 119: Emergency supply logistics — At least, exactly, and an open numerical ceiling
At least 16 relief consignments arrived at the distribution depot.
Tempting but unsafe target meaning: Exactly 16 relief consignments arrived at the distribution depot, no more.
Commitment-preserving target specification: The verified count is 16 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a vehicle held at a transfer point. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in signed receiving notes and transport status messages increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of update the distribution bulletin and produce telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 120: Emergency supply logistics — Most, all, and majority without universal completion
Most relief consignments arrived at the distribution depot.
Tempting but unsafe target meaning: All relief consignments arrived at the distribution depot, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a vehicle held at a transfer point. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger update the distribution bulletin and cause telling recipients that undelivered goods are already available. Compare the target against signed receiving notes and transport status messages. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 13 — Publishing permissions
Scene: The rights editor must approve the print-ready edition. The relevant material is documented through rights-holder correspondence and image-level licence notes. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: permission that covers print but not digital use. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 121: Publishing permissions — Some, all, and an open upper bound
Some licensed illustrations received written reuse permission.
Tempting but unsafe target meaning: Only some licensed illustrations received written reuse permission; others definitely failed.
Commitment-preserving target specification: A verified positive subset of licensed illustrations received written reuse permission; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in rights-holder correspondence and image-level licence notes may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the rights editor intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including permission that covers print but not digital use. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to approve the print-ready edition and create publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 122: Publishing permissions — Or, and, and an unwarranted exclusive choice
The rights editor may authorize a direct rights clearance or a replacement image search.
Tempting but unsafe target meaning: The rights editor must select exactly one of a direct rights clearance and a replacement image search, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a direct rights clearance might precede a replacement image search, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a direct rights clearance followed by a replacement image search. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to approve the print-ready edition, with the risk of publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 123: Publishing permissions — May, permission, and unasserted performance
The rights editor may request written clarification.
Tempting but unsafe target meaning: The rights editor is required to request written clarification now.
Commitment-preserving target specification: The rights editor has permission to request written clarification; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the rights editor from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the rights editor decides not to request written clarification because the evidence in rights-holder correspondence and image-level licence notes is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for approve the print-ready edition; a false permission can equally compromise control and cause publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 124: Publishing permissions — Sometimes, often, and frequency without an invented rate
The rights editor sometimes checks licensing restrictions before layout approval.
Tempting but unsafe target meaning: The rights editor always checks licensing restrictions before layout approval.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the rights editor checks licensing restrictions before layout approval and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind approve the print-ready edition and overlook publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 125: Publishing permissions — Not all, none, and scope of negation
Not all licensed illustrations received written reuse permission.
Tempting but unsafe target meaning: No licensed illustrations received written reuse permission.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 12 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to approve the print-ready edition and either conceal useful progress or cause publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 126: Publishing permissions — Only, focus, and exclusion of alternatives
Only the cover illustration received written reuse permission.
Tempting but unsafe target meaning: Even the cover illustration received written reuse permission; perhaps many others did too.
Commitment-preserving target specification: The cover illustration satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the cover illustration, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the rights editor believe other items are eligible when deciding to approve the print-ready edition, creating publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 127: Publishing permissions — Even, expectation, and surprising inclusion
Even the historical illustration received written reuse permission.
Tempting but unsafe target meaning: Only the historical illustration received written reuse permission; every alternative was rejected.
Commitment-preserving target specification: The historical illustration satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the cover illustration also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to approve the print-ready edition and may cause publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 128: Publishing permissions — Few, a few, and evaluative scarcity
Few licensed illustrations received written reuse permission.
Tempting but unsafe target meaning: A comfortably sufficient number of licensed illustrations received written reuse permission.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 12 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the rights editor expected before approve the print-ready edition. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 129: Publishing permissions — At least, exactly, and an open numerical ceiling
At least 12 licensed illustrations received written reuse permission.
Tempting but unsafe target meaning: Exactly 12 licensed illustrations received written reuse permission, no more.
Commitment-preserving target specification: The verified count is 12 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with permission that covers print but not digital use. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in rights-holder correspondence and image-level licence notes increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of approve the print-ready edition and produce publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 130: Publishing permissions — Most, all, and majority without universal completion
Most licensed illustrations received written reuse permission.
Tempting but unsafe target meaning: All licensed illustrations received written reuse permission, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include permission that covers print but not digital use. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger approve the print-ready edition and cause publishing protected material beyond the permission granted. Compare the target against rights-holder correspondence and image-level licence notes. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 14 — Exhibition accessibility
Scene: The accessibility reviewer must approve the visitor route. The relevant material is documented through contrast measurements and visitor-path observations. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a label behind reflective protective glass. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 131: Exhibition accessibility — Some, all, and an open upper bound
Some exhibition labels met the revised contrast standard.
Tempting but unsafe target meaning: Only some exhibition labels met the revised contrast standard; others definitely failed.
