Statement-verification questions do not ask whether a sentence sounds plausible. They ask what the source actually allows you to say. A statement can be confirmed, contradicted, or left unresolved. Those three outcomes are not shades of confidence. They are different evidence states.
Across English examinations, learners encounter formats such as True / False / Not Given and Yes / No / Not Given. The labels vary in what they point to: one family typically concerns information in the text, while another may focus on the writer’s views or claims. Yet the deep reasoning problem is stable. The learner must compare one statement with one source position and decide whether the source supports it, opposes it, or simply does not settle it.
This guide establishes SETC’s global/no-location owner for that three-state evidence judgement. It does not replace Reading Performance, Argument Evaluation, Multiple-Choice Performance, Matching Performance or Ambiguity Resolution. Those owners supply comprehension, reasoning quality, option comparison, shared-option mapping and interpretation control. Statement Verification owns the three-way classification itself.
Iona anchors evidence scope and source fidelity. Maren rewrites statements into clean propositions so hidden differences in quantifier, time, cause, comparison or certainty become visible. Leonie controls stopping: once the source clearly confirms, clearly contradicts or clearly fails to settle the claim, the learner should record the state and move.
The statement-verification chain
DEFINE THE STATEMENT → LOCATE THE RELEVANT SOURCE REGION → NORMALISE THE PARAPHRASE → TEST SCOPE / TIME / OWNER / DEGREE / RELATION → CONFIRM? → CONTRADICT? → OTHERWISE NOT GIVEN → VERIFY THE EXACT STATE → MOVE.
The chain matters because many wrong answers arise before the final label is chosen. A learner may compare the wrong paragraph, import outside knowledge, overlook one quantifier, reverse cause and effect, confuse past with present, or assume that lack of confirmation automatically means false. The final three-way decision is only as reliable as the proposition comparison beneath it.
1. True means the source supports the complete statement
A statement is confirmed when the source supports the proposition as stated. “Most participants preferred the new design” matches a source saying “nearly three-quarters chose the new design over the old one.” Vocabulary differs; meaning aligns.
Complete support matters. If the statement adds a cause, time, condition or universal scope not present in the source, confirmation may fail even when the core topic matches.
2. False means the source supports the opposite or an incompatible claim
False is not “I cannot find proof”. False requires contradiction. If the statement says the programme began in 2024 and the source explicitly says it began in 2022, the source contradicts the statement.
The contradiction can also involve scope, direction or degree. Statement: “All participants completed the trial.” Source: “Several participants withdrew before completion.” The universal claim is contradicted because the source establishes exceptions.
3. Not Given means the source does not settle the statement
Not Given is not a weaker False. It is an evidence gap. If a passage says a museum opened in 1998 but never states why the opening was delayed, a statement claiming the delay was caused by funding may be plausible, yet neither confirmed nor contradicted.
The mature learner resists the urge to complete the source with common sense. The classification is about what the text establishes, not what would make the story sensible.
4. False and Not Given are separated by contradiction
Ask one decisive question: What in the source would make the statement impossible or incompatible? If you can identify that evidence, False becomes available. If the source merely fails to mention the statement, Not Given remains stronger.
This distinction should be trained explicitly because “not confirmed” is the most common path to overusing False.
5. Yes and No still require source-owned viewpoint
In writer-view or claim tasks, Yes means the statement agrees with the writer’s position, No means it conflicts with that position, and Not Given means the writer’s position is not established.
Do not replace the writer with a quoted source, interviewee or public opinion. Ownership remains part of the proposition.
6. The statement should be rewritten as a proposition
Long statements can hide several claims. Rewrite mentally: who or what, does what, under what condition, at what time, to what degree, for what reason.
“The researchers abandoned the second method mainly because it was too expensive” contains actor, action, ordinal method, principal cause and degree word “mainly”. Every part can affect the verification state.
7. Paraphrase matching comes before label selection
Source and statement may share almost no vocabulary. “The scheme was discontinued after operating costs became unsustainable” can confirm “The project ended because running it became too expensive.”
Paraphrase Performance supplies the equivalence skill. Statement Verification uses it to compare propositions rather than words.
8. Shared vocabulary can conceal contradiction
Statement: “Higher prices increased attendance.” Source: “Attendance fell after prices increased.” The same nouns appear, but the relationship is opposite.
Keyword overlap should never be treated as confirmation. Relationship direction is evidence-bearing.
9. All, most, some and none create different claims
Quantifiers are high-value verification words. “Most students improved” does not confirm “all students improved”. “Some participants objected” does not contradict “most supported the plan”; both can be true simultaneously.
Maren highlights quantifier before comparison. A small word can change the entire truth condition of the statement.
10. “Not all” does not mean “none”
Source: “Not all households saw lower bills.” Statement: “No households saw lower bills.” The statement is not confirmed. Depending on the wording, the source may imply at least some did, which would contradict “none”.
Scope logic should be explicit enough that the learner does not turn partial failure into universal failure.
11. “Only” can reverse the verification state
Source: “The trial was conducted in three cities.” Statement: “The trial was conducted only in three cities.” If the source never excludes other locations, “only” may create an unsupported exclusivity claim rather than a confirmed one.
Not every numerical match licenses an exclusivity word.
12. “Mainly” asks for ranking, not mere presence
If the statement says a project failed mainly because of cost, the source must make cost the principal factor or clearly dominant explanation. A paragraph merely mentioning cost among several reasons does not necessarily confirm “mainly”.
Ranking words require ranking evidence.
13. Modality controls certainty
May, might, could, likely, will and must are not interchangeable. Source: “The change may reduce waiting time.” Statement: “The change will reduce waiting time.” The statement strengthens uncertainty into certainty and is not confirmed.
Whether it is contradicted or Not Given depends on whether the source positively rejects certainty or simply expresses lower confidence. This is precisely where careful state classification matters.
14. Possibility does not equal actuality
“The method could improve accuracy” does not confirm “the method improved accuracy”. Possibility and realised outcome are different propositions.
If the source later reports no improvement, False becomes available. If outcome is never reported, Not Given is more appropriate.
15. Prediction does not equal observation
Source: “Researchers expect demand to rise next year.” Statement: “Demand has risen.” Tense and epistemic status differ.
A forecast cannot confirm an observed past or present outcome. Time and evidence status both matter.
16. Past and present states must remain separate
Source: “The company once relied on paper records but now uses a digital system.” Statement: “The company uses paper records.” If the statement describes current practice, the source contradicts it.
One fact can be true historically and false now. Chronology belongs inside the proposition.
17. “Initially” and “finally” are decisive markers
A speaker or writer can hold several positions over time. A statement about initial belief should be compared with the initial state; a statement about final position with the updated state.
Do not collapse sequence into one timeless opinion.
18. Conditions are part of the verified claim
Source: “The device performs reliably if humidity remains low.” Statement: “The device performs reliably.” Removing the condition broadens the claim.
If no evidence establishes performance under other conditions, the unconditional statement is not confirmed. Whether the source contradicts it depends on what is said about higher humidity.
19. Necessary and sufficient conditions differ
“A permit is required to enter” means permit possession is necessary; it does not guarantee entry if other requirements exist. Statement: “Anyone with a permit can enter” may be unsupported.
Verification should preserve the logical direction of conditions.
20. Cause and correlation must not be merged
Source: “People who exercised more reported higher wellbeing.” Statement: “Exercise caused higher wellbeing.” The statement adds causality beyond association.
Argument Evaluation explains evidence strength; Statement Verification decides whether the source claim and statement claim are equivalent. They are not.
21. Cause direction can be reversed
Source: “Low rainfall caused reservoir levels to fall.” Statement: “Low reservoir levels caused rainfall to decline.” Same concepts, reversed mechanism, clear contradiction.
