Advanced examination reasoning is not only about finding supporting evidence. It is also about stress-testing a rule. One valid exception can defeat an absolute claim; one boundary value can expose a broken model; one control condition can show that a causal story was too strong. This workshop develops that skill across English, Mathematics and Science.
It extends the claim calibration in Vol 0066, the Science alternative-explanation work in Vol 0068, the Mathematics feasibility work in Vol 0071 and the assumption audit in Vol 0073.
For 2027 school candidates, use the official K300 English, K310 Mathematics and K326/K327/K328 combined Science syllabuses, plus the SEAB G3 directory, for assessment requirements. Every case below is original teaching material.
A counterexample tests an absolute claim
A universal statement such as “all”, “always”, “never” or “for every” can be disproved by one valid case that satisfies the stated conditions and violates the conclusion. The counterexample must fit the claim’s domain; an irrelevant exception proves nothing.
Boundary cases test where a rule begins and ends
Zero, one, maximum capacity, minimum allowed value, equality thresholds and limiting conditions often reveal whether a rule has been interpreted correctly. Boundary cases are not tricks. They are structured tests of scope.
Do not use one example to prove a universal rule
One supporting example can illustrate a claim, but it cannot usually prove a statement about every case. Counterexamples and supporting examples have asymmetric power: one valid counterexample can refute an absolute; one supporting example cannot establish universality.
The domain controls whether a counterexample is valid
If a mathematical claim concerns positive integers, a negative number is not a valid counterexample. If a Science claim is restricted to a stated condition, an example outside that condition does not refute it. Read the boundary before testing the rule.
English absolute claims are high-risk
Words such as everyone, always, no one, completely and impossible invite boundary testing. Ask whether the source actually supports such strength. If one supported exception exists, the absolute wording fails.
English case 1: ‘all students prefer digital notes’
A fictional survey of 40 respondents finds 31 prefer digital notes and 9 prefer paper. The claim “all students prefer digital notes” fails immediately because the source itself contains respondents who do not. A calibrated claim may describe the majority of respondents instead.
English case 1: do not overcorrect
The failed absolute does not mean digital notes are unpopular. The evidence still shows 31 of 40 respondents preferred them. Counterexample reasoning narrows the claim rather than reversing it automatically.
English case 2: ‘online learning never works’
One credible example of a successful online-learning situation is enough to refute the universal statement “never works”. It is not enough to prove online learning works equally well for everyone or every task.
English case 3: ‘the writer is completely opposed’
If a passage criticises a policy but acknowledges one benefit, the word completely may be too strong. The concession acts as a boundary on the writer’s stance. Attitude labels should match the whole text.
English case 4: ‘this proves the programme caused the improvement’
An alternative explanation can function like a counter-test of the causal claim. If another factor changed at the same time, the data do not isolate the programme as the only cause. This does not prove the alternative caused the result either.
English counterarguments are not random objections
A useful counterargument tests the actual claim at its vulnerable boundary. If the thesis says a policy benefits all students, a relevant counterexample is a group for whom the mechanism does not apply. An unrelated disadvantage does not directly test universality.
English concessions can define safe scope
Instead of defending an absolute, a writer can concede a boundary and preserve the core claim: “Group study can improve idea generation, although students who need quiet concentration may benefit from individual work for some tasks.”
English visual-text boundary case
A poster says “Free entry” but the small print states “registration required”. The condition does not contradict free entry; it limits what the audience must do before attending. Boundary reading prevents false either-or interpretations.
English listening boundary case
A speaker says “I usually cycle, except when it rains.” The exception disproves “always cycles” but supports “usually cycles”. The learner should preserve both the main pattern and the boundary.
English oral boundary case
If asked whether school uniforms are useful, a strong response can state a general view and identify one condition where the recommendation changes. Boundary cases can make Spoken Interaction more mature without making it indecisive.
Mathematics universal statements need mathematical counterexamples
If someone claims “the square of every number is greater than the number”, test 0 and numbers between 0 and 1. For x = 0.5, x² = 0.25, which is smaller. The original statement is false over real numbers.
Mathematics case 1: odd plus odd
The claim “odd + odd is odd” is refuted by 3 + 5 = 8. One valid counterexample is enough. The correct general rule can then be investigated separately.
Mathematics case 2: multiplication makes bigger
The statement “multiplying a positive number by another positive number makes it larger” fails for multipliers between 0 and 1. For example, 10 × 0.5 = 5. Domain and multiplier size matter.
Mathematics case 3: squaring makes bigger
For x > 1, squaring increases a positive value. For 0 < x < 1, squaring makes it smaller. The boundary x = 1 leaves it unchanged. One rule becomes three regions.
