PSLE Cross-Subject Guide: Correlation or Causation?

PSLE Learner’s Guide · Vol 0215 · Correlation, causation and alternative explanations

Find the exact decision this guide helps you make

This guide helps a Primary 6 learner decide whether two changing quantities merely vary together or whether the evidence supports a causal claim by checking time order, controlled comparisons, mechanism and alternative explanations. The short rule is state the pattern first, then test time order, control, mechanism and alternatives before using causal language. You will learn the mechanism, see worked examples, repair common misconceptions, practise in stages and use a compact examination routine.

This is an examination-performance guide, not a replacement for full topic teaching. Use the PSLE Learning Guide for the wider sequence and the eduKate learning hubs for deeper subject instruction.

For full topic teaching, examples and wider progression, continue to the PSLE Learning Guide for the wider English, Mathematics and Science performance sequence. This volume owns the narrower examination-performance decision described above.

A careful note about the current examination year

This article is written for the 2026 PSLE preparation context. SEAB’s official 2026 formats page lists English Language as subject 0001, revised Mathematics as 0008 and revised Science as 0009. Format details can change, so confirm any paper-specific timing, component or administrative instruction with the current official SEAB 2026 source and your school. The reasoning routine taught here is deliberately broader: it strengthens how you read evidence, make a decision and check an answer.

Why this mistake happens

When two things rise together, it is tempting to say one caused the other. The pattern may be real and still have several explanations: the first factor may affect the second, the second may affect the first, a third factor may influence both, or the association may appear only in a small sample. English passages can imply causation through word choice, Mathematics graphs can show association without mechanism, and Science investigations need a fair comparison before a causal conclusion. Strong reasoning preserves the pattern while limiting the explanation to what the evidence earns.

The mind likes a fast label. A familiar word, operation, result or rule triggers an answer before the conditions have been checked. That shortcut sometimes works in easy practice, so it becomes a habit. In a demanding question, however, a nearby distractor preserves the surface pattern while changing one important condition. The learner then performs a correct-looking move on the wrong object. The repair is to slow only the decision point: identify what is being linked or measured, preserve its labels, test the conditions, and then continue at normal speed.

Think of the method as a chain with four locks. Object: what exact idea or quantity is under discussion? Relationship: how are the relevant parts connected? Condition: what must be true for the move to be valid? Claim: what is the strongest conclusion the evidence actually supports? A weak answer often skips from object to claim. A strong answer makes the two middle locks visible.

The one-line operating rule

State the pattern first, then test time order, control, mechanism and alternatives before using causal language. Say it before the first practice set. During timed work, shorten it to four cue words you can recall without stopping.

The step-by-step method

Step 1: Name the two variables or events without building causation into their labels

Name the two variables or events without building causation into their labels. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

At this stage, do not decorate the answer with extra facts. Ask one diagnostic question: “What evidence would make this step wrong?” If you can name a possible failure, check it now. If the failure is absent, continue. This negative check prevents confidence from replacing proof.

Step 2: Describe the observed pattern, direction and scope using the actual evidence

Describe the observed pattern, direction and scope using the actual evidence. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

Step 3: Check whether the proposed cause occurs before the effect

Check whether the proposed cause occurs before the effect. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

Step 4: Ask whether a fair comparison changed the proposed cause while controlling important alternatives

Ask whether a fair comparison changed the proposed cause while controlling important alternatives. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

Step 5: Look for a plausible mechanism and evidence from repeated or independent tests

Look for a plausible mechanism and evidence from repeated or independent tests. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

Step 6: Write the strongest justified conclusion: association, supported causal effect under tested conditions or unresolved explanation

Write the strongest justified conclusion: association, supported causal effect under tested conditions or unresolved explanation. Do this on the page, not only in your head. A brief arrow, label, unit, underline or condition note makes the decision inspectable. If you later find an error, you can repair the exact step instead of restarting the entire answer.

Misconceptions to diagnose and repair

Misconception 1

Tempting move: Turning an upward trend into “X causes Y” without a controlled comparison. This feels efficient because it removes a decision. It is unsafe because the question may have changed the reference, unit, interval, definition or strength of claim.