Use arrows when necessary. Source relation must survive intact.
22. Sequence is not necessarily cause
Source: “The announcement was followed by a fall in sales.” Statement: “The announcement caused sales to fall.” Unless causality is stated or strongly established, the statement overreaches.
If the source offers another cause, False may be justified. If cause is not settled, Not Given is stronger.
23. Comparison statements need the same criterion
Source: System A is cheaper but less reliable than B. Statement: “System A is better than B.” “Better” is underdefined unless the criterion is supplied.
If the source never gives an overall judgement, the broad comparison may be Not Given even though component comparisons are stated.
24. Superlatives require a complete comparison set
Source compares A with B. Statement says A is the cheapest of all available systems. The source may confirm A cheaper than B without supporting the global superlative.
Comparison universe is part of scope.
25. Ownership errors can turn Yes into Not Given or No
A writer quotes an expert saying the policy is ineffective. If the writer does not endorse that view, a statement “The writer believes the policy is ineffective” may be Not Given. If the writer explicitly rejects the expert, No becomes appropriate.
Who owns the proposition is part of what must be verified.
26. Mentioning an idea does not equal supporting it
Source: “Some commentators argue that automation always reduces jobs; the evidence, however, is mixed.” Statement: “The writer believes automation always reduces jobs.” The writer reports the claim and qualifies it. The reported idea cannot be assigned automatically to the writer.
Attribution verbs and contrast markers matter.
27. Agreement can be partial
Writer: “The goal is worthwhile, but the proposed method is too expensive.” Statement: “The writer supports the proposal.” This is ambiguous unless “proposal” means goal or method.
Ambiguity Resolution may be required before verification. Once scope is defined, stance can be classified accurately.
28. Concession does not erase the main stance
“Although the sample is small, the findings are encouraging.” Statement: “The writer sees the findings as promising.” The concession limits confidence but does not reverse the positive direction.
Yes/No tasks require hierarchy of stance, not word counting.
29. Silence about motive usually means Not Given
A source says a company closed one branch after revenue fell. Statement: “The branch closed because executives wanted to reduce staff.” Unless that motive is supplied or contradicted, the claim remains Not Given.
Plausible motive is not textual evidence.
30. Silence about a detail does not contradict it
Source: “The expedition left in June.” Statement: “The expedition included twelve scientists.” If participant count is never given, Not Given. The absence of twelve is not evidence that there were not twelve.
This simple case should be mastered early because it exposes the core logic cleanly.
31. Absence of mention can matter only when the source explicitly closes the set
Source: “The only causes identified were cost and delay.” Statement: “Poor training was also identified as a cause.” Here the source explicitly defines an exhaustive set, so the additional cause is contradicted.
“Not mentioned” becomes contradiction only when the text itself establishes completeness.
32. Lists are not always exhaustive
Source: “Reasons included cost, staffing and timing.” “Included” often signals examples rather than a closed list. Statement adding another possible reason may be Not Given rather than False.
Look for only, all, exclusively, the three reasons, or other completeness markers before treating omission as contradiction.
33. Definitions can create direct contradiction
Source defines “resilience” as recovery after disruption. Statement: “Resilience means avoiding all disruption.” The definition conflicts directly.
Definition questions should compare necessary conceptual features, not broad positivity.
34. Category membership can be verified or contradicted
Source: “Mangroves are flowering plants adapted to saline coastal environments.” Statement: “Mangroves are a type of flowering plant.” True. Statement: “Mangroves are fungi.” False if the category distinction is explicit.
Category verification is often straightforward when definitions are direct.
35. A statement can be partly true and therefore still false
Statement: “The trial lasted six weeks and included 200 participants.” Source: six weeks, 150 participants. Half matches; the compound statement as a whole is incompatible.
Conjunction means both components must survive. One contradicted component can make the complete statement False.
36. A statement can be partly supported and partly unstated
Statement: “The trial lasted six weeks and was funded by the government.” Source confirms six weeks but says nothing about funding. The full statement is not fully confirmed.
If no funding contradiction exists, the unresolved component usually pushes the whole claim toward Not Given rather than True.
37. Compound statements should be split before classification
Use A AND B, A BECAUSE B, A WHILE B or another relation to expose subclaims. Then compare each component and the connector.
This prevents one true phrase from masking one unsupported or contradicted component.
38. Negation can hide the real proposition
Statement: “The researchers did not reject the method.” Source: “The method remained under consideration.” These may align, but careful logic is needed: not rejecting does not necessarily mean accepting.
Rewrite negative statements into the cleanest possible logical form before comparing them.
39. Double negatives should be normalised
“It is not uncommon” generally means reasonably common, but degree can remain softer than “common”. Do not simplify double negatives so aggressively that modality or register strength changes.
Normalisation should preserve meaning, not merely reduce cognitive load.
40. Comparative direction must be preserved
Source: “A is less expensive than B.” Statement: “B costs more than A.” True through equivalent comparison. Statement: “A costs more than B.” False through reversal.
Paraphrase recognition includes relational inversion where meaning remains equivalent.
41. Relative change and absolute level differ
Source: “Costs fell by 20% but remained higher than in neighbouring regions.” Statement: “Costs are low.” The source reports decline, not necessarily low absolute level.
Trend and level should not be confused.
42. Averages do not describe every individual
Source: “Average completion time fell to ten minutes.” Statement: “Every participant finished within ten minutes.” The average does not establish a universal individual limit.
Statistical aggregation and individual claims operate at different scopes.
43. Percentages require denominator control
“60% of surveyed users preferred A” does not confirm “60% of all users prefer A” unless the sample is explicitly equated with the entire population.
Population scope belongs inside verification.
44. “Twice as many” and “twice as much” need the right base
Source: 20 chose A and 10 chose B. Statement: “Twice as many chose A as B.” True. Statement: “A was chosen 200% more often than B.” That wording has a different mathematical relation and deserves careful checking.
Do not let numerical familiarity replace exact comparative meaning.
45. Range statements need endpoint control
Source: “Participants were aged 18 to 25.” Statement: “No participants were older than 25.” True if the range is exhaustive. Statement: “Most participants were 25.” Not Given unless distribution is supplied.
Range gives boundaries, not frequencies within the range.
46. Minimum and maximum statements are asymmetric
Source: “Applicants must be at least 18.” Statement: “Applicants can be older than 18.” True unless another upper bound exists. Statement: “All applicants are exactly 18.” False or unsupported depending on source context.
Threshold language should be translated precisely.
47. Approximation should not be sharpened into exactness
Source: “About 200 people attended.” Statement: “Exactly 200 people attended.” The statement is more precise than the evidence.
If the source’s approximation allows 200 as one possibility but does not establish exactness, Not Given can be more appropriate than True. The key is whether exact equality is actually asserted.
48. Examples do not automatically establish frequency
A passage gives two examples of delays caused by weather. Statement: “Weather is the most common cause of delay.” Frequency ranking is not established by examples alone.
Do not infer “most common” from narrative prominence.
49. Repeated mention does not necessarily mean importance ranking
A factor can be discussed at length because it is complicated, not because it is the main cause. Statements containing “main”, “most important” or “primary” need explicit or clearly implied ranking evidence.
Text length is not a ranking metric.
50. Writer attitude can be mixed
A writer may welcome a technology’s benefits while opposing its current regulation. Statement: “The writer supports the technology.” That broad statement may be too vague unless the text’s stance can be represented at that level.
Decompose attitude by object: support what, oppose what, under which condition?
51. Rhetorical questions can reveal stance
“How many warnings do we need before action becomes unavoidable?” strongly suggests the writer believes action is already justified. A statement about urgency can be supported even without a declarative sentence.
