English Matching Performance | Headings, Information, People & Features

Matching questions do not ask only whether one option is right. They ask whether a whole network of targets and options has been mapped correctly. One wrong early assignment can distort later choices, especially when options are shared, reused, restricted or deliberately similar.

English matching tasks appear in many forms: paragraph-to-heading matching, statement-to-paragraph matching, person-to-opinion matching, feature-to-entity matching, speaker-to-view matching, sentence-ending matching and other shared-option formats. The surface changes, but the deep performance problem remains stable. The learner must understand each target, understand the option set, identify what kind of relation creates a valid match, locate evidence, preserve ownership, obey reuse rules and verify that the final set remains globally coherent.

This guide establishes SETC’s global/no-location owner for English matching performance. It does not replace Multiple-Choice Performance, Reading Performance, Source Synthesis, Information-Transfer Performance or component-specific matching pages elsewhere in the eduKate estate. Those owners supply meaning, evidence and format detail. Matching Performance owns the mapping layer across a shared option space.

Iona anchors evidence, ownership and scope. Maren makes the matching relation explicit: main idea, location, speaker stance, feature, cause, example, category, completion or another task-defined link. Leonie controls allocation, collision resolution and whole-set verification so one difficult target does not consume the option map or trigger unnecessary reshuffling.

The matching-performance chain

READ THE MATCHING RULE → DEFINE EACH TARGET → MAP THE OPTION SET → IDENTIFY THE MATCH RELATION → LOCATE EVIDENCE → TEST OWNERSHIP / SCOPE → APPLY REUSE RULES → RESOLVE COLLISIONS → VERIFY THE WHOLE SET → COMMIT.

The chain differs from ordinary multiple choice because each decision can affect the remaining option landscape. In one-use-only tasks, assigning Option C to Target 1 removes C from later consideration. In reuse-allowed tasks, the same option may legitimately match several targets. In heading matching, several headings may fit local vocabulary while only one captures the paragraph’s organising idea. In person matching, one speaker may express several opinions while another shares one of them. The learner therefore needs local evidence and global set control at the same time.

1. Read the matching rule before reading the content

Matching formats differ critically in whether options can be reused, whether some options remain unused, whether one option must match every target, whether multiple targets can share one option and whether the task is one-to-one or many-to-one. These are not clerical details. They change the logic of the entire solution space.

Leonie’s first move is procedural: identify the allocation rule. If reuse is allowed, never eliminate an option merely because it has already been used. If reuse is forbidden, every secure match changes the remaining candidate set. If extra options exist, do not force every option into service.

2. A target is a receiver job

A paragraph number, person name, statement, feature label or sentence stem is not merely an item identifier. It defines what kind of relation must be matched. Paragraph-to-heading asks for organising meaning. Statement-to-paragraph asks where specific information is located. Person-to-view asks who owns a proposition. Sentence-ending matching asks which completion produces the correct grammatical and semantic proposition.

Maren’s question is: What exactly must be true for this target and option to belong together? The answer to that question determines the evidence test.

3. Build an option map before assigning anything

Shared options should be read as a system. What categories exist? Which headings are broad? Which are narrow? Which people hold similar positions? Which sentence endings differ only in cause, degree or time?

A quick option map prevents repeated full rereading. It also reveals designed contrasts. If three headings concern causes, solutions and consequences, the task may be testing paragraph function rather than topic vocabulary. If two endings differ only in “because” versus “although”, the logical relation becomes the discriminating axis.

4. Options are propositions or functions, not labels

A heading such as “Unexpected benefits of the change” contains several components: benefits, they are unexpected, and they result from or concern a change. A paragraph mentioning one expected benefit does not satisfy the complete heading.

Likewise, a person-matching option such as “believes cost matters more than speed” is a relational claim. The person must own that comparison, not merely mention cost and speed separately.

5. Matching headings requires main-idea hierarchy

A heading should normally capture the organising job or central meaning of the paragraph, not one vivid example. Learners who choose headings through keyword overlap often match a memorable noun while missing the paragraph’s purpose.

Ask what proposition explains why the paragraph contains its examples. If several sentences describe failed early attempts before presenting a successful redesign, the paragraph may be about “How repeated failure led to a new approach” rather than “Problems with the original design”. The heading should fit the paragraph’s movement, not only its first half.

6. Paragraph topic and paragraph job are not identical

Two paragraphs can share the same topic while doing different work. One may define a technology; another may criticise its limitations; another may describe adoption. A heading naming only the technology can feel relevant to all three and therefore be too broad.

Iona asks: what is this paragraph doing with the topic? Definition, example, contrast, historical development, problem, response, evidence, limitation, prediction or another function often differentiates near headings.

7. First sentences are clues, not guaranteed headings

Some paragraphs state the main idea early. Others begin with an example, transition, concession or background detail and reveal the controlling point later.

Do not assign a heading after one sentence unless the paragraph is structurally obvious. Read enough to identify the main semantic centre. A distractor heading often matches the opening but fails the paragraph’s later turn.

8. Last sentences can reveal the controlling point

A paragraph may develop several details and end with a conclusion such as “For this reason, the project was redesigned completely.” That final statement can reveal that the paragraph’s true job is transition from evidence to redesign.

Heading decisions should follow the whole paragraph architecture. Opening and ending often provide the clearest hierarchy signals, but neither should be used mechanically.

9. Repeated vocabulary can be a distractor in heading matching

A heading repeats the word “cost”, and the paragraph uses “cost” three times. Yet the paragraph’s actual point may be how costs were reduced, while another heading says “A method for lowering expenses”.

Semantic relationship outranks lexical frequency. Read the proposition expressed around repeated words rather than counting them.

10. Broad headings are attractive because they fit too much

“Challenges of urban transport” may fit three paragraphs discussing cost, congestion and maintenance. A stronger heading may identify the specific challenge and paragraph job.

If one heading could apply equally well to several targets, it may be intentionally broad. Compare specificity and coverage before assigning.

11. Narrow headings fail when they capture only one example

A paragraph argues that remote work changes office design, commuting and recruitment, using reduced commuting as one example. Heading “Less commuting” is accurate but too narrow. The central idea concerns wider organisational change.

Heading matching is therefore a compression problem: choose the option that captures maximum central content without expanding beyond the paragraph.

12. Headings can differ by time frame

“Early attempts to solve the problem” and “Current solutions to the problem” may share vocabulary. Chronology distinguishes them.

Look for dates, tense, historical markers and development language. A paragraph can mention present technology while mainly explaining its origins. The heading should match the dominant time perspective.

13. Headings can differ by stance

“Why the scheme should expand” and “Why expansion may be premature” address the same policy. Modal and evaluative language decide.

Language-for-Effect and Argument Evaluation can supply stance interpretation; Matching Performance decides which heading owns that stance globally within the option set.

14. Matching information asks location, not summary

A statement-to-paragraph task may ask where a specific example, reason, date, comparison or claim appears. The target statement can describe one small part of a paragraph.

This differs from heading matching. A paragraph can match a statement without that statement being central. Learners who use “main idea” for every matching task apply the wrong relation.

15. Restate the target statement before scanning

Convert “a reference to an unexpected consequence of the policy” into a search target: find a consequence, it was unexpected, and it resulted from the policy.

This prevents scanning for “policy” alone. All components of the statement should be represented somewhere in the matching source region.

16. Paraphrase recognition is central to matching information

Target: “a decline that occurred despite increasing demand.” Source: “Interest continued to rise, yet participation fell.” No words match except perhaps broad concepts, but the relationship is equivalent.

Paraphrase Performance supplies semantic equivalence. Matching Performance uses it to locate the correct paragraph among several sharing the same topic.

17. One paragraph can match several information statements

If the rules allow reuse, a dense paragraph may contain a date, example and criticism that correspond to three separate statements. Do not assume one target per paragraph merely because headings often work that way.

Reuse rules must remain active throughout. Leonie should never eliminate a paragraph from consideration unless the format explicitly makes matches one-to-one.

