How to perform in the new G2 SEC Mathematics examination at an advanced level includes knowing when a statement that looks true has not yet been established. One successful example can show that something is possible; it cannot prove a rule for every case. One counterexample, however, can show that an “always” claim is false. Boundary values can also reveal where a method stops being valid.
This forty-third Learner’s Guide is a boundary and counterexample clinic. The central habit is: test the claim, not only the calculation. When a statement uses words such as always, never, for every value, at least, at most, increasing or decreasing, ask what happens at the edge of the domain and what simple example would challenge the claim.
For 2027, SEAB lists G2 Mathematics as K210 under the Singapore–Cambridge Secondary Education Certificate. Use the SEAB 2027 G2 syllabus directory for current official details. The examples here are original eduKateSengkang practice designed to strengthen reasoning, not official specimen questions.
One example can confirm a case, not a universal rule
Suppose a student tests 3 + 5 = 8 and 5 + 3 = 8, then says addition is commutative. The examples are consistent with the rule, but the general property is not established merely because two cases worked.
In examination reasoning, examples are useful for exploring a claim, finding patterns and checking plausibility. They do different work from a general justification.
A counterexample has asymmetric power
If someone claims “every even number is divisible by 4”, one counterexample—6—is enough to show the claim is false.
This asymmetry is powerful. To refute an “all” statement, you do not need to test every possible case. You need one valid case inside the stated domain that breaks the rule.
The claim words
- always;
- never;
- every;
- all;
- for any value;
- must;
- cannot;
- at least;
- at most;
- increases as;
- decreases as.
These words should trigger a boundary-and-counterexample mindset because they define how broad the statement is.
Start by identifying the domain
A counterexample is useful only if it belongs to the domain of the claim. A claim about positive integers cannot be refuted with a negative decimal unless the domain actually includes it.
Before testing a statement, write or think: what values are allowed?
Boundary case one: zero
Zero is often a useful test because it sits at a boundary between positive and negative and behaves differently in multiplication, division and percentage contexts.
Do not use zero automatically. Ask whether zero belongs to the domain and whether the operation is defined there.
Boundary case two: one
One can reveal whether a pattern that grows for larger values still behaves as claimed at the smallest positive integer.
It is also useful in ratios, powers and multiplicative relationships because multiplying by one preserves a value.
Boundary case three: negative values
A method or inequality pattern that seems obvious for positive numbers can behave differently for negatives.
For example, multiplying an inequality by a negative quantity reverses its direction. Testing a simple negative value can expose a memorised rule that ignored this condition.
Boundary case four: equality
Words such as “at least” and “at most” include the boundary. Testing the exact boundary value can distinguish ≥ from > or ≤ from <.
This connects directly to Vol 0035: Constraints and Feasibility.
Boundary case five: largest or smallest allowed value
If a context limits a quantity, the edge value can reveal whether a formula or conclusion remains valid throughout the allowed range.
This is especially useful in capacity, time and integer problems.
Counterexample clinic one: averages
Claim: “The average of two group means is always the combined mean.”
Counterexample: one group has one student with mean 10, another has nine students with mean 20. The simple average of the group means is 15, but the combined mean is (10 + 180) ÷ 10 = 19. Group size matters.
What the counterexample teaches
The counterexample does more than show the claim is false. It identifies the missing condition: averaging group means directly works when the groups have equal weight.
A good counterexample should lead to a refined rule whenever possible.
Counterexample clinic two: percentages
Claim: “If a price falls by 20% and then rises by 20%, it returns to the original price.”
Counterexample: start at $100. A 20% fall gives $80. A 20% rise on $80 gives $96. The percentage base changed.
Refine the percentage claim
The correct principle is not that opposite percentage changes cancel. They cancel only under special relationships between the multipliers, not merely because the numerical percentages are equal.
The counterexample reveals the role of the changing base.
Counterexample clinic three: speed
Claim: “The average of two speeds is always the average speed for the whole journey.”