Language function can carry viewpoint evidence.
52. Sarcasm can reverse literal wording
“Another brilliant decision” after a clear policy failure may express criticism rather than praise. Literal vocabulary alone would invert the Yes/No judgement.
Language-for-Effect and Ambiguity Resolution should establish intended stance before verification.
53. Unreliable narrators complicate factual verification
If the task asks what the narrator says, verify against the narrator’s claim. If it asks what actually happened in the story world, contradictory later evidence may matter more.
Receiver job defines the level of truth being tested.
54. Hypothetical conditions define a temporary truth world
A source says, “Assume the material expands when cooled.” Statement verification inside that hypothetical must follow the supplied rule even if real-world knowledge differs.
Source-defined worlds override external priors for the task.
55. Outside knowledge is neither confirmation nor contradiction
If you know from elsewhere that a statement is factually wrong but the source says nothing about it, the answer remains based on the source. External truth does not convert Not Given into False.
This is one of the most important discipline rules in evidence-bound reading.
56. The source can itself contain uncertainty
“The decline may have resulted from weather or reduced demand.” Statement: “Weather caused the decline.” The source lists weather as one possibility, not the established cause.
Possibility inclusion does not equal causal confirmation.
57. The source can explicitly reject certainty
“It is impossible to determine which factor was decisive.” Statement: “Cost was the decisive factor.” This is stronger than mere absence; the source explicitly denies unique determination.
The statement is contradicted by the source’s epistemic limit.
58. Verification should be local and targeted
Once the relevant source region is found, compare only the dimensions that could change the state: quantifier, owner, time, condition, causality, degree, comparison or existence.
Do not repeatedly reread the whole passage when the final question is whether “some” became “all”.
59. Confidence comes from state separation
High confidence is justified when the source directly supports or contradicts the proposition. Lower confidence is appropriate when paraphrase is complex or scope is subtle. Yet the learner still needs a final classification.
Calibration asks whether another check is likely to reveal confirmation, contradiction or genuine absence.
60. The first statement-verification operating model
WHAT EXACTLY DOES THE STATEMENT CLAIM? → WHERE IS THE RELEVANT SOURCE EVIDENCE? → IS THE MEANING EQUIVALENT? → DOES ANY SOURCE DETAIL CONTRADICT IT? → IF NOT, DOES THE SOURCE ACTUALLY SETTLE IT? → CONFIRMED / CONTRADICTED / NOT GIVEN → VERIFY SCOPE → MOVE.
Statement verification is disciplined boundary control. The learner must be willing to say True when wording changes but meaning survives, False when one small qualifier makes the claim incompatible, and Not Given when the source simply does not cross the evidential gap.
Technical Field Manual: Training Three-State Verification Until the Boundaries Become Automatic
The core system becomes dependable only when learners repeatedly separate confirmation, contradiction and absence across many kinds of wording. The following drills isolate the most common state-boundary failures: quantifiers, modality, chronology, conditions, causality, ownership, comparison, compound statements, statistical scope, viewpoint and outside knowledge. The aim is not to memorise tricks. It is to make proposition comparison fast enough that the final True / False / Not Given or Yes / No / Not Given decision becomes a consequence rather than a guess.
F1. Three-column evidence sorting
Give one statement and three source snippets: one confirms it, one contradicts it and one discusses the topic without settling it. Sort the snippets into CONFIRM / CONTRADICT / DOES NOT SETTLE. This removes label anxiety and trains the deeper evidence states directly.
F2. False-versus-Not-Given minimal pairs
Use paired sources differing by one sentence. Version A omits the disputed detail; Version B explicitly gives the opposite detail. The same statement should be Not Given in A and False in B. Contradiction becomes an added evidence relation rather than mere non-confirmation.
F3. Statement decomposition
Split a statement into actor, action, object, time, cause, condition, scope and degree where present. Mark which components the source confirms. Compound or heavily modified statements become easier once every hidden requirement is visible.
F4. Paraphrase normalisation
Rewrite both statement and source into plain propositions before comparing. “Operating expenses became unsustainable” becomes “running the system became too expensive”. This helps the learner compare meaning after lexical decoration has been removed.
F5. Keyword contradiction drill
Create statement and source with identical nouns but opposite relationships: prices rose → attendance fell versus prices rose → attendance rose. Shared vocabulary has almost no evidential value unless relation survives.
F6. Quantifier ladder
Use none, some, many, most, almost all and all. Ask which source claims confirm, contradict or fail to settle each statement. Scope differences become concrete and reusable.
F7. “Not all” logic drill
Compare “not all succeeded”, “none succeeded”, “some failed”, “some succeeded” and “most succeeded”. Ask what each entails and what remains open. This prevents partial and universal claims from collapsing together.
F8. “Only” exclusivity drill
Source says trials occurred in London and Paris. Compare statements “trials occurred in London and Paris” and “trials occurred only in London and Paris”. The second adds exclusivity. Learners decide whether the source actually closes the location set.
F9. Ranking-word drill
Train mainly, primarily, most important, largest, smallest, first and final against sources that merely mention several factors. Presence is separated from priority.
F10. Modality ladder
Arrange could, may, likely, will and must across identical content. Ask when one source modality confirms another and when the statement becomes stronger than evidence. Certainty becomes a dimension of proposition equality.
F11. Possibility-versus-result drill
Source says an intervention could improve outcomes. Later versions either report improvement, report no improvement or give no result. The same statement “outcomes improved” should become True, False or Not Given depending on final evidence.
F12. Prediction-versus-observation drill
Separate expect, predict, forecast and estimate from observed, found, recorded and measured. Learners identify whether the source describes future expectation or realised outcome before classifying statements.
F13. Past-present timeline
Build sources with once, previously, now, currently and no longer. Statements differ only in time reference. Timeline control becomes automatic.
F14. Initial-final stance drill
A writer or speaker changes position. Ask separate statements about the initial and final views. One person can make both statements true at different time points.
F15. Conditionality deletion
Remove “if humidity stays low”, “provided funding continues” or “when supervised” from source claims and compare with unconditional statements. Dropping one condition broadens the proposition.
F16. Necessary-versus-sufficient drill
Use requires, is enough to, only if, if, provided that and guarantees. Ask which logical direction each expression establishes. This protects against turning requirements into guarantees.
F17. Cause-arrow drill
Convert source and statement into arrows before classifying. Rainfall ↓ → reservoir ↓ versus reservoir ↓ → rainfall ↓. Reversal becomes visually obvious.
F18. Correlation-versus-cause drill
Pair observational association with causal statements. Add a second version where the source explicitly says one factor causes the other. The learner practises moving between overreach and confirmation depending on evidence form.
F19. Sequence-versus-cause drill
Use after, followed by and subsequently without causal verbs. Ask whether causal statements are established. Temporal proximity stops masquerading as explanation.
F20. Comparison inversion
Convert “A is cheaper than B” into equivalent “B is more expensive than A” and non-equivalent “A is more expensive than B”. Relational paraphrase becomes easier to verify.
F21. Criterion control
Source compares cost and reliability separately. Statements use broad words such as better, worse or preferable. Learners identify when overall judgement is not supplied.
F22. Superlative-universe drill
Source compares only A and B; statement calls A “the cheapest available option”. Mark the hidden universe “all available options” and check whether the source covers it.
F23. Writer-versus-source ownership
Provide a paragraph where the writer quotes two external voices and offers no clear endorsement. Statements about “the writer believes…” remain Not Given unless writer stance is actually established.
F24. Writer-rejection drill
Repeat the previous source but add “This interpretation is unconvincing.” A statement attributing the rejected view to the writer now becomes No rather than Not Given.