Commitment-preserving target specification: A verified positive subset of exhibition labels met the revised contrast standard; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in contrast measurements and visitor-path observations may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the accessibility reviewer intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a label behind reflective protective glass. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to approve the visitor route and create claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 132: Exhibition accessibility — Or, and, and an unwarranted exclusive choice
The accessibility reviewer may authorize an on-wall legibility test or a prototype review.
Tempting but unsafe target meaning: The accessibility reviewer must select exactly one of an on-wall legibility test and a prototype review, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, an on-wall legibility test might precede a prototype review, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes an on-wall legibility test followed by a prototype review. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to approve the visitor route, with the risk of claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 133: Exhibition accessibility — May, permission, and unasserted performance
The accessibility reviewer may require another accessibility audit.
Tempting but unsafe target meaning: The accessibility reviewer is required to require another accessibility audit now.
Commitment-preserving target specification: The accessibility reviewer has permission to require another accessibility audit; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the accessibility reviewer from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the accessibility reviewer decides not to require another accessibility audit because the evidence in contrast measurements and visitor-path observations is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for approve the visitor route; a false permission can equally compromise control and cause claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 134: Exhibition accessibility — Sometimes, often, and frequency without an invented rate
The accessibility reviewer sometimes checks lighting conditions before sign-off.
Tempting but unsafe target meaning: The accessibility reviewer always checks lighting conditions before sign-off.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the accessibility reviewer checks lighting conditions before sign-off and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind approve the visitor route and overlook claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 135: Exhibition accessibility — Not all, none, and scope of negation
Not all exhibition labels met the revised contrast standard.
Tempting but unsafe target meaning: No exhibition labels met the revised contrast standard.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 19 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to approve the visitor route and either conceal useful progress or cause claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 136: Exhibition accessibility — Only, focus, and exclusion of alternatives
Only the entrance label met the revised contrast standard.
Tempting but unsafe target meaning: Even the entrance label met the revised contrast standard; perhaps many others did too.
Commitment-preserving target specification: The entrance label satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the entrance label, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the accessibility reviewer believe other items are eligible when deciding to approve the visitor route, creating claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 137: Exhibition accessibility — Even, expectation, and surprising inclusion
Even the backlit label met the revised contrast standard.
Tempting but unsafe target meaning: Only the backlit label met the revised contrast standard; every alternative was rejected.
Commitment-preserving target specification: The backlit label satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the entrance label also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to approve the visitor route and may cause claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 138: Exhibition accessibility — Few, a few, and evaluative scarcity
Few exhibition labels met the revised contrast standard.
Tempting but unsafe target meaning: A comfortably sufficient number of exhibition labels met the revised contrast standard.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 19 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the accessibility reviewer expected before approve the visitor route. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 139: Exhibition accessibility — At least, exactly, and an open numerical ceiling
At least 19 exhibition labels met the revised contrast standard.
Tempting but unsafe target meaning: Exactly 19 exhibition labels met the revised contrast standard, no more.
Commitment-preserving target specification: The verified count is 19 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a label behind reflective protective glass. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in contrast measurements and visitor-path observations increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of approve the visitor route and produce claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 140: Exhibition accessibility — Most, all, and majority without universal completion
Most exhibition labels met the revised contrast standard.
Tempting but unsafe target meaning: All exhibition labels met the revised contrast standard, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a label behind reflective protective glass. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger approve the visitor route and cause claiming full accessibility while a critical label remains unreadable. Compare the target against contrast measurements and visitor-path observations. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 15 — Digital transaction processing
Scene: The operations analyst must publish the settlement status. The relevant material is documented through immutable transaction IDs and reconciliation reports. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: a delayed reversal posted after the reporting cut-off. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 141: Digital transaction processing — Some, all, and an open upper bound
Some transaction records matched the settlement report.
Tempting but unsafe target meaning: Only some transaction records matched the settlement report; others definitely failed.
Commitment-preserving target specification: A verified positive subset of transaction records matched the settlement report; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in immutable transaction IDs and reconciliation reports may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the operations analyst intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including a delayed reversal posted after the reporting cut-off. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to publish the settlement status and create presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 142: Digital transaction processing — Or, and, and an unwarranted exclusive choice
The operations analyst may authorize a full ledger review or an exception-only audit.
Tempting but unsafe target meaning: The operations analyst must select exactly one of a full ledger review and an exception-only audit, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a full ledger review might precede an exception-only audit, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a full ledger review followed by an exception-only audit. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to publish the settlement status, with the risk of presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 143: Digital transaction processing — May, permission, and unasserted performance
The operations analyst may open a reconciliation case.
Tempting but unsafe target meaning: The operations analyst is required to open a reconciliation case now.
Commitment-preserving target specification: The operations analyst has permission to open a reconciliation case; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the operations analyst from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the operations analyst decides not to open a reconciliation case because the evidence in immutable transaction IDs and reconciliation reports is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for publish the settlement status; a false permission can equally compromise control and cause presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 144: Digital transaction processing — Sometimes, often, and frequency without an invented rate
The operations analyst sometimes reconciles settlement identifiers before close.
