20 Correlation and Causation Vocabulary Words

Top 20 Vocabulary Words for Children | Vol 120: Distinguish Correlation, Causation and Confounding Factors. When swimming-pool visits and ice-pop purchases rise together on warm days, does swimming cause children to buy ice pops? Not necessarily. Temperature might influence both. The observation is an association; the proposed cause needs more evidence. This guide teaches twenty words that help children spot the difference.

Suggested ages 10–14. All datasets, schools, people and numerical examples in this guide are fictional and are not the results of an actual study. The exercises help children read graphs, analyse statements, write accurate conclusions and practise the Fencing Method. The aim is scientific reasoning, not advanced statistical calculation or advice about real-life behaviour.

Find the section you need: 1. The fair headline that jumped too far · 2. A map of data and causal vocabulary · 3. What the dataset actually contains · 4. Reading patterns without inventing causes · 5. Other explanations and study design · 6. Bias, coincidence and defensible claims · 7. Original story: The Blueberry School Fair · 8. The Fencing Method for responsible conclusions · 9. Twelve worked evidence-reading problems · 10. Ten misconceptions repaired · 11. Seven-day parent plan · 12. FAQs, statistical sources and next reading

The fair headline that jumped too far

Children encounter causal claims in stories, advertisements, social posts and classroom research. ‘People who study more often get better scores, so studying must be the only reason’ looks plausible but can miss important context. The pattern might still be valuable; the problem is announcing a definite causal explanation before the evidence can distinguish other possibilities.

A correlation describes a pattern among values. A cause is a stronger claim about how changing one factor actually affects another. The direction of a scatter plot’s dots does not reveal a causal mechanism. The same two numbers may rise together because of an unmeasured third factor, because the causal direction is reversed or partly because of coincidence.

In the example that follows, the class has just four afternoons of fictional data. The limited sample and the lack of a controlled design do not justify a definite real-world claim. Good English can preserve that truth: ‘In this sample, pool visits were associated with higher ice-pop sales; the relationship could be influenced by temperature.’

A map of data and causal vocabulary

Reader’s question	Words to use	What evidence to inspect
What was recorded?	variable, observation, dataset, sample, population	Rows, units, selection method and target group
What pattern appears?	association, correlation, positive correlation, negative correlation, scatter plot	Direction, shape and exceptions
What might explain it?	causation, confounding factor, alternative explanation, reverse causation	Mechanism, direction and shared influences
How could we investigate?	intervention, comparison group, selection bias	Design, comparability and sampling
What can we conclude?	coincidence, uncertainty, causal claim	Strength, limitations and wording

The fictional afternoon table

Afternoon	Temperature °C	Pool visits	Ice pops sold
A	22	9	10
B	24	12	15
C	29	25	30
D	31	28	35

The invented numbers show all three quantities increasing in this tiny sample. They do not reveal the full year, any actual customer behaviour or a tested causal intervention. Temperature may be one common influence; special school events, price, availability and time of day may also matter. A good student identifies plausible hypotheses without pretending the dataset has already tested them.

What the dataset actually contains

Begin with data definitions. Understanding the unit of observation and the population of interest prevents a child from claiming that four invented afternoons represent an entire year.

Word	Meaning	Natural phrase	Example
1. variable	A characteristic or quantity that can differ between observations, such as temperature or attendance.	measure a variable	The class recorded temperature and number of ice pops sold each afternoon.
2. observation	A recorded event, measurement or case used as evidence.	make an observation	One row represented a single afternoon at the fictional school fair.
3. dataset	An organised collection of information, often as rows and columns.	read a dataset	The invented dataset contained four rows of temperatures and sales.
4. sample	The set of observations selected from a larger group of interest.	four-day sample	Four afternoons formed a very small sample of possible fair days.
5. population	The full group of individuals, events or situations the question is intended to describe.	target population	The class wanted to discuss all school-fair afternoons across the year.

