Science Education Systems · Article 26. Maya, Jia Jun, Hana and Ethan remain fictional Punggol learners. This article follows the causality layer: how Science distinguishes things that happen together from things that help make other things happen.
The 50-second parent route
Correlation asks whether two things vary together.
Causality asks whether one contributes to producing a change in the other.
The route is:
observation → association → candidate cause → mechanism → controlled comparison → alternative explanations → counterfactual → replication → convergence → calibrated causal claim
The central question is not merely:
What changed?
It is:
What made it change, and how do we know?
This article extends How Science Experiment Design Works, How Science Data Interpretation Works and How Scientific Argumentation Works.
1. Humans are causality-seeking machines
The light flickers.
A bus arrives late.
A plant droops.
A child gets a higher mark.
We immediately ask why.
This instinct is powerful.
It is also dangerous because plausible stories can appear before evidence.
2. Sequence is not enough
A happened.
Then B happened.
That does not prove A caused B.
Temporal order can be necessary for many causal claims, but it is not sufficient.
3. Correlation is not causation
Two variables rise together.
Maybe X causes Y.
Maybe Y causes X.
Maybe Z causes both.
Maybe the association is partly coincidental.
Scientific causality begins by keeping these alternatives alive long enough to test them.
4. Confounders are hidden alternative causes
Students who sleep more score higher.
Does sleep cause the entire difference?
Perhaps.
But study habits, health, stress, family routines or prior attainment may also differ.
A confounder is a factor linked to both the proposed cause and outcome that can create or distort an association.
5. Controlled experiments reduce alternative explanations
When one factor is deliberately changed and relevant others are controlled, causal inference becomes stronger.
This is why fair-test logic is so important in school Science.
The learner is not merely following a procedure.
They are protecting a causal claim.
6. Control groups answer a counterfactual question
What would have happened without the intervention?
We cannot observe the exact same sample both treated and untreated at the same moment.
A control group approximates that missing comparison.
Causal reasoning is fundamentally counterfactual.
7. The counterfactual is the world that did not happen
If the plant received more light, what would have happened if it had not?
If the medicine was given, what would have happened without it?
If the circuit changed, what would have happened if it stayed the same?
Experiments create comparison groups because the true counterfactual cannot be observed directly for the same individual system.
8. Mechanisms strengthen causal explanations
A statistical association says X and Y move together.
A mechanism explains the process connecting them.
Light affects photosynthesis.
Resistance affects current under relevant circuit conditions.
Temperature affects particle motion within a model.
Mechanisms make causal claims more coherent.
9. But a plausible mechanism is not proof
Ethan can invent a clever mechanism for almost anything.
The mechanism must still face evidence.
Plausibility generates hypotheses.
Evidence tests them.
10. Primary 3 causality begins with simple condition–outcome links
Change one relevant condition.
Observe the result.
Keep other important conditions similar.
The child learns that causal claims need fair comparison.
11. Primary 4 causality adds process
Condition.
process.
outcome.
The learner should move beyond “because it changed” toward a short mechanism.
12. Primary 5 causality becomes systemic
One change can produce several downstream effects.
A reduced input affects a process, which affects a system response, which affects the observed outcome.
Causal chains lengthen.
13. Primary 6 causality must survive unfamiliar contexts
The story changes.
The same causal structure remains.
Students must identify which variable changed and which mechanism explains the result.
14. Secondary Science formalises causality
Independent variables.
dependent variables.
controls.
rates.
feedback.
mechanisms.
quantitative models.
causal reasoning becomes increasingly explicit.
15. Maya’s causality weakness is first-story bias
She sees the first plausible cause and stops.
Her repair:
name one alternative explanation before deciding.
16. Jia Jun’s causality weakness is keyword causation
He writes “friction” or “photosynthesis” and assumes the causal chain is complete.
His repair:
state what the concept does in this specific system.
17. Hana’s causality weakness is fear of claiming anything
She notices that perfect certainty is impossible.
Then refuses to identify a strong causal relationship even after a well-controlled design.
Her repair is calibrated confidence.
18. Ethan’s causality weakness is alternative overload
He keeps every possible cause alive indefinitely.
