Science // Scientific Inquiry

Correlation vs Causation

Two things happening together doesn't mean one caused the other — recognise the difference between correlation and causation before drawing conclusions

Ages
8 to 10
Type
Thinking skill
Builds on
2 topics
Leads to
2 topics

Quick check

If the learner noticed that children who eat breakfast tend to do better at school, would they understand that breakfast might not be the cause — there could be another reason both things are true?

Ask this to see whether the topic is secure.

Topic // Mastery

What mastery looks like

What a learner can do once this topic is secure.
  • developmental changes in children's recognition of evidence relevance to causal explanations
  • causal learning research

Topic // Map

Where this fits

What to learn first and what this topic opens up, with the reason for each link. Required links come first.
  • RequiredMust come first
  • HelpfulHelps, but not essential

Before this

2 topics
  • Could there be another explanation?Required

    Scientific Inquiry // Ages 7 to 9

    Recognising that correlation is not causation requires the habit of generating alternative explanations — the correlation/causation distinction is a specific case of asking 'is there another explanation?'

  • Describing Rules & PatternsHelpful

    Learning to Learn // Learning to Learn // Ages 8 to 9

    Evaluating whether a pattern is truly causal requires the universal generalisation habit — asking whether the rule you think you've spotted actually holds across cases

You are here

Correlation vs Causation

Ages 8 to 10 // Thinking skill

Leads to

2 topics
  • Science Can Be RevisedRequired

    Scientific Inquiry // Ages 9 to 11

    Understanding that scientific knowledge is provisional requires seeing concrete examples of how apparent patterns and causal claims get revised — the correlation/causation distinction is one key mechanism

  • Evidence Supporting IdeasHelpful

    Scientific Inquiry // Ages 9 to 11

    Evaluating the strength of scientific evidence requires recognising when apparent correlations might not be causal — a key dimension of evidence quality

Topic // Research-informed

Teaching notes

From the Marble Skill Taxonomy. These notes have not been validated in classrooms, so treat them as a starting point for your own judgement.

The approach

For ages 8–10, teach that two things changing together (correlation) does not prove one causes the other. Learners practise resisting the causal leap, proposing an alternative explanation (a hidden common cause, coincidence, or reversed causal direction), and recognising that a controlled/fair test — changing one variable while holding everything else constant — is what the curriculum treats as establishing cause. The anchor example is ice-cream sales and drowning (or sunburn) both rising in hot weather, with temperature as the hidden common cause.

Common misconceptions

  • Assuming that because two things rise and fall together, one must cause the other.
  • Never considering a hidden third factor that drives both variables.
  • Ignoring the possibility that the causal arrow runs the other way.
  • Treating a strong or striking correlation as automatically causal.
  • Thinking observation of a pattern alone (without a controlled test) can prove causation.
  • Believing coincidental correlations can't happen in 'real' data.

Useful representations

  • Two lines rising and falling together (correlation) with a hidden third variable driving both
  • The confounder diagram: A ← C → B (common cause) vs A → B (true cause)
  • Reversed-arrow case (B causes A) and coincidence case
  • Control-test view: hold the confounder fixed, vary A alone, watch B not follow
  • Scatter plot whose trend flattens once the confounder is controlled

Teaching practices

  • Use vivid, genuinely spurious correlations (ice-cream sales/drownings, shoe size/reading age) so the trap is felt, then reveal the hidden common cause.
  • Teach three escape routes from a correlation — hidden common cause, coincidence, or reversed direction — as things to check before claiming cause.
  • Link causation to the fair test: only holding everything else constant and changing one variable can establish that A causes B.
  • Have children propose the confounder themselves ('what else could make BOTH go up?') rather than being told it.
  • Distinguish 'good evidence for a pattern' from 'good evidence for a cause' explicitly when reporting findings.
  • Frame the correlation/causation check as a special case of 'could there be another explanation?' to connect the habits.

Activities to try

  1. 1

    Bust a fake cause by manipulating a system

    Why it works: A correlation can be produced by a hidden common cause, a reversed causal direction, or coincidence; holding the hidden factor fixed and changing only one variable is what tests whether a true cause exists.

    1. Watch two variables — ICE-CREAM SALES and DROWNING ACCIDENTS — plotted side by side as you scrub a month-slider; notice both lines rise together through summer and fall together in winter.
    2. State the tempting but wrong conclusion the data invites ('it looks like ice cream causes drowning').
    3. Reveal a hidden covered slider; find it is TEMPERATURE, then drag it and observe BOTH lines move together because of it.
    4. Switch to CONTROL mode: pin temperature at a fixed value, then raise and lower ice-cream sales on their own, and watch whether the drowning line follows or stays flat.
    5. Optionally switch to a scatter-plot view and watch how holding the hidden factor (temperature) fixed flattens the apparent trend between ice-cream sales and drowning.
  2. 2

    Name the confounder or the test that settles it

    Why it works: Recognising the causal leap and being able to name a plausible confounder (or the test that would decide) is distinct from merely sensing a claim is wrong.

    1. For a given vivid correlation (ice cream / drowning, shoe size / reading age), decide whether it shows that A causes B or whether something else could explain the link.
    2. Sort a set of claims into 'correlation only' versus 'evidence of a real cause'.
    3. Match each correlation to its most likely explanation: a hidden common cause, a reversed direction, or coincidence.
    4. For each case, describe a fair test, or explain why one is hard or impossible (for example, you cannot change someone's shoe size), that would help check whether A really causes B.
  3. 3

    Interrogate a bold causal headline

    Why it works: Challenging a causal claim means actively testing it against the three escape routes — common cause, reversed direction, coincidence — and insisting on a controlled test before accepting cause.

    1. Read a bold causal claim (e.g. 'children with bigger feet are better at spelling — so big feet cause good spelling').
    2. Ask probing questions out loud or in writing: Could something else cause both? Which way does the arrow point? Could this be coincidence?
    3. Propose a hidden common cause (e.g. older/bigger children have both larger feet and more spelling practice).
    4. Describe how you would actually test whether ice cream affects sunburn so the weather can't fool you.
    5. State what it would take to convince you that A really causes B, not just goes with it.

Science // Scientific Inquiry

More in Scientific Inquiry

All 30 topics