Word 1: variable

Meaning boundary: A variable is not necessarily a cause; it is something measured or classified that may change. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Name one variable without claiming why it changes. Hide the model sentence and ask for a different observation or headline containing “measure a variable”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Use clear units or categories. A label such as ‘warm’ is less precise than temperature in degrees Celsius when numerical comparisons matter. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 2: observation

Meaning boundary: An observation describes what was recorded, not all unseen influences or the reason an outcome occurred. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain what one row of a table represents. Hide the model sentence and ask for a different observation or headline containing “make an observation”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: The same observed association may support different causal stories. Separate what is seen from what is inferred. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 3: dataset

Meaning boundary: A tidy table is not automatically a representative or accurate source, and adding numbers does not prove causation. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Identify the fields and number of observations. Hide the model sentence and ask for a different observation or headline containing “read a dataset”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: The dataset’s collection method, scope and definitions matter as much as the numbers it displays. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 4: sample

Meaning boundary: A sample is not automatically the entire population, and a small or biased sample may mislead. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain why four afternoons cannot stand for all seasons. Hide the model sentence and ask for a different observation or headline containing “four-day sample”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Representative selection matters; selecting only hot days or only successful fairs can distort the patterns. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 5: population

Meaning boundary: Population is defined by the research question; it does not always mean human residents. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Define whether the question concerns four afternoons, one school year or all nearby schools. Hide the model sentence and ask for a different observation or headline containing “target population”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: A conclusion about a broad population needs evidence that supports extending beyond the observed sample. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Quick retrieval: cover the table and show a changed scatter plot or a different four-row dataset. Ask for one accurate pattern description and one separate causal hypothesis that has not yet been established.

Reading patterns without inventing causes

The next five terms name the observed relationship. Positive and negative refer to the direction of an association, not to whether it is morally good or bad.

Word	Meaning	Natural phrase	Example
6. association	A relationship in which values or categories occur together in a patterned way.	observed association	The hotter fictional afternoons tended to show more ice-pop sales.
7. correlation	A statistical description of how two numerical variables vary together, often referring to linear association.	positive correlation	Temperatures and sales rose together in the small example.
8. positive correlation	A pattern where higher values of one variable tend to accompany higher values of the other.	positive relationship	As the fictional temperature increased, ice-pop purchases also increased.
9. negative correlation	A pattern where higher values of one variable tend to accompany lower values of the other.	negative relationship	Warmer afternoons were paired with fewer hot-drink sales in another invented table.
10. scatter plot	A graph showing pairs of numerical observations as points on two axes.	examine a scatter plot	Each point plotted one afternoon’s temperature against the number of drinks sold.

Word 6: association

Meaning boundary: Association alone does not establish that one variable causes another. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Name two measurements that move together without saying why. Hide the model sentence and ask for a different observation or headline containing “observed association”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: The pattern could reflect a common cause, reversed direction, selection effects or chance. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 7: correlation

Meaning boundary: Correlation does not, by itself, prove cause and effect, and a linear coefficient may miss a curved pattern. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain what the data show before inventing a mechanism. Hide the model sentence and ask for a different observation or headline containing “positive correlation”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: A correlation coefficient can be sensitive to outliers and should be interpreted with a scatter plot and suitable measurement units. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 8: positive correlation

Meaning boundary: Positive correlation does not mean that every individual observation increases or that one variable necessarily causes the other. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Describe the overall direction without using the word always. Hide the model sentence and ask for a different observation or headline containing “positive relationship”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Outliers and natural variability can coexist with an upward trend. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 9: negative correlation

Meaning boundary: Negative correlation is not proof that the first variable directly reduced the second. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain the overall downward pattern rather than reading every point as exact. Hide the model sentence and ask for a different observation or headline containing “negative relationship”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: A negative association can arise through a third factor or through a measurement process that varies across groups. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 10: scatter plot

Meaning boundary: A scatter plot displays an association pattern but does not label the correct causal direction automatically. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Identify what the horizontal and vertical axes represent. Hide the model sentence and ask for a different observation or headline containing “examine a scatter plot”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Axis scales and unusual points affect the appearance of patterns. Check the data before assuming a straight-line relationship. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Other explanations and study design

These words move from observation toward a proposed explanation. A child should learn that every arrow labelled causes needs justification, not merely a relationship in a graph.