F25. Partial-agreement grid
Map writer stance across goal, method, timing and cost. Statements target one dimension at a time. Mixed stance becomes decomposable rather than ambiguous.
F26. Concession hierarchy
Use although, while, despite and nevertheless to separate acknowledged limitation from controlling stance. Yes/No judgements improve when main and concessive propositions are distinguished.
F27. Silence-versus-contradiction drill
Give statements about motive, price, location and number that the source never mentions. Then add explicit opposite details one at a time. Not Given changes to False only when contradiction enters.
F28. Exhaustive-list drill
Compare “reasons included A and B” with “the only reasons were A and B”. Test statements adding C. List completeness becomes explicit evidence.
F29. Definition-boundary drill
Give a source definition and several statements that add, remove or reverse defining features. Conceptual verification becomes precise rather than associative.
F30. Compound-statement splitting
Mark A, B and connector in statements containing and, because, although, while or so that. Classify each subclaim before deciding the whole statement.
F31. Half-true / half-false drill
Source confirms A and contradicts B. Statement says A and B. One contradicted component invalidates the conjunction.
F32. Half-true / half-unstated drill
Source confirms A and says nothing about B. Statement says A and B. The unresolved component prevents full confirmation without creating contradiction.
F33. Negation normalisation
Rewrite “did not reject”, “not uncommon”, “cannot be ruled out” and similar constructions without losing degree. The learner becomes less vulnerable to grammatical surface complexity.
F34. Average-versus-individual drill
Give average values and statements about every participant, the typical participant or one participant. Aggregation scope becomes a standard verification check.
F35. Sample-versus-population drill
Source says 60% of surveyed users; statements alternate between surveyed users and all users. Denominator control becomes explicit.
F36. Trend-versus-level drill
Costs fall but remain high; satisfaction rises but remains below target. Statements distinguish direction of change from absolute status.
F37. Range-versus-distribution drill
Source gives minimum and maximum age. Statements ask about most common age, average age and whether anyone exceeds the range. Boundary evidence is separated from internal distribution.
F38. Approximate-versus-exact drill
Use about, approximately, nearly and around against exact-number statements. The learner practises refusing unjustified precision.
F39. Frequency-ranking drill
Give several examples without frequency data, then add explicit “most common” evidence in a second version. Statements about prevalence change state only when ranking evidence appears.
F40. Rhetorical-question stance
Use rhetorical questions that strongly imply writer position and genuine questions that remain open. Learners classify function before Yes/No judgement.
F41. Sarcasm and literal reversal
Use identical positive words in sincere and sarcastic contexts. Verification follows intended stance, not dictionary polarity alone.
F42. Hypothetical-world drill
Give a fictional system with explicit rules that contradict real-world knowledge. Statements are judged inside the defined source world only.
F43. Outside-knowledge abstinence
Use familiar facts deliberately omitted from the source. Ask the learner to label statements about those facts. Not Given discipline strengthens when outside truth is tempting.
F44. Source-uncertainty drill
Use “may be A or B”, “cannot determine”, “evidence is inconclusive” and “likely but not certain”. Statements that force unique certainty are tested against explicit epistemic limits.
F45. Local-reread protocol
When uncertain, state the exact disputed dimension before rereading: owner, time, all/some, cause/sequence, likely/certain. Search for discriminating evidence rather than rereading the whole paragraph passively.
F46. Not-Given stopping rule
After locating the correct source region and checking likely paraphrase, ask whether further nearby text contains confirmation or contradiction. If neither exists and the source region is exhausted, stop. This prevents Not Given from becoming endless search.
F47. False proof requirement
No False answer is accepted in practice unless the learner can quote or paraphrase the contradictory source claim. This destroys the habit “I did not see it, so it must be false”.
F48. True proof requirement
No True answer is accepted merely because the statement seems reasonable. Learners identify the source proposition that fully supports it. Confirmation becomes evidence-based rather than intuition-based.
F49. State-change editing
Take one statement and edit one word at a time until its state changes from True to False to Not Given. Quantifier, time, condition, owner or certainty words become visibly powerful.
F50. Statement-author drill
Give a source paragraph and ask the learner to write one confirmed, one contradicted and one genuinely unresolved statement. Designing the three states demonstrates whether the distinction is understood deeply enough to generate.
F51. Same-source mixed-state set
Use one paragraph to support six statements distributed across all three states. Learners cannot rely on passage-level impressions; every statement requires local proposition comparison.
F52. Yes/No ownership rotation
Use writer, expert, critic and institution voices inside one passage. Statements alternate whose view is being tested. Attribution control becomes part of verification rather than a separate afterthought.
F53. Twelve-week progression
Weeks 1–2: three states, contradiction proof, silence discipline and paraphrase. Weeks 3–4: quantifiers, modality, chronology and conditions. Weeks 5–6: causality, comparison, ownership and mixed stance.
Weeks 7–8: compound statements, negation, statistics and ranges. Weeks 9–10: writer views, rhetorical stance, hypothetical worlds and outside-knowledge control. Week 11: timed mixed passages. Week 12: unseen independent verification with no mechanism labels.
F54. Six-week compressed progression
Week 1: diagnose False/Not Given confusion and keyword matching. Week 2: scope, modality, time and condition. Week 3: cause, comparison and compound claims. Week 4: writer ownership and stance. Week 5: statistics, outside knowledge and timed local verification. Week 6: fresh mixed test and targeted retest.
F55. Personal verification profile
Track recurring errors: False overuse, Not Given avoidance, quantifier loss, certainty inflation, time drift, causal overreach, ownership confusion, statistical scope or outside-knowledge intrusion. Profile-led practice is more efficient than repeating generic T/F/NG sets.
F56. Confidence-by-evidence-state
After difficult items, state whether confidence comes from direct confirmation, direct contradiction or exhausted source with no settlement. Not Given confidence can be as legitimate as True/False confidence when search boundaries are controlled.
F57. Timed verification set
Use mixed statements with one deliberately difficult Not Given item. Apply a time ceiling and parking rule. The learner practises containment without guessing prematurely.
F58. Error-log reconstruction
For each wrong item record statement proposition, source proposition, missed dimension and correct state. The correction lives at the semantic level rather than as a memorised answer letter.
F59. Independent mastery test
Give a fresh passage containing factual claims, quoted sources, mixed writer stance, data, conditions and one genuinely unresolved issue. The learner receives no hints about which mechanism each statement tests.
Mastery is shown when every state can be justified through source relation rather than test-taking intuition.
F60. The field-manual operating question
When uncertain, ask: What exact source proposition would have to exist to make this statement True, and what exact source proposition would make it False? If neither exists in the relevant source region, Not Given becomes the principled state rather than a fallback.
Iona checks the source boundary. Maren exposes the statement’s truth conditions. Leonie stops the search once the evidence state is settled. That is the discipline that turns three labels into one coherent verification system.
Applied Verification Atlas: Sixty Worked True, False, Not Given and Yes, No, Not Given Cases
The atlas uses short source–statement pairs so the evidence boundary stays visible. Every case asks two questions: what does the source actually establish, and what extra condition would be needed to move the statement into another state? The goal is not to memorise answers. It is to recognise the relation that makes each classification inevitable.
Atlas 1. Direct confirmation
Source: “The library opens at 9 a.m. on weekdays.” Statement: “The library opens at nine in the morning on weekdays.” State: True. The wording changes but time and condition remain identical.
Atlas 2. Direct contradiction
Source: “The library opens at 9 a.m. on weekdays.” Statement: “The library opens at 8 a.m. on weekdays.” State: False. The source supplies an incompatible opening time.