Tempting but unsafe target meaning: The operations analyst always reconciles settlement identifiers before close.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the operations analyst reconciles settlement identifiers before close and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind publish the settlement status and overlook presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 145: Digital transaction processing — Not all, none, and scope of negation
Not all transaction records matched the settlement report.
Tempting but unsafe target meaning: No transaction records matched the settlement report.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 84 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to publish the settlement status and either conceal useful progress or cause presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 146: Digital transaction processing — Only, focus, and exclusion of alternatives
Only the refund record matched the settlement report.
Tempting but unsafe target meaning: Even the refund record matched the settlement report; perhaps many others did too.
Commitment-preserving target specification: The refund record satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the refund record, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the operations analyst believe other items are eligible when deciding to publish the settlement status, creating presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 147: Digital transaction processing — Even, expectation, and surprising inclusion
Even the reversed transaction matched the settlement report.
Tempting but unsafe target meaning: Only the reversed transaction matched the settlement report; every alternative was rejected.
Commitment-preserving target specification: The reversed transaction satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the refund record also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to publish the settlement status and may cause presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 148: Digital transaction processing — Few, a few, and evaluative scarcity
Few transaction records matched the settlement report.
Tempting but unsafe target meaning: A comfortably sufficient number of transaction records matched the settlement report.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 84 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the operations analyst expected before publish the settlement status. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 149: Digital transaction processing — At least, exactly, and an open numerical ceiling
At least 84 transaction records matched the settlement report.
Tempting but unsafe target meaning: Exactly 84 transaction records matched the settlement report, no more.
Commitment-preserving target specification: The verified count is 84 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with a delayed reversal posted after the reporting cut-off. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in immutable transaction IDs and reconciliation reports increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of publish the settlement status and produce presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 150: Digital transaction processing — Most, all, and majority without universal completion
Most transaction records matched the settlement report.
Tempting but unsafe target meaning: All transaction records matched the settlement report, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include a delayed reversal posted after the reporting cut-off. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger publish the settlement status and cause presenting a partly reconciled ledger as fully balanced. Compare the target against immutable transaction IDs and reconciliation reports. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Laboratory 16 — Marine species observation
Scene: The survey coordinator must release the survey findings. The relevant material is documented through field notes, imagery and verification signatures. This local evidence frame, rather than an abstract dictionary definition, supplies the domain against which quantities and alternatives are interpreted. The reviewer keeps a specific edge case in view: two similar species with different identification markers. The following variations are independent experiments, not a claim that a real organisation uses all ten formulations.
Diagnostic 151: Marine species observation — Some, all, and an open upper bound
Some recorded sightings were independently verified.
Tempting but unsafe target meaning: Only some recorded sightings were independently verified; others definitely failed.
Commitment-preserving target specification: A verified positive subset of recorded sightings were independently verified; the rest are not settled by this sentence.
Mechanism and hidden question. The archived evidence in field notes, imagery and verification signatures may establish at least one positive case without proving either that the other cases failed or that every case succeeded. If the survey coordinator intentionally chose a weaker quantifier because the audit is incomplete, asserting a negative about the remainder confuses incomplete knowledge with negative evidence. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the final audit later confirms that every relevant record or unit succeeded, including two similar species with different identification markers. The source statement can remain true. The proposed ‘only some’ target, interpreted as explicitly excluding all, would then be false. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An inflated exclusion can distort the decision to release the survey findings and create overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 152: Marine species observation — Or, and, and an unwarranted exclusive choice
The survey coordinator may authorize a photographic recheck or a second independent identification.
Tempting but unsafe target meaning: The survey coordinator must select exactly one of a photographic recheck and a second independent identification, never both.
Commitment-preserving target specification: Either named option is permitted; whether the combination is barred needs additional context.
Mechanism and hidden question. The source offers two alternatives relevant to the work, but offering alternatives is not by itself a rule that one cancels the other. In particular, a photographic recheck might precede a second independent identification, or an urgent review might require both. A target-language connective with a compulsory exclusive reading should not be selected merely because English says ‘or’. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider an authorised plan that includes a photographic recheck followed by a second independent identification. Check whether the procedure actually forbids that plan. If not, the target’s exclusion is added rather than translated. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false either-or requirement can obstruct the work needed to release the survey findings, with the risk of overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 153: Marine species observation — May, permission, and unasserted performance
The survey coordinator may request another observer review.
Tempting but unsafe target meaning: The survey coordinator is required to request another observer review now.
Commitment-preserving target specification: The survey coordinator has permission to request another observer review; the statement does not establish a duty or actual event.
Mechanism and hidden question. The relevant task is to distinguish an option available to the survey coordinator from a mandatory instruction. If the source is a rule of permission, the target must not turn it into a requirement. If the surrounding text instead discusses uncertainty about whether an event will happen, the modal must be analysed anew as epistemic possibility; neither reading can be decided from a dictionary entry alone. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine the survey coordinator decides not to request another observer review because the evidence in field notes, imagery and verification signatures is satisfactory. If non-performance is allowed, a mandatory rendering contradicts the source permission. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A false duty consumes resources or changes responsibility for release the survey findings; a false permission can equally compromise control and cause overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 154: Marine species observation — Sometimes, often, and frequency without an invented rate
The survey coordinator sometimes compares observation notes against the specimen guide.