Counterexample: travel equal distances at 30 km/h and 60 km/h. The whole-journey average is 40 km/h, not 45 km/h, because the slower stage lasts longer.
Refine the speed claim
The simple arithmetic mean of the two speeds works when the two speeds are maintained for equal times, not automatically for equal distances.
The missing condition is weighting.
Counterexample clinic four: algebra
Claim: “If x² = 9, then x = 3.”
Counterexample: x = -3 also gives x² = 9. The original statement loses one solution.
The algebra lesson
When reversing an operation, ask whether the inverse step produces more than one possibility. Do not let a familiar positive example erase another valid value.
Context may later exclude one root, but the algebraic solution set should be found first.
Counterexample clinic five: fractions
Claim: “A larger denominator always means a smaller fraction.”
Counterexample: 9/10 is larger than 1/2 even though 10 is the larger denominator. The numerator changed as well.
Refine the fraction claim
For fractions with the same positive numerator, a larger positive denominator gives a smaller positive fraction. The original claim omitted a condition.
Counterexample thinking teaches the learner to search for the missing condition instead of memorising an exception list.
Counterexample clinic six: graphs
Claim: “If a graph is rising, the y-value is positive.”
Counterexample: a rising line can lie entirely below the x-axis. Rising describes change as x increases; positive describes position above zero.
The graph lesson
Different graph features answer different questions: sign, gradient, intercept, maximum, minimum and domain should not be blended.
A useful counterexample separates two concepts that a learner has accidentally treated as identical.
Counterexample clinic seven: probability
Claim: “If an event happened several times in a row, it is less likely to happen next.”
For independent repeated trials with unchanged probability, the next-trial probability does not change merely because of the previous run. The claim needs a mechanism connecting trials before the history changes the next probability.
The probability lesson
Past outcomes can matter when the sample space changes, such as drawing without replacement. They need not matter when the trials remain independent.
The condition is not “previous results never matter”; it is whether the experiment changes.
Counterexample clinic eight: geometry
Claim: “Any quadrilateral with four equal sides is a square.”
A rhombus with non-right angles provides a counterexample. Four equal sides are not enough; a square also has right angles.
The geometry lesson
Counterexamples help distinguish defining conditions from properties that are shared by a wider family.
A diagram should be used to test the definition, not judged by appearance alone.
Counterexample clinic nine: data
Claim: “Two data sets with the same mean have the same distribution.”
Counterexample: 10, 10, 10 and 0, 10, 20 both have mean 10 but very different spread.
The data lesson
A single summary statistic cannot preserve every feature of a data set.
Use the statistic appropriate to the question and remember what information it compresses away.
Counterexample clinic ten: direct proportion
Claim: “If y increases when x increases, y is directly proportional to x.”
Counterexample: y = x + 5 increases with x but is not directly proportional to x because the ratio y/x is not constant and the relationship does not pass through the origin.
The proportionality lesson
Increase together is weaker than direct proportion. The relationship needs the defining structure, not merely the same trend direction.
The simplest counterexample rule
When refuting a claim, choose the simplest valid counterexample you can find.
Simple numbers make the logic visible and reduce the chance that arithmetic obscures the reason the claim fails.
Do not use an invalid counterexample
If the statement is restricted to positive integers, x = 0 may not be allowed. If a denominator must be non-zero, a zero-denominator “counterexample” is not inside the function’s domain.
The first step is always domain control.
A counterexample is not a proof of the opposite
Finding one case where “all A are B” fails does not prove “no A are B”.
It proves only that the universal claim is false. Keep the logical conclusion no stronger than the evidence.
Boundary cases can confirm definitions
A boundary value may not refute a rule, but it can reveal whether equality is included and whether the expression remains defined.
Testing x = 0, x = 1 or an endpoint is a quick way to inspect a formula before trusting a general statement.