Atlas 3. Missing closing time
Source: “The library opens at 9 a.m. on weekdays.” Statement: “The library closes at 6 p.m.” State: Not Given. The source says nothing about closing time.
Atlas 4. Most versus all
Source: “Most participants completed the course.” Statement: “All participants completed the course.” State: Not confirmed. If “most” is used in its normal non-universal sense, the statement is incompatible with the source’s limited scope and should be treated as False in a task where “most” excludes all. The important lesson is to preserve the quantifier exactly.
Atlas 5. Some versus none
Source: “Some residents opposed the plan.” Statement: “No residents opposed the plan.” State: False. “Some” establishes at least one opponent, directly contradicting “none”.
Atlas 6. Some does not settle most
Source: “Some residents opposed the plan.” Statement: “Most residents opposed the plan.” State: Not Given. The source establishes existence, not majority.
Atlas 7. Exclusive “only” unsupported
Source: “The survey included schools in Singapore and Malaysia.” Statement: “The survey included schools only in Singapore and Malaysia.” State: Not Given unless the source closes the location set elsewhere. Mentioning two places does not prove exclusivity.
Atlas 8. Exclusive set directly contradicted
Source: “The survey included schools in Singapore, Malaysia and Thailand.” Statement: “The survey included schools only in Singapore and Malaysia.” State: False. Thailand is an explicit counterexample to the exclusivity claim.
Atlas 9. Main cause stated
Source: “Although staffing and timing caused difficulties, cost was the principal reason the project ended.” Statement: “The project ended mainly because of cost.” State: True. The source ranks cost as principal.
Atlas 10. Main cause merely mentioned
Source: “The project faced problems involving cost, staffing and timing.” Statement: “The project’s main problem was cost.” State: Not Given. The source lists factors without ranking them.
Atlas 11. Possibility confirmed as possibility
Source: “The intervention may reduce recovery time.” Statement: “The intervention could reduce recovery time.” State: True in meaning: both present reduction as a possibility rather than an established result.
Atlas 12. Possibility inflated to certainty
Source: “The intervention may reduce recovery time.” Statement: “The intervention will reduce recovery time.” State: Not Given or incompatible depending on the task’s strict treatment of modality. The source does not establish certainty. What matters is that it cannot be marked True.
Atlas 13. Predicted outcome versus observed outcome
Source: “Analysts expect demand to rise next year.” Statement: “Demand rose this year.” State: Not Given. Prediction about next year does not settle an observed outcome this year.
Atlas 14. Past state contradicted by current statement
Source: “The office used paper files until 2024, when it switched completely to digital records.” Statement: “The office currently uses paper files.” State: False. Current practice is explicitly digital.
Atlas 15. Historical statement preserved
Source: “The office used paper files until 2024.” Statement: “The office once used paper files.” State: True. “Once” correctly places the claim in the past.
Atlas 16. Conditional claim preserved
Source: “The device is reliable when humidity remains below 40%.” Statement: “The device is reliable in low-humidity conditions.” State: True if “low humidity” accurately represents the stated threshold in context.
Atlas 17. Condition removed
Source: “The device is reliable when humidity remains below 40%.” Statement: “The device is reliable in all conditions.” State: Not Given unless the source explicitly reports failure above 40%; the source establishes one reliable condition, not universal reliability.
Atlas 18. Condition explicitly violated
Source: “The device is reliable below 40% humidity but frequently fails above that level.” Statement: “The device is reliable in all conditions.” State: False. The source gives direct counterevidence.
Atlas 19. Necessary condition mistaken for guarantee
Source: “Applicants need a valid permit to enter the site.” Statement: “Anyone with a valid permit is guaranteed entry.” State: Not Given. The permit is necessary, but the source does not state it is sufficient.
Atlas 20. Correlation converted to cause
Source: “Students who slept longer tended to score higher.” Statement: “Longer sleep caused higher scores.” State: Not Given. Association does not establish causal direction.
Atlas 21. Causal statement confirmed
Source: “The researchers concluded that the new cooling system caused the reduction in failure rates.” Statement: “Failure rates fell because of the new cooling system.” State: True.
Atlas 22. Cause reversed
Source: “Low rainfall caused reservoir levels to fall.” Statement: “Falling reservoir levels caused lower rainfall.” State: False.
Atlas 23. Sequence presented as cause
Source: “Sales fell shortly after the announcement.” Statement: “The announcement caused sales to fall.” State: Not Given unless the passage establishes the causal link elsewhere.
Atlas 24. Comparative equivalence
Source: “Model A is cheaper than Model B.” Statement: “Model B costs more than Model A.” State: True. The comparison is reversed grammatically but preserved logically.
Atlas 25. Broad “better” unsupported
Source: “A is cheaper; B is more reliable.” Statement: “A is better than B.” State: Not Given. No overall criterion of “better” is supplied.
Atlas 26. Superlative overreach
Source: “A is cheaper than B.” Statement: “A is the cheapest system on the market.” State: Not Given. The comparison universe is much larger than the source.
Atlas 27. Writer quotes without endorsement
Source: “Dr Lee argues that the policy is too expensive. Other evidence points in several directions.” Statement: “The writer believes the policy is too expensive.” State: Not Given. Dr Lee owns the claim; writer endorsement is not established.
Atlas 28. Writer explicitly rejects quoted view
Source: “Dr Lee argues the policy is too expensive, but this objection ignores the substantial long-term savings.” Statement: “The writer agrees that the policy is too expensive overall.” State: No. The writer challenges the cost objection.
Atlas 29. Writer cautiously supports
Source: “The early results are promising, though more evidence is needed before nationwide expansion.” Statement: “The writer views the early results positively.” State: Yes. Qualification does not erase positive evaluation.
Atlas 30. Writer enthusiasm overstated
Source: “The early results are promising, though more evidence is needed.” Statement: “The writer is completely convinced the programme should expand immediately.” State: No if the writer explicitly withholds immediate expansion; the statement conflicts with the stated caution.
Atlas 31. Motive missing
Source: “The company closed its northern branch after revenue declined.” Statement: “Executives closed the branch to reduce staff numbers.” State: Not Given. The source gives timing and context, not the stated motive.
Atlas 32. Motive contradicted
Source: “The branch was closed solely because the building was unsafe; staffing levels were unchanged.” Statement: “The branch closed to reduce staff numbers.” State: False.
Atlas 33. Open list does not exclude extras
Source: “Problems included cost, staffing and timing.” Statement: “Poor communication was also a problem.” State: Not Given. “Included” does not close the set.
Atlas 34. Closed list excludes extras
Source: “The three problems were cost, staffing and timing.” Statement: “Poor communication was also a problem.” State: False. The source defines an exhaustive three-item set.
Atlas 35. Definition confirmed
Source: “Resilience is the capacity to recover after disruption.” Statement: “Resilience involves recovery after disruption.” State: True.
Atlas 36. Definition contradicted
Source: “Resilience is the capacity to recover after disruption.” Statement: “Resilience means preventing all disruption.” State: False. Prevention and recovery are not equivalent.
Atlas 37. Compound statement half false
Source: “The trial lasted six weeks and included 150 participants.” Statement: “The trial lasted six weeks and included 200 participants.” State: False. One conjunct is directly contradicted.
Atlas 38. Compound statement half unstated
Source: “The trial lasted six weeks.” Statement: “The trial lasted six weeks and was funded by the government.” State: Not Given. The funding component is unresolved.
Atlas 39. Negative statement preserved
Source: “The committee kept the proposal under consideration.” Statement: “The committee did not reject the proposal.” State: True if continued consideration logically excludes rejection in the source context.
Atlas 40. “Did not reject” does not mean “approved”
Source: “The committee kept the proposal under consideration.” Statement: “The committee approved the proposal.” State: Not Given. Non-rejection does not establish approval.