Tempting but unsafe target meaning: The survey coordinator always compares observation notes against the specimen guide.
Commitment-preserving target specification: The behaviour occurs on at least some occasions, without an assertion of universal regularity.
Mechanism and hidden question. The utterance describes a recurring activity, not a guaranteed step performed on every occasion. Substituting a target-language expression meaning ‘always’ can turn a descriptive practice into a false process assurance. Substituting ‘rarely’ may impose an opposite evaluative stance. Without a published frequency denominator, numerical equivalence is not justified. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Compare two ordinary work cycles, one in which the survey coordinator compares observation notes against the specimen guide and another in which an authorised alternative check is used. Both can fit the source, but not the proposed ‘always’ target. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Readers relying on a universal process claim may misjudge the diligence behind release the survey findings and overlook overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 155: Marine species observation — Not all, none, and scope of negation
Not all recorded sightings were independently verified.
Tempting but unsafe target meaning: No recorded sightings were independently verified.
Commitment-preserving target specification: At least one relevant item did not satisfy the predicate; the statement does not say that zero items satisfied it.
Mechanism and hidden question. The expression negates a universal claim. It does not negate each individual positive case. A natural target should preserve the possibility of substantial success alongside one or more exceptions. The critical problem is the scope of negation, not merely the polarity of a single word. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose 26 items succeeded and exactly one comparable item failed. The source ‘not all’ is true. A target that says ‘none’ is false. Repeat the exercise with reversed counts to prove that the difference is structural. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. An all-or-nothing error can invalidate the decision to release the survey findings and either conceal useful progress or cause overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 156: Marine species observation — Only, focus, and exclusion of alternatives
Only the rare-species sighting was independently verified.
Tempting but unsafe target meaning: Even the rare-species sighting was independently verified; perhaps many others did too.
Commitment-preserving target specification: The rare-species sighting satisfied the property and other contextually relevant alternatives did not.
Mechanism and hidden question. The important semantic work is not the isolated word ‘only’ but the location of its focus. In this utterance the focused item is the rare-species sighting, not the action or the person producing the report. The exact alternatives depend on context; it would be excessive to claim that every conceivable object in the world is excluded. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Introduce one additional relevant item that also satisfied the property. If the domain is unchanged, the literal only-statement is now undermined. A target that merely adds surprise still allows the second case and has weakened the source. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A lost exclusion can make the survey coordinator believe other items are eligible when deciding to release the survey findings, creating overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 157: Marine species observation — Even, expectation, and surprising inclusion
Even the night-time sighting was independently verified.
Tempting but unsafe target meaning: Only the night-time sighting was independently verified; every alternative was rejected.
Commitment-preserving target specification: The night-time sighting satisfied the property, although it was a notable or unexpected case among salient alternatives.
Mechanism and hidden question. Even typically depends on a scale of expectations or noteworthiness; it does not ordinarily assert that the named item is the only successful one. The source may imply that more ordinary cases also succeeded, but the precise alternative set and strength of that implication need contextual confirmation. In the target, a particle, special word order or short paraphrase may be needed to convey the surprise. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Imagine that the rare-species sighting also succeeded and nobody considers that surprising. The even-statement remains coherent; the proposed only-statement becomes false. This isolates the direction of the error. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Mistaking unexpected inclusion for exclusivity hides the broader evidence needed to release the survey findings and may cause overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 158: Marine species observation — Few, a few, and evaluative scarcity
Few recorded sightings were independently verified.
Tempting but unsafe target meaning: A comfortably sufficient number of recorded sightings were independently verified.
Commitment-preserving target specification: The successful number is small relative to the relevant expectation; the wording often frames that scarcity as significant.
Mechanism and hidden question. Unlike a bare numeric count, ‘few’ carries a contextual judgement about how many would normally be enough. One project may treat 26 successful cases as a reassuring sample; another may regard that number as dangerously inadequate. A direct counterpart of ‘a few’ can preserve a positive minimum while erasing the source’s scarcity framing. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Ask what level of completion the survey coordinator expected before release the survey findings. If the count falls well short of that standard, the target must not imply comfortable adequacy. If the standard is unknown, flag it instead of inventing one. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. Reassuring wording can hide an inadequate sample, with the downstream danger of overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 159: Marine species observation — At least, exactly, and an open numerical ceiling
At least 26 recorded sightings were independently verified.
Tempting but unsafe target meaning: Exactly 26 recorded sightings were independently verified, no more.
Commitment-preserving target specification: The verified count is 26 or greater; the source fixes the floor, not the ceiling.