His repair:
rank alternatives by evidence and mechanism.
19. Dose-response patterns can strengthen causal reasoning
As exposure increases, the outcome changes systematically.
This can support causality when other explanations are controlled.
But dose-response alone is not decisive.
The design and mechanism still matter.
20. Temporality matters
For X to cause Y, X usually must occur before Y.
If the supposed cause appears after the outcome, the causal story needs revision.
21. Strength of association can matter
A large, consistent effect may be harder to explain away than a tiny unstable one.
But strong association can still arise from bias or confounding.
Effect size is one clue, not the whole case.
22. Consistency across studies matters
Different groups.
different settings.
similar relationship.
Replication can strengthen confidence that the cause is not an artefact of one study.
23. Specificity can sometimes help
If one exposure is tightly linked to one distinctive outcome, causal inference may strengthen.
But many real causes have multiple effects, and many outcomes have multiple causes.
Specificity is useful in some cases, not a universal rule.
24. Biological plausibility can help
Does the proposed cause fit established scientific knowledge?
If not, stronger evidence may be needed.
But new discoveries can reveal mechanisms previously unknown.
Plausibility should guide scrutiny, not become a barrier to genuine novelty.
25. Natural experiments can support causality when controlled trials are impossible
Sometimes researchers cannot ethically or practically assign exposures.
Natural events or policy changes can create comparison conditions.
Causal reasoning then depends heavily on design and assumptions.
26. Randomisation is powerful because it distributes unknown confounders probabilistically
In suitable experiments, random assignment helps make groups comparable on both measured and unmeasured factors on average.
This strengthens causal inference.
Students need not master advanced statistics to understand the principle.
27. Blinding can reduce expectancy effects
If participants or researchers know which condition is which, expectations can influence behaviour or measurement.
Blinding is one way to reduce some biases.
28. Placebo controls can separate treatment effects from expectation and context
In medicine, an inactive comparison can help estimate how much improvement is attributable to the treatment itself rather than expectation, natural recovery or other influences.
This is causal design in practice.
29. Ethical constraints shape causal evidence
We cannot randomly assign harmful exposures merely to prove causation.
Researchers must sometimes rely on observational evidence, natural experiments and converging methods.
See How Scientific Ethics Works.
30. Causal inference can be strong without one perfect experiment
Mechanism.
observational consistency.
natural experiments.
dose-response.
replication.
multiple methods.
Converging evidence can build a strong causal case.
31. Reverse causation is a common trap
People who are ill take more medicine.
Does medicine cause illness?
The direction may be reversed.
Temporal sequence and study design matter.
32. Selection effects can create false causal stories
Who enters the study?
Who stays?
Who drops out?
If participation differs systematically, observed relationships can be distorted.
33. Measurement error can weaken causal inference
If the supposed cause is measured poorly, the relationship can be blurred or biased.
Measurement quality therefore sits upstream of causal confidence.
See How Scientific Measurement Works.
34. Misclassification can distort causality
If participants or objects are placed in the wrong categories, comparisons become unreliable.
Classification and causality are connected.
35. Causal chains can have mediators
X affects M.
M affects Y.
M is a mediator in the pathway.
Understanding intermediate mechanisms helps explain how causes produce effects.
36. Moderators change the strength of causal effects
A treatment works better for one group than another.
Temperature changes the effect of another variable.
Context matters.
Causes often operate conditionally rather than universally.
37. Necessary and sufficient causes are different
A necessary condition must be present for an outcome.
A sufficient condition can produce the outcome under stated conditions.
Many real scientific causes are neither simply necessary nor sufficient by themselves.
38. Multiple causation is normal
Plant growth depends on light, water, nutrients, temperature, genetics and more.
Exam answers sometimes isolate one factor because the experiment controls the others.
Real systems can have many interacting causes.
39. Feedback complicates causality
A affects B.
B then affects A.
Simple one-way arrows are no longer enough.
This leads directly to systems thinking.
See How Scientific Systems Thinking Works.
40. Intervention is one test of causal understanding
If we believe X causes Y, changing X should alter Y in a predictable way under relevant conditions.
Interventions connect causal models to action.
