Master Art of Translation | The Comparative Frame Fidelity System — How Baselines, Denominators, Reference Groups and Thresholds Keep Claims Exact

How do you translate comparative statements accurately when a source says one result is faster, a rate is lower, a group is more likely, an outcome improved by 12%, or a limit is no more than five? Comparative translation is not simply translating “more,” “less,” “better” or “than.” It means preserving the baseline, reference group, denominator, measurement, period, conditions, direction and degree of certainty that make a comparison true. This article offers a practical comparative-equivalence method for translators, localization teams, editors and reviewers working between any languages.

Readers searching for how to translate comparisons, percent versus percentage points, translation of statistics, comparative adjectives across languages, translating thresholds such as at least and up to, or avoiding misleading translated charts often encounter a concealed problem. A target sentence can keep every visible number while changing what the numbers compare. “Thirty per cent lower than last year” is not “thirty percentage points lower,” “twice as likely” is not “two percentage points more likely,” and “best among tested models” is not “best available.” A translation may be perfectly fluent and still invite the wrong decision.

The Comparative Frame Fidelity System establishes a repeatable method for preserving comparative claims: extract both sides of the comparison, identify the measure and its population, freeze the baseline and relevant time interval, distinguish relative from absolute changes, keep ranking and threshold qualifiers attached, reconstruct the relationship naturally in the target language, and test it against a concrete counterexample. The method applies to literary and everyday language as well as reports, interfaces, policies and research, without pretending that a fluent English paraphrase is a certified translation into every other language. It is a specialist child of the Master Art of Translation architecture, rather than a replacement for vocabulary learning or English grammar.

Architecture reference: EDKSG-TRANS-MASTER-WORLD-1420. This chapter owns the cross-language fidelity of comparative frames. It does not teach statistics as an independent discipline, replace specialist legal or scientific translation, or repeat the earlier figurative-frame chapter. All examples are invented English-language meaning glosses intended to show what must survive when a real language pair is selected.

The quick answer: a twelve-field comparison card

Compared entity A: what item, group, activity or outcome is being described?
Comparator B: what other item, group, past value, target or hypothetical world supplies the reference?
Measure: faster in time, lower in cost, higher in concentration, more likely by probability, or something else?
Base and denominator: what does the percentage, rate, index or average divide by?
Direction: more, less, higher, lower, nearer, farther, equal, not inferior or incomparable?
Magnitude: exactly how much, including units and whether the change is absolute or relative?
Time frame: which interval, reference year, exposure window, follow-up period or reporting edition?
Population and selection: which participants, products, observations or exclusions enter the comparison?
Scope: does only, all, some, at least, no more than or among tested alternatives restrict the assertion?
Uncertainty: is the source definitive, estimated, provisional, statistically uncertain or illustrative?
Authority: who approved the comparison, evidence, definition and published source version?
Reader consequence: what would the target reader conclude or do if the comparison were slightly wrong?

Contents and reading route

The hidden problem: an invisible comparator
How comparison frames carry meaning
The 24-pass repeatable translation method
Failure modes and immediate repairs
Worked comparative-translation diagnostics
Transfer workshops and practice
Frequently asked questions
Research, limitations and next routes

The hidden problem: a number does not disclose what it compares

Comparison is a relation, not a decorative adjective. In an ordinary conversation, somebody may say that the new route is shorter; a listener supplies an understood earlier route. In a report, that same missing comparator can invalidate a claim. Source writers frequently expect readers to recover comparators from headings, tables, methods sections, prior sentences or shared practical knowledge. Translators working segment by segment encounter a grammatical adjective with insufficient context and fill the gap with a reasonable-looking guess.

A translator who changes “the response rate doubled” into “twice as many people responded” may alter the measured object. A rate may be the number of responses divided by invitations, eligible participants, delivered messages or completed contacts. Twice the rate does not prove twice the count if the denominator changed. The problem is not arithmetic competence alone. It is the loss of an argument slot in the source claim: rate of what, among whom, over what period, relative to what baseline?

Ranking statements hide a different trap. “The most accurate of the three tested systems” contains a closed reference class. Dropping “of the three tested systems” creates an open-world superlative. A reader may take it to mean the best system anywhere. Adding that sweeping promise is not legitimate translation. The same error appears when “one of the highest-scoring groups” becomes “the highest-scoring group,” or when a tie becomes a sole winner.

Boundaries often matter more than central values. “At least 50” includes 50; “more than 50” excludes it. “Below 50” excludes 50, while “50 or below” includes it. A localized interface that changes a comparison operator may unintentionally change eligibility, price, access or safety. A reviewer should write down the boundary case before approving a natural-sounding target sentence. If the boundary case changes, meaning changed.

The same numerical value can indicate different evidence. An estimated value of 10 with a wide uncertainty interval is not the same claim as a directly counted 10. A 20% increase in a rare risk may be a very small absolute change. A 20-percentage-point increase is an additive difference between proportions. Translation does not authorize the writer to choose whichever framing feels more vivid. If the reader needs an explanation, label the explanation as such and preserve the source assertion separately.

Some comparisons are not numerical at all. “This interpretation is less accusatory” compares pragmatic force, not character count. “Her second apology was warmer” compares interpersonal tone, not an approved temperature scale. “The new design feels more open” is a reported or aesthetic judgement, not proof of a measured performance gain. The translation must first classify whether the claim is measurable, perceptual, argumentative or relational. Otherwise an expressive adjective may be elevated into a quantitative fact.

Cross-language grammar can compel choices that the source leaves implicit. A target language may require explicit comparative particles, adjective agreement, scalar constructions, classifiers or natural word-order changes. Such changes can be faithful. The invariant is not grammatical shape; it is the comparison relation that a competent reader reconstructs. A target should say what the source supports, no more and no less. If grammatical naturalness forces a distinction absent from the source, the missing distinction is a query, not permission to invent it.

Comparison claims are especially vulnerable to the reporting environment. A chart label may define the denominator but disappear in an exported summary. A translated headline may lose the footnote “among respondents.” An accessibility label may verbalize a bar but not its reference year. A promotional excerpt may detach “estimated” from a number. The same claim must be reviewed at the level of the complete reader experience, not just in an isolated segment window.

Three questions diagnose most failures. First, what would make the source claim true or false? Second, what would make the target claim true or false? Third, can you imagine a world in which one is true and the other false? If you can, you have found a semantic divergence requiring repair or an explicitly authorized adaptation. This counterexample test is more reliable than saying that two expressions sound approximately equivalent.

Do not use a bare numerical error score as the whole quality decision. ISO 5060:2024 addresses human evaluation of translation output using error types and configured ratings, and ISO 11669:2024 addresses translation-project specifications and risk analysis. Neither should be described as certifying the claims on this webpage. Their practical relevance is that a comparison error needs an agreed task specification, a defensible evaluation rule and escalation proportional to consequence. A minor stylistic divergence and a threshold reversal are not interchangeable.

The mechanism: thirty-six comparison invariants

1. Declared comparator

A translation of “improved” must recover the earlier version, previous year, control group, plan or expectation that supplies the comparison.

When no comparator is stated, retain the omission if the target language permits it; otherwise ask for contextual evidence.

A useful working annotation for declared comparator records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps declared comparator a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

2. Implicit comparator

An author can assume readers remember a comparison class from earlier discourse.

The translator must trace that class instead of selecting the nearest convenient noun.

A useful working annotation for implicit comparator records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps implicit comparator a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

3. Comparison dimension

“Better” may describe accuracy, cost, durability, enjoyment or accessibility.

Specify the dimension only when the source context establishes it, not when the translator simply prefers a measurable interpretation.

A useful working annotation for comparison dimension records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps comparison dimension a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

4. Difference versus ratio

An additive gap and a multiplicative ratio have different truth conditions.

Use a small numerical counterexample to test whether the target preserved the operation.

A useful working annotation for difference versus ratio records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps difference versus ratio a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

5. Relative percentage

“20% lower” usually expresses a difference relative to a reference value.

Record the original reference and calculate a sample check before rewriting.

A useful working annotation for relative percentage records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps relative percentage a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

6. Percentage points

A movement between two percentages is not automatically a relative percent change.

Keep the words “percentage points” or their accurate target equivalent when that is the actual unit.

A useful working annotation for percentage points records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps percentage points a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

7. Denominator identity

A rate represents events or cases relative to a stated eligible count, period or exposure.

Changing the denominator changes the reported phenomenon even if the numerator survives.

A useful working annotation for denominator identity records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps denominator identity a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

8. Numerator identity

Responses, completed cases, attempts, people and events are not interchangeable counts.

Never promote a count of actions into a count of individuals without evidence.

A useful working annotation for numerator identity records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps numerator identity a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

9. Population frame

“Among surveyed users” and “among all users” refer to different collections.

Keep the inclusion rule and recruitment or observation limits attached.

A useful working annotation for population frame records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps population frame a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

10. Baseline year

An index, change or growth estimate may use a nominated year or earlier observation.

Preserve the year, period and rebasing note; do not assume the latest date is the reference.

A useful working annotation for baseline year records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps baseline year a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

11. Reference period

Weekly, monthly, annual and lifetime rates describe different time exposures.

Transfer duration and aggregation along with the quantity.

A useful working annotation for reference period records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps reference period a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

12. Comparator status

A historical value, target, peer average and hypothetical counterfactual play different epistemic roles.

Identify which one the source invokes and keep that role apparent.

A useful working annotation for comparator status records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps comparator status a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

13. Lower-is-better reversal

A decrease in error can be improvement while a decrease in survival would not be.

Preserve polarity relative to the measured objective, not an adjective like “down” alone.