Counterexample versus special case
A special case is simply one particular example. A counterexample is a special case that contradicts a universal claim.
The distinction matters because not every unusual example refutes anything. The example must directly violate the statement being tested.
Test the exact wording
Claim: “A square has four equal sides.” A non-square rhombus does not refute that claim because the statement does not say only squares have four equal sides.
Claim: “Every quadrilateral with four equal sides is a square.” Now the rhombus is a valid counterexample. Counterexample reasoning depends on the exact logical direction.
Always, sometimes, never
A useful classification task is to decide whether a statement is always true, sometimes true or never true within a stated domain.
The learner should support “sometimes” with at least one true case and one false case. “Always” requires general justification beyond examples. “Never” requires a general reason why no valid case can work.
Always/sometimes/never clinic: odd and even
Statement: “The sum of two odd integers is even.” This is always true. One way to justify it is to write the odd integers as 2a + 1 and 2b + 1. Their sum is 2(a + b + 1), which is even.
Examples can help discover the rule, but the algebra shows why every pair of odd integers follows it.
Always/sometimes/never clinic: products
Statement: “The product of two numbers greater than 1 is greater than both numbers.” If the domain is positive real numbers greater than 1, the statement is true. If the domain were merely positive numbers, 0.5 × 0.5 = 0.25 shows a different behaviour.
A domain restriction can turn a false broad claim into a true refined claim.
Boundary clinic: inequalities
Suppose the solution to a condition is x ≥ 4. Test x = 4 first. If it satisfies the original condition, equality belongs. Then test a nearby value below 4 and above 4 to see whether the direction makes sense.
Boundary testing is not a substitute for solving, but it is a useful check on the final set.
Boundary clinic: piecewise practical decisions
A delivery service charges one fee up to a weight threshold and another fee above it. The exact threshold value is a high-risk case.
Do not assume the fee jumps before or after the boundary. Read whether the wording says “up to”, “less than”, “more than” or “at least”.
Boundary clinic: whole-number decisions
A capacity calculation may produce 6.0 exactly in one case and 6.01 in another. The first needs six units; the second may require seven if partial units are impossible.
Testing just above the boundary helps the learner see why practical rounding is not ordinary nearest-number rounding.
Boundary clinic: graph intervals
A graph may show a model only for 0 ≤ x ≤ 20. Values outside that interval are not automatically valid predictions.
The endpoint x = 20 can also behave differently from interior values if the context changes there. Domain boundaries are part of the model.
Boundary clinic: denominators
An algebraic expression with a variable in the denominator excludes values making that denominator zero.
A simplified expression may look harmless after cancellation, but the original restriction can remain relevant. Keep track of where the original expression was defined.
Boundary clinic: square roots
If a real-number expression contains a square root, the quantity under the root may need to satisfy a non-negative condition.
Testing the boundary where the radicand becomes zero can reveal the endpoint of the allowed domain.
Boundary clinic: geometry
A triangle becoming “flat” at the boundary of a side-length condition is a useful way to understand why strict inequality may be required.
A boundary diagram can reveal when an ordinary figure turns into a degenerate case that no longer satisfies the intended geometric definition.
Boundary clinic: probability
Probability boundaries are 0 and 1. They are not ordinary interior values: 0 describes an impossible event under the model, 1 a certain event.
Testing these endpoints can reveal whether a proposed formula or interpretation behaves sensibly.
Counterexample clinic: correlation
Claim: “Whenever two quantities increase together, one causes the other.”
A counterexample can use two quantities that both rise because of a third factor. The mathematical lesson is broader: association in data does not automatically determine causal direction.
Counterexample clinic: linear-looking data
Claim: “Three points on a graph that appear nearly straight prove the relationship is linear everywhere.”
The points are compatible with many possible curves. A small observed range can look approximately linear even when the wider relationship is not.
Counterexample clinic: symmetry
Claim: “If a shape has one line of symmetry, it must be an isosceles triangle.”