Mechanism and hidden question. This case is primarily a semantic bound, not an optional implicature. The wording permits counts above the stated number even if an ordinary reader assumes it is the current minimum known to the speaker. Reports often use a lower bound while further evidence is arriving, as can happen with two similar species with different identification markers. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Suppose the final record in field notes, imagery and verification signatures increases the count by three. The ‘at least’ statement remains true; an ‘exactly’ translation becomes false. A target that drops the boundary operator is not merely shorter; it changes the data. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. The false ceiling can alter the interpretation of release the survey findings and produce overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Diagnostic 160: Marine species observation — Most, all, and majority without universal completion
Most recorded sightings were independently verified.
Tempting but unsafe target meaning: All recorded sightings were independently verified, without exception.
Commitment-preserving target specification: A majority of the relevant group satisfied the condition; universal completion is not asserted.
Mechanism and hidden question. ‘Most’ communicates a high proportion relative to the relevant set, but it does not entitle the translator to upgrade the claim to total coverage. Context may make the difference operationally crucial: an unverified minority could include two similar species with different identification markers. The translator should preserve the relevant noun-phrase restriction and avoid adding a numeric percentage unless the original supplies one. The translator must determine whether a stronger statement is entailed, merely suggested, or excluded. Treating those categories as one can turn a natural-looking sentence into a materially different instruction or report.
Adversarial diagnostic. Consider a group where almost every unit passes but one item fails. This situation can satisfy the source’s most-claim, while refuting the all-claim. Check whether the target remains true under that exception. A successful bilingual candidate should pass this test when read without the original text in view; otherwise the source-to-target comparison has found a commitment mismatch.
Practical consequence and repair. A universal overclaim can prematurely trigger release the survey findings and cause overstating the certainty of an uncertain population record. Compare the target against field notes, imagery and verification signatures. If the source does not settle the stronger proposition, keep the target open or seek clarification. If the genre requires a fully explicit decision, add the decision only after its owner authorises it; translation itself must not manufacture the missing evidence.
Part VI — From diagnostics to operational translation quality
A four-column commitment ledger
For every consequential scalar expression, create a compact ledger with four columns: source expression, asserted proposition, plausible contextual inference, and prohibited target additions. An example row might read: ‘some checked’ | at least one checked | in ordinary reporting perhaps not all | do not assert exactly how many unchecked. Another row might read: ‘only batch A cleared’ | batch A cleared and salient other batches not cleared | the distinction matters for a release decision | do not translate only as even. The ledger turns a vague editing conversation into a specific audit. The translator can point to the place where a target overstates certainty or changes a threshold.
The ledger is especially valuable when a translation passes through several hands. A first translator may correctly leave an upper bound open. A copy editor, looking only for readability, may replace the cautious formulation with a cleaner absolute. An automated model may produce yet another strongly worded paraphrase. Without a recorded commitment, the team has no stable reference point. The ledger does not eliminate editorial judgement; it identifies the boundary that judgement must not cross without new evidence.
Two-reader review instead of word-by-word approval
Use one reviewer with access to source and target, and a separate target reader who sees the target first. The bilingual reviewer records omissions, additions, numerical scope, focus and modality. The target-only reader reports what an ordinary competent reader would think was asserted and what was merely hinted at. When the two readings disagree, investigate the wording. This is not a mechanical pass/fail algorithm: target interpretation is context-sensitive, and a test reader may lack specialist background. However, disagreement is useful evidence of a hidden pragmatic problem.
A quality reviewer should avoid calling every non-literal difference an error. A different target construction may preserve the same relation between weaker and stronger alternatives more faithfully than the word-for-word candidate. When the text is high stakes, choose reviewers competent in the domain and the language pair, and verify the source brief. Translation competence includes knowing when the evidence is insufficient and requesting an author clarification.
Where standards help, and where they do not
The <a href=”https://www.iso.org/standard/80701.html” rel=”noopener”>ISO 5060:2024 translation-output evaluation standard</a> describes an approach to human evaluation of translation output, including errors and their weighting. It covers human translation and machine-translation output; it is not a magic dictionary of individual pragmatic solutions. A scalar drift that changes optionality into obligation or partial results into universal results should be classified by consequence, not only by the visual number of changed words. The separate <a href=”https://www.iso.org/standard/59149.html” rel=”noopener”>ISO 17100:2015 translation-services requirements</a> addresses translation-service processes and competences. Consult the current official standard and the applicable contract rather than treating this guide as a certification of any provider.
Standards do not decide whether an individual ‘or’ means exclusive choice in a particular dispute. They help organise responsibilities, specifications, revision and evaluation. A professional project still needs source-owner decisions for ambiguous policies, contextual judgments by language specialists, and evidence-aware quality assurance. For a small internal translation, a miniature version of the same discipline is useful: specify the audience, keep the original accessible, document uncertain terms, and retest the important commitments before release.
When localisation and message formatting complicate quantities
Grammar-sensitive interface strings add another problem. The Unicode <a href=”https://unicode.org/reports/tr35/tr35-numbers.html” rel=”noopener”>CLDR plural rules specification</a> defines locale-dependent plural categories such as one, few, many and other, using minimal-pair evidence about grammatical forms. These categories do not themselves mean the pragmatic English expressions ‘a few’, ‘few’, ‘most’ or ‘all’. A translation system that equates a CLDR category named ‘few’ with an English statement meaning ‘a small, disappointing number’ has confused grammatical message selection with scalar semantics.