Atlas 41. Average versus every individual
Source: “Average completion time was ten minutes.” Statement: “Every participant finished within ten minutes.” State: Not Given. The mean does not define each individual value.
Atlas 42. Population scope widened
Source: “60% of surveyed users preferred the new design.” Statement: “60% of all users prefer the new design.” State: Not Given. Survey sample and total population are different denominators.
Atlas 43. Trend versus absolute level
Source: “Costs fell by 20% but remained higher than the regional average.” Statement: “Costs are now low.” State: False if “higher than regional average” clearly conflicts with the source’s defined notion of low; otherwise the broad adjective may be underdefined. The safe lesson is that a falling trend does not itself prove a low level.
Atlas 44. Range confirms upper bound
Source: “Participants were aged 18 to 25.” Statement: “No participant was older than 25.” State: True if the stated range is exhaustive.
Atlas 45. Range does not give mode
Source: “Participants were aged 18 to 25.” Statement: “Most participants were 25.” State: Not Given. Range provides boundaries, not frequency distribution.
Atlas 46. Approximate number versus exact number
Source: “About 200 people attended.” Statement: “Exactly 200 people attended.” State: Not Given. Approximation does not establish exact equality.
Atlas 47. Examples do not prove “most common”
Source: gives several weather-related delay examples. Statement: “Weather is the most common cause of delay.” State: Not Given unless frequency evidence or explicit ranking is supplied.
Atlas 48. Rhetorical question reveals stance
Source: “How many more warnings do we need before action becomes unavoidable?” Statement: “The writer believes action is already justified.” State: Yes, assuming the question functions rhetorically in context.
Atlas 49. Genuine question leaves stance open
Source: “What is the best way to finance the scheme?” followed by several options and no preference. Statement: “The writer supports public taxation as the best method.” State: Not Given.
Atlas 50. Sarcasm reverses surface praise
Source: after a costly failure, writer says “Another brilliant decision.” Statement: “The writer approves of the decision.” State: No if surrounding context clearly establishes sarcasm.
Atlas 51. Hypothetical rule controls truth
Source: “For this model, assume pressure falls as temperature rises.” Statement: “Within the model, higher temperature is associated with lower pressure.” State: True even if the relationship differs from a learner’s outside expectations.
Atlas 52. Outside knowledge cannot create False
Source: says nothing about the capital of a country. Statement: gives a capital the learner knows is wrong. State: Not Given for a source-bound task. External factual knowledge is not source contradiction.
Atlas 53. Source lists possibilities
Source: “The decline may reflect weather, reduced demand or both.” Statement: “Weather was the cause of the decline.” State: Not Given. Weather is possible, not uniquely established.
Atlas 54. Source explicitly denies unique cause
Source: “The evidence cannot determine which factor caused the decline.” Statement: “Weather was the decisive cause.” State: False. The source explicitly contradicts decisiveness.
Atlas 55. “At least” preserves open upper range
Source: “Applicants must be at least 18.” Statement: “Applicants may be older than 18.” State: True in the ordinary reading because the rule sets a minimum rather than an exact age, unless another source restriction closes the upper range.
Atlas 56. Minimum does not give maximum
Source: “Applicants must be at least 18.” Statement: “Applicants cannot be older than 40.” State: Not Given. No upper bound is supplied.
Atlas 57. “Twice as many” confirmed
Source: 20 participants chose A and 10 chose B. Statement: “Twice as many participants chose A as B.” State: True.
Atlas 58. Percentage wording altered
Source: A received 20 responses and B received 10. Statement: “A received 200% more responses than B.” State: False under standard percentage-change arithmetic: 20 is 100% more than 10, though it is 200% of 10.
Atlas 59. Partial writer support
Source: “The goal is necessary, but the proposed timetable is unrealistic.” Statement: “The writer supports the goal.” State: Yes. Statement: “The writer supports the proposed timetable.” State: No. Object of stance must be preserved.
Atlas 60. The clean Not Given endpoint
Source: provides a project’s date, location, cost and outcome but never names the director. Statement: “The project director was Dr Chen.” State: Not Given. Once the relevant project description is exhausted and no ownership evidence appears, further searching is low value.
Atlas Return: Every State Has a Different Proof Burden
True needs positive equivalence: the source supports the complete statement. False or No needs incompatibility: the source supplies an opposite, exception, different owner, different time, different degree or explicit rejection. Not Given needs disciplined absence: after locating the relevant source region, neither support nor contradiction settles the proposition.
The strongest learner does not ask which label feels safest. She asks what proof burden has actually been met. That habit makes three-state verification transferable across factual information, writer viewpoints, data, comparisons, conditions and unfamiliar topics.
Advanced Verification Laboratory: Edge Cases, Learner Profiles and Independent Mastery
The field manual trains the components and the atlas shows clean examples. This final laboratory focuses on cases where strong learners still hesitate: statements that are too strong but not explicitly denied, partial writer agreement, implicit exhaustiveness, logical implications, numerical language, grammatical negation, source uncertainty and the exact point at which continued searching stops being useful.
The guiding discipline is conservative but not timid. Do not mark True without support. Do not mark False without contradiction. Do not choose Not Given merely because paraphrase is difficult. Each state must earn itself through the relation between statement and source.
A1. Stronger statement, weaker source
Source: “The policy is likely to reduce traffic.” Statement: “The policy will reduce traffic.” The statement increases certainty. If the source does not reject the possibility of failure, the safest three-state judgement is usually Not Given rather than True. The key principle is that weaker evidence does not automatically contradict the stronger statement; it simply fails to establish it.
A2. Source explicitly denies certainty
Source: “The policy may reduce traffic, but its effect cannot be predicted with confidence.” Statement: “The policy will definitely reduce traffic.” Here the statement is not merely unsupported; the source explicitly rejects that certainty. False becomes stronger.
A3. Weaker statement, stronger source
Source: “All participants completed the programme.” Statement: “Some participants completed the programme.” The weaker existential claim is logically supported by the universal claim if at least one participant existed. True. Learners should understand logical entailment, not only lexical equivalence.
A4. Universal claim and empty-set caution
Formal logic can make universal statements vacuously true over empty groups, but examination passages usually describe real entities. Do not import abstract edge-case logic that the text does not invite. Interpret natural-language quantifiers in their ordinary contextual use.
A5. “Few” is not “none”
Source: “Few residents objected.” Statement: “No residents objected.” False because “few” normally means a small number greater than zero. Statement: “Most residents objected.” False in ordinary usage because “few” directly conflicts with a majority.
A6. “A few” and “few” differ pragmatically
“A few succeeded” often highlights some success; “few succeeded” highlights scarcity. Both imply more than zero but carry different emphasis. Verification should preserve factual quantity first and pragmatic framing second.
A7. “Almost all” excludes all in ordinary exam reading
Source: “Almost all participants completed the test.” Statement: “All participants completed it.” False in normal usage because “almost all” signals at least one exception. Statement: “Most participants completed it.” True because almost all entails most.
A8. “At most” creates an upper bound
Source: “At most 20 people can enter.” Statement: “More than 20 people may enter.” False. Statement: “Exactly 20 people will enter.” Not Given; capacity ceiling does not establish actual attendance.
A9. “At least” creates a lower bound
Source: “At least 50 responses were received.” Statement: “Fewer than 50 responses were received.” False. Statement: “Exactly 50 responses were received.” Not Given because the total may be higher.
A10. “Up to” does not state the actual value
Source: “The battery lasts up to ten hours.” Statement: “The battery always lasts ten hours.” False or Not Given depending on whether “up to” is treated as explicitly allowing shorter duration; in ordinary usage it does, so “always ten” conflicts with the stated maximum framing.