Mathematics case 4: reciprocal size
For x > 1, 1/x lies between 0 and 1. For 0 < x < 1, the reciprocal is greater than 1. Boundary reasoning around x = 1 exposes the change.
Mathematics case 5: inequality sign
After dividing an inequality by a negative number, the sign reverses. Test a simple value from the proposed solution region to catch a forgotten reversal.
Mathematics case 6: capacity
If a bus holds at most 40 people, 40 is feasible and 41 is not. Testing the boundary converts an inequality into a concrete check.
Mathematics case 7: break-even threshold
At the break-even value, the compared models should be equal. Just below and above it, the preferred option should switch in the expected direction. Testing three boundary-adjacent values checks both algebra and interpretation.
Mathematics case 8: probability boundary
Probabilities lie from 0 to 1 inclusive. Values exactly 0 and 1 represent impossible and certain events. A calculated 1.02 violates the domain immediately.
Mathematics case 9: triangle condition
Three positive lengths form a non-degenerate triangle only if each pair sums to more than the third. A boundary equality produces a degenerate straight-line case, not an ordinary triangle.
Mathematics case 10: percentage
A percentage can exceed 100% when one quantity is more than the reference quantity. The claim “percentages can never exceed 100%” is false. Context controls whether >100% is meaningful.
Mathematics case 11: mean between extremes
For positive weights or ordinary unweighted mean, the mean lies between the minimum and maximum values. An output outside the observed range is a counter-signal that the calculation or interpretation needs checking.
Mathematics case 12: graph intercept
A cost model with a positive fixed fee should have a positive y-intercept. Testing x = 0 reveals whether the model preserved the fixed term.
Mathematics case 13: direct proportion
For direct proportion y = kx, x = 0 gives y = 0. A model with non-zero intercept is not direct proportion even if its graph is a straight line.
Mathematics case 14: inverse proportion
For y = k/x, x = 0 is not in the domain. Testing the boundary reminds the learner that not every algebraic substitution is permissible.
Mathematics case 15: scale
A scale factor of 1 should leave lengths unchanged. Area factor is 1² and volume factor 1³. This trivial case can test whether a formula behaves sensibly.
Mathematics case 16: percentage change zero
If original and new values are equal, percentage change should be 0%. If the formula returns something else, the base or subtraction is wrong.
Science universal claims need scientific boundaries
A statement such as “higher temperature always increases rate” may fail outside the mechanism’s valid range or in biological enzyme contexts where high temperatures disrupt structure. The relevant syllabus context determines the boundary.
Science case 1: catalyst
A catalyst increases reaction rate in the relevant model but is not a reactant supplying additional product. The claim “a faster reaction always produces more final product” can fail when the limiting amount is unchanged.
Science case 2: light intensity and photosynthesis
Increasing light can raise photosynthesis rate over a range, but a plateau can occur when another factor limits the process. The counterexample is not that light never matters; it reveals a boundary to the simple rule.
Science case 3: exercise and heart rate
Exercise often raises heart rate, but the exact response depends on intensity, individual condition and timing. A universal claim about one fixed value for everyone is not supported by the mechanism.
Science case 4: cooling
A hotter object may cool faster initially because the temperature difference is larger, yet it can still remain hotter than another object later. Rate and amount boundaries prevent simplistic conclusions.
Science case 5: floating
The statement “heavy objects sink” is refuted by large ships. Mass alone is not the criterion. Density, buoyancy and displaced fluid matter. The counterexample directs attention to the correct model.
Science case 6: conductors
A claim that “metals are always cold” confuses material property with temperature. A metal object can be hot. What differs under room conditions is often heat-transfer rate and sensation.
Science case 7: plant growth
More fertiliser does not imply indefinitely more growth. Beyond an appropriate range, other limitations or harmful effects may appear. The mechanism has conditions.
Science case 8: reaction completion
A reaction graph can plateau because a limiting reactant is exhausted. Adding more of an excess reactant may not change the final amount. This tests whether the learner identified the actual limiting condition.
Science case 9: measurement anomaly
One anomalous reading can challenge the claim that every measurement followed a smooth pattern, but it does not automatically disprove the underlying scientific relationship. First evaluate measurement quality and repeat evidence.
Science case 10: control result
If a control condition shows the same effect as the treatment, the claim that the treatment alone caused the effect is weakened. The control acts as a direct test of an alternative explanation.
Science case 11: zero input
Testing zero can reveal background effects. If a sensor reads a non-zero value when the input should be absent, the learner should consider offset or background rather than assuming every reading is target signal.