18. One statement should normally map to the best source region

Several paragraphs may discuss the same issue, but one contains the specific relation the target describes. “A criticism based on cost” does not match every paragraph that mentions cost; it matches where cost functions as criticism.

Match relation + content, not content alone.

19. Examples and general claims must stay distinct

Target: “an example of a community adapting to water shortage.” Paragraph A gives a general theory of adaptation; Paragraph B describes one town changing crop choices. B matches the example target.

Information matching often hinges on discourse role: example, definition, cause, criticism, prediction, comparison. The source paragraph must contain the requested role, not merely the topic.

20. Matching people requires proposition ownership

A passage may contain several researchers, interviewees, writers or speakers. Matching tasks ask which person holds a view, reports a finding, proposes a solution or expresses a concern.

Map person → propositions before matching statements. Names alone are weak anchors because the passage may report one person discussing another’s idea.

21. Mentioning an idea is not owning it

“Chen rejects Patel’s claim that cost is the main barrier.” Chen mentions the cost claim but does not own it. Patel owns the claim; Chen owns the rejection.

Person matching must track stance and attribution. Source Synthesis supplies provenance discipline; Matching Performance applies it across a shared name set.

22. Agreement can make two people look identical

Two experts may both support a policy but for different reasons. If the target says “supports because of lower cost,” only the expert who gives cost as the reason matches.

Ownership includes proposition structure: stance + reason + condition. Shared stance does not imply shared match.

23. Conditional views must retain their condition

A speaker supports expansion only if independent testing succeeds. A target saying “supports immediate expansion” does not match, even though the broad direction is positive.

Conditions are part of proposition ownership. Matching statements must preserve them.

24. People can change their minds

An interviewee may describe an earlier opinion and then explain a revised view. If the target asks current position, match the final stance. If it asks who “initially opposed the plan,” the earlier stance becomes relevant.

Person + time is the full ownership key when views change.

25. Pronouns can obscure person ownership

A paragraph names several people and later uses “she”, “they” or “the researcher”. Resolve reference before assigning a target statement.

Ambiguity Resolution operates upstream. A matching answer built on the wrong antecedent can contaminate the whole person map.

26. Matching features asks entity-property binding

Options may be cities, technologies, species, programmes or theories. Targets describe properties: “uses the least energy”, “was developed first”, “requires specialist staff”, “works best in dry conditions”.

Create an entity-property matrix. A correct match binds the property to the right entity, not merely to the right topic cluster.

27. Nearby properties can migrate between entities

A paragraph compares System A and B in adjacent sentences. Learners often remember the property but lose which system owns it.

Use explicit labels or a small matrix during difficult sets. Entity binding is more important than raw fact memory.

28. Comparative properties require a comparison set

“Cheapest” means lowest cost among the compared entities, not simply “cheap”. “Most reliable” may be inferred only if the source provides enough comparative evidence.

Matching features often tests ranking language. Preserve the comparison universe and criterion.

29. Matching sentence endings is proposition construction

A sentence stem plus ending must form one grammatical and semantically accurate proposition. The correct ending is not merely the phrase most related to the source topic.

Read the complete sentence after every serious candidate. Grammar, logical relation, source meaning and scope all have to survive.

30. Grammar can eliminate endings before source comparison

If the stem ends “because…”, an ending functioning as a result may be logically impossible. If it ends “is likely to…”, an ending beginning with a noun phrase may fail grammar.

Use grammar as a first compatibility filter, then verify meaning. A grammatical ending is not automatically correct, but an ungrammatical one can often be rejected immediately.

31. Logical relation can eliminate endings

Stem: “The researchers rejected the early model because…” Ending A gives a reason; Ending B gives what happened afterward. Both may describe true facts, but only the causal completion fits the relation encoded by “because”.

Sentence-ending matching is information transfer through grammar plus logic.

32. Near endings often differ by scope

One ending says “works in coastal areas”; another says “works only in coastal areas”. If the source says coastal use is common but not exclusive, “only” invalidates the second.

Scope words remain high-value discriminators inside matching sets.

33. Near endings often differ by certainty

“May reduce costs” and “will eliminate costs” can share the same topic. Modality determines fit. Match source certainty exactly enough for the receiver.

Language Control and Argument Evaluation provide the semantic distinction; Matching Performance applies it across the ending bank.

34. Sentence-ending tasks can contain unused options

Extra endings are often deliberately plausible. Do not force every phrase into a stem. The global set can remain complete while some options are unused if the instructions allow it.

Unused options are not evidence of an error unless the task contract says every option must be used.

35. One-use constraints create a matching board

When each option may be used once, secure matches change the remaining board. But do not make an uncertain early match simply to reduce choices.

Use confidence tiers. Lock strong matches; pencil or mentally hold moderate matches; delay weak collisions. Constraint propagation should amplify evidence, not replace it.

36. Reuse-allowed tasks require independent evidence

If Option B can match several targets, using it once should not reduce its later probability. Learners trained on one-use formats often make this error automatically.

Reset the mental model: each target must be tested independently against the shared option set, even if repeated answers look visually uncomfortable.

37. Extra options exist to create plausible collisions

An unused heading may fit a small detail in one paragraph but not the whole paragraph. An unused person statement may be plausible but owned by nobody. Extra options increase discrimination demand.

Do not search for a home for every option. Search for the best option for every target.

38. Collision means two targets want the same restricted option

In one-use tasks, if Targets 2 and 5 both seem to require Option D, at least one local decision is wrong—or the rule was misunderstood.

Do not randomly move one. Compare the two target-option fits directly. Which match has stronger evidence? Which target has an alternative that fits nearly as well? Global collision reveals where to inspect locally.

39. Collision resolution should begin with the weaker match

If D is directly supported for Target 2 but only vaguely plausible for Target 5, keep D with Target 2 and reopen Target 5. Strong evidence should not be sacrificed merely to create a tidy set.

Leonie protects high-confidence anchors while using remaining constraints to guide the uncertain items.

40. Global set verification catches local errors

After all matches are tentatively filled, inspect the pattern under the actual rules. Has a one-use option been duplicated? Has a reuse-allowed option been rejected merely because it was used before? Does one heading remain unused appropriately? Does one target have no plausible support?

Whole-set review should identify structural anomalies, not trigger indiscriminate reshuffling.

41. Set coherence is a check, not primary evidence

“This arrangement uses every option exactly once” is not proof unless the task requires that structure. A neat permutation can still be semantically wrong.

Local evidence remains primary. Global consistency can expose mistakes but should not manufacture matches unsupported by the source.

42. Solve the easiest anchors first when the format benefits from it

In one-use matching, obvious anchors can reduce the remaining set and reveal contrasts. In reuse-allowed information matching, target order may matter less; solving easier items first can still build a map of the text.

However, do not spend excessive time scanning all targets for the easiest one if the task naturally follows source order. Strategy should fit format, not become ritual.

43. Hard targets should be parked without poisoning the set

If two headings remain plausible for one paragraph, mark both mentally and continue. Later assignments may resolve the collision, especially under one-use constraints.

Parking is disciplined postponement, not avoidance. Record the exact uncertainty so later evidence can close it.

44. Repeated rereading should become targeted comparison

When stuck between Heading B and F, do not reread the paragraph without a question. State: B says paragraph is mainly about cause; F says it is mainly about consequence. Then reread for which relation controls the paragraph.

Targeted rereading is faster because it seeks discriminating evidence rather than general familiarity.

45. Matching tasks punish local keyword heuristics repeatedly

Because options are shared, one broad keyword can attract the learner across several targets. “Education” may appear in three headings and four paragraphs.

Repeated topic overlap is often intentional. The task asks what distinguishes the targets within the shared topic. Relationships, function, stance and scope are the discriminators.

46. Matching can test negative evidence

A heading may require a problem, but the paragraph contains only successful outcomes. Even without a direct contradictory sentence, absence of the required function weakens the match.

Be careful: absence is useful only when the option’s essential component should be visible if it were truly the paragraph’s main job. Do not overuse “not mentioned” as proof in dense texts.