A11. Permission and actuality differ
Source: “Visitors may enter the archive with permission.” Statement: “Visitors enter the archive.” Not Given. Permission does not establish that entry actually occurred.
A12. Obligation and compliance differ
Source: “Staff must submit reports weekly.” Statement: “Staff submit reports every week.” Not Given unless compliance is reported. Rules describe required behaviour, not necessarily actual behaviour.
A13. Intention and action differ
Source: “The company plans to open two new branches.” Statement: “The company opened two new branches.” Not Given. Intent does not confirm completed action.
A14. Attempt and success differ
Source: “The team tried to reduce noise levels.” Statement: “Noise levels were reduced.” Not Given unless outcome is supplied. Effort is not result.
A15. Refusal and inability differ
Source: “The company refused to release the data.” Statement: “The company was unable to release the data.” Not Given. Choice and capacity are distinct explanations.
A16. “No evidence of X” does not always mean “evidence of not-X”
Source: “Researchers found no evidence that the treatment caused harm.” Statement: “The treatment caused no harm.” The source reports absence of evidence, not necessarily proof of absence. Unless the study design supports the stronger conclusion, Not Given is safer.
A17. Explicit evidence of absence changes the state
Source: “Continuous monitoring showed that none of the participants experienced the specified adverse effect.” Statement: “No participants experienced that adverse effect.” True. Here absence is directly observed and stated.
A18. “Failed to prove” does not mean “disproved”
Source: “The study failed to prove that the method improves memory.” Statement: “The study proved that the method does not improve memory.” False as a representation of the study’s conclusion; failure to establish benefit is not proof of no benefit.
A19. “Could not rule out” preserves possibility
Source: “Researchers could not rule out measurement error.” Statement: “Measurement error was definitely responsible.” Not Given. Remaining possible is not established cause.
A20. Presupposition can carry evidence
Source: “The company stopped exporting the product in March.” Statement: “The company exported the product before March.” Normally True because “stopped exporting” presupposes prior exporting, unless the context unusually cancels that presupposition.
A21. “Again” implies a previous occurrence
Source: “The system failed again on Tuesday.” Statement: “The system had failed before Tuesday.” True in ordinary usage. Statement: “The same component caused both failures.” Not Given unless the cause is stated.
A22. “Still” implies continuation
Source: “The store is still closed.” Statement: “The store was closed before the present moment.” True under ordinary interpretation. Statement: “The store will remain closed tomorrow.” Not Given.
A23. “Even” signals expectation, not exact probability
Source: “Even experienced technicians made the error.” Statement: “Experienced technicians were expected to be less likely to make the error.” This may be a reasonable implication, but task designers vary in how much pragmatic inference they expect. Use the local passage and official task style; do not overextend one discourse marker.
A24. Counterfactual condition reveals non-actuality
Source: “If the team had received funding, the second phase would have begun.” Statement: “The second phase began.” False if the counterfactual clearly implies funding was not received and the phase did not begin.
A25. Hypothetical future remains open
Source: “If funding arrives next month, the team will begin phase two.” Statement: “Phase two will begin next month.” Not Given because the condition’s fulfilment is not established.
A26. “Unless” encodes a condition
Source: “The event will proceed unless heavy rain is forecast.” Statement: “The event will proceed regardless of weather.” False because the source names a weather condition that can prevent it.
A27. Writer reports consensus without personal stance
Source: “Most researchers in the field favour regulation.” The writer provides no evaluation. Statement: “The writer favours regulation.” Not Given. Reporting consensus is not automatically self-alignment.
A28. Writer uses endorsement verb
Source: “The strongest account is Patel’s explanation that incentives matter more than awareness.” Statement: “The writer prefers Patel’s explanation.” Yes. The evaluative phrase assigns writer stance.
A29. Writer concedes fact but rejects conclusion
Source: “Costs have risen, but this does not justify abandoning the programme.” Statement: “The writer believes rising costs justify ending the programme.” No. Agreement about the fact of rising cost does not imply agreement with the proposed consequence.
A30. Writer accepts conclusion for another reason
Source: “The scheme should end, not because it is costly, but because it no longer achieves its purpose.” Statement: “The writer wants the scheme to end.” Yes. Statement: “The writer wants it to end because of cost.” No.
A31. Percentage points versus percent change
Source: support rose from 40% to 50%. Statement: “Support increased by 10 percentage points.” True. Statement: “Support increased by 10%.” False under standard mathematical wording; the relative increase is 25%.
A32. Median does not equal mean
Source: “The median income was $40,000.” Statement: “The average income was $40,000.” Not Given if “average” means mean. Statistical measure names matter.
A33. Highest value versus largest increase
Source: Region A ends highest, Region B rises most over the period. Statement: “Region A experienced the greatest increase.” False if the data explicitly show B’s larger change. Final level and change magnitude are different comparisons.
A34. Rate versus count
Source: City A has more accidents in total; City B has a higher accident rate per 1,000 residents. Statement: “City A is more accident-prone.” Not Given unless the criterion is defined. Counts and rates answer different questions.
A35. Per-capita language matters
Source: Country X emits more total carbon; Country Y emits more per person. Statement: “Residents of X emit more carbon on average.” False if the per-capita data show Y higher.
A36. “Doubled” versus “increased by 200%”
Source: value rose from 50 to 100. Statement: “The value doubled.” True. Statement: “The value increased by 200%.” False in standard percentage-change language; it increased by 100% and became 200% of the original.
A37. Relative ranking with ties
Source: A and B both scored 90, the highest score. Statement: “A scored higher than B.” False. Statement: “A had one of the highest scores.” True.
A38. “More than half” versus “half or more”
Source: exactly 50% agreed. Statement: “More than half agreed.” False. Statement: “At least half agreed.” True. Boundary words decide.
A39. “No later than” includes the boundary
Source: applications must arrive no later than Friday. Statement: “Applications arriving on Friday are permitted.” True unless a time-of-day restriction changes the rule. “No later than” includes Friday.
A40. “Before Friday” excludes Friday
Source: applications must arrive before Friday. Statement: “Applications arriving on Friday are permitted.” False under ordinary temporal interpretation.
Learner Profiles: Twelve Ways Three-State Verification Breaks
Profile 1. The False Overuser
If a statement cannot be found, this learner marks False. The repair is a contradiction-proof rule: every False must point to source evidence incompatible with the statement. If no contradiction can be named, reopen Not Given.
Profile 2. The Not-Given Avoider
This learner feels every statement must be decided by common sense, so missing details are forced into True or False. Repair with familiar facts deliberately omitted from source texts. The learner must accept that source-bound uncertainty is a valid final state.
Profile 3. The Keyword Confirmer
Shared nouns create automatic True responses even when cause, direction or scope changes. Repair with high-overlap contradictions where every content word repeats but the relation reverses.
Profile 4. The Quantifier Skipper
All, most, some, only and none are treated as small words. Repair with state-change editing: change only the quantifier and observe the label change.
Profile 5. The Certainty Inflater
May becomes will; association becomes cause; likely becomes certain. Repair with modality ladders and evidence-strength matching. The learner must preserve the source’s epistemic ceiling.
Profile 6. The Timeline Flattener
Past and current states are merged. A company that once used paper is treated as still using it. Repair with explicit initial / transition / current timelines.
Profile 7. The Cause Inventor
Sequence, correlation and plausible mechanisms are upgraded to cause. Repair with arrow diagrams and a rule: causal language in the statement requires causal evidence in the source.
Profile 8. The Ownership Drifter
Quoted views migrate to the writer. Repair with proposition-owner labels and Yes/No rotation among writer, expert, critic and institution voices.