The <a href=”https://cldr.unicode.org/translation/getting-started/plurals” rel=”noopener”>Unicode CLDR plural translation guidance</a> illustrates why natural strings require language-specific plural forms. The <a href=”https://www.w3.org/International/questions/qa-i18n/” rel=”noopener”>W3C explanation of internationalization and localization</a> reminds developers that localisation includes formatting and design concerns beyond substituting words. A robust application should keep quantity-bearing messages as complete, localisable units, preserve their placeholders and inequalities, and test rendered examples with numbers outside the expected range. Translators should never ‘repair’ a broken source-string architecture by inventing semantic bounds.
An audit rubric with observable evidence
Assertion accuracy
Can the reviewer state exactly what must be true without depending on a vague paraphrase? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Quantity bounds
Do the open and closed numerical boundaries survive, including a denominator and its restrictions? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Alternatives
Does the target preserve which stronger expressions would have been relevant in context? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Modality
Are permission, ability, possibility, obligation and certainty distinguished where the document requires? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Negation
Does negation apply to the same predicate and quantified group? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Focus
Do only and even select the same alternative or unexpected member? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Inference openness
Have defeasible inferences stayed defeasible unless the source explicitly commits to them? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Audience interpretation
Will a competent target reader draw a comparably safe practical conclusion? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Traceability
Can the team name the document, speaker, evidence and clarification supporting its decision? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Revision integrity
Did the final editorial pass preserve all earlier approved boundaries? Mark the result as supported, ambiguous or drifted rather than relying on a numerical score without an explanation. When ambiguous, quote the smallest relevant source span and record which alternative readings remain possible. A reviewer should be able to reproduce the decision using the same context and evidence; if the test cannot be repeated, the issue is not ready to be closed.
Part VII — Transfer workshops for new tasks and language pairs
Transfer workshop 1: A formal policy
Extract permission and obligation markers; write examples in which an authorised action does not happen. Decide whether the target permits the same non-performance. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 2: A research abstract
Compare all hedges, majority statements and minimum counts to the evidence tables. A polished abstract must never claim universal effects where the study measured only a subset. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 3: A customer interface
Test ‘or’ buttons by following both paths and checking whether the design truly bars combination. Ensure the target does not invent a choice restriction. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 4: A safety bulletin
Keep reported uncertainty distinct from verified facts. A false certainty can change behaviour even if every translated noun is correct. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 5: A classroom dialogue
Give the learner a some/all sentence followed by ‘in fact all’. Ask which interpretation is contradicted and which is merely surprising. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 6: A negotiation email
Underline only and even, locate the focus, and identify whom the language excludes or unexpectedly includes before editing tone. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 7: A literary narrative
Preserve an unreliable narrator’s cautious quantifiers rather than replacing them with omniscient certainty; ambiguity may be part of the voice. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 8: An audiovisual subtitle
Check whether reduced space has erased a necessary not, only, maybe or at least. Rework surrounding wording before sacrificing commitment. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 9: A product manual
Treat at least, no more than and exactly as quantitative conditions and test numeric counterexamples independent of word choice. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 10: A public dataset
Protect denominators, sampling frames and missing records; never present a sample percentage as a universal population finding. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 11: An automated chatbot
Audit whether the bot uses permissive or mandatory instructions, particularly when summarising a longer policy. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 12: A multilingual survey
Ensure some, none, all and other choices are comparable; changed response scales destroy comparability before statistical analysis. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 13: A legal-adjacent notice
Seek qualified review of obligations and exclusions. Do not infer binding intent from conversational heuristics alone. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 14: A translation-memory segment
Compare the new source against the old one for changes to only, even, a number or negation before accepting a 99-percent match. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 15: A learner’s reading journal
Record a source sentence, literal meaning, conversational inference and competing target paraphrases; revisit after a day without the source. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Transfer workshop 16: A publication summary
Restore one crucial uncertainty term after compression and ask whether the shortened summary remains true under the source’s counterexamples. Work through three passes. First, isolate the exact claim and the relevant alternative set. Second, draft a target-language expression without copying the source word order by habit. Third, invent a case where the tempting stronger interpretation fails while the source remains true, then ask a target reader to explain the result. Record what changed during revision, because a successful method should produce knowledge that survives the next document rather than a one-time fix.
Part VIII — A four-week practice sequence
Week one — Truth conditions
Create twenty pairs using some/all, at least/exactly and not all/none. For each pair, write a situation that proves the sentences differ. Avoid translating until you can say precisely what changes. Assess progress with explanations rather than raw speed. The useful outcome is not a high count of completed sentences but a growing ability to predict exactly when a fluent translation makes a stronger, narrower or more certain claim than the source. Retain examples where you changed your own mind after considering a counterexample; those often teach more than the obvious successes.