47. Matching confidence should be target-specific

One match may be direct, another inferred, another unresolved between two options. Do not let uncertainty about one item lower confidence in secure anchors.

Mark mental confidence by separation strength. This helps whole-set review focus where rechecking can actually change the outcome.

48. A one-use constraint can raise confidence—but only after evidence

If four of five matches are strongly established in a one-to-one task, the final unused option may determine the last target. That is legitimate constraint propagation when earlier matches are secure.

But if the earlier matches are weak, the apparent “last option” can lock in a chain of mistakes. Constraint power depends on anchor quality.

49. Do not reshuffle secure answers to satisfy intuition

Repeated use of the same letter in reuse-allowed tasks or an unused attractive option can feel suspicious. These feelings are not evidence.

Change a match only when new textual, grammatical, ownership or rule evidence appears.

50. Matching errors have distinct mechanisms

Rule error: reuse or one-use instructions misunderstood.
Target error: wrong relation assumed for the task type.
Keyword match: shared vocabulary replaces semantic fit.
Hierarchy error: example matched where main idea is required.
Ownership error: right proposition assigned to wrong person or entity.
Scope error: condition, degree or time removed.
Collision error: restricted option assigned twice without resolution.
Constraint overreach: remaining-option logic overrides weak local evidence.
Reuse suppression: a valid repeated option is rejected because it was used before.
Overchecking: secure anchors are reshuffled without new evidence.

51. Matching review should ask why the wrong option looked right

If a learner chose Heading C because one phrase repeated, the repair is keyword control. If she chose Person B because B discussed the idea while Person A actually endorsed it, the repair is proposition ownership. If she duplicated a one-use option, the repair is allocation control.

Error review becomes transferable when it names the mapping mechanism rather than the topic.

52. Paragraph mapping can support heading matching without replacing it

Brief paragraph-job notes—defines problem, gives example, evaluates limitation—can reduce repeated rereading. But annotation should remain light enough that the learner still compares actual paragraph meaning with headings.

PSLE-specific passage-mapping pages own their level-specific system. This global owner uses the general principle only as a support for matching decisions.

53. Person maps should separate claim, reason and condition

Instead of writing “Ahmed = supports policy”, record “supports policy because low cost, if safeguards added”. This richer proposition prevents later target statements from collapsing nuanced views into broad labels.

The map should remain compact; its purpose is discrimination, not transcription.

54. Feature maps should use entity rows and property columns

For several technologies, a small grid of cost / speed / condition / limitation can externalise the relationships most likely to be matched.

This is particularly useful when facts are dense and similar. The grid reduces cross-entity contamination while preserving comparison.

55. Sentence-ending maps should focus on grammar plus relation

Mark each ending as cause, result, condition, purpose, description or another role where useful. Then compare only logically compatible endings before source verification.

This reduces the shared option set without resorting to topical guessing.

56. Matching performance needs timed practice only after mapping accuracy stabilises

Untimed practice should first teach target definition, option structure and evidence. Timing is then introduced to test whether learners preserve those operations under pressure.

Speed should come from smaller candidate sets and faster discrimination, not from skipping rule reading or assigning headings after the first keyword.

57. Matching tasks reward selective externalisation

A tiny option map or person grid can save working memory when several similar candidates interact. But full transcription slows the task.

Externalise only the dimensions that discriminate: paragraph job, person stance, entity property, unresolved finalist. Everything else can remain in the source.

58. Easy matches should become anchors, not assumptions

A direct phrase or unmistakable main idea can be locked provisionally. Yet whole-set verification should still confirm no rule conflict or stronger competing fit exists.

An anchor is evidence-backed. An assumption is merely convenient. The distinction matters when later collisions appear.

59. The correct final set should be explainable locally

After using global constraints, the learner should still be able to justify each match from the source. “It was the only option left” is insufficient if no semantic evidence supports the pair.

Constraint propagation narrows; evidence confirms.

60. The first matching-performance operating model

WHAT DOES THE TARGET REQUIRE? → WHAT DOES EACH RELEVANT OPTION REALLY MEAN? → WHAT SOURCE EVIDENCE BINDS THEM? → DOES OWNERSHIP / SCOPE / TIME FIT? → IS REUSE ALLOWED? → DO OTHER MATCHES CREATE A COLLISION? → RESOLVE THE WEAKEST LINK → VERIFY EVERY LOCAL MATCH → VERIFY THE SET → MOVE.

Matching performance is a constrained mapping problem. Reading supplies meaning. Paraphrase supplies equivalence. Source Synthesis supplies ownership. Ambiguity Resolution separates near interpretations. Multiple Choice supplies distractor comparison. Matching adds one distinctive demand: several targets and options coexist inside one shared space, so local decisions must survive global constraints.

Technical Field Manual: Training Matching as a Whole-Set Decision System

The core article establishes the logic of matching. This field manual turns that logic into deliberate practice. Each drill isolates one failure point: instruction rules, target typing, heading hierarchy, information location, person ownership, feature binding, sentence-ending grammar, collision resolution, reuse control or whole-set verification. The drills should eventually be mixed so learners must select the correct operation without being told which mechanism is active.

F1. Rule rewrite

Before solving, rewrite the allocation rule in plain English: “Each heading once; two headings unused”, “paragraph letters may repeat”, “each person can be used more than once”, “each sentence ending only once”.

This turns instructions into operational constraints. Learners who skip this step often apply the wrong matching mathematics despite understanding the text.

F2. Rule-switch drill

Use the same option set under two different rules: reuse allowed versus reuse forbidden. Ask how the strategy changes.

The learner experiences that option removal is logical only when the task authorises it. Matching habits should follow the current contract, not the previous worksheet.

F3. Target-type labelling

Label targets MAIN IDEA, LOCATION, OWNER, PROPERTY, COMPLETION, CAUSE, EXAMPLE or another relation before comparing options.

This prevents one universal “find the similar phrase” method from being applied to tasks that demand different evidence relationships.

F4. Option decompression

Take a compact heading such as “A surprising obstacle to expansion” and unpack it: obstacle; related to expansion; surprising or unexpected.

Every component must fit. A paragraph about an expected obstacle does not match merely because “expansion” appears repeatedly.

F5. Option-family mapping

Group headings or statements by contrast: causes / effects / solutions; positive / negative / conditional views; early / current / future stages.

The group map exposes the dimensions the option writer is using to create distractors.

F6. Paragraph-job compression

After each paragraph, write no more than five words describing its job: defines problem; gives historical origin; challenges old theory; presents new evidence.

Then compare those job labels with headings. This creates an intermediate representation less vulnerable to copied vocabulary.

F7. Topic-versus-job pairs

For each paragraph write topic and job separately: “solar power / explains storage problem”; “solar power / reports falling cost”.

The drill teaches why headings with the right topic can still be wrong when they misrepresent what the paragraph does.

F8. Opening-versus-whole test

Show only the first sentence and ask for a provisional heading. Then reveal the full paragraph and require an update.

Learners see how examples, concessions and later turns can make early heading choices unstable.

F9. Final-sentence hierarchy test

Use paragraphs where the last sentence reframes earlier details. Ask which heading fits before and after reading the conclusion.

This trains whole-paragraph hierarchy without turning the final sentence into a mechanical answer key.

F10. Broad-heading rejection

Create one heading that could fit three paragraphs broadly and three narrower headings that capture each paragraph’s actual function.

Learners must explain why broad relevance is weaker than specific central fit.

F11. Narrow-heading rejection

Use one vivid example inside a paragraph and create a heading around it. Compare with a heading capturing the broader organising proposition.

The exercise trains level control: detail is not automatically main idea.

F12. Heading time-axis drill

Use headings differing only by period: origins, early attempts, current practice, future challenge. Mark temporal language in the paragraph.

Chronology becomes an explicit heading discriminator.

F13. Heading stance-axis drill

Give headings sharing the same topic but differing by support, criticism, uncertainty or conditional approval.

Learners identify modal verbs, concessions and evaluative language before choosing.

F14. Keyword decoy drill

Design a heading containing the paragraph’s most repeated noun but the wrong relation. The correct heading uses different vocabulary.