A useful working annotation for lower-is-better reversal records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps lower-is-better reversal a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

14. Rank domain

“Top” requires a ranked collection and a stated criterion.

An accurate target cannot silently broaden a small comparison set into the world.

A useful working annotation for rank domain records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps rank domain a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

15. Tie and partial order

Equal ranks and incomparable measures can produce multiple nonunique leaders.

Do not convert “joint first” into “sole winner” or incomparable dimensions into one league table.

A useful working annotation for tie and partial order records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps tie and partial order a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

16. Ordinal versus cardinal

A rank of second and a score of two are different data types.

Translate label and value according to their roles; ranking position is not distance.

A useful working annotation for ordinal versus cardinal records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps ordinal versus cardinal a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

17. Threshold inclusivity

At least, more than, up to, fewer than and no more than differ at the endpoint.

Test the boundary itself, not just a typical value above or below it.

A useful working annotation for threshold inclusivity records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps threshold inclusivity a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

18. Range endpoints

Intervals may be inclusive, exclusive, approximate or open-ended.

Translate the inclusion convention and units, using a note if typography alone is insufficient.

A useful working annotation for range endpoints records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps range endpoints a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

19. Equality frame

“As effective as” need not prove statistically equivalent performance in a trial.

Preserve the source’s claim type and do not upgrade ordinary equality language to formal equivalence certification.

A useful working annotation for equality frame records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps equality frame a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

20. Noninferiority

A formal noninferiority conclusion depends on a defined margin and design.

Do not paraphrase it as superiority or absolute equality.

A useful working annotation for noninferiority records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps noninferiority a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

21. Approximation

About, roughly, nearly, up to and at least encode different uncertainty and bounds.

Do not erase the distinction for rhetorical force.

A useful working annotation for approximation records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps approximation a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

22. Distribution versus average

The mean can change while many individual outcomes do not, and vice versa.

Avoid turning a summary statistic into a property of every member of the population.

A useful working annotation for distribution versus average records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps distribution versus average a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

23. Median versus mean

A median splits ordered observations; an arithmetic mean sums and divides.

Do not treat them as interchangeable equivalents for “average” without source metadata.

A useful working annotation for median versus mean records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps median versus mean a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

24. Rate versus risk

Events per exposure and probability over an interval are not identical.

Check denominator and follow-up window before writing a reader-friendly gloss.

A useful working annotation for rate versus risk records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps rate versus risk a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

25. Risk ratio versus odds ratio

These ratios use different denominators and may diverge materially.

Never replace a technical ratio label because it sounds easier to understand.

A useful working annotation for risk ratio versus odds ratio records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps risk ratio versus odds ratio a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

26. Change versus difference

A cross-sectional gap and a within-group change have different comparison axes.

Preserve whether the same entity is compared across time or separate entities at one time.

A useful working annotation for change versus difference records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps change versus difference a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

27. Causation versus contrast

A difference between groups does not alone establish what produced it.

Translate observational comparisons without attaching unauthorized causal verbs.

A useful working annotation for causation versus contrast records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps causation versus contrast a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

28. Sample versus universe

A study finding in a sample may not generalize to an entire population.

Keep sample qualifiers and uncertainty distinct from the target audience.

A useful working annotation for sample versus universe records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps sample versus universe a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

29. Controlled conditions

“Under test conditions” narrows the applicability of a comparison.

Do not remove the environment or protocol when converting a sentence into a headline.

A useful working annotation for controlled conditions records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps controlled conditions a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

30. Metric direction

Scores, losses, delays, costs, satisfaction and uncertainty use different scales and goals.

Preserve the name and unit of the measure and explain why higher or lower matters only with source support.

A useful working annotation for metric direction records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps metric direction a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

31. Missing data

The absence of a comparator may result from unmeasured, withheld or inapplicable values.

Never interpret missing as zero or create a confident comparison from a blank cell.

A useful working annotation for missing data records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps missing data a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

32. Unit normalization

Per person, per kilometre, per hour and per transaction normalize numbers differently.

Reconstruct the expression as a quotient and test the denominator before localizing the unit format.

A useful working annotation for unit normalization records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps unit normalization a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

33. Precision and rounding

A rounded displayed number can hide a small but important threshold gap.

Keep the original precision policy and do not assert more significant digits than the source warrants.

A useful working annotation for precision and rounding records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps precision and rounding a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

34. Scoping qualifiers

Only, among, for eligible cases and in this analysis limit a claim’s comparison universe.

Put qualifiers close enough to their target expressions that their scope stays clear.

A useful working annotation for scoping qualifiers records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps scoping qualifiers a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

35. Statistical uncertainty

Confidence intervals, error bars and model assumptions qualify comparisons.

Do not turn “no clear difference detected” into “proven identical.”

A useful working annotation for statistical uncertainty records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps statistical uncertainty a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

36. Source and target layers

A translated comparison may span headline, chart, alt text, callout and supporting table.

Audit the complete reader view for consistent denominator and reference class.

A useful working annotation for source and target layers records the source phrase, the nearby evidence that establishes the comparison, the target reading at risk, and a counterexample showing how the meaning could drift. Rather than checking whether a target word looks like a dictionary match, ask a reviewer to reconstruct both alternatives and explain the reference class from the translated passage alone. If that reconstruction needs information visible only in the source, the target’s surrounding material is incomplete.

Editorial decision: resolve a missing fact with the source owner or an approved reference, not a translator’s speculation. If the source itself is vague, maintain that degree of vagueness where the target permits it; if an action depends on the ambiguity, hold the affected phrase for clarification. This keeps source and target layers a meaning-preservation problem rather than an excuse to rewrite the author’s evidence.

The repeatable method: twenty-four passes from source to target

Pass 1: Freeze source identity

Record exact document, version, date, figure and section; a good translation of an obsolete comparison is still the wrong release.

Required output: A source snapshot with stable identifier. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Review method: test the draft against a concrete imagined reader choice. If the target sounds natural but permits a reader to select a different comparator, denominator, threshold or degree of certainty, the pass is not complete. Record the disagreement as an evidence question or meaning issue; do not resolve it by replacing the precise sentence with a smoother but broader promise.

Pass 2: State reader question

Write the decision the reader expects to make: which is larger, by how much, among which items, or inside which limit.

Required output: One plain-language reader question. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 3: Extract the compared entities

Name item A and item B, even when B is implied by a pronoun, earlier paragraph or diagram.

Required output: Two entity labels and referent traces. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 4: Identify the measure

Decide whether the text compares probability, amount, frequency, ranking, speed, experience, cost or another property.

Required output: A measure statement with units if defined. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 5: Capture the baseline

Locate past period, reference condition, control, plan, target or theoretical comparator.

Required output: A baseline record with source evidence. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 6: Capture numerator and denominator

For rates and percentages, record the counted events and the eligible observations or exposure.

Required output: A quotient specification. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 7: Classify mathematical relation

Mark additive change, relative change, ratio, ordinal comparison, inequality, equivalence, association or rhetorical contrast.

Required output: A relation type and a simple truth test. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 8: Preserve population inclusion

List eligibility, exclusions, sampled units and selection method when they qualify the assertion.

Required output: A reference-class note. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 9: Preserve period and exposure

Mark start and end dates, recurrence, follow-up, per-unit exposure and whether times overlap.

Required output: A temporal measurement card. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 10: Lock boundary operators

Write each less-than, at-least, at-most or between condition in plain language; test equality at endpoint.

Required output: A threshold table. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 11: Preserve direction of good and bad

Identify what makes improvement meaningful; lower latency can be good while lower coverage is not.

Required output: A metric-polarity note. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 12: Mark evidence status

Separate observation, model estimate, forecast, opinion, hypothesis and certified conclusion.

Required output: An evidential-status label. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 13: Identify uncertainty

Record approximation, confidence interval, margin, missing evidence or small sample language.

Required output: A disclosure note. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 14: Map qualifiers and scope

Attach only, among, in this test, for those eligible and other limiting language to the correct proposition.

Required output: A qualifier-to-claim diagram. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 15: Read surrounding graphics

Check axes, legend, footnotes, accessibility description, table headings and unit labels.

Required output: A cross-asset inventory. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 16: Write a source-language expansion

Restate the comparison in an unambiguous diagnostic sentence without pretending that expansion is publication copy.

Required output: A reference paraphrase. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 17: Construct target alternatives

Produce at least two natural target options if the mapping is difficult; explain what each preserves or loses.

Required output: A comparison of candidates. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 18: Run numeric counterexamples

Use imaginary values that distinguish ratios from differences, rates from counts and inclusive from exclusive bounds.

Required output: An explicit failing counterexample. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 19: Run reference-class counterexamples

Imagine the claim applied to outsiders, untested cases or a different time period; ask whether the target excludes them.

Required output: A class-boundary test. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 20: Review with target-reader task

Show only the target plus required chart/context and ask readers what they would compare or act upon.

Required output: A reader-interpretation record. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 21: Perform independent bilingual revision

A reviser compares source and target including evidence and visuals; fluency alone is insufficient.

Required output: An annotated meaning check. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 22: Assign risk and ownership

Escalate high-consequence thresholds, eligibility, medical, safety, financial or legal claims to the competent owner.

Required output: A named approval decision. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 23: Package the release

Check final layout, localized numerals, decimal separators, alt text, transcript, excerpts and metadata.

Required output: A cross-channel parity check. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Pass 24: Maintain a regression test

Preserve a boundary case and comparator definition so a future update cannot reintroduce the same defect.

Required output: A reusable test case and change log. Before moving on, a second person should be able to point to the source evidence, repeat the comparison without changing its scope, and identify a plausible misleading translation that this pass is intended to prevent.