Week two — Alternatives and context
Use brief conversations. Change what the speaker knows, what the listener asked and which stronger alternatives are relevant. Observe how the same literal statement can support different inferences. Assess progress with explanations rather than raw speed. The useful outcome is not a high count of completed sentences but a growing ability to predict exactly when a fluent translation makes a stronger, narrower or more certain claim than the source. Retain examples where you changed your own mind after considering a counterexample; those often teach more than the obvious successes.
Week three — Target-language formulations
Work with a competent bilingual reader. Translate the examples naturally, then compare target-only interpretations. Collect false equivalents, misleading particles and focus-placement risks in a small termbase. Assess progress with explanations rather than raw speed. The useful outcome is not a high count of completed sentences but a growing ability to predict exactly when a fluent translation makes a stronger, narrower or more certain claim than the source. Retain examples where you changed your own mind after considering a counterexample; those often teach more than the obvious successes.
Week four — Real-world review
Select anonymised, permitted short documents from several genres. Run the twenty-four gates, create a commitment ledger, classify errors by practical consequence and repeat the full review after a separate editing pass. Assess progress with explanations rather than raw speed. The useful outcome is not a high count of completed sentences but a growing ability to predict exactly when a fluent translation makes a stronger, narrower or more certain claim than the source. Retain examples where you changed your own mind after considering a counterexample; those often teach more than the obvious successes.
Part IX — Reader questions and practical answers
FAQ 1. Does some always mean some but not all?
No. A common conversational inference is not a universal entailment. A contextually stronger reading may be natural, but the literal claim ordinarily remains compatible with all. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 2. What is scalar implicature in simple language?
It is an inference a listener may draw when a speaker uses a weaker expression despite a relevant stronger alternative. The chosen wording, context and assumptions about the speaker matter. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 3. Why does this affect translation accuracy?
A target can preserve dictionary words while altering what is definitely asserted, what is merely suggested and what remains possible. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 4. Is or always exclusive?
No. Or may allow both alternatives in its ordinary truth-conditional use. Some contexts, definitions and conventions impose exclusivity, which must be investigated rather than assumed. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 5. Does only mean even?
No. Only selects and excludes alternatives; even presents a case as notable or unexpected. Confusing them can reverse which cases appear eligible. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 6. Can at least be translated as exactly?
Not without a specific source-authorised restriction. At least allows a larger count; exactly excludes it. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 7. Why is not all different from none?
Not all permits positive cases alongside at least one exception, while none excludes every positive case in the relevant domain. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 8. Should a translator spell out every implied meaning?
No. Unnecessary explanation can assert an inference the source left open. Explain when the purpose requires it and evidence permits it. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 9. How do I test a quantifier translation quickly?
Use a counterexample: imagine every member succeeds, or one fails, or more than the stated minimum is found. Compare which statements remain true. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 10. Is this only about English?
No. The tests concern meaning across languages. Their surface realisation must be verified for each target language and genre. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 11. Does few mean zero?
Not necessarily. Few typically indicates a small number relative to expectation, and can be true with some positive cases. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 12. Why can a few sound more positive than few?
The expressions often frame the same rough number differently: a few highlights a positive available subset; few can signal scarcity. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 13. Can machine translation detect implicatures?
Sometimes systems produce natural interpretations, but neither grammatical fluency nor plausible paraphrase proves a correct pragmatic commitment. Human checks remain important. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 14. Is back-translation sufficient?
No. A reverse translation may reproduce vocabulary while missing target-language pragmatic expectations. Use direct logical tests and target-reader review. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 15. Can a word be literally right but pragmatically wrong?
Yes. The core problem of this chapter is that the target can invite an unjustifiably stronger inference. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 16. Is an implicature always cancellable?
Many conversational implicatures are defeasible, but classifications vary by phenomenon. Do not use one simple cancellation test to settle all meaning categories. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 17. What does focus change?
Focus changes the alternatives being compared. With only or even, moving the focus may change who is excluded or what is surprising. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 18. How are plural rules related to quantity?
Plural rules select grammatical message forms in particular locales. Their category names are not substitutes for semantic quantifiers such as most or at least. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 19. Should a warning or medical instruction rely on implicature?
Consequential instructions should minimise ambiguity and follow authorised source meanings. Use a qualified domain reviewer rather than guessing omitted obligations. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 20. How should a translator handle an ambiguous source?
Identify the competing readings and ask the responsible source owner when the difference can affect decisions. Do not disguise uncertainty with polished wording. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 21. How can a learner practise without a second language?
Practise same-language commitment paraphrases and counterexamples first. Then test real target-language formulations with competent readers. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 22. Can most mean all?
A most-statement need not exclude an all-true world under every theoretical treatment, but it does not itself license the translator to assert all. Preserve the source’s level of commitment. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 23. What about ‘probably’ and ‘possibly’?
They communicate different evidential strengths or likelihood judgements, depending on context. Avoid upgrading possibility to confident prediction. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 24. How does an editor avoid introducing mistakes?