This weakens lexical matching and strengthens semantic paraphrase recognition.

F15. Information-target decomposition

Target: “a reference to a failed attempt caused by lack of funding.” Break into failed attempt + cause = lack of funding.

Scan for all components, not one noun. The paragraph must contain the requested relation, not just shared topic vocabulary.

F16. Paraphrase-location drill

Write target statements using no words that appear in the source paragraph while preserving meaning.

Learners must locate through semantic equivalence, reducing dependence on exact lexical overlap.

F17. Role-of-information drill

For matched information, label the source function: example, cause, definition, prediction, criticism, comparison, evidence or consequence.

This helps when several paragraphs mention the same subject but use it for different discourse jobs.

F18. Reuse-allowed paragraph drill

Build a paragraph containing three targetable facts and allow its letter to be reused. Learners must match all three statements to the same paragraph when evidence requires it.

The drill removes the mistaken belief that repeated letters look suspicious.

F19. One-use paragraph drill

Use a different task where each paragraph receives one heading only. Learners compare how elimination becomes legitimate only under the new rule.

Alternating F18 and F19 builds rule sensitivity.

F20. Person proposition map

Create a compact row for each person: stance / reason / condition / time. Fill only what discriminates.

This prevents names from becoming empty labels and makes near-opinion statements easier to compare.

F21. Mention-versus-own drill

Write sentences where Person A quotes or rejects Person B. Ask who owns the embedded proposition and who owns the reaction.

This is one of the highest-value person-matching drills because reported ideas often create attractive ownership errors.

F22. Shared-stance different-reason drill

Three people all support the same plan for different reasons. Target statements specify the reason.

Learners must preserve proposition detail rather than matching broad positive stance.

F23. Conditional-opinion drill

Give speakers who support an idea only under different conditions. Target options omit, preserve or alter those conditions.

Conditions become ownership-bearing evidence rather than optional nuance.

F24. Opinion-update drill

A speaker changes position after new evidence. Ask separate targets about initial and final view.

The learner records person + time, preventing abandoned views from contaminating current stance.

F25. Pronoun-to-person drill

Use paragraphs with several names followed by pronouns. Require antecedent resolution before any matching decision.

This creates a clean interface between Ambiguity Resolution and person matching.

F26. Feature matrix

Create rows for entities and columns for cost, speed, limitation, condition, date or another relevant property.

Fill only high-value distinctions. Dense feature matching becomes a binding problem rather than a memory test.

F27. Property-migration trap

Present two entities in one sentence and attach separate properties. Later ask which feature belongs to which entity.

The drill targets cross-entity contamination directly.

F28. Comparative feature drill

Ask who is cheapest, fastest or most stable when the source gives pairwise comparisons rather than one explicit ranking.

Learners construct the comparison set before assigning the target.

F29. Sentence-ending grammar filter

Hide the source and eliminate endings that cannot grammatically complete each stem.

Then reveal the source and decide among the grammatical survivors. Grammar reduces the candidate set but does not replace semantic verification.

F30. Sentence-ending relation filter

Label endings cause, result, purpose, condition, description or comparison. Match only endings compatible with the stem’s logical connector.

This turns sentence completion into relation matching rather than topical association.

F31. Ending scope contrast

Use near endings differing by only, some, mainly, always or under certain conditions.

Small modifiers become the central discriminators.

F32. Ending certainty contrast

Compare may, likely, will and must endings against source modality.

Learners practise matching evidential strength as part of the proposition.

F33. One-use option board

Write targets vertically and options horizontally. Cross out only secure assignments. Leave uncertain options active for several targets.

This externalises constraint propagation without pretending provisional matches are final.

F34. Reuse option board

Use the same visual board but never cross out options after use. Mark evidence independently for every target.

Alternating one-use and reuse boards develops procedural flexibility.

F35. Collision diagnosis

Deliberately create two targets tentatively assigned the same restricted option. Ask which match has stronger local evidence and which target has the better alternative.

Collision becomes a diagnostic signal rather than a reason to reshuffle randomly.

F36. Anchor-strength grading

Label matches strong, moderate or weak based on evidence separation.

Only strong anchors should drive elimination aggressively. Weak anchors remain candidates, not constraints.

F37. Last-option test

In one-to-one practice, leave one target and one option at the end. Ask whether the learner can still find positive evidence for the final match.

This trains the rule: remaining option suggests; local evidence confirms.

F38. Extra-option tolerance

Give more headings than paragraphs and make two distractors highly attractive. Learners must finish comfortably with unused options.

The drill removes the urge to “use everything”.

F39. Repeated-answer tolerance

In reuse-allowed tasks, design a set where one option correctly matches three targets.

Learners experience that repeated letters can be normal and evidence-based.

F40. Parking protocol

When two candidates remain, write both beside the target and state the unresolved difference: cause versus consequence, support versus conditional support, example versus main idea.

Continue. Later matches or rereading can resolve the exact uncertainty without restarting the whole task.

F41. Targeted reread

Before rereading, state what evidence would distinguish the finalists. Search only for that relation.

This converts rereading from repetition into information gain.

F42. Negative-evidence check

For a candidate heading, ask whether its essential feature appears at all. A heading about criticism needs criticism, not merely the same topic.

Use absence cautiously and only when the missing component is central to the option’s meaning.

F43. Whole-set anomaly scan

After tentative completion, scan only for rule violations, impossible duplicates, unsupported orphan targets and suspiciously weak anchors.

Do not re-solve every secure item. Whole-set review should target anomalies.

F44. Local-proof reconstruction

Select random final matches and require one sentence of justification from the source.

If a match can only be justified as “the option was left over”, reopen it.

F45. Easy-anchor timing

Mix obvious and difficult targets. Learners should lock genuinely direct matches quickly and spend comparison time where separation is small.

Time allocation follows uncertainty and potential constraint value.

F46. Hard-target containment

Place one intentionally ambiguous target early. Require parking after a sensible effort.

The exercise protects the rest of the set from one difficult collision.

F47. Matching-error ledger

Record errors under rule, target relation, keyword, hierarchy, ownership, scope, collision, reuse, grammar or overchecking.

The ledger should reveal repeated mechanisms across different topics and matching formats.

F48. Distractor-author drill

Give a paragraph and ask the learner to write one broad heading, one narrow heading and one keyword-copy heading that are all tempting but wrong.

Designing distractors strengthens hierarchy and relation awareness.

F49. Person-distractor authoring

Write a statement that attributes Person A’s reason to Person B while preserving the shared topic.

Learners see how ownership can be corrupted without changing vocabulary.

F50. Ending-distractor authoring

Create an ending that is grammatically correct and factually true elsewhere but logically wrong for the stem.

This teaches why complete proposition fit matters.

F51. Same-text multi-format drill

Use one passage for heading matching, information matching and feature matching.

The learner experiences that source meaning stays constant while the matching relation changes.

F52. Same-relation new-domain drill

Practise matching headings across ecology, technology, history and social issues while keeping the main-idea relation constant.

Transfer becomes independent of topic familiarity.

F53. Listening speaker matching

Use three speakers with overlapping vocabulary but different stance, reasons and conditions.

The learner maps final proposition ownership rather than keywords or voice familiarity.

F54. Visual-feature matching

Match features to diagrams, advertisements or locations where icons and labels define properties.

The same entity-property binding principle survives when evidence is multimodal.

F55. Twelve-week progression

Weeks 1–2: rule reading, target types, option decompression and paragraph-job mapping. Weeks 3–4: heading hierarchy, information location and paraphrase. Weeks 5–6: person ownership and feature matrices.

Weeks 7–8: sentence endings, grammar and logical relations. Weeks 9–10: one-use/reuse constraints, collisions and parking. Week 11: mixed timed sets. Week 12: unseen transfer without labelled matching mechanisms.

F56. Six-week compressed progression

Week 1: diagnose rule, keyword and hierarchy errors. Week 2: headings and information. Week 3: people and features. Week 4: sentence endings and option-set logic. Week 5: collisions, timed parking and whole-set verification. Week 6: fresh mixed matching test.