Word	Meaning	Natural phrase	Example
11. causation	A relationship in which changing one factor actually helps bring about a change in another, under relevant conditions.	evidence of causation	The teacher asked what evidence would show a promotion caused more purchases.
12. confounding factor	A factor related to both the putative cause and outcome that can distort the relationship being studied.	consider a confounding factor	Hot weather might influence both swimming-pool visits and ice-pop sales.
13. alternative explanation	Another account of an observed pattern that is compatible with available evidence.	test an alternative explanation	A special school event might explain higher sales on one afternoon.
14. reverse causation	A situation where the apparent outcome may instead influence the supposed cause.	consider reverse causation	A class considered whether high reading confidence might lead to more voluntary reading.
15. intervention	A deliberate change made to a condition or factor so that its effects can be studied or achieved.	introduce an intervention	The fair might test a clearly defined new sign at one stall.

Word 11: causation

Meaning boundary: Causation is a stronger claim than correlation. Observational association can be compatible with many causal explanations. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Write a sentence using ‘may be associated’ before upgrading to ’causes’. Hide the model sentence and ask for a different observation or headline containing “evidence of causation”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Causal questions require mechanism, study design, appropriate comparisons and consideration of other explanations. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 12: confounding factor

Meaning boundary: A confounder is not merely any extra variable; it must relate in a way that can mislead interpretation of the focal association. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain why temperature could matter to both observed activities. Hide the model sentence and ask for a different observation or headline containing “consider a confounding factor”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: The class is proposing a plausible explanation, not proving the precise influence from four fictional rows. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 13: alternative explanation

Meaning boundary: An alternative should fit the observed facts and be testable where possible; inventing anything imaginable is not equally useful. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Name a second plausible reason for high sales. Hide the model sentence and ask for a different observation or headline containing “test an alternative explanation”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Good reasoning compares explanations against data rather than choosing the most entertaining story. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 14: reverse causation

Meaning boundary: Reverse causation concerns possible direction of effect, not merely a weak association. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Write which variable might be causing which other variable. Hide the model sentence and ask for a different observation or headline containing “consider reverse causation”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: The direction of correlation cannot be determined just by placing one variable on the horizontal axis. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 15: intervention

Meaning boundary: An intervention alone does not make a study convincing; other differences can occur between groups or times. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: State exactly what changed in the hypothetical experiment. Hide the model sentence and ask for a different observation or headline containing “introduce an intervention”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Real experiments involving people require appropriate ethics and consent. For children, use harmless paper simulations. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Bias, coincidence and defensible claims

The final group concerns research quality and claims. Interventions, comparison groups, who was sampled and the role of coincidence all affect how much confidence a conclusion deserves.

Word	Meaning	Natural phrase	Example
16. comparison group	A group or condition used as a reference when evaluating an intervention or outcome.	use a comparison group	The paper experiment compared two imaginary identical stalls, one with a new sign.
17. selection bias	Systematic distortion caused by which observations or participants are included or excluded.	identify selection bias	Counting only the busiest fair afternoons could exaggerate typical sales.
18. coincidence	A pattern or event occurring together without a demonstrated underlying connection.	possible coincidence	Two unrelated totals happened to match on one invented day.
19. uncertainty	The degree to which available information leaves multiple possibilities or outcomes unresolved.	acknowledge uncertainty	The four observations were too few to support a confident causal conclusion.
20. causal claim	A statement asserting that one factor produces or changes another.	evaluate a causal claim	The headline claimed that a swimming club had caused ice-pop sales to rise.