Keep a commitment ledger and require justification before strengthening a hedge, broadening a group, adding only, deleting not or fixing an upper bound. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 25. Can prose style be changed freely?
Style can change when meaning and function survive. The translator may restructure grammar, but must preserve decisive exclusions, limits, qualifications and expected inferences. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 26. How do professional standards help?
Standards give useful structures for evaluation and process responsibilities, but do not replace contextual judgment about the intended meaning of a specific sentence. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 27. Does the method apply to multilingual UX?
Yes. Test alternative choices, plural forms, placeholders and numerical boundaries using rendered target messages and realistic user actions. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 28. How is this chapter different from vocabulary?
Vocabulary explains word meanings and acquisition. This chapter owns the translation decision about which asserted and implied boundaries survive cross-language transfer. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 29. How does this differ from negative-question replies?
A negative-question chapter solves response polarity and echo correction. This one solves stronger alternatives and the implications created by choosing weaker expressions. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
FAQ 30. What is the most useful final question?
Ask: could the target reader reasonably act on a stronger, weaker or more exclusive claim than the source author was willing to make? If yes, revisit the translation. For consequential work, always bring the surrounding source context back into view before treating the answer as a fixed rule.
Part X — A compact glossary for practical use
Assertion. The proposition put forward as true by an utterance under a given interpretation. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Entailment. A relation where one proposition’s truth guarantees another’s truth. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Scalar implicature. A context-dependent inference associated with selecting a weaker expression from salient alternatives. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Alternative set. The relevant competing statements or focal items a listener considers. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Exhaustivity. The interpretation that relevant alternatives outside a selected option are excluded. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Domain restriction. The contextually selected group over which a quantifier ranges. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Focus. The element contrasted with relevant alternatives, marked by syntax, particles or prosody. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Modal force. The difference between possibility and necessity, or permission and obligation, in a given modal use. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Epistemic modality. Language about knowledge, evidence, likelihood or certainty. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Deontic modality. Language about permission, requirement or normative status. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Defeasibility. The capacity of an inference to be withdrawn or overridden with additional information. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Lower bound. A minimum amount without a required ceiling. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Upper bound. A maximum amount without a required floor. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Scope. The structural region that an operator such as negation, only or a quantifier affects. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Target-only review. Review in which a competent target reader assesses the target text without leaning on source wording. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Commitment ledger. A record distinguishing source assertions, contextual inferences, uncertain readings and target prohibitions. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Counterfactual diagnostic. A carefully invented alternative scenario used to expose a meaning mismatch. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Pragmatic enrichment. Interpretive contribution beyond minimal compositional semantics under a context and theory. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Inclusive disjunction. An or-relation that permits both alternatives to hold. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Exclusive disjunction. An or-relation that allows exactly one of the relevant alternatives. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Quantifier drift. A translation error that changes number, coverage, bound, group or the relationship of a quantity to its domain. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Inference inflation. Treating a weak, defeasible suggestion as an explicit or necessary assertion. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Inference loss. A target version that fails to communicate an important contrast or pragmatic effect supported by source context. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Revision gate. A defined test that must pass before a translation is released. In a translation audit, use this term only after checking how the relevant construction works in the actual source and target contexts; the label alone does not choose the target wording.
Research and source-checking routes
The foundational and professional resources for further checking are Stanford Encyclopedia of Philosophy: Implicature, The Role of Alternatives in Language (Frontiers in Communication), ISO 5060:2024 on translation-output evaluation, ISO 17100:2015 on translation services, Unicode CLDR number and plural rules, CLDR Plurals & Units, and W3C internationalization/localization guidance. Check current editions and language-specific resources before applying any technical recommendation. Examples in this article are original teaching scenarios; none is a verified statement about a real organisation.
Next routes: keep the translation problem in the correct architecture
Return to Master Art of Translation | The Complete System for Moving Meaning Between Languages for the full source-analysis, drafting, revision and quality-assurance architecture. The immediately preceding Negative-Question Answer Fidelity System supplies a companion method for yes/no responses, echo correction and polarity reversals. Continue inside this translation tree when a known word becomes unreliable because it crosses languages.
Use the protected Vocabulary Learning Hub when the bottleneck is learning, remembering or distinguishing the meanings of quantity words and phrase families before translation. Use the existing How English Works owner when the task is English-side interpretation, sentence form, discourse and comprehension without a cross-language decision. These sibling routes answer different questions; none of their owners needs to be rewritten to support the present method.
Final principle: preserve what can be true, not just what sounds plausible
The decisive question in scalar translation is not ‘Which target expression resembles the source word?’ It is ‘What can remain true of the world if this sentence is true, and what does the intended reader reasonably infer in this context?’ A faithful version should not convert permission into obligation, possibility into certainty, a subset into a universal set, a negated universal into none, or unexpected inclusion into exclusion. It should not forcibly spell out an inference merely because the target sentence is easier to read that way. Skilled translation preserves the limits of a speaker’s knowledge as carefully as it preserves the facts the speaker knows. When that discipline becomes routine, fluency serves meaning rather than replacing it.