Science case 12: maximum range
At instrument saturation, additional physical change may not produce larger recorded values. The measurement boundary can make a real process look like a plateau.
EMS boundary type 1: domain
Who or what is the rule about? Students, respondents, positive integers, measured temperatures, registered candidates? A counterexample outside the domain is invalid.
EMS boundary type 2: time
Does the rule apply today, over one interval, usually or always? A single time window cannot automatically support a permanent claim.
EMS boundary type 3: condition
Does the rule require fixed temperature, constant speed, matched audience, or no competing factor? Remove the condition and the conclusion may fail.
EMS boundary type 4: threshold
At equality, maximum, minimum or zero, behaviour can change. Thresholds often separate regions where different conclusions apply.
EMS boundary type 5: representation
A rule may be true for one quantity but not another: total versus average, amount versus rate, respondents versus population. Counterexamples can expose silent category changes.
EMS boundary type 6: wording strength
Most is not all; may is not will; greater than is not greater than or equal to. Boundary words define the rule and must survive paraphrase.
How to build a valid counterexample
Step 1: restate the claim exactly. Step 2: identify its domain and conditions. Step 3: choose a case inside those conditions. Step 4: show that the conclusion fails. Step 5: state what the counterexample disproves—and what it does not.
Do not overgeneralise from the counterexample
If one case disproves “all”, it does not prove “none”. The correct repair may be “not all”, “under some conditions”, or a more precise rule. Counterexample reasoning should narrow claims rather than flip them blindly.
Use boundary cases before complicated cases
Zero, one, equality, maximum capacity and simple fractions often expose errors faster than complex examples. Start with the cheapest decisive test.
Use ordinary cases after boundary cases
A formula can behave correctly at x = 0 but still fail elsewhere. Boundary checks are diagnostic, not complete proof. After passing an edge case, test a typical interior value too.
Supporting examples and proof
Several supporting examples increase confidence but do not prove a universal mathematical statement. Formal reasoning is needed where proof is required. Keep empirical checking and proof distinct.
Counterexample and proof complement each other
A counterexample settles falsity of a universal statement immediately. If no counterexample is found, the statement is not automatically proven. The learner may need algebraic reasoning, theorem conditions or another proof method.
English rebuttal workshop
Take an absolute claim from an argumentative paragraph. Build one relevant counterexample, then rewrite the thesis with a narrower scope that survives the exception. This creates stronger argument rather than simple contradiction.
Mathematics conjecture workshop
Generate a pattern from several examples, then test zero, one, negative values, fractions and boundary conditions within the domain. A conjecture becomes stronger when it survives deliberate attempts to break it.
Science model workshop
State a simple model, then ask when another factor becomes limiting, when the instrument saturates, or when the mechanism’s assumptions fail. Boundary thinking turns memorised rules into conditional models.
Independent task A
Refute or defend: “Every positive number becomes larger when squared.” Give a valid case and state a corrected rule.
Independent task B
A school survey finds 70% of respondents prefer Option A. Refute or defend: “All students prefer Option A.” Explain the population boundary.
Independent task C
A reaction runs faster with smaller particles but reaches the same final gas volume. Which absolute claim about rate and final amount does this challenge?
Independent task D
A cost model says Plan X is cheaper for n 40. Test n = 39, 40 and 41 conceptually. What is the role of the boundary?
Independent task E
A listener says, “I usually walk unless it rains.” Which word prevents the conclusion “the speaker always walks”?
Independent task F
A sensor reads 0.4 units even when the target source is absent. What assumption about zero input has failed, and what might the control reading be used for?
Worked feedback A
Take x = 0.5. Squaring gives 0.25, so the universal claim is false over positive real numbers. A corrected statement is that for x > 1, x² > x; at x = 1 they are equal; for 0 < x < 1, x² < x.
Worked feedback B
The claim fails because respondents are not automatically the whole student population, and 70% is not all. A safe statement is that 70% of the surveyed respondents preferred A.
Worked feedback C
It challenges the claim that a faster reaction necessarily produces more final product. Rate can differ while final amount remains the same under the stated limiting conditions.
Worked feedback D
The boundary 40 separates decision regions. Testing values on both sides checks whether the inequality interpretation matches the model; at the boundary both plans are equal.
Worked feedback E
Usually and unless define the pattern and exception. The source explicitly leaves room for non-walking cases.
Worked feedback F
The assumption that zero target input gives zero reading has failed. The control may reveal background or offset that should be investigated or accounted for according to the task.
Error log: invalid counterexample
Record why the example fell outside the claim’s domain. Practise restating the domain before selecting a test case.