Failure modes: what goes wrong and how to repair it

Failure 1: Wrong comparison class

A claim among tested products becomes a claim about every product.

Restore the tested set and put the restriction near the superlative.

Diagnostic discipline: find the exact source phrase, write down what a target reader would now believe, and construct a minimal counterexample that makes that belief diverge from the original. Preserve existing source evidence rather than supplementing it with plausible but unverified facts. When the consequence involves eligibility, compliance, payment, treatment or safety, hold the claim for an appropriately qualified source owner before release.

Failure 2: Relative change mistaken for points

A 10% rise from 50% is rendered as a 10-point rise to 60%.

Keep the reference percentage and distinguish the resulting 55% from 60%.

Failure 3: Percentage points mistaken for relative growth

A movement from 20% to 25% is described as five per cent growth.

State five percentage points, or the correct relative change of 25% if separately authorized.

Failure 4: Rate replaced by count

A doubled response rate becomes twice as many responses.

Restore eligible invitations or the relevant base.

Failure 5: Average turned into universal

The average duration falls, so a target claims every user is faster.

Keep summary language and avoid every/all.

Failure 6: Median silently replaced

A skewed distribution’s median is labeled the mean.

Confirm statistic type and use the approved label.

Failure 7: Reference year lost

An index shows 130 against a defined base but target claims 130% growth over last year.

Restore the base year and index convention.

Failure 8: Time periods mixed

A quarter-on-quarter comparison becomes year-on-year.

Reattach the original interval.

Failure 9: Population broadened

“Among respondents” becomes “people generally.”

Maintain sampling restriction and uncertainty.

Failure 10: Experimental condition erased

Performance under a specified load is advertised as unconditional.

Restore operating environment and workload.

Failure 11: Threshold inclusive/exclusive swap

“At least 18” becomes “older than 18.”

Check the exact age 18 boundary.

Failure 12: Upper limit misread

“Up to five” is interpreted as exactly five.

Preserve maximum rather than guarantee.

Failure 13: Minimum confused with recommendation

“No fewer than three checks” becomes a suggestion of three.

Keep mandatory minimum where the source makes it binding.

Failure 14: Rank and score confused

Rank 1 is interpreted as score 1, although lower rank means better placement.

Label ordinal ranking separately.

Failure 15: Ties erased

Joint first is rendered as sole leader.

Restore equal position.

Failure 16: Superiority fabricated from noninferiority

A noninferiority result is described as superior performance.

Preserve the formal conclusion and its limitations.

Failure 17: No difference mistaken for equality

Failure to detect a difference becomes proof that outcomes are identical.

Keep the scope of the evidence.

Failure 18: Correlation promoted to cause

A higher observed rate is said to be caused by a particular intervention.

Remove unauthorized causal explanation.

Failure 19: Metric swapped

Response time improvement is described as increased throughput.

Retain the measured dimension; they need not move together.

Failure 20: Negative scale inverted

A lower error score is translated as a worse result.

Check what higher and lower mean for the specified metric.

Failure 21: Zero used for missing

A missing table entry becomes a measured zero.

Mark missing/inapplicable separately.

Failure 22: Rounding hides a boundary

A precise value just below a requirement is rounded and announced as compliant.

Use the governing unrounded value for the decision.

Failure 23: Uncertainty stripped

An estimate is stated without its qualifier.

Restore uncertainty status near the quantity.

Failure 24: Confidence interval reversed

An interval estimating effect uncertainty is reported as a range containing individual outcomes.

Describe the estimate and interval’s actual role.

Failure 25: Odds ratio relabeled

A technical odds ratio is called a risk ratio.

Retain ratio type or add a clearly separated nontechnical explanation.

Failure 26: Denominator changes silently

Per-visit rates are expressed as per-person rates.

Track unit and denominator.

Failure 27: Currency or unit normalization hidden

Two costs with different time coverage are presented as directly equal.

Normalize with stated assumptions or preserve the mismatch.

Failure 28: Category boundary lost

A statement applying to completed cases becomes all started cases.

Restore completion criterion.

Failure 29: Exception stripped

“All except two controls” becomes “all controls.”

Keep the exception’s membership and count.

Failure 30: Indirect comparison overstated

A model estimates A relative to C and B relative to C, but the target states a direct A versus B trial.

Protect the comparison path.

Failure 31: Chart legend conflicts

Translated label reverses series colours or time-order legend.

Verify against the actual plotted series.

Failure 32: Footnote detached

A table caption loses a qualification visible only in the source figure note.

Carry the footnote or keep the claim appropriately restricted.

Failure 33: Comparative negation lost

“Not less reliable” becomes “less reliable.”

Test polarity and equality cases.

Failure 34: Double negative simplified incorrectly

A phrase excluding “not uncommon” becomes “rare.”

Restate the allowable range before smoothing.

Failure 35: “Twice as much” ambiguity

An increase by two times is confused with a final total of two times.

Require explicit equation or source clarification.

Failure 36: “More than half” compressed

A strict majority becomes “half or more.”

Test an exact 50% case.

Failure 37: “Every other” confused

A cadence of alternate days becomes every two days starting from a different reference.

Preserve schedule anchor and permitted interpretation.

Failure 38: Rank among available candidates

A lowest-priced eligible option becomes globally lowest-priced.

Keep availability and eligibility qualifiers.

Failure 39: Comparator changes in a quotation

A speaker compares with “our earlier version,” but a reporter changes to a competitor’s version.

Retain quoted speaker and referential context.

Failure 40: Statistical significance overstated

A significant difference is reported as practically important.

Preserve the statistical result without inventing decision magnitude.

Failure 41: Practical value understated

A large operational change is dismissed because the formal statistical test is uncertain.

Carry both effect magnitude and evidential limit.

Failure 42: Target condition omitted

“Better at low loads” becomes “better” everywhere.

Reattach the condition before publication.

Failure 43: Visual scale read as meaning

Truncated axes exaggerate visual gap and the translation adds dramatic language.

Trust values and axis annotations, not chart impression.

Failure 44: Comparable cohorts confused

An older cohort and newer cohort differ in composition; target implies same participants.

Describe cohorts separately.

Failure 45: Cumulative versus annual

An accumulated total is reported as annual flow.

Recover the measurement period.

Failure 46: Inverse percentage error

A change in failure from 2 to 1 per thousand is called a 1% decrease.

Calculate the correct absolute and relative interpretations.

Failure 47: Conditional rank

A top performer among low-cost options becomes top performer overall.

Preserve the conditioning category.

Failure 48: Preference confused with accuracy

Survey preference is described as objectively higher quality.

Label the measured subjective response.

Failure 49: Sample-size omission

A tiny comparison loses its sample size and becomes a universal claim.

Retain the study frame and its uncertainty.

Failure 50: Projected versus observed

An estimated future reduction becomes an achieved reduction.

Keep prediction modality and date.

Failure 51: Rebased index undocumented

A chart is rebased to 100 but target presents 100 as a measured physical unit.

Label index base and interpretation.

Worked comparative translation diagnostics: invented cases, explicit tests

Every case below is a constructed meaning test, not a certified bilingual translation or a factual claim about a real company, programme, study or person. Each illustrates how a target-language expression could alter a source comparison despite sounding fluent. The reference wording is the semantic objective a translator must convey naturally in the actual target language; it is not an instruction to copy English word order.

Diagnostic 001: Newsletter response rate

Source meaning and setting. A bulletin reports that the response rate grew from 10% to 15% among delivered invitations. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says responses increased by five per cent. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain five percentage points and a 50% relative increase only if separately explained. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Find the invariant first. The source supplies a particular comparison relation. The unsafe line does not simply choose a different synonym; it would make the text true in a different set of circumstances. That is the core definition of this error. A reviewer should list both compared entities, the measure, the reference base and every qualifier, then point to the target words that carry each item.

Worked diagnostic and acceptance test. Test 100 delivered messages: counts rise from 10 to 15, showing why the measures differ. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Use a target-only reading test: hide the source, present the completed target with its ordinary supporting context, and ask which comparison a careful reader would calculate or believe. If the answer differs, the target needs a semantic repair even if its sentence structure is elegant. This exercise is transferable: repeat it with a different source and target language, avoiding automatic assumptions about one language’s comparative particles, number formatting or adjective order. The translation succeeds only when the relation survives the language change.

Diagnostic 002: Warehouse errors

Source meaning and setting. The error rate fell by a quarter from an earlier rate of 8 per hundred completed orders. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says errors fell by 25 percentage points. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. The new rate is six per hundred, two percentage points lower than eight. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Separate data from interpretation. The first task is to distinguish an observed or reported quantity from an editorial inference about it. A comparison may be correct within its stated data frame and false when generalized to a new population or time. Preserve the exact evidence and then choose grammatical forms that sound natural to the intended audience.

Worked diagnostic and acceptance test. Check both 8-2 and 8×0.75. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Ask a second reviewer to construct a different but source-consistent data table that would contradict the unsafe wording. Such a counterexample proves the phrases are not equivalent; it avoids an argument about which phrasing merely feels more compelling. This exercise is transferable: repeat it with a different source and target language, avoiding automatic assumptions about one language’s comparative particles, number formatting or adjective order. The translation succeeds only when the relation survives the language change.

Diagnostic 003: Reading speed

Source meaning and setting. The new workflow is 20% faster at a fixed passage length under an agreed timing definition. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims the elapsed time is 20% shorter without verifying what faster means. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve the original speed definition or explicitly calculate time from the specified measure. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Treat the comparison as an equation. The words in a comparative claim can often be expanded into a structured relationship: A and B, measure M, operator, denominator, time and restrictions. That formal expansion is for diagnosis, not for public display. Once the invariant is visible, the target language may reorder, inflect or paraphrase freely without losing the essential comparison.