F57. Personal matching profile

Track whether the learner fails through keyword matching, broad headings, narrow headings, ownership drift, scope loss, grammar, one-use constraint errors, reuse suppression or overchecking.

Different error profiles require different drills even when the final score is identical.

F58. Confidence-by-separation

Rate each difficult match by how clearly the winner separates from the nearest rival.

This helps calibration and tells Leonie which items deserve review.

F59. Independence test

Give a mixed matching section without labels such as “heading” or “ownership” beyond the real task instructions. The learner must infer the relation, manage the option set and verify globally.

Mastery is flexible mapping under constraints, not success on one familiar matching worksheet.

F60. Field-manual operating question

When a match looks plausible, ask: What exact relation binds this target to this option, and what evidence would make the nearest rival fail?

Iona protects the evidence. Maren names the relation. Leonie protects the board. Once those three operations become automatic, matching stops feeling like a puzzle of letters and becomes what it really is: structured semantic allocation.

Applied Matching Atlas: Fifty Worked Collisions Across Headings, Information, People, Features and Endings

The atlas turns the matching system into concrete decisions. Each case contains at least one attractive near-match. The task is to state why the correct pairing fits more completely, why the rival fails, and whether any global constraint changes the remaining option board.

Atlas 1. Heading trap: topic match, wrong job

Paragraph: several early electric-car designs are described, each followed by the technical reason it failed. Heading A: “The history of electric cars.” Heading B: “Why early designs repeatedly failed.” Both share the topic. B captures the paragraph’s controlling job: failed attempts plus reasons.

Atlas 2. Heading trap: one vivid example

Paragraph explains how remote work changes recruitment, office size and commuting, using shorter commutes as one illustration. “Less commuting” is a true detail and a weak heading. “How remote work reshapes organisations” covers the central movement.

Atlas 3. Heading trap: broad option

Three paragraphs all discuss city transport. Heading “Problems in urban transport” could apply broadly to all of them. One paragraph specifically explains why maintenance costs rise as systems age. A more precise heading about ageing infrastructure is the stronger match.

Atlas 4. Heading trap: opening sentence decoy

A paragraph opens by describing high construction cost, then spends most of its space showing how modular design reduced that cost dramatically. Heading “The high price of construction” matches the opening; “A design that lowered construction cost” matches the paragraph as a whole.

Atlas 5. Heading trap: later turn

The first half praises a new policy; the second half presents evidence that benefits disappear after six months. Heading “Early success, later disappointment” captures the paragraph’s movement better than “Benefits of the new policy”.

Atlas 6. Heading trap: expected versus unexpected

The paragraph reports that a cheaper material also turned out to be stronger, surprising the researchers. Heading “Advantages of the cheaper material” misses the surprise component. “An unexpected strength advantage” fits more completely.

Atlas 7. Heading trap: past versus present

A paragraph discusses nineteenth-century methods and ends by noting they are no longer used. “Current production methods” is topic-adjacent and time-wrong. “Early production techniques” preserves chronology.

Atlas 8. Heading trap: criticism versus description

A paragraph explains a model and then argues its assumptions are unrealistic. Heading “How the model works” captures exposition but not the controlling evaluative turn. “Why the model’s assumptions fail” better matches the paragraph’s job.

Atlas 9. Heading trap: solution versus problem

The paragraph begins with water shortage, but most of it describes how recycled water reduced the shortage. A problem heading fits context; a solution heading fits the paragraph’s central function.

Atlas 10. Heading collision under one-use rules

Two paragraphs seem to fit “A practical solution”. Paragraph A gives several proposed ideas; Paragraph B describes one solution actually implemented. The stronger anchor belongs to B. Reopen A and compare it with the remaining heading “Ideas considered before action”.

Atlas 11. Information matching: example versus theory

Target: “an example of a town adapting to rising temperatures.” Paragraph A explains adaptation theory. Paragraph B describes one town installing shaded walkways. B matches because the target asks for an instance, not the general principle.

Atlas 12. Information matching: cause relation

Target: “a fall caused by higher prices.” Several paragraphs mention prices and falling demand. The match belongs where the causal relation is explicit or strongly supported, not merely where both nouns appear.

Atlas 13. Information matching: unexpected consequence

Target contains two requirements: consequence + surprise. A paragraph describing a predictable consequence is incomplete. The correct paragraph says planners “did not anticipate” the outcome or otherwise signals unexpectedness.

Atlas 14. Information matching: one paragraph reused

Paragraph C contains a date, a criticism and one named example. Three targets ask for exactly those elements. Under reuse-allowed rules, C can legitimately answer all three. Repetition is correct because the evidence repeats, not because the answers look balanced.

Atlas 15. Information matching: same topic, different function

Paragraphs A and D both discuss training. Target: “a criticism of current training”. A describes how training works; D argues it ignores real conditions. D owns the critical function.

Atlas 16. Information matching: paraphrase without keywords

Target: “participation declined even though interest increased.” Source: “Applications continued to rise, yet fewer people completed the programme.” The match depends on contrastive meaning, not repeated vocabulary.

Atlas 17. Information matching: definition versus example

Target asks where “resilience is defined”. One paragraph describes a resilient community; another states resilience means recovering after disruption. The definition target belongs to the second.

Atlas 18. Information matching: criticism based on cost

Several paragraphs mention cost. Only one uses cost as the reason an approach should be rejected. The target requires criticism + basis = cost, not cost mention alone.

Atlas 19. Person matching: mention versus ownership

“Ahmed rejects Lina’s view that speed matters most.” Target: “believes speed is the main priority.” Lina matches. Ahmed owns the rejection, not the embedded claim.

Atlas 20. Person matching: same stance, different reason

Both Mei and Harun support expansion. Mei cites lower cost; Harun cites improved access. Target specifies “supports expansion because it reduces expense”. Mei matches despite both speakers being positive.

Atlas 21. Person matching: conditional support

Priya supports the plan only if safety testing is independent. Target says “supports immediate implementation”. That is too strong. Another target “would support implementation after independent testing” preserves the condition and matches.

Atlas 22. Person matching: initial versus final view

Jon originally opposed the scheme, then changed his mind after seeing data. Target “initially opposed” matches Jon; target “now supports” also matches Jon if reuse is allowed. Person + time determines the proposition.

Atlas 23. Person matching: pronoun ambiguity

Two researchers are named, followed by “she later questioned the result”. Reference must be resolved before the statement is assigned. The option set cannot repair a wrong antecedent.

Atlas 24. Person matching: quote ownership

A journalist quotes Dr Lee saying the policy is “premature”, then writes that officials disagree. Target “calls the policy premature” matches Dr Lee, not the journalist or officials.

Atlas 25. Person matching: partial agreement

Speaker A agrees with the goal but rejects the method. Target “opposes the goal” does not match; target “objects to the proposed method” does. Dimension-specific stance matters.

Atlas 26. Feature matching: nearby property migration

System A is cheap but slow. System B is fast but costly. Target “has the lowest cost” belongs to A; “offers the highest speed” to B. Dense comparison can swap properties if entity labels are not maintained.

Atlas 27. Feature matching: condition matters

Technology C works efficiently only in dry climates. Target “works well in all climates” is wrong despite strong performance being mentioned. Target “depends on dry conditions” matches the actual constraint.

Atlas 28. Feature matching: comparative property

A is cheaper than B; B is cheaper than C. Target “most expensive” belongs to C even if the source never says “C is most expensive” directly. The comparison chain supports the ranking.

Atlas 29. Feature matching: development date

Three methods are described with dates. Target “was developed first” asks for temporal ranking, not merely a date. Build the comparison set before assigning.

Atlas 30. Feature matching: one entity owns several features

Reuse-allowed rules permit System D to match “lowest energy use”, “requires specialist staff” and “best for small sites” if all three properties are supported. Do not distribute features artificially to create variety.

Atlas 31. Sentence ending: grammar eliminates first

Stem: “The researchers decided to…” Ending A begins with a noun phrase; Ending B begins “repeat the trial”. B is grammatically viable before source meaning is even checked.