Many non-triangular shapes have a line of symmetry. The counterexample exposes that the claim confuses one property with a complete classification.
Counterexample clinic: sequences
Claim: “If the first three terms increase by 2, the sequence must continue increasing by 2 forever.”
A sequence can be defined to match the first three terms and then change. Early pattern recognition is useful, but the generating rule or additional information matters.
Counterexample clinic: rounding
Claim: “Rounding each value before adding always gives the same total as adding first and rounding once.”
Counterexample: 1.46 + 1.46 = 2.92, which rounds to 2.9 to one decimal place. Rounding each first gives 1.5 + 1.5 = 3.0.
The rounding lesson
Intermediate rounding can change later results. Keep sufficient precision unless the question instructs otherwise.
One counterexample is enough to reject the universal claim that early rounding never matters.
Counterexample clinic: percentage points
Claim: “An increase from 20% to 30% is a 10% increase.”
The increase is 10 percentage points, but relative to the original 20%, it is a 50% increase. The words “percent” and “percentage points” answer different comparisons.
Counterexample clinic: area and perimeter
Claim: “A shape with larger perimeter always has larger area.”
Different rectangles can refute this. A long thin rectangle can have a larger perimeter but smaller area than a more compact rectangle.
The geometry lesson
Measurements that seem related may not determine one another uniquely. Counterexamples help identify which information is insufficient.
Counterexample clinic: same area
Claim: “Rectangles with the same area have the same perimeter.”
A 4 by 4 square has area 16 and perimeter 16. A 2 by 8 rectangle also has area 16 but perimeter 20.
Counterexample clinic: same perimeter
Claim: “Rectangles with the same perimeter have the same area.”
A 5 by 5 square and a 1 by 9 rectangle both have perimeter 20, but areas 25 and 9.
The counterexample search strategy
- Write the domain.
- Identify the strongest word in the claim.
- Choose the simplest edge or unusual case.
- Calculate or reason exactly.
- Check that the case truly satisfies the claim’s starting conditions.
- State precisely what the counterexample disproves.
This keeps the search efficient and logically disciplined.
The boundary search strategy
- Identify the allowed interval or condition.
- Test the exact endpoint.
- Test one value just inside.
- Test one value just outside if meaningful.
- Compare the results with the wording.
Boundary testing is especially useful in inequalities, domains and practical constraints.
The refined-rule habit
After finding a counterexample, ask: what extra condition would make the rule true?
This turns refutation into concept building. The learner moves from “the rule is wrong” to “the rule needs this boundary or condition”.
Refined rule example: group means
False broad claim: “Average the group means to get the combined mean.”
Refined rule: direct averaging of group means works when the groups carry equal weight; otherwise the combined mean must weight by group size.
Refined rule example: denominators
False broad claim: “Larger denominator means smaller fraction.”
Refined rule: for positive fractions with the same numerator, a larger positive denominator gives a smaller value.
Refined rule example: speed
False broad claim: “Average speed is the mean of stage speeds.”
Refined rule: the arithmetic mean works for stages of equal duration; in general, use total distance divided by total time.
The minimal-pair drill
Create two nearly identical statements, one true and one false. Ask what single condition changes the verdict.
This is the claim-level equivalent of the method contrast in Vol 0039.
The true-false-repair drill
Give a statement. If false, provide a counterexample and repair the statement. If true, explain why the tested examples are not enough and supply a general reason where appropriate.
The emphasis is reasoning, not guessing true or false.
The domain-first drill
Give several statements without domains. Ask how the truth value could change under integers, positive reals or another allowed set.
This builds the habit that mathematical claims live inside domains.
The graph-claim drill
Show a graph and ask several claims: positive, increasing, directly proportional, constant gradient, above a threshold.
The learner marks which feature of the graph supports or refutes each claim. This separates position from change and local from global behaviour.