Worked diagnostic and acceptance test. If speed is 1.2 times baseline, task duration is about five-sixths, not four-fifths. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Check both a common numerical value and a boundary or exceptional value. Easy values sometimes conceal ambiguity; the equality point, an excluded group or a zero denominator usually exposes an incorrect operator or assumed population. This exercise is transferable: repeat it with a different source and target language, avoiding automatic assumptions about one language’s comparative particles, number formatting or adjective order. The translation succeeds only when the relation survives the language change.

Diagnostic 004: Battery capacity

Source meaning and setting. An equipment sheet states capacity is 10% higher than the previous model at the same test condition. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says battery life is 10% longer in all uses. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep the tested capacity comparison and its condition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Protect the reference class. Every “more,” “less,” “first,” “highest” or “unchanged” refers to some set of relevant alternatives. Readers use surrounding context to infer that set. If translation widens or shrinks it, a statement can shift from modest and accurate to sweeping and unjustified, despite retaining every salient noun.

Worked diagnostic and acceptance test. A higher capacity does not fix identical runtime under a different power draw. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Test the statement in a hypothetical new context: introduce another candidate, another period or a missing observation. Would the source still stand while the target fails? If so, restore the explicit or recoverable original limitation. This exercise is transferable: repeat it with a different source and target language, avoiding automatic assumptions about one language’s comparative particles, number formatting or adjective order. The translation succeeds only when the relation survives the language change.

Diagnostic 005: Travel duration

Source meaning and setting. A brochure says Route A takes twelve minutes less than Route B under the observed timetable. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says Route A is twelve per cent quicker. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve the twelve-minute absolute difference. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Preserve evidence strength. A grammatical comparison may express a measurement, model estimate, perception, forecast or provisional inference. A translator should not strengthen one kind of statement into another by removing hedges, adding precision, or describing an association as a proven mechanism.

Worked diagnostic and acceptance test. Compare an 80-minute route with a 20-minute route to see why the percentage would vary. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

For review, separately mark what the source asserts, what its evidence suggests and what remains uncertain. The target should let its reader make those distinctions without consulting a source-language appendix unavailable to them. This exercise is transferable: repeat it with a different source and target language, avoiding automatic assumptions about one language’s comparative particles, number formatting or adjective order. The translation succeeds only when the relation survives the language change.

Diagnostic 006: Energy per task

Source meaning and setting. A prototype uses 30% less energy per completed task than the reference under an identical workload. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says total electricity consumption fell by 30%. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep the per-task denominator and workload. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Total tasks may rise enough to increase total energy despite lower energy per task. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 007: Cost per attendee

Source meaning and setting. An event budget shows lower spending per attendee despite a higher total expense. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the event was cheaper overall. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Translate the normalized cost and distinguish total expense. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Vary attendance from fifty to two hundred while preserving total budgets. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 008: Water use intensity

Source meaning and setting. A facility measures litres per unit produced rather than total litres. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says overall water extraction decreased. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep production-normalized intensity and avoid claiming absolute volume change. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Increase output sufficiently to show intensity declining while total water rises. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 009: Survey satisfaction

Source meaning and setting. Seventy per cent of respondents preferred version A, out of respondents who completed both trials. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says seventy per cent of all customers preferred A. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Restore the respondents who completed both trials. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Nonrespondents may have unknown preferences. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 010: Course completion

Source meaning and setting. The completion percentage is calculated from learners who began the programme, not those merely enrolled. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target labels it completion among all registrants. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Use the source entry criterion. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Test cases where many registrants never began. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 011: Clinic queue

Source meaning and setting. The median wait decreased from twenty to fifteen minutes in observed appointments. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says average waiting time fell by five minutes. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep median and observed appointments. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A single extreme delay can change the mean without moving the median. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 012: Product scores

Source meaning and setting. One item is ranked first out of five evaluated items under one rubric. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it the highest-scoring item worldwide. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Restrict the conclusion to the five-item evaluation and named rubric. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Introduce an untested sixth item with a higher score. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 013: Shared lead

Source meaning and setting. Two designs tied for the highest tested usability score. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says one design won the test outright. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve a joint first place. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. If both score 92, neither is the sole leader. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 014: Cheapest qualified

Source meaning and setting. An offer is the least expensive among currently available qualifying packages. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it is the cheapest package anywhere. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep both availability and qualification. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Imagine an excluded package with a lower price. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 015: Top quartile

Source meaning and setting. A researcher reports a result within the top quarter of a specified distribution. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes the item as ranking number one. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep quartile membership without implying the maximum. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Many results can share the upper quartile. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 016: Percentile

Source meaning and setting. A score is at the 80th percentile in the stated reference sample. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says 80% of questions were correct. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve percentile and sample. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A percentile indicates relative standing, not raw accuracy. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 017: First of two

Source meaning and setting. A prototype ranks first in a two-candidate pilot. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes a broad tournament champion. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep the pilot’s small reference class. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The candidate may not have competed with any other alternatives. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 018: Rank score inversion

Source meaning and setting. A quality-control table assigns rank 1 to the best result but higher defect scores to worse results. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls the result with score 1 best without checking which column is rank. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep rank and metric score distinct. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Test a case with rank 1 and defect score 12. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 019: Historic baseline

Source meaning and setting. An index value of 115 is measured relative to a base period assigned 100. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says a 115% increase since last month. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep index level and reference period. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Index 115 indicates 15% above its designated base, not 115% growth. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 020: Rebased series

Source meaning and setting. Two published graphs both say index 120 but use different base years. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target combines them as a direct level comparison. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. State the reference period of each; rebase or do not directly compare. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The same physical series can show different index values under different bases. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 021: Quarterly growth

Source meaning and setting. A report compares the current quarter with the immediately preceding quarter. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says growth was year-on-year. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep quarter-on-quarter basis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A seasonal cycle can make the two measures move oppositely. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 022: Fiscal versus calendar

Source meaning and setting. A comparison uses one fiscal year but the translated chart labels calendar year. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target invites comparison with a different twelve-month interval. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain the source fiscal period and its actual dates if approved. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Shift a quarter’s observations to demonstrate mismatch. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 023: Rolling average

Source meaning and setting. A graph shows a rolling seven-day average and not daily counts. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls the plotted point an exact one-day measurement. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve averaging window. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A single spike can be smoothed away in the seven-day series. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 024: Cumulative since launch

Source meaning and setting. A dashboard reports total signups since launch. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target labels total as signups this week. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve cumulative period. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The cumulative total cannot be treated as a weekly flow. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 025: Twelve-month risk

Source meaning and setting. A hypothetical risk estimate covers events within twelve months of entry. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes lifetime risk. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep observation horizon. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Late events may occur outside the measured interval. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 026: Exposure hours

Source meaning and setting. A safety note reports incidents per thousand operating hours. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says incidents per thousand units. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep hours as denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One unit may operate one hour or thousands. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 027: Population rate

Source meaning and setting. A public notice gives incidents per one hundred thousand residents. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls the same number a percentage. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain scaling factor and denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Ten per hundred thousand is not ten per hundred. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 028: Cases per test

Source meaning and setting. A diagnostic laboratory tracks positives per completed valid test. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says positives per unique person. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep tests rather than people. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One person can have multiple tests. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 029: Tickets per user

Source meaning and setting. A software report counts support tickets per active account. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target refers to tickets per purchase. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep account denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One account may place many purchases. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 030: Return visits

Source meaning and setting. A site measures visits per visitor, not visitors per visit. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reverses the quotient. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain numerator and denominator order. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. With two visitors making six visits, ratios are three and one-third respectively. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 031: Risk ratio

Source meaning and setting. A hypothetical trial reports risk in group A as 1.5 times the risk in group B. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says group A has 1.5 percentage points more risk. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain a ratio unless the two absolute risks are known. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Risks 2% and 3% imply one percentage point, yet ratio 1.5. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 032: Odds ratio

Source meaning and setting. A paper states an odds ratio of two for a specified comparison. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target replaces “odds” with “probability”. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve odds ratio as a distinct statistic. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Probabilities of 0.2 and 0.333 have odds 0.25 and 0.5. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 033: Absolute risk reduction

Source meaning and setting. A fictional treatment reduces event frequency from 4% to 3%. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says a one per cent relative reduction. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep one percentage point absolute reduction, or 25% relative reduction as a separate calculation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Recompute both with 100 and 1000 observations. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 034: Risk difference subgroup

Source meaning and setting. A report gives an effect only for an identified subgroup meeting a defined criterion. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target generalizes to every participant. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep subgroup label and eligibility. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A different subgroup may show no effect. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 035: Diagnostic sensitivity

Source meaning and setting. A test comparison concerns sensitivity for a designated condition and reference procedure. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims overall accuracy improved. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve sensitivity rather than overall accuracy. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. False positives and prevalence also affect total accuracy. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 036: False positive rate

Source meaning and setting. An algorithm has a lower false positive rate at a specified decision threshold. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it makes fewer errors of every type. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep false-positive type and threshold. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. False negatives may rise while false positives fall. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 037: Precision metric

Source meaning and setting. A model’s precision rises in the test set. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target substitutes recall. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep precision and its numerator/denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A model can increase precision while missing more true cases. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 038: Recall metric

Source meaning and setting. The source reports recall among actual relevant items. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports accuracy among all items. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve recall’s reference set. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Most irrelevant items can inflate accuracy without increasing recall. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 039: Uncertainty interval

Source meaning and setting. An estimate is accompanied by an interval representing uncertainty about the effect. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says every individual outcome lies in that interval. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain interval’s statistical interpretation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An individual may lie far outside an interval about the mean. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 040: No clear difference