Error log: overcorrected claim
If you moved from “all” to “none” after one exception, rewrite the conclusion as “not all” and identify what the evidence still supports.
Error log: boundary ignored
Record the equality, zero or maximum condition that changed the rule. Practise a new question where the boundary appears in a different context.
Error log: one supporting example treated as proof
Separate illustration from proof. Ask what reasoning would be required to establish the statement for every case.
Repair route
Begin with absolute English claims and simple mathematical rules where one counterexample is easy to find. Make the learner state exactly what has been disproved.
Stabilisation route
Mix true universal rules, false universal rules and conditional rules. The learner should not assume every task contains a counterexample. Sometimes the correct result is that the proposed edge case still fits.
Extension route
Use scientific models with multiple boundary conditions and mathematical statements with restricted domains. Ask the learner to repair the rule after finding where it fails.
What progress should look like
Progress is visible when the learner checks domain before claiming a counterexample, tests edge cases without overgeneralising, and rewrites broken rules into more precise conditional statements.
Final operating rule
When a statement feels too neat, ask where it stops being true. Test zero, one, equality, extremes, exceptions and the stated domain. If one valid case breaks an absolute claim, narrow the rule—do not simply reverse it.
EMS counterexample and boundary-case checklist
- state the claim exactly
- identify its domain and conditions
- test zero, one, equality or relevant extremes
- choose a case inside the domain
- show precisely where the conclusion fails
- do not turn one exception into the opposite universal
- repair the rule with a narrower condition
- separate counterexample from formal proof
Continue the G3 EMS route
Return to the G3 SEC Learner’s Guide hub for the full EMS route and numbered series.
Advanced counterexample and boundary laboratory
Counterexample laboratory: test the strongest word first
If a sentence contains every, always, never, must or impossible, test that word before rewriting the rest. Absolute language creates a clear burden. A single valid exception can show that the wording, rather than the entire idea, is the problem.
Counterexample laboratory: keep the repaired claim useful
After finding an exception, do not retreat into vagueness. Replace “all” with the largest scope the evidence still supports. Replace “always” with a condition or frequency term. Precision should preserve as much truth as possible.
Counterexample laboratory: distinguish ‘not always’ from ‘never’
If one case breaks “always”, the justified conclusion is often “not always”. It does not follow that the opposite happens every time. This distinction is crucial in English argument, Science explanation and mathematical conjecture.
Counterexample laboratory: test hidden quantifiers
A statement may be universal even without the word all. “Multiplying makes numbers bigger” implies a general rule. Ask what domain and conditions are silently assumed. Then test fractions, zero or negatives if they belong to the domain.
Counterexample laboratory: test equality separately
Many rules change at equality. Greater than is not the same as greater than or equal to. At a break-even point two costs match; at a capacity limit the value may still be feasible; at a turning point a local gradient may be zero.
Counterexample laboratory: test zero separately
Zero often removes an effect or exposes an intercept. A direct-proportion graph should pass through the origin. A fixed-fee cost model should not. Zero can reveal whether the learner understood the relationship.
Counterexample laboratory: test one separately
One is a powerful multiplicative boundary. Multiplying by one leaves a value unchanged; powers at one behave differently from values above or below it. This can expose overgeneralised ‘bigger’ or ‘smaller’ rules.
Counterexample laboratory: test maximum and minimum
If a model has a capacity, probability bound or allowed range, test the endpoint itself and one value beyond it. This distinguishes inclusive from exclusive constraints.
Counterexample laboratory: test extreme but valid cases
An extreme case can make a hidden assumption visible. A very small positive multiplier, a near-zero denominator or an almost-full capacity may show why a rule needs a condition. The case must still be valid within the domain.
Counterexample laboratory: avoid impossible edge cases
Do not test a geometric length with a negative value if the claim is only about physical lengths. Do not use an unregistered subject to challenge a statement about a candidate’s own papers. Edge cases only count when they satisfy the original conditions.
Proof versus counterexample
One counterexample proves a universal statement false. No finite list of supporting examples proves it true. If the question asks for proof, use valid reasoning across the whole domain rather than a collection of examples.
Pattern versus theorem
A sequence of examples can suggest a conjecture. The next step is to search for a counterexample and then prove or refine the rule. Pattern recognition is the beginning of mathematical reasoning, not the final guarantee.
Science model versus law-like wording
School Science often uses simplified models under stated conditions. Stress-testing those conditions helps the learner avoid translating a useful model into an unlimited claim. The model may remain valid in its intended range.