Atlas 32. Sentence ending: cause versus result

Stem: “The programme was cancelled because…” One ending gives the reason funding disappeared; another describes what happened afterward. Only the causal ending fits the relation.

Atlas 33. Sentence ending: scope difference

Endings: “works in coastal regions” versus “works only in coastal regions”. Source says coastal use is common but inland examples exist. The unrestricted ending matches; “only” overstates.

Atlas 34. Sentence ending: certainty difference

Endings: “may reduce costs” versus “will remove all costs”. Source says the effect is possible and partial. The first preserves both modality and scope.

Atlas 35. Sentence ending: true elsewhere

An ending states a true fact from another paragraph but does not complete the current stem’s proposition. Matching requires stem-ending fit plus source support, not truth anywhere in the text.

Atlas 36. Sentence ending collision

Two stems appear to need Ending F in a one-use task. Stem 2 receives direct source support; Stem 4 merely shares topic vocabulary. Lock F to Stem 2 and reopen Stem 4.

Atlas 37. One-use anchor propagation

Four of five heading matches are direct. The remaining heading becomes highly likely for the last paragraph. Positive paragraph evidence still confirms the match; constraint propagation accelerates but does not replace proof.

Atlas 38. Weak-anchor cascade

An uncertain early heading is treated as fixed. Three later matches are forced around it, creating a tidy but wrong set. Whole-set review should identify that the original anchor had weak evidence and reopen it first.

Atlas 39. Reuse suppression error

Paragraph B already matched one information statement, so the learner refuses B for another despite the rules allowing reuse. This is a procedural error, not a reading error.

Atlas 40. Extra-option anxiety

One unused heading looks sophisticated, so the learner reshuffles secure answers to force it into the set. If extra options are allowed, unused does not mean unresolved. Leave the distractor unused.

Atlas 41. Repeated-answer anxiety

Reuse-allowed task yields B, B, D, B. The pattern looks suspicious but all three B matches have direct evidence. Letter distribution is not a semantic constraint.

Atlas 42. Parking resolves later

Paragraph 3 is between headings A and E. The learner parks both. Later Paragraph 5 matches A decisively under one-use rules, leaving E for Paragraph 3. Rechecking confirms E’s local fit. Parking allowed global evidence to mature.

Atlas 43. Targeted reread after collision

Two headings differ only in cause versus consequence. The learner rereads Paragraph 4 specifically asking whether the paragraph explains why an event happened or what happened because of it. The relation—not the topic—resolves the match.

Atlas 44. Negative evidence

Heading requires criticism of a method. Paragraph contains only neutral description and one positive result. Absence of criticism weakens the heading enough to reject it when a more descriptive option exists.

Atlas 45. Listening speaker matching

Three speakers all mention price. Speaker A says price is decisive, B says quality matters more, C says price matters only for first-time buyers. Target statements must preserve stance, comparison and condition, not merely the shared noun.

Atlas 46. Listening opinion update

A speaker first prefers online learning, then after describing isolation says a blended model is better. Target “final preference” maps to blended learning. The abandoned early view remains a distractor.

Atlas 47. Visual feature matching

A map uses icons to show wheelchair access at Sites A and C. Target “accessible by wheelchair” matches A and C if reuse or multi-select rules allow. Text-only scanning would miss the evidence.

Atlas 48. Multimodal entity collision

Site B’s text mentions family facilities; Site C’s image shows a playground. Target “designed especially for children” may require integrating wording and image rather than assuming any child-related feature is equivalent. Match the exact evidence relation.

Atlas 49. Whole-set anomaly

After completion, one paragraph has no plausible heading while another weakly holds two. The anomaly suggests an earlier misassignment. Reopen the weakest local match, not every answer.

Atlas 50. Local proof after global resolution

Constraint logic leaves Heading G for Paragraph 6. The learner checks Paragraph 6 and finds direct evidence for G’s proposition. The match is now supported both globally and locally. This is the ideal endpoint: constraints narrow; evidence closes.

Atlas Return: Matching Is Relationship Control Across a Shared Space

Across all fifty cases, the visible answer is only a letter or label. The real decision is relational. A paragraph is not matched because it contains the same noun; it is matched because its organising job fits the heading. A person is not matched because a topic is mentioned; the person owns the proposition. An entity is not matched because a property appears nearby; the property is bound to that entity. A sentence ending is not matched because it is true somewhere; it completes this stem grammatically and semantically.

The shared option set adds a second layer. Local matches interact under reuse rules, one-use constraints, extra options and collisions. Strong performance keeps both scales active: local evidence decides what belongs together; global constraints help reveal where a local decision deserves another look.

Advanced Matching Laboratory: Constraint Reasoning, Learner Profiles and Independent Mastery

The field manual trains individual mechanisms and the atlas shows common collisions. This final laboratory focuses on the difficult boundary cases that separate competent matching from reliable matching under examination conditions. The learner must keep two scales active at once: the local scale of evidence binding one target to one option, and the global scale of rules governing how all matches coexist.

The central danger is that global structure can become seductive. A neat one-to-one pattern can look convincing even when one local match is weak. Equally, a learner can become so committed to local evidence that she ignores a clear rule collision revealing an earlier mistake. Strong performance lets local evidence lead and lets global constraints audit.

A1. Partial match is not complete match

A heading says “How falling prices encouraged wider adoption”. The paragraph says prices fell, but adoption remained unchanged. One component matches and the relationship does not. The heading fails because the causal consequence is absent.

Advanced matching should treat options as bundles of conditions. Topic, relation, scope and time all contribute. One correct component cannot rescue a broken proposition.

A2. Two headings can both be true at different levels

A paragraph explains several environmental problems caused by tourism, with coral damage as the longest example. “Damage to coral reefs” is true at detail level; “Environmental costs of mass tourism” is true at paragraph level.

Heading matching asks which option represents the organising level. The narrower heading loses because it cannot account for the paragraph’s other evidence.

A3. One heading can fit two paragraphs for different reasons

“A failure of planning” seems plausible for Paragraph 2 because delays were not anticipated and Paragraph 4 because budgets were underestimated. In a one-use task, compare which paragraph presents planning failure as the central idea rather than a side detail.

The collision is useful. It forces a direct comparison between two local fits instead of treating each assignment independently.

A4. Hidden comparison words can decide a heading

Heading: “A more reliable alternative”. The paragraph need not contain “more reliable”; it may compare System A’s repeated failures with System B’s stable performance. The comparative relation is distributed across evidence.

Matching is not restricted to phrase-level equivalence. Whole-paragraph propositions can imply the heading relation through contrast.

A5. Negative headings require a genuinely negative centre

Heading: “Why the approach should be abandoned.” A paragraph mentioning one drawback does not automatically match if its conclusion remains strongly supportive.

Stance hierarchy matters. Acknowledged limitation and overall rejection are different paragraph jobs.

A6. Concession can disguise the true heading

Paragraph: “Although expensive at first, the system saves money over ten years and is now preferred by most operators.” A heading focused on high cost may match the concession; a heading about long-term financial advantage matches the controlling proposition.

Concessive structure often places a tempting distractor early while the main clause carries the paragraph’s real direction.

A7. Matching information can require inference, not quotation

Target: “a reference to concern about fairness”. Source: “Critics noted that low-income residents would pay the same fee despite having fewer alternatives.” The word fairness may never appear, yet unequal burden supplies the concept.

The match is valid if the inference is close and task-appropriate. Information matching does not demand exact wording unless the format explicitly says so.

A8. But inferred matches need tighter control than direct ones

If one paragraph explicitly criticises unfairness and another merely describes different impacts, the explicit paragraph is the stronger match for a target naming fairness.

Rank evidence strength. A direct semantic statement can outrank a plausible inference when both compete for one target.

A9. Information location can be distributed across two sentences

Sentence one describes a new material; sentence two says “this reduced energy use unexpectedly”. Target: “an unforeseen efficiency benefit”. The match depends on linking the referent and effect across sentences.