Source meaning and setting. A comparison finds no statistically detectable difference under its design. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the two outcomes are proven identical. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve absence of sufficient evidence for difference. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An underpowered comparison may not detect a real gap. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 041: Equivalence design

Source meaning and setting. A study explicitly tests an equivalence margin fixed in advance. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target drops the margin and calls treatments universally identical. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep formal margin and domain of measured outcomes. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Outcomes can be equivalent within a margin without being the same in every way. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 042: Noninferiority

Source meaning and setting. An evaluation concludes an option is not inferior beyond a predeclared margin. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it more effective. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Use “noninferior” according to the actual design and approved explanation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A less effective option can still meet the noninferiority criterion. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 043: Confidence score

Source meaning and setting. A classifier reports an internal model score of 0.8 rather than calibrated probability. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says an event has an 80% real-world chance. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep score label or require calibration evidence. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A score may not correspond to empirical event frequency. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 044: Estimated budget

Source meaning and setting. A projected cost is around eighty units with a declared contingency. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls eighty the final audited expenditure. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve projection and approximation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Actual spending may later differ. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 045: Forecast interval

Source meaning and setting. A chart shows a central forecast plus upper and lower possible paths. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports only the top path as inevitable. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep central forecast and uncertainty band. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Test a lower outcome still permitted by the source. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 046: Qualitative more likely

Source meaning and setting. A character says “more likely” without a quantified probability. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target invents a 70% chance. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve qualitative relative judgement. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Many possible probabilities satisfy the comparison. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 047: A little warmer

Source meaning and setting. A review says the second response feels a little warmer than the first. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target converts this to a scientifically measured emotional improvement. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve subjective perception and scale restraint. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Two readers may reasonably differ in tone judgement. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 048: Fewer questions

Source meaning and setting. A narrator says the revised essay raises fewer questions than the draft. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says all questions have been answered. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep relative decrease without a zero claim. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The revision can leave several unresolved questions. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 049: Wider scope

Source meaning and setting. A specification covers more features than the earlier draft. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it covers every requested feature. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve comparative expansion. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. More is not necessarily complete. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 050: Not as harsh

Source meaning and setting. A critique says one description is not as harsh as another. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the description is kind. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep comparative relation without absolute positive judgement. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Both descriptions can be harsh while one is less so. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 051: Less certain

Source meaning and setting. A scientist says the second hypothesis is less certain than the first. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target labels it false. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve degree of confidence rather than truth value. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Both hypotheses may remain plausible. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 052: Nearly twice

Source meaning and setting. An experimental quantity is nearly twice the baseline, but not exactly twice. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states exactly double. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep approximate relation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A factor of 1.9 is near two without equal two. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 053: At least double

Source meaning and setting. A policy goal requires at least twice the current capacity. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states precisely twice and forbids exceeding it. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve a minimum bound. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A triple increase still meets the stated minimum. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 054: More than half

Source meaning and setting. A committee notes support from strictly more than half its voting members. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target allows exactly half. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep strict majority. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A 5-to-5 split is not more than half. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 055: Half or more

Source meaning and setting. A rule requires support from half or more of an eligible panel. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target requires a strict majority. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve equality inclusion. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Exactly half should pass the source boundary. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 056: No more than ten

Source meaning and setting. A procedure caps submissions at ten. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says fewer than ten. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Include ten as permitted. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Try precisely ten submissions. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 057: Fewer than ten

Source meaning and setting. A membership criterion says fewer than ten sessions may be missed. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target permits ten. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Exclude ten as a disqualifying case. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Test nine and ten separately. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 058: Between endpoints

Source meaning and setting. A specified temperature interval is inclusive at both ends. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target uses wording that excludes endpoints. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep inclusive bounds. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A reading at either endpoint must be classified consistently. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 059: Upper bound up to

Source meaning and setting. A brochure offers a rebate of up to twenty units under specified criteria. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target guarantees every applicant twenty. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep maximum and eligibility. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An applicant receiving five still fits the source. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 060: Minimum time

Source meaning and setting. An instruction requires the item to remain in place at least fifteen minutes. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target changes to a maximum of fifteen minutes. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve required lower bound. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An allowed twenty-minute period shows the difference. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 061: Maximum delay

Source meaning and setting. A service notice says no more than two days under defined conditions. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims delivery exactly in two days. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep upper bound, conditions and possibility of earlier delivery. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A one-day delivery is compliant. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 062: Tolerance range

Source meaning and setting. A specification permits values within a stated tolerance around a nominal value. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes the nominal value as exact and mandatory. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep tolerance and units. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An allowed boundary value need not equal the nominal center. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 063: Opt-in threshold

Source meaning and setting. Eligibility begins once an account has completed at least three qualifying sessions. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target counts nonqualifying sessions. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve both threshold and qualification. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Two qualifying plus one excluded session should not pass. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 064: Age endpoint

Source meaning and setting. A fictional rule says applicants aged eighteen or older. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says strictly older than eighteen. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Include the eighteenth birthday under the source definition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Test age exactly eighteen. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 065: Date cutoff

Source meaning and setting. Submissions are accepted through the stated final day. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says only before that day. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve whether the final date is included. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A timestamp on the final day separates readings. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 066: Conditional discount

Source meaning and setting. The higher discount applies only after both eligibility conditions are satisfied. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target permits either condition alone. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep conjunction and threshold. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Check an applicant satisfying just one condition. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 067: Conditional upgrade

Source meaning and setting. The upgrade is available if either of two independent criteria applies. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target requires both criteria. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep disjunction and defined scope. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One satisfied criterion should suffice. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 068: Exceptional group

Source meaning and setting. All selected candidates except those with incomplete records were included. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states every selected candidate was included. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain exception condition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One incomplete record creates a valid excluded case. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 069: Only completed trials

Source meaning and setting. A performance result summarizes runs that finished without cancellation. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target includes incomplete runs in the denominator. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve completion criterion. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Cancellation-heavy conditions may change the reported score. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 070: Best under load

Source meaning and setting. A device is fastest when tested at a named low-load setting. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target asserts it is fastest at any load. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve operating condition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. At high load a rival could outperform it. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 071: Lowest in region subset

Source meaning and setting. The article calls an amount lowest among the surveyed outlets in one reporting cohort. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says lowest across all markets. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep cohort restriction. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An unsurveyed outlet may offer less. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 072: Most preferred in survey

Source meaning and setting. An option receives the most votes among four options in a particular poll. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it objectively superior. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain survey preference rather than quality judgment. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A respondent may prefer convenience over measured performance. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 073: More reliable at night

Source meaning and setting. A route comparison applies to nighttime operations under a defined dataset. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target extends the result to daytime. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep time-of-day qualification. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A daytime schedule may reverse the ordering. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 074: Highest rated by group

Source meaning and setting. A workshop has the highest rating among respondents who attended the full course. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target extends to people who left early. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep respondent and attendance frame. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Those who quit may report different views. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 075: Price before tax

Source meaning and setting. An item costs less than the competitor before applicable taxes and service charges. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims a lower final checkout amount. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve pre-tax price basis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Different taxes or fees can reverse total cost. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 076: Per-month plan

Source meaning and setting. An annual subscription has a lower equivalent monthly cost when paid upfront. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the user can pay that lower amount each month. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Distinguish billing cadence from normalized monthly cost. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A lower effective monthly price may require a full-year advance payment. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 077: Exchange rates

Source meaning and setting. A comparison uses values converted at an expressly named historical exchange rate. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target presents them as current market values. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep conversion date and method. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Exchange movements may alter current equivalents. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 078: Nominal versus real

Source meaning and setting. A report compares amounts adjusted for price changes. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls them unadjusted nominal amounts. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep inflation-adjusted basis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Nominal increases can conceal real declines. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 079: Household versus person

Source meaning and setting. A report compares expenditure per household. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it spending per individual. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve household denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Two households of different sizes need not align with per-person estimates. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 080: Emissions intensity

Source meaning and setting. A chart reports emissions per unit of output. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports absolute emissions. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep production denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Output may grow while intensity falls. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 081: Water quality indicators

Source meaning and setting. A graph compares concentrations of a substance, not total pollutant mass. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says overall discharged mass is lower. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain concentration and sample basis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Different water volumes change total mass independently. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 082: Study samples

Source meaning and setting. Group A has a higher observed mean in a sample of twelve than Group B in a sample of ten. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target generalizes to their entire populations. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep sample size and evidential limit. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Sampling variation can reverse a population-level conclusion. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 083: Effect after adjustment

Source meaning and setting. A report gives a comparison adjusted for declared covariates. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target presents the value as an unadjusted difference. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain adjusted-analysis label. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Raw group averages can differ from the adjusted estimate. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 084: Rate standardized

Source meaning and setting. A study reports an age-standardized rate under a specified reference population. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes a crude observed rate. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep standardization procedure. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Different group age structures produce different crude rates. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 085: Mean change versus final level

Source meaning and setting. A group improved by five points on average during training. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says final scores averaged five points. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep change in score, not outcome level. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Starting scores could be high or low. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 086: Within-person comparison

Source meaning and setting. Participants are compared with their own pre-test measurements. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls them a comparison with a different group. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain paired design. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Group differences can conflict with within-person changes. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 087: Peer comparison

Source meaning and setting. An organization compares its results with other organizations in the same cohort. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it an earlier-year self-comparison. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep peer-group identity. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A group can outperform peers while declining over time. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 088: Benchmark threshold

Source meaning and setting. A model exceeds a minimum benchmark on one test set. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims all its results exceed that benchmark. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Restrict test set and measure. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Another dataset may have lower accuracy. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 089: Chart two series