Word 16: comparison group

Meaning boundary: A comparison group helps interpretation only when the groups are meaningfully comparable in relevant respects. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Name what should be kept comparable between two groups. Hide the model sentence and ask for a different observation or headline containing “use a comparison group”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Differences in location, time or visitor type may affect results even if one group is labelled ‘control’. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 17: selection bias

Meaning boundary: Selection bias is not just a small sample; even a large sample can be unrepresentative when selection is distorted. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Explain why surveying only customers at an ice-pop stall misses noncustomers. Hide the model sentence and ask for a different observation or headline containing “identify selection bias”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Before generalising, ask who or what was absent from the data collection. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 18: coincidence

Meaning boundary: Coincidence is a possible explanation, not a proven conclusion whenever evidence is weak. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Suggest what additional observations could make a chance explanation less convincing. Hide the model sentence and ask for a different observation or headline containing “possible coincidence”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Random variation can produce apparent patterns, particularly in small data sets or when many comparisons are searched. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 19: uncertainty

Meaning boundary: Uncertainty is not a synonym for uselessness or error; it expresses limits that guide further checking. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Name one additional fact that would reduce uncertainty. Hide the model sentence and ask for a different observation or headline containing “acknowledge uncertainty”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Statistical uncertainty, measurement error and causal uncertainty can differ. A careful article does not claim that more data automatically solves confounding. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Word 20: causal claim

Meaning boundary: A causal claim needs stronger support than the fact that two quantities rose together. Ask the learner whether this term describes the measured data, the relationship visible in a plot, a possible causal story or the study’s limitations. Those are separate jobs, and one label should not stand in for all of them.

Independent use: Rewrite a causal headline as an accurate observational statement. Hide the model sentence and ask for a different observation or headline containing “evaluate a causal claim”. The child should be able to defend why this precise expression fits the evidence rather than copying a dramatic causal verb.

Further reasoning: Be especially careful when a claim concerns health, education or public policy; a correlation-based headline can misdirect action. Invite a second explanation of the same imaginary data and ask what additional observation, controlled comparison or methodological information could distinguish it. A defensible statement may still leave several possibilities open.

Original story: The Blueberry School Fair

The school, people, sales counts and temperatures in this story are invented teaching material, not a real experiment.

In an invented class project at Blueberry School, four afternoons were recorded at a fictional fair. The children counted ice pops sold and visits to a nearby pool. Their teacher gave them a small dataset that included a third column for afternoon temperature. None of these numbers came from real children, businesses or weather reports.

On the cooler afternoons, fewer pupils visited the pool and fewer ice pops were sold. On warmer afternoons, both numbers were higher. A student named Ravi immediately wrote a headline: ‘Swimming causes ice-pop purchases.’ His friend Nadia asked him to point to the part of the dataset that proved the direction of cause.

Ravi could not. The table showed an association between visits and purchases, not that one produced the other. Nadia noticed the temperature column: hot weather could plausibly increase interest in both swimming and cold treats. She labelled temperature a potential confounding factor rather than an established explanation.

The teacher drew a scatter plot. Each point stood for a single afternoon. The upward direction indicated positive association, but the class discussed why a four-point graph was not evidence of a universal law. An unusual school event, the day of the week or measurement mistakes might influence the observed values.

The children changed the question. Rather than asking ‘Did swimming cause ice-pop sales?’, they asked ‘What evidence would be needed to evaluate whether a new stall sign changed sales?’ The teacher described a harmless paper experiment in which two otherwise comparable fictional stall cards differed only in their displayed sign.

Nadia explained why comparison mattered. If one stall card attracted more customers but also had a better location or longer opening hours, the sign would not be the only plausible cause of the difference. Labelling one card ‘experimental’ would not magically erase those other factors.

Ravi then introduced the phrase reverse causation using another example. Perhaps children who already felt confident reading chose to read more often, instead of reading time alone creating the confidence. The teacher said direction can be complicated and may involve influences both ways.

The class rewrote its original headline: ‘In our fictional four-day dataset, pool visits and ice-pop sales rose together on warmer afternoons. Temperature is one possible explanation, but the chart does not establish causation.’ The sentence was not as dramatic. It was much more informative.