Local source regions can carry the target relation jointly. One-sentence scanning may miss the evidence.

A10. Person ownership can be nested

“Rivera agrees with Chen’s prediction but rejects Chen’s explanation.” Rivera owns agreement with the predicted outcome and disagreement with the causal account. Chen owns both the original prediction and explanation.

Targets must specify which proposition layer they ask about. Broad “agrees with Chen” may be too imprecise if the passage separates outcome and explanation.

A11. Reported speech can shift ownership subtly

A writer says “According to Malik, the pilot failed because staff were undertrained.” The failure explanation belongs to Malik unless the writer independently endorses it later.

Person matching should preserve attribution boundaries even when the sentence appears inside another writer’s paragraph.

A12. Agreement words can conceal different degrees

Speaker A “strongly supports” expansion; Speaker B “would accept a limited trial”; Speaker C “sees potential but wants more evidence”. All lean positive. Targets distinguishing strong support, conditional acceptance and cautious interest require degree-sensitive matching.

Positive versus negative is often too coarse for advanced person matching.

A13. One person can own contradictory views at different times

An expert may say she once believed remote work reduced productivity but now thinks the evidence shows no loss. A target naming past belief and another naming current belief can both match her under reuse rules.

Time is not decoration; it is part of proposition ownership.

A14. Feature matching can require transitive comparison

A uses less energy than B; B uses less than C. Target: “highest energy use”. C matches through relational inference even if no sentence directly says “C uses the most”.

Matching features can therefore require constructing a comparison order rather than copying one property statement.

A15. Comparative chains can break under different criteria

A is cheaper than B, while B is faster than C. These statements do not establish whether C is most expensive or A is slowest. Comparison must remain criterion-specific.

Feature matrices should keep cost, speed and reliability in separate columns precisely to prevent cross-axis inference.

A16. Feature targets can ask absence as well as presence

Target: “does not require external power”. The source may describe Systems A and B as powered and System C as manual. C matches through contrast, even if “does not require external power” is not written verbatim.

Negative properties must be supported, not guessed from silence. Manual operation provides the evidence.

A17. Sentence endings can be grammatically compatible and semantically impossible

Stem: “The new material was selected because…” Two endings are grammatically valid. One gives lower weight; another gives increased cost. If the source identifies weight as the decision reason, grammar cannot distinguish, but semantic ownership can.

Grammar is a filter, not a final judge.

A18. Sentence endings can be semantically related but grammatically illegal

An ending may contain the right idea but produce “decided to the cheaper option”. It cannot match a stem requiring a verb phrase.

Matching endings requires the completed sentence, not the isolated option, to be valid.

A19. Shared endings can create false symmetry

Several stems concern cost, and several endings mention cost. The set is designed so each stem requires a different cost relation: cause, limitation, result, comparison.

Topic similarity increases the need to identify grammatical and logical relation.

A20. One-use constraints create dependency chains

If Target 1 securely uses A, Target 2 narrows to B/C. If Target 2 securely uses B, Target 3 narrows to C/D. This is legitimate only because every earlier assignment has local evidence.

Think of constraint propagation as a bonus created by confidence, not a substitute for confidence.

A21. A weak first anchor can contaminate the entire chain

When Target 1 is assigned A from keyword overlap alone, every later candidate set shrinks incorrectly. By the final item, the learner feels forced into a bizarre last option.

The bizarre final result is not necessarily the last item’s problem. Trace dependency backward to the earliest weak anchor.

A22. Global inconsistency should trigger backward diagnosis

If no option fits Target 6 and one earlier match was only moderate confidence, reopen that earlier match first. Do not indiscriminately erase strong evidence-backed assignments.

Matching review becomes dependency tracing rather than random shuffling.

A23. Reuse-allowed tasks have no depletion effect

In a person-matching task where names may repeat, Person B matching Target 1 should not make B “less likely” for Target 4. Probability does not decrease simply because an answer has been used.

Every target receives an independent evidence test unless the rules create explicit dependence.

A24. Extra options increase decoy density, not task obligation

If eight headings exist for five paragraphs, three can remain unused. Their presence is designed to make broad or partial matches attractive.

The learner’s obligation is complete target coverage, not complete option consumption.

A25. Matching can be many-to-one

Several statements can map to one paragraph, several views can map to one speaker, several features can map to one entity. Many-to-one structure is common and legitimate where reuse is allowed.

Learners should stop expecting visually balanced answer distributions.

A26. Matching can be one-to-one

Heading matching often creates one heading per paragraph under explicit rules. Here each secure assignment reduces the remaining option board.

The key is procedural flexibility: know which mathematical structure this task actually uses.

A27. Some matching tasks are effectively multi-select

A target may ask which people share a view or which features apply to an entity. If the instructions allow several answers, membership testing replaces winner-take-all assignment.

Never import a single-answer assumption into a task whose output structure permits multiple matches.

A28. Matching under unfamiliar topics should rely more on relation than background

A learner may know nothing about marine carbon capture but can still distinguish paragraphs that define a process, report a limitation and present a solution.

Unfamiliar-Topic Performance supports the temporary model; Matching Performance uses paragraph job, ownership and relation to map options.

A29. Familiar topics can create false matches through prior knowledge

A learner knows electric vehicles reduce tailpipe emissions and matches a statement about emissions to a paragraph discussing battery cost simply because the topic is EVs.

Prior knowledge can predict likely content but cannot substitute for local evidence binding target to source.

A30. Matching should survive paraphrase density

A strong task may avoid repeating almost any target wording in the source. Learners should still match through proposition equivalence, hierarchy and discourse function.

The ultimate skill is semantic mapping under lexical change.

Learner Profiles: Twelve Ways Matching Performance Breaks

Profile 1. The Keyword Matcher

This learner selects the heading, paragraph or person sharing the most obvious words. Paraphrased correct options look risky; copied distractors look safe.

Repair with keyword-decoy drills and proposition paraphrase. Require the learner to state relation and job before judging lexical overlap.

Profile 2. The Topic-Not-Job Reader

The learner can state what a paragraph is about but not what it does with the topic. Broad headings feel correct across multiple paragraphs.

Repair with topic/job pairs: “renewable energy / explains storage problem”, “renewable energy / reports new adoption”. The second half becomes the heading discriminator.

Profile 3. The Detail Hugger

Concrete examples feel safer than abstract main ideas, so narrow headings win too often.

Repair with hierarchy mapping: claim → reasons → examples. The learner must identify the level the matching task requests before choosing.

Profile 4. The Broad-Heading Lover

General headings feel comprehensive and intellectually safe. Specific central meanings are rejected as “too narrow”.

Repair by comparing how much unrelated material the broad heading leaves unexplained. A good heading covers the paragraph tightly, not vaguely.

Profile 5. The Ownership Drifter

This learner remembers ideas but loses who said, believed, rejected or qualified them. Person matching collapses under dense attribution.

Repair with person proposition maps and mention-versus-own drills. Names should become linked to stance, reason, condition and time.

Profile 6. The Feature Migrator

Properties move between nearby entities. The learner remembers that one system was cheaper and another faster but swaps which was which.

Repair with entity-property matrices. Binding is the weak link, not fact recognition.

Profile 7. The Reuse Suppressor

Once an option is used, the learner refuses to use it again even when the rules explicitly allow reuse.

Repair by alternating one-use and reuse tasks and verbalising the rule before every set. Procedural control must become context-sensitive.

Profile 8. The One-Use Violator

The learner focuses locally and assigns the same restricted heading or ending twice. Global board state is not tracked.

Repair with option boards and whole-set anomaly scans. Local accuracy must coexist with allocation rules.

Profile 9. The Collision Shuffler

When two targets want the same one-use option, this learner moves answers randomly until the board looks neat.

Repair by comparing local evidence strength. Keep the stronger anchor; reopen the weaker target. Collision is a diagnostic question, not a demand for symmetry.

Profile 10. The Last-Option Guesser

The learner assigns the final unused option without checking whether it actually fits the last target.

