One successful example can create false confidence. A learner tries one Mathematics case, the shortcut works, and the shortcut becomes a rule. A character performs one kind act, so the learner concludes the character is always generous. One Science trial matches an explanation, so the learner treats the explanation as proven. In each case, the first example may be correct. The mistake is expanding it farther than the evidence allows.
This volume develops one cross-subject PSLE habit: one example does not prove a rule. When you notice yourself moving from “this worked here” to “this always works”, test a second case chosen to challenge the claim. The second case should not be random. It should probe the boundary, alternative interpretation, different base, different starting condition or changed context that could make the rule fail.
The goal is not to become doubtful about everything. Many school rules are valid because they are taught, defined or supported by broader reasoning. The purpose of the second-case test is to separate an established rule from a pattern you inferred from one example. It is especially useful when you invent a shortcut, generalise from a passage detail, or turn one investigation result into a universal claim.
An example can support without proving
A correct example shows that a rule can work in that case. It does not automatically show that the rule works in every relevant case.
The strength of the conclusion should match the number, variety and logic of the examples.
A counterexample can defeat an overbroad rule
If someone claims a rule always works, one valid case where it fails is enough to disprove the universal claim.
You do not need ten failures to challenge an always statement.
Choose a second case that stresses the rule
A useful retest changes the feature most likely to matter: parity, zero, equality, a different base, a different character context or an alternative scientific condition.
Repeating a nearly identical example may only confirm the same narrow case.
Boundary cases are especially informative
Try zero, one, equal values, maximum or minimum values when those are legal.
A shortcut that works for ordinary values can fail at a boundary.
Reversed cases can expose direction errors
Swap before and after, larger and smaller, giver and receiver, or cause and effect when the task permits.
If the rule breaks when direction reverses, the original shortcut may have been directional rather than general.
Different contexts test transfer
A rule that works with money should still work with lengths or groups if the underlying mathematics is the same.
If it depends on the story noun, the learner may have memorised a surface pattern.
English traits need patterns, not one isolated act
A single action can support a temporary state or situational motive without proving a permanent character trait.
Use repeated behaviour or especially diagnostic choices before making broad character claims.
English vocabulary clues can be local
One word meaning inferred from context may not carry unchanged into another sentence.
Meaning depends on usage, collocation and context.
Mathematical shortcuts need structural reasons
A shortcut is trustworthy when you can explain why it follows from the relationship, not only because two examples gave the right answer.
Test the shortcut on a deliberately different legal case before adopting it.
Science explanations need discriminating evidence
One result may fit several mechanisms.
A second condition that produces different predictions can separate them more effectively than repeating the first result.
Repeated trials answer reliability questions
Repeating one setup can show whether a result is consistent.
It does not automatically test whether a different explanation would also fit.
A second case can preserve the same answer
Not every change should alter the result.
An irrelevant change is useful because it tests whether the learner changes answers just because the surface changed.
Generalisation needs scope words
Some, often, under these conditions and in this passage are weaker than all, always and never.
Use the narrowest wording that the evidence justifies.
A proven rule and an observed pattern are different
School Mathematics may give a rule that applies under stated conditions. Science may establish a relationship under a fair comparison. English may define a grammar structure.
Do not demote an established rule to guesswork, but do check whether the current case satisfies its conditions.
The second case should answer a specific doubt
Do not generate another example merely to stay busy.
State what failure the new case is testing: boundary, base, order, source, scope, measurement or mechanism.
A six-step second-case routine
- State the rule or conclusion you are tempted to generalise.
- Identify the condition under which it worked the first time.
- Choose a second case that changes one important feature.
- Predict what the rule says should happen.
- Solve or inspect the second case independently.
- Keep, narrow or reject the rule based on whether the reasoning survives.
Thirty worked second-case tests
English: one apology
A character apologises once after hurting a friend.
The likely generalisation error is one act becomes permanent trait. The apology may support regret or responsibility in that moment, but it does not by itself prove the character is always considerate.
Look for later behaviour: does the character repair the harm, repeat it, or act differently when another difficult choice appears? The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one refusal
A character refuses help once.
The likely generalisation error is automatic independence. The refusal might show independence, pride, embarrassment or urgency depending on context.
Test a second situation where help is offered under different conditions. A broad trait needs a pattern. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one repeated word
A writer repeats a word in one paragraph.
The likely generalisation error is repetition always means emphasis. It may create emphasis, cohesion, rhythm or simply reflect the topic.
Check another repeated word in context before treating all repetition as the same effect. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one vocabulary clue
A word seems to mean ‘quiet’ in one sentence.
The likely generalisation error is fixed meaning across contexts. The inferred meaning may be local or one sense among several.
Test the word in a second sentence or use nearby context before generalising. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one contrast connector
A sentence with ‘but’ contains a reversal.