Science anomaly versus counterexample
An anomalous measurement is not automatically a counterexample to a scientific relationship. First decide whether the reading is reliable and whether it truly satisfies the same conditions. Measurement uncertainty and model failure are different possibilities.
Science control as boundary test
A control condition can test what happens when the intended factor is absent or at a reference state. If the same effect appears, the causal claim needs revision. The control is a designed counter-test, not merely an extra group.
English concession versus counterexample
A concession can acknowledge a boundary before an opponent provides it. “Online learning can be effective for independent tasks, although some practical skills need supervised equipment” turns a fragile absolute into a defensible position.
English rebuttal after a counterexample
Do not deny a valid exception. Explain why the exception narrows rather than destroys the main argument, or revise the claim. A strong rebuttal respects evidence.
Oral boundary testing
In Spoken Interaction, the examiner may introduce an exception: “What about students who cannot afford the device?” Treat it as a boundary test of your position. Respond by adjusting conditions rather than repeating the original claim unchanged.
Listening boundary testing
Words such as except, unless, only and usually define exceptions. If notes lose these words, the learner may accidentally turn a conditional rule into an absolute one.
Visual-text boundary testing
Eligibility boxes, deadlines and exclusions often function as boundaries. A bold headline may give the main invitation; the small condition tells who, when or under what circumstances it applies.
Mathematics boundary-value proof check
After solving an inequality such as x ≤ 7, test x = 7 and a nearby value such as 8. This quickly reveals whether the endpoint and direction were interpreted correctly.
Mathematics domain repair
If a formula fails at x = 0 because division by zero occurs, the repaired rule should state x ≠ 0. Do not say the formula is useless; state its valid domain.
Mathematics piecewise repair
If a rule changes after a threshold, use separate expressions or cases. A piecewise model is often the correct repair when one formula cannot cover the whole domain.
Mathematics integer repair
If a continuous solution gives 12.4 buses, the practical answer must respect whole numbers and direction of need. Boundary testing around 12 and 13 resolves the decision.
Mathematics probability repair
If an event probability calculation exceeds 1, do not round it down. The range violation proves something earlier is wrong. Inspect the sample space or event combination.
Mathematics percentage repair
If a percentage greater than 100% appears, do not reject it automatically. Ask whether the quantity can exceed the reference. Percentage bounds depend on context.
Cross-subject case: ‘more is always better’
English can challenge the claim rhetorically, Mathematics can test a counterexample with diminishing or constrained returns, and Science can identify a limiting factor. The shared reasoning is boundary awareness, not identical subject content.
Cross-subject case: ‘faster means better’
English asks what better means, Mathematics asks which criterion is optimised, and Science distinguishes rate from final amount. A counterexample often appears as soon as the evaluation criterion is made explicit.
Cross-subject case: ‘same result means same process’
Two methods can reach the same answer or final amount through different routes. English sources can agree for different reasons, Mathematics models can intersect at one point, and Science processes can share an endpoint. Equal outcome does not prove identical mechanism.
Cross-subject case: ‘different numbers mean contradiction’
Different units, denominators, time windows or quantities can produce different numbers without disagreement. Before searching for a counterexample, confirm that the claims are truly about the same thing.
Boundary-case question design
When reviewing practice, create one ordinary case and one boundary case for the same rule. This checks whether the learner understood the condition or merely copied the procedure.
Changed-detail retest
After correcting a rule, change one detail that moves the case across the boundary. If the learner continues using the old method unchanged, the condition has not been internalised.
Counterexample log
Record the broken claim, valid counterexample, domain, and repaired rule. This is more useful than writing ‘careless’ beside the answer because it captures the exact boundary that was missed.
False counterexample log
Also record examples that looked like counterexamples but failed because they violated the domain. Learning why an exception does not count is part of mastering the rule.
Teacher feedback question
Ask: “What condition would make my rule valid?” This often produces a more useful repair than simply asking for the correct answer. It turns feedback into a boundary statement the learner can transfer.
Parent support question
Ask the learner to explain one example where the rule works and one where it does not. The aim is not to catch the child out, but to make the condition visible in ordinary language.
Exam-time use
Counterexample searching should be proportional to the task. A direct calculation does not need an extended philosophical audit. Use boundary checks when a claim, model or answer feels too broad, implausible or condition-sensitive.
Final counterexample standard
The learner has mastered this skill when they can break a false absolute with one valid case, defend a true rule by respecting its domain, and repair an overbroad statement without overcorrecting into the opposite extreme.
G3 Learner’s Guide navigation: ← Vol 0076 · Master Hub · Vols 0001–0077 · Worked-reasoning route.