Repair: return to the line, number, setup or statement that supplies the condition. Mark it with a short label. Then rebuild only the affected link. Keep all earlier work that remains valid. The purpose of checking is targeted correction, not anxious repetition.

Misconception 2

Tempting move: Ignoring a third factor that could influence both variables. This feels efficient because it removes a decision. It is unsafe because the question may have changed the reference, unit, interval, definition or strength of claim.

Misconception 3

Tempting move: Using a single anecdote as proof of a general causal relationship. This feels efficient because it removes a decision. It is unsafe because the question may have changed the reference, unit, interval, definition or strength of claim.

Misconception 4

Tempting move: Rejecting a real association merely because causation is not established. This feels efficient because it removes a decision. It is unsafe because the question may have changed the reference, unit, interval, definition or strength of claim.

Misconception 5

Tempting move: Treating sequence alone as sufficient proof of cause. This feels efficient because it removes a decision. It is unsafe because the question may have changed the reference, unit, interval, definition or strength of claim.

Eight worked cases

Worked case 1: Ice-cream sales and heat illness

Scenario. Both are higher on hotter days.

The tempting error. Claiming ice-cream causes heat illness. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. Temperature is a plausible third factor: it increases demand for cold food and also affects heat exposure. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. State the association and identify weather as an alternative explanation. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Check. Re-read the command word, substitute the conclusion into the original situation and ask whether every noun, number and comparison still has the same referent or unit. If the answer is stronger than the evidence, narrow it.

Delayed transfer. Compare days with similar temperature but different sales. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 2: Practice and score improvement

Scenario. Pupils who practise more often have higher scores.

The tempting error. Concluding every extra practice session directly produces the score difference. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. Prior knowledge, support and motivation may influence both practice and scores. A matched or controlled study would strengthen causation. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Keep the useful association while limiting the causal claim. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Use before-and-after data with a comparison group and reassess. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 3: Plant light experiment

Scenario. Identical plants receive different light durations while water, soil and time are controlled; growth differs repeatedly.

The tempting error. Saying the graph alone proves the mechanism. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. The controlled manipulation supports a causal effect of tested light duration on measured growth under those conditions. The biological mechanism needs its own evidence. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Separate causal result from deeper mechanism. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Change water amount with light and identify the confound. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 4: Narrative sequence

Scenario. A character wears a lucky badge and later wins a race.

The tempting error. Treating earlier occurrence as cause. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. Training, conditions and performance provide alternatives. The story may reveal belief or motif without proving physical causation. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Explain what the sequence suggests about the character before claiming real cause. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Add evidence that the badge changes behaviour and trace an indirect mechanism. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 5: Graph with reversed possibility

Scenario. Screen time and tiredness increase together.

The tempting error. Assuming screen time must cause tiredness. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. Late-night tiredness could also encourage passive screen use, or workload could influence both. Direction remains unresolved without design evidence. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. List plausible directions before selecting one. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Use time-stamped intervention data to strengthen one direction. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 6: Mathematical functional rule

Scenario. A table is generated by y = 3x for chosen x values.

The tempting error. Saying every real-world pair with this pattern proves x physically causes y. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. Inside the defined mathematical model, y is determined from x. A real-world causal interpretation requires knowledge beyond numerical fit. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Separate rule dependence in a model from physical causation. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Fit the same points with a different rule outside the observed range. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 7: Controlled intervention

Scenario. One group uses a new revision routine and a comparable group keeps the old routine; assignment and conditions are controlled and the result repeats.

The tempting error. Calling causation absolutely proven for every learner. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. The design supports a causal effect for the tested population and implementation, but scope and individual variation remain. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Use causal language with explicit conditions. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Change the population or duration and plan a transfer test. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Worked case 8: No association

Scenario. A careful dataset shows no clear relationship within the measured range.

The tempting error. Concluding the variables can never influence one another. The response may contain relevant knowledge, yet it does not answer the precise decision because one label or condition has been lost.

Work the mechanism. No detected association in this range may reflect weak effect, limited variation or measurement sensitivity. Write the decisive evidence beside the step where it is used. This makes the reasoning readable to another person and prevents a later sentence from quietly changing the meaning.

Answer form. Report the result and scope without universalising. Notice that the answer carries both the conclusion and the reason that licenses it. It is complete without becoming a list of everything you know.