The counterexample error ledger
- counterexample outside the domain;
- example fails to contradict the exact claim;
- one supporting example treated as proof;
- counterexample used to claim the opposite universal statement;
- boundary excluded or included incorrectly;
- pattern generalised beyond observed range;
- missing condition not identified after refutation;
These errors reveal whether the weakness is calculation or logical scope.
The 20-minute reasoning session
Spend five minutes testing five claims with simple examples. Spend ten minutes finding counterexamples for false universal claims. Spend five minutes rewriting the false claims with conditions that make them accurate.
This practice strengthens reasoning without requiring a full paper.
A four-week boundary-and-counterexample build
Week 1 — domains and boundaries
Use inequalities, denominators, roots and practical constraints.
Week 2 — numerical counterexamples
Use averages, percentages, ratios and rounding.
Week 3 — graphical and geometric claims
Separate graph features, area/perimeter and classification conditions.
Week 4 — timed mixed reasoning
Place claims inside unfamiliar contexts and require a quick test before the learner accepts them.
Use representation switching
Use Vol 0031. A claim that looks plausible in words may become obviously false when graphed, tabulated or drawn.
Use constraints and feasibility
Use Vol 0035. Boundaries matter not only in abstract algebra but also in real-world decisions.
Use checking systems
Use Vol 0019. A counterexample is one more kind of checking: it tests a claim rather than only a numerical result.
Use the Mathematics index for concept repair
If the learner cannot construct or interpret the counterexample because the topic is not understood, return to the Complete Mathematics Index.
Reasoning tools work only when the underlying mathematics is available.
The PSLE bridge
The earlier habit Represent Before You Calculate remains useful. At G2, the learner can use representations not only to solve but also to challenge claims.
Use Examination Craft
For timed decision-making and checking, continue through the Examination Craft hub. Boundary and counterexample checks should remain quick enough to help rather than become a separate essay.
Final rule
Do not trust a mathematical claim because several friendly examples agree with it.
Read the domain. Test the boundary. Search for the simplest counterexample. If the claim fails, identify the missing condition and refine the rule. The strongest learner is not the one who accepts patterns fastest; it is the one who knows how to test whether the pattern deserves trust.
Boundary testing in equations
A solved equation should be checked against any excluded values from the original expression. A value produced by later algebra may be invalid because the original denominator was zero or because the contextual variable could not take that value.
The final solution set belongs to the original problem, not merely to the last simplified line.
Boundary testing in inequalities
When the final answer is x < 5 or x ≤ 5, substitute the boundary into the original condition. This reveals whether equality belongs.
Then test one convenient value on the accepted side and one on the rejected side. The three checks make direction and boundary visible.
Boundary testing in graphs
A graph may change behaviour at an intercept, turning point or endpoint. Test claims locally and globally.
A graph can be increasing over one interval and decreasing over another. A statement about “the graph” may need an interval before it becomes accurate.
Boundary testing in geometry
Check whether the extreme case still forms the intended shape. A triangle can approach a flat configuration as one side approaches the sum of the other two.
The boundary shows why some geometric conditions use strict inequalities rather than equality.
Boundary testing in optimisation
If a question asks for the maximum or minimum feasible value, the optimum often occurs near a constraint boundary.
After finding a candidate, verify that it remains inside every condition and that moving beyond it would violate at least one requirement.
Boundary testing in discrete contexts
For counts, test the two neighbouring integers around a calculated threshold.
If 6.2 buses are required, six and seven are the useful boundary candidates. Six fails capacity; seven is the minimum feasible whole number.
Counterexamples and calculator output
A calculator can generate examples quickly, but it does not decide whether an example is logically relevant.
The learner still has to know the domain, the claim and the condition the example must violate.
Counterexamples and diagrams
A rough diagram can expose a geometry counterexample faster than algebra. Draw a rhombus that is not a square, or two equal-area rectangles with different perimeters.
The diagram should then be supported by exact properties or dimensions so the counterexample is not based only on appearance.