Source meaning and setting. A localized line chart switches the legend labels for two series. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target narrative says the wrong group improved. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Match legend against actual data series. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Follow a single point through legend, colour and axis. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 090: Bar chart truncated axis

Source meaning and setting. A chart visually magnifies small differences by starting its vertical axis above zero. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target copy says one value is several times larger. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Use actual numbers rather than perceived bar height. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A difference of 95 versus 100 is about 5.3% relative to 95. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 091: Logarithmic scale

Source meaning and setting. A line chart’s vertical axis uses a logarithmic scale. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target commentary treats equal vertical spacing as equal additive differences. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep log-scale description. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Equal spacing can represent equal ratios. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 092: Error bars

Source meaning and setting. A plot includes error bars reflecting a defined uncertainty measure. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The translated alt text omits them and describes point estimates as exact. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve uncertainty semantics in accessible description. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A wide bar may affect interpretation. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 093: Stacked percentages

Source meaning and setting. A stacked chart shows components as shares of a whole within each period. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says component percentages are shares of the global total. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep per-period denominator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Each stack can sum to 100% independently. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 094: Chart footnote

Source meaning and setting. A footnote excludes incomplete observations from a comparative table. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The translated headline drops the exclusion. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Carry the qualification near the claim or preserve accessible footnote linkage. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Adding incomplete cases can change the rate. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 095: Heatmap normalization

Source meaning and setting. A heatmap scales colour intensity separately within each row. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target description compares colours across rows as absolute values. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve per-row normalization. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The same shade need not mean the same raw value. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 096: Conditional average

Source meaning and setting. A report averages only purchases above a stated minimum size. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls the figure average purchase size for all purchases. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain conditioned sample. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Many small purchases could lower the overall mean. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 097: Multi-metric trade-off

Source meaning and setting. Design A is faster but less accurate than Design B. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls A unconditionally better. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve two dimensions and trade-off. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A reader prioritizing accuracy may reasonably choose B. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 098: Pareto front

Source meaning and setting. Several options are not dominated across two metrics. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says one option is the unique best. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Explain nondominance without fabricating one total order. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. No choice need win both criteria. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 099: Different weights

Source meaning and setting. A combined rank uses a weighting formula declared in a methods note. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target omits weights and treats rank as intrinsic. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep weighting basis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Different weights can reverse order. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 100: Conditional correlation

Source meaning and setting. A positive association appears only within the observed subgroup. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target asserts it across pooled groups. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep conditioning and analysis group. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Pooling may eliminate or reverse the association. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 101: Causation language

Source meaning and setting. A report observes higher use and higher satisfaction together. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says higher use caused satisfaction. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve association without causal claim. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A third factor could influence both. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 102: Indirect network comparison

Source meaning and setting. Two alternatives are each compared with a shared control, without a direct comparison. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes a direct head-to-head trial. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep indirect methodology. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The indirect estimate depends on assumptions absent from a direct experiment. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 103: Reported perception

Source meaning and setting. A participant says a design “felt safer.” This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states that accidents were objectively reduced. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Retain participant perception as perception. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Feeling safe need not prove fewer accidents. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 104: Expected versus actual

Source meaning and setting. A forecast says production may be 15% higher next year. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says production already rose 15%. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep forecast, year and possibility. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The predicted increase may not occur. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 105: Preliminary findings

Source meaning and setting. An early analysis suggests one approach could be marginally more efficient. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target declares a proven large advantage. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve provisionality and small estimated degree. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Final data may change the conclusion. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 106: Negative comparative

Source meaning and setting. “Not less accessible” is used, without claiming measurable superiority. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says “more accessible.” An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve nonworsening claim without adding improvement. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Equal accessibility satisfies source but not “more.” A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 107: Double comparison

Source meaning and setting. The sentence says A is not as much more expensive than B as C is. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target collapses it into A cheaper than B. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Parse the nested difference first. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A may remain more expensive than B while having a smaller premium than C. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 108: Contrastive only

Source meaning and setting. “Only the second method improves the median” excludes the first method. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the second method improves only the median. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep scope of “only” on the method. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The second method may improve other measures too. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 109: Counterfactual baseline

Source meaning and setting. A simulation compares emissions to a modeled no-change scenario, not last year’s observed value. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target labels it year-over-year emissions. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep counterfactual reference. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Observed history need not equal hypothetical baseline. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 110: Historical control

Source meaning and setting. A retrospective comparison relies on an earlier cohort with different operating conditions. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it simultaneous random assignment. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep design and historical nature. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Time-related confounders may matter. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 111: Censored follow-up

Source meaning and setting. A study reports time-to-event comparison with participants having incomplete follow-up. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says every participant was followed until the event. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep censoring and study design. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Some participants may exit without an event. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 112: Precision significant digits

Source meaning and setting. A result is given as about 2.3 units after rounding. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports 2.30000 as if measured to five decimals. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep meaningful precision. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Extra zeros can imply an unsupported measurement resolution. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 113: Boundary rounding

Source meaning and setting. A value 4.995 is displayed as 5.00 under a visual rounding convention but evaluated against a strict minimum of 5.00 using raw data. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it certainly passes. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve decision based on unrounded value. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Displayed equality does not establish compliance. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 114: Missing comparator

Source meaning and setting. A report omits the earlier value but writes “higher than before.” This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target inserts an invented numerical increase. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve qualitative comparison and request data before numeric expansion. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Many different earlier values satisfy “higher.” A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 115: Misleading chart title

Source meaning and setting. A chart uses “improvement” for a decrease in time, not an increase in score. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target labels the plotted time values as gains in score. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Connect evaluative title to actual metric. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Lower seconds and higher points are different variables. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 116: Quote scope

Source meaning and setting. A respondent says the second draft is “more convincing to me.” This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports everyone found it objectively stronger. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep attributed individual judgement. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A single individual’s preference does not establish universal agreement. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 117: Model conditions

Source meaning and setting. A simulation predicts more favorable outcomes only if a specified demand assumption holds. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target omits the assumption and promises the outcome. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep conditional model and forecast. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. If demand differs, the result need not follow. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 118: Completeness restriction

Source meaning and setting. “Most complete among the editions examined” is rendered as “the complete edition.” This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target changes a relative ranking into an absolute completeness claim. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve examined set and relative nature. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Even the best examined edition can omit material. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 119: Rhetorical comparison

Source meaning and setting. An essay calls an argument “closer to the point” without providing a numeric metric. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target assigns a scored accuracy improvement. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve rhetorical positioning as qualitative comparison. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. No recognized scale may exist for “closer.” A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 120: Sensitivity analysis

Source meaning and setting. A result holds under several but not every tested model specification. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the result is robust to any analysis. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep range of sensitivity checks and exceptions. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An untested model could change the finding. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 121: Subset exclusion

Source meaning and setting. A dashboard performance claim excludes maintenance downtime from measured availability. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target includes downtime in its interpretation of the same percentage. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep exclusion criteria. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Availability can be lower if maintenance is included. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 122: Consecutive versus cumulative

Source meaning and setting. An achievement requires five consecutive qualifying attempts. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says five attempts in total. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain adjacency condition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Five attempts separated by failures do not satisfy consecutive rule. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 123: Pre-post net change

Source meaning and setting. A report says the net average improved even though some cases deteriorated. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target asserts every case improved. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve aggregate statistic and heterogeneous outcomes. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An improving mean can coexist with individual declines. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 124: Ordinal rating

Source meaning and setting. A 1-to-5 satisfaction scale is compared by median rating. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target treats rank intervals as equal physical units. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep ordinal semantics or disclose modelling assumptions. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Difference between ratings 1 and 2 need not equal 4 and 5. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 125: Ratio of ratios

Source meaning and setting. A document compares the proportional changes in two groups rather than their final raw amounts. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target switches to raw amount comparison. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve the ratio-of-ratios structure. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Groups can have different baselines and equal proportional changes. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 126: Absolute volume versus share

Source meaning and setting. An audience share falls while the absolute audience grows because the whole market expands. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says fewer people watched. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep share and total distinct. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A declining slice of a growing total can still be larger. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 127: The excluded zero

Source meaning and setting. A statistical summary omits zeros according to a stated rule. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims zeros were included in the average. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain missing/zero handling rule. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Including zero observations changes the mean. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 128: Averages over different windows

Source meaning and setting. One performance figure averages seven days and another averages thirty days. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target presents them as strictly comparable same-window averages. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Label windows before ranking. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Short-term variation may make rankings unstable. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 129: Unit scale

Source meaning and setting. One table expresses amounts in thousands and another in individual units. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target copies numbers without scaling unit headings. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve magnitude multiplier. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The value “5” can stand for five or five thousand. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 130: Per-capita versus per-household

Source meaning and setting. Two regions are compared on households, not population size. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it per-capita comparison. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep unit of analysis. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Different household sizes alter the population-normalized result. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 131: Alternative hypothesis

Source meaning and setting. A study says group differences are compatible with a proposed explanation but not decisive. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls the proposed explanation proven. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep evidential compatibility. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A different mechanism may fit the same difference. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 132: Not worse than target

Source meaning and setting. A process meets a predetermined maximum defect rate. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it outperformed every competitor. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep compliance with target distinct from ranking. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Many competing processes may also meet the target. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 133: Quality vs affordability

Source meaning and setting. A buyer’s shortlist ranks packages by affordability given minimum quality. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims the cheapest is highest quality. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep constraint and optimizing criterion distinct. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The best affordable option need not top quality alone. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 134: Benchmark withdrawn