Delayed transfer. Expand the range and improve measurement precision. Do this variation after a gap, not immediately, so retrieval rather than imitation drives the solution.

Practice progression: from recognition to transfer

Round 1: mark only

Use six short items. Do not answer them. Mark the object, reference, unit, interval, condition or rule that controls the decision. The goal is accurate noticing without the burden of full writing.

Record the result as one of four codes: N for noticing error, M for method error, I for interpretation error, and C for checking error. Practise the weakest code with a fresh item. Do not repeat an entire worksheet when only one decision needs repair.

Round 2: explain the choice

Solve four items and add one sentence beginning “This is valid because…”. If the sentence merely repeats the answer, the condition is still hidden.

Round 3: repair a wrong answer

Take four plausible wrong answers. Identify the earliest faulty link, preserve the correct work and rewrite only what must change. This develops error diagnosis rather than answer copying.

Round 4: mixed questions

Mix this job with two older skills. Remove headings that reveal the topic. Decide which routine applies before solving. Selection is part of performance.

Round 5: timed set

Use a small timed set with a fixed check budget. Mark a cue at the decision point, finish the answer, then perform one positive check and one rejection check.

Round 6: delayed transfer

Return after two or three days with different surface details. Before seeing notes, write the one-line rule from memory and solve. Compare the new error pattern with the earlier one.

A compact practice log

Question source and date
Decision cue you marked
Your first answer in one line
Earliest wrong link, if any
Corrected rule in twelve words or fewer
Date for delayed transfer
Whether the new item was solved without notes

A useful log is small enough to maintain. It should help you predict the next mistake, not become a scrapbook of completed pages.

Checking that tests meaning, not neatness

A check should have a target. First run a positive test: identify the evidence, calculation or condition that supports your answer. Then run a rejection test: name the closest tempting alternative and the exact reason it fails. This pair is more powerful than rereading the same line while hoping to feel certain.

Use a reverse check where possible. Reconstruct the original amount, replace a connector with its expanded meaning, map a conclusion back to the data, or apply the repaired rule to the apparent exception. Reverse checking exposes lost labels and unsupported leaps because the answer has to return to its source.

Stop when the check has answered a real question. Repeatedly changing a sound answer because it feels unfamiliar is not checking. If the object, relationship, condition and claim all remain consistent, keep the answer and move on.

Examination control

When time is tight, use the cue ORCC: Object, Relationship, Condition, Claim. Write only the smallest useful mark—an arrow, a unit, a time interval, a boundary or a qualifying word. The mark should prevent the known error without slowing the rest of the paper.

Read the command word and identify the decision.
Mark the object or quantity that must keep its identity.
State the relationship or operation in plain language.
Check the condition that makes it valid.
Give the claim at the strength supported by evidence.
Use one positive and one rejection check, then stop.

If you become stuck, do not produce a longer answer immediately. Shorten the task: write what is known, what must be decided and what condition separates the two best options. A precise partial structure often unlocks the next step and also makes any partial credit more defensible.

A seven-day delayed-transfer plan

Day 1: learn the one-line rule and complete two worked cases with notes.
Day 2: solve four short items without notes; classify errors by N, M, I or C.
Day 3: mix two new items with two older PSLE skills and hide the topic labels.
Day 4: explain one case aloud to a parent, tutor or imaginary learner in under two minutes.
Day 5: complete a timed mini-set and use the ORCC marks only at decision points.
Day 6: rest this skill or practise a different subject so memory has to consolidate.
Day 7: attempt one unfamiliar transfer item, then write the smallest rule revision needed.

Success on Day 7 matters more than fluency immediately after reading. Delayed transfer shows whether you can retrieve the method, recognise its job under new surface details and control it without a model beside you.

Guidance for parents and tutors

Ask for the decision, not a performance speech. Useful prompts are: “What exactly are you deciding?”, “Which word, unit or condition controls it?”, “What is the closest wrong option?” and “How can you check without starting again?” These questions keep ownership with the learner while making thinking visible.