Counterexamples and tables
Tables are useful for claims about numerical patterns. Generate several cases, but remember that supporting cases do not establish a universal rule.
Use the table to search for a failing case or to discover a condition that might matter.
Counterexamples and graphs
Graphs can expose broad claims about sign, monotonicity, intersections or proportionality. A curve can show immediately that “always increasing” fails on one interval.
Graphical evidence is especially helpful when the statement is about behaviour across a domain.
The smallest-failing-case heuristic
When a claim is false, the simplest counterexample often has the highest teaching value. Start with 0, 1, -1, a small fraction or the smallest allowed integer.
Simple values reduce arithmetic noise and make the logical reason for failure easier to explain.
The edge-and-middle test
For an interval-based claim, test one boundary value, one interior value and one value near the opposite boundary.
This does not prove the claim, but it is an efficient way to detect obvious failure before investing in more formal reasoning.
The reverse-claim warning
Claim: “If a shape is a square, it has four equal sides.” The reverse statement “If a shape has four equal sides, it is a square” is not automatically true.
Many examination errors come from reversing a valid implication. Test the reversed statement separately.
The converse clinic
Whenever a rule is written as “if A, then B”, ask whether “if B, then A” has also been established.
A counterexample may show that the converse fails even when the original statement is valid.
The necessary-versus-sufficient habit
A condition can be necessary without being sufficient, or sufficient without being necessary.
Four equal sides are necessary for a square but not sufficient by themselves. A square needs additional angle conditions.
The classification drill
Give the learner several properties and ask which shapes must have them, may have them or cannot have them.
This builds set-based reasoning and reduces the tendency to treat one shared property as a complete definition.
The false-rule repair
After refuting a claim, rewrite it with the missing condition included.
This is the most important final step. It transforms a mistake into a more precise mathematical rule rather than leaving only a list of exceptions.
The counterexample quality check
- Inside the stated domain?
- Satisfies the claim’s starting conditions?
- Actually violates the conclusion?
- Simple enough to explain clearly?
- Does not rely on an accidental calculation error?
If all five are yes, the counterexample is useful.
The boundary quality check
- Exact boundary identified?
- Equality inclusion checked?
- Nearby values tested appropriately?
- Domain still respected?
- Contextual constraints preserved?
Boundary checks should remain connected to the original question rather than become abstract side calculations.
The claim-repair notebook
Keep a short record of false broad rules that appeared in practice and the condition that repaired each one.
Examples might include equal-time averages, equal-weight group means, same-numerator fractions and directly proportional relationships through the origin.
The advanced reasoning standard
An advanced learner can do more than calculate an answer. They can challenge an overbroad rule, produce a valid counterexample, identify the domain and state the condition that makes the corrected rule work.
This is not about turning every question into formal proof. It is about refusing to confuse a pattern with a law.
Final perspective
Mathematics becomes more reliable when claims are treated as objects that can be tested.
Ask where the claim lives, what happens at the boundary and whether one simple case can break it. If the claim survives examples, do not assume it is proven; if a counterexample breaks it, refine the rule. That habit protects method selection, checking and interpretation across the paper.
One final counterexample rule
When a claim survives several examples, resist the temptation to upgrade repeated agreement into proof. Ask instead what feature all your examples share. You may have tested only positive values, equal groups, familiar shapes or interior points while the claim fails at a boundary or under a different valid case.
When a counterexample succeeds, do not stop with “false”. State the exact condition the example exposes. The most useful mathematical correction is a narrower rule that explains both why the original examples worked and why the counterexample did not.
A useful final habit is to ask whether the claim would still survive if the numbers, signs, shapes or group sizes changed within the allowed domain. If one valid change breaks it, you have found the boundary of the rule. Record that boundary explicitly so the corrected principle becomes reusable in later problems.
The corrected rule should now explain both the successful examples and the counterexample that exposed the missing condition.