Repair with the last-option test: global elimination may propose the answer; local evidence must still confirm it.

Profile 11. The Grammar-Only Ender

In sentence-ending tasks, any grammatical completion feels plausible. The learner underweights source meaning and logical relation.

Repair with grammar-first, meaning-second filters. Grammar can eliminate; semantic evidence selects.

Profile 12. The Over-Checker

Secure anchors are repeatedly reopened because one unused option looks attractive or the answer pattern feels strange.

Repair with confidence-by-separation and evidence-required change rules. A match changes only when new semantic, grammatical, ownership or rule evidence appears.

Full-System Stress Tests

Stress Test 1. Headings plus information on the same text

First match one heading to each paragraph. Then match detailed information statements to paragraphs with reuse allowed. The learner must switch relation: main idea in the first task, local information location in the second.

Errors reveal whether the learner truly understands target type or merely applies one familiar matching routine.

Stress Test 2. People with overlapping views

Use four speakers who agree on the goal but differ on reasons, conditions and degree. Matching statements should force proposition-level ownership.

The test is passed when broad stance no longer dominates finer evidence.

Stress Test 3. Features under dense comparison

Describe four systems in alternating sentences with cost, speed, reliability and conditions interleaved. Targets ask comparative and absolute properties.

The learner may use a compact matrix but must avoid full transcription. The test measures binding under density.

Stress Test 4. Sentence endings with grammar and scope traps

Give endings that are topically similar, several grammatically possible, but different in causality, condition and certainty.

The learner must use stem grammar to reduce candidates and source meaning to finish the decision.

Stress Test 5. One-use collision chain

Construct a heading task where one weak early choice creates three later conflicts. The learner should identify the earliest low-confidence assignment rather than repair the final collision only.

This tests dependency diagnosis and protection of strong anchors.

Stress Test 6. Reuse-allowed repeated answer pattern

Design answers B, B, B, D, B legitimately. Learners must resist distribution intuition and maintain independent evidence checks.

This directly tests whether procedural rules outrank psychological discomfort.

Stress Test 7. Unfamiliar topic matching

Use a technical passage on an unfamiliar domain. Headings differ by paragraph function; person statements differ by stance; features differ by explicit comparison.

The learner should lean on structure, ownership, paraphrase and relation rather than background knowledge.

Stress Test 8. Listening speaker matching with corrections

Speakers revise one position mid-turn, quote another speaker and use similar keywords. Targets ask final views.

The learner must preserve final semantic state and speaker ownership while the audio continues.

Self-Study Protocol

For every uncertain or wrong match, record five things: target type, option chosen, why it looked plausible, what exact relation failed, and what evidence made the correct option stronger. If one-use or reuse rules affected the board, note that separately.

Retest the same mechanism on an unrelated text. A broad-heading error in science should be tested again in history or culture. An ownership error in written interviews should be retested in listening. Transfer matters more than familiarity with the original correction.

During review, avoid rewriting the entire passage. Create only the smallest representation that would have prevented the error: paragraph job, person proposition, feature matrix, ending relation or collision note.

Teacher and Tutor Protocol

When a learner gives a wrong match, ask first: “What relation were you trying to match?” If the answer is “same topic”, the problem may already be visible. Then ask: “What does the target require beyond topic?” and “Which source evidence proves that relationship?”

For collision errors, do not announce which answer to move. Ask which of the competing matches has stronger evidence and which earlier assignment was least secure. This teaches dependency diagnosis rather than answer rearrangement.

If the learner cannot understand the paragraph itself, route upstream to Reading. If she understands it but cannot compress its main job, train heading hierarchy. If she understands each person but loses attribution, train ownership. Matching feedback should remain mechanism-specific.

Parent Guide: How to Help Without Solving the Match

Ask: “What is this paragraph mainly doing?” “Who actually owns that idea?” “What does this option say that the other one does not?” “Can this answer be used again according to the instructions?” “If two questions want the same answer, which match is stronger?”

These questions teach the decision process without requiring parents to know the correct letter in advance. Avoid saying “choose the one with the same words” or “you have already used B, so it cannot be B again” unless the actual rules justify that claim.

Final Mastery Audit: Fifty Checks for Matching Performance

Can the learner state whether options may be reused?
Can she identify whether extra options may remain unused?
Can one-use and reuse rules be applied without mixing them?
Can each target be classified by the relation it requires?
Can a complex heading be unpacked into its essential components?
Can options be grouped by cause, effect, solution, stance, time or another discriminating axis?
Can paragraph topic be separated from paragraph job?
Can an opening-sentence decoy be resisted?
Can later paragraph turns update a provisional heading?
Can broad headings be rejected for fitting too much?
Can narrow headings be rejected for fitting only an example?
Can temporal distinctions among headings be preserved?
Can stance distinctions among headings be preserved?
Can repeated vocabulary be ignored when the relation is wrong?
Can matching information be treated as location rather than summary?
Can target statements be decomposed before scanning?
Can paraphrase be recognised without shared keywords?
Can example, definition, criticism, prediction and comparison roles be distinguished?
Can one paragraph be reused when rules permit?
Can one statement be matched to the best source region rather than any topical region?
Can proposition ownership be tracked across people?
Can mentioning an idea be distinguished from endorsing it?
Can shared stance but different reasons be separated?
Can conditional opinions retain their conditions?
Can initial and final views remain distinct?
Can pronoun reference be resolved before person matching?
Can entity-property binding survive dense comparison?
Can comparative chains be built without mixing criteria?
Can sentence endings be filtered grammatically?
Can sentence endings be filtered by logical relation?
Can scope words distinguish near endings?
Can modality distinguish near endings?
Can true-but-wrong-location endings be rejected?
Can strong anchors be distinguished from convenient assumptions?
Can one-use constraints reduce the board only after secure matches?
Can reuse-allowed options remain available after earlier use?
Can extra options remain unused without causing panic?
Can repeated answer letters remain acceptable when evidence supports them?
Can collisions be resolved by comparing local evidence strength?
Can the weakest match be reopened first?
Can difficult targets be parked without contaminating the set?
Can targeted rereading search for discriminating evidence?
Can negative evidence be used cautiously?
Can a whole-set anomaly trigger backward diagnosis?
Can the final unused option be locally verified rather than merely accepted?
Can every final match be justified from source evidence?
Can matching operate under unfamiliar topics?
Can matching operate across written, spoken and visual sources?
Can timing increase without sacrificing rule reading and target definition?
Can the learner finish a mixed matching set independently without relying on keyword matching or answer-pattern superstition?

Final Return: Matching Is Local Evidence Inside a Global Constraint System

The visible action in a matching task is small: write a letter, choose a name, attach a heading, complete a sentence. The invisible work is larger. The learner must understand what kind of relationship the task asks for, interpret each option as a proposition or function, locate discriminating evidence, preserve ownership and scope, and then respect the rule governing how all matches coexist.

That is why matching deserves its own performance owner. Multiple choice asks which one option wins for one question. Information transfer asks how meaning enters a constrained receiver. Matching asks how several targets and options are allocated across one shared semantic space. The answers can interact. A secure match can reduce later possibilities. A weak match can poison the set. Reuse rules can make repetition correct. Extra options can remain deliberately unused.

Iona keeps each local match accountable to evidence. Maren identifies the exact relation that makes target and option belong together. Leonie manages the shared board, collisions, parking and stopping. Together they protect the final standard: local evidence first, global consistency second, and no match accepted merely because the letters happen to fit.

THE SET CAN SUGGEST WHERE TO LOOK. ONLY THE TEXT CAN PROVE THE MATCH.

Iona asks whether the evidence binds this option to this target. Maren asks what relation creates the match. Leonie asks how the assignment changes the remaining board and whether rechecking can still improve it. Together they turn matching from visual guesswork into controlled mapping.

Place matching inside the wider reading task. Reading and Comprehension — find the broader skill route; Secondary English Comprehension Examination Performance — connect matching practice to passage reading, evidence and answer control.

English Matching Performance | Match Headings, Information, People, Features and Sentence Endings Under Test Conditions