Before they left, the teacher asked whether all correlations are meaningless. Nadia answered no: patterns can be important clues and help make predictions. The mistake was not noticing an association; the mistake was treating that association alone as proof of a specific cause.

Repairing the headline with evidence

Version	Sentence	Assessment
Overclaim	Swimming causes ice-pop sales.	Not established by this dataset
Observed pattern	Higher pool visits accompanied higher ice-pop sales across four fictional afternoons.	Supported description
Competing explanation	Temperature might influence both activities.	Plausible, not proved
Cautious conclusion	The dataset suggests association but cannot establish which variable causes another.	Defensible statement

Invite the child to rewrite the headline in no more than twenty-five words. A good answer names an observed association, a sample limit and one alternative explanation. The phrase ‘does not prove’ is useful when it indicates a specific missing step in reasoning.

The Fencing Method for responsible conclusions

Fence	Statement	Reason it remains defensible
1	Four fictional afternoons were observed.	Names sample size
2	More pool visits accompanied higher ice-pop sales.	Describes association
3	Temperatures were also higher on those afternoons.	Adds recorded context
4	Temperature may influence both variables and could confound the relationship.	Labels a plausible alternative
5	This sample does not demonstrate that pool attendance causes purchases.	Limits the causal claim

The Fencing Method builds a sentence through supported layers. A child who adds ‘therefore warm days definitely make everyone buy ice pops’ has gone beyond the evidence. Replacing definite causes with a clearly stated hypothesis improves the reasoning, rather than making the writer sound uncertain for no reason.

Pair	Difference	Diagnostic question
Variable / cause	Something measured versus a factor with demonstrated effect	Does the table prove the causal role?
Sample / population	Observed cases versus target group	How far can four rows be generalised?
Correlation / causation	A co-varying pattern versus an effect relationship	What alternatives remain?
Confounder / coincidence	Systematic common influence versus chance occurrence	What third factor might affect both?
Positive / negative correlation	Same-direction versus opposite-direction tendencies	What happens to the other variable as one rises?
Intervention / comparison group	Deliberate change versus reference condition	What exactly differs between conditions?

Twelve worked evidence-reading problems

Worked exercise 1: Temperatures and ice-pop sales rise together in a four-day table.

Try independently first. Identify what was recorded, how the observations were selected and whether the statement concerns association or a causal effect. A good answer may be short if it distinguishes the relevant ideas.

Model reasoning: Describe a positive association or correlation in the small sample. Misleading shortcut: Say heat definitely caused every purchase. Why the shortcut fails: Direction of movement is evidence of association, not complete causal proof.

Transfer to a new example by changing the dataset or proposed explanation. Ask your child what a stronger design or a different observation might help settle. Do not treat a paper simulation as evidence about real customers or real students.

Worked exercise 2: Pool visits and ice-pop sales both increase on warm days.

Model reasoning: Temperature is a plausible confounding factor that could influence both. Misleading shortcut: Claim visiting the pool must cause ice-pop sales. Why the shortcut fails: A common cause can produce an association between other variables.

Worked exercise 3: Temperatures rise while hot-drink sales fall in the imaginary data.

Model reasoning: Describe a negative association. Misleading shortcut: Say every person will always buy fewer hot drinks. Why the shortcut fails: Patterns may contain individual exceptions and alternative explanations.

Worked exercise 4: A graph has four points moving approximately upward.

Model reasoning: It illustrates a possible upward relationship, but it is a very small dataset. Misleading shortcut: Claim the pattern proves a universal law. Why the shortcut fails: Sample size and representativeness matter.

Worked exercise 5: The sample includes only the busiest Saturdays.

Model reasoning: Selection bias may overstate typical activity. Misleading shortcut: Call the selected days representative without checking. Why the shortcut fails: The sampling method can distort generalisation.

Worked exercise 6: A student finds a strong correlation between two unrelated tallies.

Model reasoning: Consider coincidence and other explanations before making a causal claim. Misleading shortcut: Say numerical strength is proof of mechanism. Why the shortcut fails: Strength of association does not settle cause.