When an error occurs, locate its earliest point. Praise accurate diagnosis even before the final answer is corrected. If the learner repeatedly loses the same label, introduce one stable notation: arrows for reference, units beside remainders, interval brackets for rates, or boxes around claim conditions. Avoid adding several new routines at once.

Build variation gradually. Begin with clean examples, then add one plausible distractor, then mix the skill with older content, and finally delay the next attempt. A learner who succeeds only when the worksheet title names the method has learned recognition, not transfer.

Keep emotional temperature low. A surprising exception or remainder is information, not proof that the learner is “bad at” a subject. The practical question is which link changed. Once that link is named, practise a small set that isolates it and then return to mixed questions.

The independent-reader test

Before you call the skill secure, explain one solution to a reader who cannot see your thoughts. Name the starting information, the decision, the condition that controls it and the reason the rejected alternative fails. If the explanation depends on “you know what I mean”, a label is still missing. Add that label to the answer, then remove any sentence that does not help the reader follow the chain.

Next, cover the worked example and rebuild the method from a fresh prompt. Do not copy the surface details. Recover the sequence of decisions. This separates genuine control from familiarity with one page. A learner who can transfer the decision routine to new names, numbers or apparatus has a more dependable examination skill than a learner who can reproduce only the model answer.

Finally, compress the method without losing its safety condition. Write one sentence of no more than fifteen words, then test that sentence against an awkward case. If the short rule gives the wrong result, restore the missing condition. Good compression keeps the mechanism available under time pressure while the awkward case prevents the cue from becoming a careless shortcut.

A timed diagnosis drill

Set aside twelve minutes and choose three unfamiliar questions that could trigger this error. Spend the first minute identifying the decision in each question without solving it. Mark the object, relationship, condition and permitted claim. This preview is not wasted time: it prevents you from entering a long method before you know what the method must preserve.

Solve the three questions in the next eight minutes. After each answer, write one tiny prediction about what a correct answer should preserve: a reference word, a whole-number boundary, a pathway, a unit, a direction or an observable result. Then compare the finished answer with that prediction. If they disagree, repair the earliest link and keep the later work only when its meaning remains valid.

Use the final three minutes to classify the errors. A noticing error means the controlling detail was never marked. A method error means the right information was present but the wrong operation or reasoning chain was used. An interpretation error means the result was calculated or observed correctly but described at the wrong level. A checking error means the answer contained a detectable contradiction that the final audit missed.

Repeat the drill a week later with different surface details and without looking at the old corrections. Improvement is not merely a higher score. Look for a shorter pause at the decision point, fewer unnecessary restarts, more stable labels and a clearer reason for rejecting the nearest wrong answer. Those are signs that the performance routine is becoming portable.

Frequently asked questions

What is correlation?

A pattern in which two variables vary together; it does not by itself establish cause.

What is causation?

A relationship in which changing one factor produces a change in another under stated conditions.

Why does time order matter?

A proposed cause must occur before its effect, though sequence alone is not enough.

What is a third variable?

Another factor that may influence both observed variables and create or alter their association.

Can a graph prove causation?

A graph can display a pattern; causal support comes from design, control, timing and mechanism evidence.

Does lack of causation mean the correlation is useless?

No. A dependable association can still support prediction or further investigation.

What is the fastest exam check?

Write pattern first, then ask: control, before, mechanism, alternative?

What should I study next?

Use the PSLE Learning Guide and subject hubs for evidence, fair tests and conclusion control.

Your next route

Use this guide until you can state the rule, solve an unfamiliar case and explain why the closest alternative fails. Then move to a deeper subject route: the English Learning Hub, Mathematics Learning Hub, Science Learning Hub, or the full learning-hubs directory. Return to the PSLE Learning Guide when you want to place this skill inside a broader revision sequence.

Continue through this four-guide sequence

This four-guide sequence moves from choosing an appropriate expression to choosing a justified conclusion. Vol 0212 matches English register to audience, purpose and relationship. Vol 0213 identifies the exact geometric structure before applying an angle fact. Vol 0214 separates pitch from loudness by linking each perception to the correct vibration feature. Vol 0215 distinguishes a shared pattern from evidence that one factor caused another.

How to Perform in PSLE | Learner’s Guide Vol 0215 | English, Mathematics and Science: Separate Correlation From Evidence of Causation