The likely generalisation error is all but-sentences have same emotional effect. But signals contrast, but the effect depends on the ideas contrasted.
Compare another sentence where the contrast is factual rather than emotional. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one kind action
A character gives up a seat.
The likely generalisation error is heroic trait generalisation. The action supports consideration, but heroic is much stronger.
Look for a second significant choice or use a narrower trait. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one late arrival
A character arrives late once.
The likely generalisation error is always unreliable. One event supports lateness in that instance, not a permanent pattern.
Check whether the passage shows repeated lateness or a specific cause. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
English: one inference question
A clue leads to the correct answer ‘worried’.
The likely generalisation error is memorised action-emotion mapping. The same action can mean something else elsewhere.
Use a second passage where the action occurs under a different context. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: divide by two after double counting
A learner lists every unordered pair twice in one problem and dividing by two works.
The likely generalisation error is divide-by-two shortcut becomes universal. The shortcut fails when a self-pair or unequal duplication pattern exists.
Test a case containing an equal pair such as 6 and 6 before generalising. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: add percentages
A 10% decrease followed by 10% increase happens to look close to no change.
The likely generalisation error is percent changes treated as additive. Successive percentages use changing bases.
Test with 100: decrease to 90, then increase 10% of 90 to 99. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: nearest whole number
A quotient of 2.4 boxes in one problem is rounded to 2 and happens to fit a maximum constraint.
The likely generalisation error is nearest rounding becomes universal. Minimum-capacity problems may require rounding up.
Test 81 pupils in 40-seat buses: 2.025 requires 3 buses. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: perimeter shortcut
One joined-shape problem works by subtracting twice the shared edge.
The likely generalisation error is shortcut applied to all line-length questions. The rule belongs to outside perimeter for the stated join.
Test a question asking for all drawn lines including the internal edge. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: ratio parts
A 2:3 ratio gives a first group smaller than half the total.
The likely generalisation error is ratio statement overgeneralised. The result depends on the part structure.
Test 3:2 and check how the comparison changes. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: one small case
A pattern formula works for n=2.
The likely generalisation error is formula accepted. One successful case is weak evidence for a general formula.
Test n=3 and a boundary case if legal. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: same answer by two routes
Two methods give the same result.
The likely generalisation error is agreement treated as proof. Both methods can inherit the same wrong assumption.
Use an independent given condition or representation that does not depend on the disputed step. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: double number line
A proportional line works for price per item.
The likely generalisation error is all cost problems treated as proportional. A fixed fee breaks the through-zero relationship.
Test a $5 booking fee plus $4 per ticket. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Mathematics: one counterexample
An always claim fails for one legal value.
The likely generalisation error is keeping the universal rule. One valid counterexample disproves the universal claim.
Narrow the rule to the conditions under which it actually works. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one cup stays warmer
An insulated cup remains warmer after ten minutes.
The likely generalisation error is insulation conclusion overextended. The result can support reduced heat transfer under the fair comparison.
It does not prove every material called insulation behaves identically in all conditions. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one plant grows taller
One plant under a changed condition ends taller.
The likely generalisation error is cause treated as proven. Other relevant variables may explain the difference.
Use a second fair comparison or controlled design before making a causal claim. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one trial matches prediction
A result matches the proposed mechanism.
The likely generalisation error is confirmation treated as proof. Alternative explanations may predict the same result.
Choose a second condition where the explanations predict differently. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one dry surface
No droplets appear on one surface.
The likely generalisation error is absence used as universal rule. No observed droplets under those conditions does not prove the surface can never develop droplets.
Change the temperature or humidity conditions in a paper retest. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one switch state
Both bulbs are on when a switch is closed.
The likely generalisation error is unique circuit inferred. Several circuit arrangements can produce that observation.
Test an open state or another switch configuration where predictions differ. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: one equal endpoint
Two samples finish at the same temperature.
The likely generalisation error is same heat transfer inferred. Starting temperatures, amounts or materials may differ.
Compare a second case with controlled starts or amounts. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Science: repeated identical setup
Three repeats give similar results.
The likely generalisation error is mechanism uniqueness inferred. Repeatability strengthens reliability, not uniqueness of explanation.
Use a second experimental condition that distinguishes mechanisms. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one successful mnemonic
A short cue helps on one question.
The likely generalisation error is cue treated as full rule. The cue may be too narrow or accidentally aligned with that example.
Test it on a changed question and explain why it still applies. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one right answer after correction
A learner gets the repaired question right.
The likely generalisation error is learning declared complete. The learner may remember the correction.
Use a delayed changed-detail retest such as Vol 0056. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one wrong retest
A learner fails one changed question.
The likely generalisation error is skill declared absent. The failure may come from arithmetic, vocabulary, copying or an ambiguous rewrite.
Inspect the reasoning before diagnosing the whole skill. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one high score
A learner scores well on one paper.