Source meaning and setting. A comparative headline refers to a benchmark later marked invalid. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target repeats it as current proof without the correction. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Require authority and version verification. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A retracted comparison cannot support a fresh unqualified statement. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 135: Stronger conclusion only within measure

Source meaning and setting. A study’s evidence is stronger for one outcome than another. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says all conclusions became certain. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep outcome-specific evidence strength. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. One robust finding does not certify every claim. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 136: Graph legend semantically overlapping

Source meaning and setting. Two line labels use “baseline” to mean different things: initial data and control model. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target merges them into one series. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Distinguish the two roles by context. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An initial measurement and reference simulation are different objects. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 137: Nearly zero

Source meaning and setting. A statement says the observed rate was close to zero but not zero. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target writes no events occurred. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep approximation and observation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A single rare event is compatible with nearly zero but not literal zero. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 138: Growth to versus by

Source meaning and setting. A target is 150% of baseline, whereas a draft says it increased by 150%. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The translation changes a final level into a growth increment. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep 150% of baseline or 50% above baseline. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Baseline 100 gives final 150, not 250. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 139: Three times larger

Source meaning and setting. The source’s wording may ambiguously mean three times as large or three times larger than baseline. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target commits to 3x or 4x without clarification. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Record unresolved multiplier ambiguity and request authorized explanation. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Baseline 10 yields 30 or 40, a material distinction. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 140: More efficient per unit

Source meaning and setting. The technology uses less input per delivered result but has higher fixed setup cost. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it is cheaper for every scale of deployment. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep marginal efficiency and fixed cost separate. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. At tiny volumes the fixed cost may dominate. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 141: Budget underspend

Source meaning and setting. An actual spend is lower than a planned ceiling but still higher than last year’s spend. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says spending fell compared with last year. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep plan-versus-actual comparator. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Two different baselines support opposite directions. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 142: Ranged quotation

Source meaning and setting. A source says a figure “could be between ten and twenty” under certain assumptions. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states a precise fifteen. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve possibility range and assumptions. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Ten and twenty remain permitted values. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 143: Lower rate, larger count

Source meaning and setting. A large group has lower event rate but more events than a small group. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says the larger group had fewer events. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep the distinction between count and rate. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Ten out of 1,000 exceeds three out of 100 in count but not rate. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 144: Rank among tested languages

Source meaning and setting. A multilingual benchmark ranks one approach highest across the language pairs that were evaluated. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says it is best for all languages. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Limit to evaluated pairs and criteria. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. An untested pair may behave differently. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 145: Difference of differences

Source meaning and setting. The estimate compares the change in group A with the change in group B. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes only group A’s raw change. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep both time differences and group contrast. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Both groups can improve while the difference-in-differences is negative. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 146: Relative reduction baseline zero

Source meaning and setting. A source’s baseline is zero, so a percentage change is undefined. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target invents an infinite increase figure. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Report count difference or query the intended statistic. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Division by zero does not provide a normal percent change. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 147: Same label different basis

Source meaning and setting. Two charts both label “completion” but one counts completed tasks and another completed users. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target combines them as one identical metric. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Maintain metric definition and unit. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A user can complete many tasks. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 148: Category-weighted average

Source meaning and setting. A score uses category weights specified in a technical appendix. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target describes it as an unweighted arithmetic average. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve weighting and appendix linkage. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Unequal weights can change combined rank. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 149: Nonmonotonic score

Source meaning and setting. A scale has an optimal middle zone where both extremes are worse. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says higher is always better. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve scale interpretation and recommended band. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A high numeric reading need not mean better performance. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 150: One-sided limit

Source meaning and setting. An uncertainty statement gives only a lower confidence bound. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target supplies an invented symmetric interval. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep one-sided direction and confidence description. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The omitted upper limit is not known from source. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 151: Independent event comparisons

Source meaning and setting. An experiment compares two kinds of events, one rare and one common. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target combines counts and calls them a single outcome. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep event definitions separate. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A change in the common event can hide an increase in the rare one. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 152: Inclusion window

Source meaning and setting. A report covers records filed from the first through the last day of a stated interval. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target inadvertently excludes the final day. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve endpoint inclusivity and clock-zone basis when material. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A valid record on the boundary day should count. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 153: Qualitative rank in fiction

Source meaning and setting. A character believes a rival is more dangerous than another without narrator endorsement. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target states this as established fact. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve focalized belief and speaker ownership. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Other characters may contradict the claim. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 154: Lower but not best

Source meaning and setting. An option is less costly than the previous plan but remains costlier than two alternatives. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target calls it cheapest. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve pairwise comparison rather than total rank. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. The source can be true while cheapest is false. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 155: Unmeasured subjective baseline

Source meaning and setting. A review says “the room feels brighter than before” without light readings. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target claims measured illumination increased by a stated percentage. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep reported impression. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A person’s perception may change without an instrument reading. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 156: Fairness group disparity

Source meaning and setting. A metric shows different error rates across evaluated groups. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target reports an overall rate and conceals disparity. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep group-specific comparisons and their definition. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. A pooled average may hide large variation. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 157: User rating change

Source meaning and setting. An app’s displayed rating rises from 4.2 to 4.4 stars with a changed set of respondents. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target says each user rated it higher. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Preserve aggregated score and respondent frame. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. New users can raise an aggregate while individual ratings are unchanged. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Diagnostic 158: Safety margin

Source meaning and setting. An engineering note reports a margin above a minimum test requirement. This is a constructed comparison frame: it should be treated as a working proposition, not as evidence that a named real-world institution achieved such a result. Before translation, identify the text’s audience and whether the comparison appears in prose, a table, a spoken statement or a visual.

Tempting but unfaithful target interpretation. The target converts that into a guarantee beyond all operating conditions. An otherwise fluent translation might express this unsafe interpretation because a qualifier was dropped, a unit was generalized, a numerical operator was replaced or an implicit comparator was guessed. The danger is that a reader would act on a proposition the source does not warrant.

Meaning-preserving repair brief. Keep tested bound and purpose. The eventual target-language wording may use different word order, a fuller explanatory phrase or a natural comparative construction, but it must retain the stated relation. Where a sentence has an unavoidable ambiguity, do not solve it by inventing data; record the unresolved choice and ask the authorized source owner.

Worked diagnostic and acceptance test. Performance outside the test envelope is unknown. A bilingual reviewer should perform that calculation or counterexample, explain the exact source-to-target divergence, and record which word, qualifier or chart label was repaired. A pass means the repaired target licenses the same comparison, at the same evidential strength and within the same population and time bounds.

Transfer workshops: apply the method instead of memorizing isolated wording

Workshop 1: Research abstract

Assignment. Find every comparison in a short abstract and annotate numerator, denominator, group, period and uncertainty. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Write a two-line counterexample for the claim with the greatest potential to mislead. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Pass criteria. A reader using only the translated artefact can identify the same alternatives, conditions, comparison direction and uncertainty as a reader using the source. An editor can trace each critical wording choice to evidence or a documented approved query, and the case can be reused as a regression test when the source or locale changes.

Workshop 2: Pricing webpage

Assignment. Compare an annual-prepay offer with a monthly-billed offer without implying identical cancellation or billing terms. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Test a reader who cancels after one month and a reader who continues twelve months. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 3: Healthcare leaflet

Assignment. Separate reported risk difference, personal risk, and general advice without supplying unauthorized treatment recommendations. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Use a hypothetical baseline and check that the reader does not infer a fabricated absolute benefit. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 4: Technical brochure

Assignment. Distinguish throughput, latency and efficiency in a claimed improvement. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Imagine a process that is faster for one input but slower in total across a batch. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 5: Employment report

Assignment. Check whether salary comparisons are hourly, monthly, full-time-equivalent or annual. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Use two hypothetical work schedules that produce different totals for equal hourly wages. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 6: Education assessment

Assignment. Explain percentile rank versus raw percentage correct without teaching either as an interchangeable score. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Make an example where a high percentile coexists with a moderate raw score. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 7: Scientific chart

Assignment. Rewrite alt text for a graph with confidence intervals and a nonzero axis. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Have a reader explain which values are measured and which are inferred. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 8: Policy eligibility notice

Assignment. Rewrite at least, more than, below and no more than into natural target-language conditions. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Classify a person exactly on each endpoint and document every expected answer. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 9: Product ranking

Assignment. Preserve “among items tested” and the criteria on which rank is based. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Add an untested competitor and check that the original statement remains honestly restricted. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 10: Annual report

Assignment. Distinguish nominal change, inflation-adjusted change and changed exchange-rate assumptions. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Use an illustrative price index to test whether the claims can move in opposite directions. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 11: Accessibility description

Assignment. Write a chart description that states series, denominator, baseline, trend and uncertainty without overstating shape. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Test with someone who only hears the description. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 12: Literary scene

Assignment. Preserve a character’s subjective comparison without turning it into the narrator’s established fact. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Ask who owns the opinion and what another character could plausibly dispute. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 13: Survey summary

Assignment. Keep response selection and sample size attached to a preference result. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Imagine strong nonresponse bias and see which target formulations remain justified. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 14: Environmental dashboard

Assignment. Distinguish emissions intensity per output from absolute emissions. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Increase production while improving intensity and compare possible totals. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 15: Website performance

Assignment. Translate a claim about median response time separately from mean latency and uptime. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Construct a distribution with rare long delays to expose median-mean confusion. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 16: Automotive testing

Assignment. Keep load, temperature, speed, fuel use and test-cycle limits in performance comparisons. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Show why a result for one specified condition does not certify every road condition. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 17: Mobile app A/B test

Assignment. Keep assignment group, conversion denominator, significance status and rollout conditions. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Vary traffic volumes while preserving conversion proportions. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 18: Finance disclosure

Assignment. Maintain period, currency, accounting basis and sign when comparing figures. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Explain how a change from negative to less negative differs from positive growth. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 19: Museum exhibit

Assignment. Compare versions of an artefact using provenance and the exact set of examined copies. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Do not turn “oldest known surviving example” into “first example ever made.” Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 20: Legal terms overview

Assignment. Translate a comparative qualification without expanding rights or excluding eligible boundary cases. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Have a qualified professional verify the decision endpoint against the controlling text. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 21: Machine translation assessment

Assignment. Distinguish automated metric, human rating, dataset scope and statistical uncertainty. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Add an untested language pair to expose unsupported “best for every language” claims. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 22: Public-information infographic

Assignment. Reconcile headline, chart, footnote, alt text and download CSV labels. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Ask a target reader to reproduce the same simple ratio from each channel. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 23: Scientific peer review

Assignment. Classify whether a comparison is direct, indirect, adjusted, observational or randomized. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Explain what the translator may label without making a methods judgement. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Workshop 24: Everyday dialogue

Assignment. Translate a nonnumeric expression such as “less awkward than before” without inserting invented measurements. Begin by making a source claim ledger: who speaks, which two alternatives are compared, the exact dimension, any baseline, the population and the implied reader task. Give the learner the source context rather than a phrase detached from its document.