41. Prediction is another test
A causal model should predict what happens when conditions change.
If predictions repeatedly fail, the model requires revision.
42. Counterexamples can reveal missing causes
The model predicts Y whenever X occurs.
Then X occurs without Y.
Maybe another condition is required.
Boundary failures improve causal models.
43. Causal diagrams can make assumptions visible
Boxes and arrows can show proposed relationships.
Once drawn, hidden confounders and mediators become easier to discuss.
Representation supports reasoning.
44. Arrows should mean something precise
An arrow should not simply mean “related.”
If it represents causation, the learner should be able to explain the mechanism or evidence supporting that direction.
45. Causality and explanation overlap
Many scientific explanations are causal.
They describe how a change propagates through a mechanism.
See How Science Explanation Works.
46. Causality and evidence overlap
The stronger the causal claim, the more demanding the evidential burden.
See How Science Evidence Works.
47. Causality and uncertainty overlap
Even strong causal evidence can leave uncertainty about effect size, mechanism details or which groups are affected most.
Scientific language should preserve those boundaries.
48. Causality and decision-making overlap
If an action changes a cause, what outcome should we expect?
Policy, medicine and engineering depend on causal models because decisions aim to change the future.
See How Science Decision-Making Works.
49. AI can generate causal stories too easily
A language model can produce a plausible explanation from correlation.
The learner should ask:
What evidence supports the direction?
What confounders exist?
Was there an intervention?
What mechanism is established?
Can the claim be replicated?
50. AI can help train causal critique
Useful prompts:
“Give me a correlation and three possible causal stories.”
“Hide one confounder and let me find it.”
“Give me a causal claim that overreaches the study design.”
“Ask me what experiment would discriminate among explanations.”
51. Parents can model causal discipline with one phrase
“How do we know that caused it?”
This is useful after:
“I wore my lucky shirt and scored well.”
“I ate this and felt better.”
“The app changed and my phone became slow.”
The question does not deny the possibility.
It asks for the evidence chain.
52. Small-group tuition can compare causal models
Three students draw different arrows for the same result.
The tutor asks:
Which arrow is supported?
Which needs another control?
Which reverses cause and effect?
Which leaves a hidden variable?
Causal reasoning becomes visible.
53. A compact causality checklist
- What is the proposed cause?
- What is the outcome?
- Did the cause occur first?
- Are the variables associated?
- What alternative explanations exist?
- Could a confounder explain both?
- Was there a controlled intervention?
- What mechanism connects cause to outcome?
- Does the effect replicate?
- Does it survive different methods?
- What conditions modify the effect?
- How strong should the causal claim be?
54. Frequently asked questions
What is the difference between correlation and causation?
Correlation means variables are associated. Causation means a change in one contributes to producing a change in another under relevant conditions.
Why are controlled experiments powerful?
They reduce alternative explanations by deliberately changing one factor while keeping relevant other conditions comparable.
What is a confounder?
A confounder is another factor related to both the proposed cause and outcome that can create or distort an observed association.
Why does mechanism matter?
A plausible, evidence-supported mechanism explains how the cause could produce the observed effect and strengthens causal coherence.
Can observational studies support causality?
Yes, especially when multiple designs, natural experiments, mechanisms and replications converge, though causal inference generally requires more caution than in well-controlled experiments.
How does causality help PSLE Science?
It supports fair-test reasoning, variable control, explanation, prediction and identifying why one changed condition affects an outcome.
How does causality change in Secondary Science?
It becomes more formal and quantitative, incorporating confounding, mechanisms, rates, feedback, experimental design and uncertainty.
55. Continue the Science Education Systems series
- How Scientific Classification Works
- How Scientific Systems Thinking Works
- How Science Decision-Making Works
Conclusion: A cause is a claim about what would change the world
Maya sees a pattern.
Jia Jun names the variable.
Hana asks for a control.
Ethan suggests another explanation.
Science requires all four.
Association.
mechanism.
comparison.
counterfactual.
replication.
uncertainty.
A causal claim is stronger than saying two things happen together.
It says that changing one part of reality would help change another.
That is why causal reasoning sits at the centre of explanation, experiment and responsible action.