The likely generalisation error is readiness guaranteed. One paper samples limited content and conditions.
Look for consistent performance across varied and delayed tasks. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one low score
A learner performs poorly once.
The likely generalisation error is global ability label. The score does not by itself identify the cause.
Classify the error patterns before choosing the repair. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Cross-subject: one example in explanation
A learner provides one relevant example.
The likely generalisation error is universal claim attached. Use the example to support the stated point, not to prove everyone or always.
Narrow the claim unless broader evidence is supplied. The second case should be chosen because it tests a specific weakness in the rule, not because it is merely another question.
Before deciding, predict what the proposed rule says about the new case. Then solve or inspect the case independently. If the outcome differs, revise the rule rather than forcing the new example to fit.
For delayed transfer, change the surface context and remove the warning that a generalisation is being tested. The learner should notice the boundary or alternative without being prompted by the title of the exercise.
Three useful second-case designs
Boundary case
Choose the smallest, largest, zero, equality or one-unit case allowed. Boundaries often expose hidden assumptions. A rectangle-pair rule that works for unequal sides may behave differently for a square. A counting shortcut may fail when an item can pair with itself.
Contrast case
Change the condition that should reverse or alter the answer. If a listening plan is accepted rather than rejected, the final location should change. If a switch moves from one branch to a shared path, more bulbs may be affected. This tests sensitivity to a decisive detail.
Irrelevant-change case
Change a name, colour or other non-causal feature. The answer should remain stable. This case prevents the learner from developing the opposite bad habit: changing the answer whenever anything on the page changes.
When you should trust a rule without inventing endless examples
Not every established school rule needs to be rediscovered from examples. If the Mathematics syllabus teaches a relationship and you understand its conditions, apply it. If a grammar rule or scientific definition is established, the job is to recognise whether the current case satisfies it. The second-case method is most valuable when the learner has invented a shortcut or made a broad claim from limited evidence.
A structural explanation is stronger than a pile of matching examples. For instance, if two quantities remain in a fixed ratio, scaling both by the same factor follows from the definition of equivalent ratios. Testing examples can build confidence, but the reason explains why the rule extends beyond the examples.
Conversely, repeated examples cannot rescue a rule that lacks the necessary conditions. Ten joined-shape problems using full-edge contact do not prove the same formula applies to overlapping figures or questions asking for all internal lines. Scope remains part of the rule.
A seven-day generalisation-control cycle
- Day 1: identify broad words such as all, always, never and must.
- Day 2: test Mathematics shortcuts with boundary cases.
- Day 3: test English trait and vocabulary claims with a second context.
- Day 4: test Science explanations with a competing mechanism or changed condition.
- Day 5: include irrelevant-change cases where the answer should stay the same.
- Day 6: delayed mixed practice without announcing which rule is under test.
- Day 7: review only the rules that failed and rewrite each with its correct conditions.
Parents and tutors: ask what would make the rule fail
When a learner says “this always works”, ask for the conditions rather than immediately disagreeing. Then ask for one legal case near the boundary. The child may discover that the rule is correct under a narrower set of conditions.
If the learner gives one example as proof, ask what a different example would need to show to challenge the claim. This turns correction into a testable decision rather than an argument about authority.
Do not demand a counterexample when the rule is already formally established and the learner’s real problem is applying it. The diagnostic question is whether the learner understands the conditions, not whether every theorem must be rediscovered from scratch.
Frequently asked questions
Does one example prove a rule?
Usually no. It shows the rule works in that example. A structural reason or broader evidence is needed for a general claim.
Can one example disprove an always-rule?
Yes, if it is a valid counterexample within the rule’s stated conditions.
Why test an irrelevant change?
Because understanding includes knowing when the answer should stay the same. Otherwise learners may overreact to surface changes.
How many examples are enough?
There is no universal number. Use examples to test boundaries and alternatives, but rely on the underlying rule or evidence when it is established.
Is a repeated Science trial the same as a new condition?
No. Repeating improves evidence about reliability; changing a condition can test whether a proposed explanation makes a different prediction.
How does this help English?
It prevents one action, one phrase or one context from being turned into a permanent trait, fixed word meaning or universal interpretation.
Official 2026 PSLE frame
SEAB’s 2026 PSLE English, Mathematics and Science syllabuses assess interpretation, application, reasoning and communication in subject-specific forms. Testing a second case is a learner strategy for checking whether an inferred rule or explanation travels beyond one example; it is not an additional official examination requirement. See the 2026 PSLE formats page.
Batch routes
Use Vol 0061 for English modifier attachment, Vol 0062 for proportional-base control and Vol 0063 for temperature-versus-heat reasoning. The wider route remains the PSLE Learning Guide.
The performance rule
One correct example can begin a hypothesis. Before you turn it into a rule, test a second case chosen to expose the boundary.