Worked exercise 7: A headline reads ‘Pool visits cause ice-pop sales’.

Model reasoning: Rewrite as ‘Pool visits and ice-pop sales rose together in this dataset’. Misleading shortcut: Keep the causal verb without evidence. Why the shortcut fails: The revised wording matches the observation.

Worked exercise 8: A child’s confidence and reading time are correlated.

Model reasoning: Consider that confidence may encourage reading, reading may aid confidence or both are influenced by other factors. Misleading shortcut: Assert only one causal direction from a scatter plot. Why the shortcut fails: Reverse causation and feedback can be plausible.

Worked exercise 9: Two stall cards differ by sign and location.

Model reasoning: Location is an alternative explanation that threatens a sign-only causal conclusion. Misleading shortcut: Assume the sign caused the entire difference. Why the shortcut fails: Meaningful comparison requires relevant conditions to be comparable.

Worked exercise 10: A paper experiment changes only the sign on otherwise identical fictional cards.

Model reasoning: This creates a cleaner conceptual comparison for assessing the sign’s possible role. Misleading shortcut: Call it a real-world causal result about buyers. Why the shortcut fails: A simulation teaches design but does not establish actual behaviour.

Worked exercise 11: A scatter plot’s vertical axis starts at an unusual value.

Model reasoning: Examine scales and source numbers before describing the effect size. Misleading shortcut: Infer a huge change solely from a dramatic-looking picture. Why the shortcut fails: Graph presentation can exaggerate perceived differences.

Worked exercise 12: The dataset has missing observations from rainy days.

Model reasoning: State the limitation and ask whether omitted days change the pattern. Misleading shortcut: Conclude the original association is perfectly general. Why the shortcut fails: Missing or selected data can affect results.

Ten misconceptions repaired

Correction 1

Mistaken claim: Correlation is the same as causation.

More accurate version: Correlation describes association; causation concerns an actual effect relationship. Explanation: The statistical questions differ.

Follow up with another plausible example. Ask your child to explain not only what makes the first statement incorrect but also which part could be repaired with cautious language and which would need new research.

Correction 2

Mistaken claim: If two things rise together, one must create the other.

More accurate version: Both could respond to another factor, influence one another or coincide. Explanation: A common pattern admits multiple explanations.

Correction 3

Mistaken claim: The variable on the horizontal axis must be the cause.

More accurate version: Axis position is a plotting choice, not proof of causal direction. Explanation: A graph has no causal arrow by itself.

Correction 4

Mistaken claim: A confounding factor is any third number in a table.

More accurate version: A confounder is related to the explanatory factor and outcome in a way that can distort their apparent relationship. Explanation: Role matters more than column count.

Correction 5

Mistaken claim: A small dataset cannot reveal any meaningful clues.

More accurate version: Small datasets can suggest questions but usually limit confident generalisation. Explanation: Evidence may be preliminary without being worthless.

Correction 6

Mistaken claim: A large dataset is automatically unbiased.

More accurate version: Systematic selection bias can remain even in large collections. Explanation: Sampling method matters.

Correction 7

Mistaken claim: An intervention automatically proves a cause.

More accurate version: Other differences and study design quality still need assessment. Explanation: One deliberate change is only part of causal investigation.

Correction 8

Mistaken claim: A comparison group is useful regardless of its differences.

More accurate version: A meaningful comparison requires relevant conditions to be sufficiently comparable. Explanation: Confounding can persist across groups.

Correction 9

Mistaken claim: A chance explanation is always the true explanation.

More accurate version: Coincidence is one hypothesis among several, not a default certainty. Explanation: Consider mechanisms and better evidence.

Correction 10

Mistaken claim: Admitting uncertainty means the writer knows nothing.

More accurate version: Uncertainty describes precisely what is not settled and what would help investigate. Explanation: Disciplined limitations improve conclusions.