Challenge test. Check whether tone, speaker ownership and comparison point survive. Produce an intentionally misleading candidate and annotate the precise boundary it crosses. Then prepare a meaning-preserving target-language draft, a one-paragraph justification and a minimal counterexample that would reject the bad draft. If the exercise involves controlled specialist claims, the language reviewer must not substitute for the relevant domain authority.

Frequently asked questions

1. What is comparative equivalence in translation?

It is preservation of the relation that makes a comparison true: the two referents, dimension, baseline, denominator, scope, direction and evidence. Surface similarity of comparative words is not enough.

For a practical check, locate the comparison’s missing argument or limiting condition in the actual source, not merely in this explanatory answer. Then test the target on a borderline case. The right wording is the one that stays faithful when the easy, typical example is replaced by an awkward but still possible one.

2. How do I translate “more than” accurately?

Check which object is greater and whether equality is excluded. A value exactly at the threshold must fail a strict greater-than condition.

3. Is “at least” the same as “more than”?

No. At least includes the boundary value; more than excludes it. Translate that difference even if both sound natural in everyday speech.

4. How do I translate “up to”?

Usually as an upper limit, not a guaranteed exact payment, quantity or achievement. Its legal or promotional interpretation depends on surrounding conditions.

5. What is the difference between percent and percentage points?

Percentage-point change is the arithmetic difference between two percentages; relative percent change divides that difference by the starting percentage.

6. Can I replace “percentage points” with “percent” to simplify a text?

Not without changing the mathematics. Explain the term separately if the audience needs help, but keep the source relationship intact.

7. Why is the denominator important?

Without it, the same numerator can mean very different event rates. Per user, per test, per transaction and per hour are different measures.

8. Is “twice as likely” the same as “twice as many”?

No. Likelihood describes a probability or frequency relative to a base; the number affected also depends on population size.

9. Can an average stand for every participant?

No. The mean or median describes a distribution summary, and individual experiences can vary widely.

10. Can “no significant difference” mean equal results?

Not by itself. Statistical non-detection is distinct from an affirmative equivalence conclusion under a suitable design.

11. Can “noninferior” be rewritten as “better”?

No. Noninferiority is a distinct formal conclusion relative to a margin, and it does not establish superiority.

12. How should superlatives be translated?

Carry the ranked reference class, measurement criterion, test conditions, and whether ties are permitted.

13. What if the comparison class is absent?

Look for context, captions and the author’s defined terms. If it remains unknown, preserve ambiguity when possible or request clarification.

14. Should translators recalculate figures?

They can use verification calculations to detect errors, but must not silently correct authoritative source figures without an approved source-owner process.

15. How do I handle rounded values?

Distinguish displayed rounding from the value used for threshold decisions. Keep the stated precision and do not invent significant digits.

16. Are rankings and ratings interchangeable?

No. Rank is position within an ordered set; rating is a score according to an instrument or scale.

17. How do I translate an index of 120?

State the base period or index convention. Under a base of 100, 120 represents 20% above that base, not automatically 120% growth from the previous period.

18. What if a target language uses different number formatting?

Apply the target locale’s recognized decimal, grouping and percent conventions, but keep the underlying numerical value and relationship unchanged.

19. Does translation permit converting units?

Only under the project brief and with verified conversions, unit context, rounding and safety implications. The reference dimension must survive.

20. What if only a chart shows the denominator?

Keep the chart and its accessible text linked to the paragraph. Do not publish a context-free excerpt that implies an unsupported comparison.

21. Does “faster” always mean less time?

It often does in ordinary conversation, but formal performance metrics may define speed differently. Recover the source’s measure before deriving time savings.

22. Can relative improvement be greater than 100%?

Depending on the quantity and definition, yes; the translator must not “fix” such a figure without checking the calculation and context.

23. What is a comparator-first workflow?

It means discovering both compared entities and the measure before drafting the target adjective, phrase or chart label.

24. What makes a comparative translation error severe?

Severity depends on consequence and specification. A changed eligibility boundary or false safety claim can matter much more than a stylistic inconsistency.

25. How does AI translation affect comparisons?

Automated systems can create fluent but incorrect baseline choices, overgeneralized claims and lost qualifiers. They need source-context retrieval and structured human review.

26. Are all comparison frames numerical?

No. Relative politeness, rhetorical strength, emotional distance and narrative judgement are comparisons whose evidence may be qualitative.

27. What if the source itself exaggerates a comparison?

Translation should not secretly rewrite the author. Record source-quality concerns and route corrections through an authorized source owner.

28. How can I test a comparative translation quickly?

Write a simple case where the source must be true and ask whether the target is also true. Then test a boundary, excluded class or changed denominator.

29. Can I use back-translation as proof?

It can reveal some problems, but a fluent back-translation may repeat the same error. Use direct evidence comparison and bilingual review.

30. What should I store for future updates?

Keep source version, approved comparator and denominator, example calculation, target wording, open queries and a regression case.

31. Does this article teach vocabulary or English grammar?

No. It owns the cross-language transfer decision. The protected Vocabulary Learning Hub and How English Works remain the appropriate destinations for those subjects.

32. Which standard should I consult for project preparation?

ISO 11669:2024 provides general guidance for translation-project specifications, needs analysis, risk and workflow. It does not itself verify a translated statistic.

33. What is ISO 5060:2024 used for?

It provides guidance for evaluating human and machine translation output with human evaluators and specified error types; it is not a claim that every example here has undergone formal certification.

34. How can readers learn this system?

Begin with a short paragraph containing one ratio, one threshold and one ranked claim. Annotate, translate and independently counterexample-test each of the three.

Research, terminology and quality boundaries

This guide reflects public documentation accessed on 4 November 2026. The cited standards and technical documents are used for terminology and production framing rather than as an endorsement or certification of this article. The worked examples are author-created exercises and should be checked against actual domain evidence before use in an operational, medical, legal, financial or safety setting.

ISO 5060:2024 — Translation services: Evaluation of translation output, general guidance. The ISO abstract describes human evaluation, error types, evaluation specifications, and sampling.
ISO 11669:2024 — Translation projects: General guidance. Its published scope addresses translation-project specifications, needs analysis, risk and workflow.
W3C Internationalization Tag Set 2.0. Its localization-quality data categories can record detected issue types and severity in multilingual content.
Unicode CLDR: Number Symbols and UTS #35, Part 3: Numbers. These describe locale-aware number and percent presentation.
Unicode CLDR: Number and Currency Patterns. The documented format patterns distinguish decimal, percentage, currency and scientific presentation.

A reviewable comparison record: reusable editorial template

Field	Question for the translator or reviewer
Source and version	Which approved source and figure version governs?
A and B	What exactly is compared with what?
Dimension	Which measurable or qualitative property is in question?
Formula	Is the relation additive, multiplicative, ordinal, conditional or descriptive?
Population	Who or what enters the denominator and ranked class?
Time	Which dates, reference years or exposure windows apply?
Boundary	Which threshold values are included or excluded?
Evidence status	Is it fact, estimate, forecast, opinion or hypothesis?
Counterexample	Which invented observation would expose a wrong translation?
Decision	What has been approved, queried or blocked?

Where to go next: preserve the architecture without duplicating other owners

The parent system for translation practice, equivalence, context and workflow is the translation root. This specialist chapter adds one particular diagnostic skill: reconstructing a comparative frame before recreating it in another language. When a comparison depends on nonliteral imagery, the preceding chapter on figurative-frame continuity is the appropriate companion. When the obstacle is learning meanings, collocations or word families, follow the protected vocabulary owner. When the obstacle is English-only grammar, sentence structure or writing, follow the protected English owner. These are links between disciplines, not editorial takeovers.

Master Art of Translation | The Complete System for Moving Meaning Between Languages — translation root and architecture.

Master Art of Translation | The Figurative Frame Continuity System — preceding child (1410), for extended metaphors, analogies and image chains.

Vocabulary Learning Hub — protected vocabulary meanings, knowledge, retrieval and application.

How English Works | The Complete Map and Learning Routes — protected English-language learning and theory.

Final release question: can a careful target-language reader reconstruct the same two alternatives, the same dimension, the same baseline, the same boundary and the same evidence status as a careful source-language reader? If the answer is yes under ordinary conditions and at the awkward endpoints, the comparative relationship has survived. If not, repair the relation before polishing the sentence. A comparison is accurate only when both sides are still comparing the same thing.