Profile 9. The Outside-Knowledge Importer
Real-world knowledge overrides source limits. Repair with passages that intentionally omit famous facts. The answer key follows the source, not the learner’s encyclopedia.
Profile 10. The Statistical Scope Loser
Average becomes every individual; sample becomes population; trend becomes level. Repair with paired statements differing only by denominator or aggregation level.
Profile 11. The Compound-Statement Blur
One correct half makes the whole statement feel true. Repair by splitting conjunctions and causal compounds before classification.
Profile 12. The Endless Searcher
Not Given feels unsafe, so the learner rereads the whole text repeatedly looking for hidden proof. Repair with source-region exhaustion and stopping rules. Once the relevant region has been checked for equivalence and contradiction, absence can be a complete decision.
Full-System Stress Tests
Stress Test 1. One paragraph, all three states
Write six statements against one paragraph: two confirmed, two contradicted, two unstated. Mix paraphrase density so keyword matching cannot predict the state. The learner must justify each classification independently.
Stress Test 2. Quantifier-only changes
Keep content nouns identical and vary some, most, all, only, at least and at most. This reveals whether scope words are being processed under time pressure.
Stress Test 3. Writer versus quoted sources
Use an article containing three expert views and one writer conclusion. Yes/No/Not Given statements rotate ownership. The learner must map voice before stance.
Stress Test 4. Data-rich verification
Use percentages, ranges, averages, rates and trends. Statements deliberately switch denominator, level and comparison basis. This tests whether statistical language survives paraphrase.
Stress Test 5. Familiar-topic outside-knowledge trap
Choose a familiar topic and omit several widely known facts. Statements about those omitted facts must remain Not Given. The task tests source fidelity against confident prior knowledge.
Stress Test 6. Unfamiliar-topic paraphrase
Use a technical topic with clear internal definitions. Statements paraphrase relationships heavily. The learner must rely on the temporary source model rather than background knowledge.
Stress Test 7. Compound claim trap
Every statement contains two components. Some pairs are true/true, true/false, true/unstated, false/unstated. The learner must classify the whole proposition after component analysis.
Stress Test 8. Timed Not-Given containment
Place one hard Not Given item early in a timed set. The learner applies a local search ceiling, marks provisionally if needed and protects later items. Performance is measured by accuracy plus absence of cascade.
Self-Study Protocol
For every wrong or uncertain item, write four lines: statement proposition, source proposition, relation between them, final state. Then name the missed dimension: scope, time, cause, owner, modality, comparison, statistics, condition, compound claim or absence.
For False answers, always record the contradictory evidence. For Not Given answers, record which part of the statement remains unsettled. For True answers, record the paraphrase bridge. These three review habits train three different proof burdens rather than one generic “find evidence” strategy.
Retest the mechanism on a new domain after delay. A quantifier error corrected in a travel passage should be retested in science, policy or history. Transfer matters more than remembering the old answer.
Teacher and Tutor Protocol
When a learner says False, ask “What contradicts it?” When she says Not Given, ask “Which exact part remains unsettled?” When she says True, ask “What source meaning is equivalent?” These three questions expose whether the classification was evidence-based.
If the learner cannot locate the relevant source region, route upstream to Reading. If she locates it but cannot recognise paraphrase, route to Paraphrase Performance. If she understands both propositions but confuses contradiction with absence, remain in this owner and train the three-state boundary directly.
Do not overteach folklore such as “extreme words are usually False” or “Not Given appears a certain number of times”. Such heuristics create brittle answer-pattern behaviour. Extreme words should trigger scope checks; distribution patterns should carry no evidential authority.
Parent Guide: How to Review Without Giving the Label
Ask: “Where does the passage support this?” “Where does it say the opposite?” “If neither, what part is missing?” These questions lead directly to the three states without requiring specialist terminology.
If a child uses outside knowledge, ask “Does the passage say that?” If the answer is no, return to the evidence boundary. If the child marks False because a detail is missing, ask “What would the source have to say to contradict it?”
Final Mastery Audit: Fifty Checks for Three-State Verification
- Can the learner define True as complete support rather than familiarity?
- Can False be distinguished from mere lack of proof?
- Can Not Given be accepted as a complete evidence state?
- Can every long statement be decomposed into proposition components?
- Can paraphrase be recognised without shared vocabulary?
- Can shared vocabulary be rejected when relations differ?
- Can all, most, some, few, none and almost all remain distinct?
- Can “only” exclusivity be noticed?
- Can main, primary and most important trigger ranking checks?
- Can may, could, likely, will and must preserve different certainty levels?
- Can possibility be separated from realised outcome?
- Can prediction be separated from observation?
- Can past, present and future states remain distinct?
- Can initial and final positions be tracked separately?
- Can conditions remain attached to claims?
- Can necessary and sufficient conditions be distinguished?
- Can cause be separated from correlation?
- Can cause direction be preserved?
- Can sequence be separated from causation?
- Can comparative direction be preserved under paraphrase?
- Can broad “better” claims be rejected when no criterion is supplied?
- Can superlatives be checked against the correct comparison universe?
- Can writer-owned claims be separated from quoted-source claims?
- Can reporting an idea be distinguished from endorsing it?
- Can partial agreement be decomposed by object or dimension?
- Can concessions be separated from the controlling stance?
- Can silence about motive remain Not Given?
- Can an explicitly exhaustive list create contradiction through omission?
- Can a non-exhaustive list remain open?
- Can definitions be verified by necessary conceptual features?
- Can compound statements be split before classification?
- Can one false conjunct invalidate a conjunction?
- Can one unstated conjunct prevent full confirmation?
- Can negation and double negatives be normalised without losing degree?
- Can averages be distinguished from individual values?
- Can sample findings be distinguished from population claims?
- Can trend be separated from absolute level?
- Can ranges be distinguished from internal distributions?
- Can approximate values be distinguished from exact values?
- Can percentage points be distinguished from percentage change?
- Can counts be distinguished from rates?
- Can thresholds such as at least and at most be interpreted correctly?
- Can permission, obligation, intention and attempt be separated from actual outcomes?
- Can “no evidence of” be distinguished from “evidence of absence”?
- Can presuppositions such as stop, again and still be used cautiously?
- Can hypothetical source rules override outside knowledge?
- Can outside knowledge be excluded from source-bound verification?
- Can explicit source uncertainty limit stronger statements?
- Can the learner stop searching once the relevant region is exhausted?
- Can every final state be justified with the correct proof burden under time pressure?
Final Return: Unsupported Is Not False
The entire task can be reduced to one discipline: do not collapse three evidence states into two. A statement can be supported. It can be contradicted. Or the source can leave it unsettled. Not Given is not indecision; it is a precise conclusion about the limits of the available text.
This is why statement verification belongs beside the other global SETC performance owners but remains distinct from them. Reading finds meaning. Paraphrase recognises equivalence. Ambiguity resolves competing interpretations. Argument Evaluation judges evidence quality. Multiple Choice compares alternatives. Statement Verification decides the source–statement relationship itself.
Iona asks what the source establishes. Maren asks what the statement requires. Leonie asks whether the evidence boundary has already been reached. Together they protect the decisive habit: confirm only what is supported, reject only what is contradicted, and leave genuinely unsettled claims exactly where the text leaves them.
TRUE NEEDS SUPPORT. FALSE NEEDS CONTRADICTION. NOT GIVEN NEEDS NEITHER.
Iona protects what the source actually establishes. Maren exposes the exact proposition being tested. Leonie prevents absence from turning into endless searching and contradiction from being confused with uncertainty. Together they protect the core distinction on which the whole task depends: unsupported is not the same as false.