Seven-day parent plan

Day 1

Day 1: Practise variable, observation, dataset, sample and population with the fictional fair table. Define what one row and the target population represent. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 2

Day 2: Study association, correlation, positive correlation, negative correlation and scatter plot. Describe pattern direction without claiming its cause. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 3

Day 3: Learn causation and confounding factor. Ask whether temperature could explain both swimming visits and ice-pop sales. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 4

Day 4: Practise alternative explanation and reverse causation. Rewrite one headline to fit evidence rather than asserting a causal direction. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 5

Day 5: Introduce intervention and comparison group using paper stall cards that differ in one stated feature. Discuss why actual people would need ethical study design. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 6

Day 6: Read selection bias and coincidence, then practise uncertainty and causal claim. Ask what further observations or designs would help. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

Day 7

Day 7: Give a fresh imaginary chart with an apparent positive pattern. Hide all definitions and ask for one observational statement, two alternative explanations and a careful conclusion. Ask for two sentences: one describes what the data show and one identifies an alternative explanation or missing piece of evidence. Close the glossary before the answer and revisit one earlier word tomorrow.

A progress record centred on reasoning

Stage	Observable answer	Useful follow-up
Recognised	Names a variable, sample and correlation.	What does one row represent?
Supported	Describes positive or negative pattern with help.	What does the plot not show?
Independent	Offers a plausible confounding explanation.	How might you test it?
Transferable	Rewrites a new misleading causal headline.	What wording fits the evidence?

The learner need not calculate a correlation coefficient to use these twenty words well. The primary goals are accurate observation, a coherent possible explanation, a fair consideration of alternatives and careful conclusions. A short analysis meeting all four is stronger than a confident claim that cannot survive one question.

FAQs, statistical sources and next reading

FAQ 1. What does correlation mean?

It is a statistical description of how two variables vary together. Positive correlation tends upward; negative tends in opposite directions. It does not by itself establish cause.

FAQ 2. Can correlation be useful without proving causation?

Yes. It may help identify patterns worth exploring or support prediction. Causal conclusions require a stronger basis.

FAQ 3. What is a confounding factor?

Another factor connected with both the proposed cause and outcome that can distort their apparent relationship. In the fictional fair example, temperature is a plausible candidate.

FAQ 4. What is reverse causation?

It is the possibility that the apparent outcome influences the variable treated as the cause. The causal direction can also involve feedback or shared causes.

FAQ 5. Why does sample selection matter?

Observations chosen only from convenient or extreme situations may not represent the larger population. That can distort conclusions.

FAQ 6. Does a controlled comparison automatically prove a result?

No. A strong design seeks comparability and considers chance, measurement, compliance and other influences. A paper exercise is only a learning model.

FAQ 7. What should children write instead of ‘X causes Y’?

When the evidence is observational, use phrases such as ‘X and Y were associated’ or ‘higher X accompanied higher Y’ and identify possible alternatives.

FAQ 8. How can a parent check transfer?

Use a new fictional dataset about umbrellas and wet pavements or reading time and scores. Ask the child to identify variables, the observed association and two causal alternatives.

Statistics references

Penn State STAT 100: Relationships Between Measurement Variables explains several reasons an association might appear, including reversed causation, confounding and coincidence. Penn State STAT 200: Describing Data discusses scatter plots, correlation and outlier limitations. Penn State STAT 500: Collecting and Summarising Data covers lurking and confounding variables. The fair example is fictional and deliberately small.

The connected eduKate vocabulary journey

Vol 89: Data, Means, Medians and Outliers · Vol 104: Online Claims, Sources and Misinformation · Vol 101: Arguments, Evidence and Counterclaims · Vocabulary OS Series Index · The Fencing Method.

Parent takeaway: the most useful question after reading a correlation is ‘What else could explain this pattern?’ A child who can name the data, describe the association, propose a confounding factor and keep the causal conclusion appropriately limited is learning the language of sound reasoning.

Top 20 Vocabulary Words for Children | Vol 120: Distinguish Correlation, Causation and Confounding Factors