A researcher reports that hours of sleep and next-day exam performance are positively correlated. What does that finding alone establish?
Choose an answer, then check it.
Psychology · Foundations
Correlation Versus Causation
On this page 9 sections
In 30 seconds
correlation A statistical relationship in which two variables tend to change together; the link may or may not be causal. Full entry → means two things tend to change together; causation A cause-and-effect relationship in which changes in one variable produce changes in another. Full entry → means one thing makes another happen. The two are not the same, and confusing them is the most common mistake in science reporting. A link between variables can be real yet have nothing to do with cause: a hidden third factor may drive both, or the arrow may point the other way. Experiments, which actively change one variable Any measurable characteristic that can take different values, such as hours of sleep, exam scores, or coffee consumption. Full entry → and watch for changes in another, are the method that can test a cause-and-effect claim.
Why this matters
Every day, headlines convert a study's correlation into a confident cause: screens damage attention, late nights wreck grades, walking improves mood. Most of those headlines are wrong, not because the studies are fake but because a link is not a cause. Learning to spot the difference changes how you read news, evaluate health advice, and argue with evidence in any course that uses research. It also protects you from your own mind: humans are pattern-seeking animals, and the pattern the brain loves to see is cause. In college writing and in professional life, the person who can say this study shows a link, not a cause, is the one who reasons clearly. This lesson gives you that move.
The college version
The two words, defined
Correlation and causation are easy to say and easy to confuse, so psychologists start with a plain working definition. A correlation is a relationship between two variables: when one changes, the other tends to change too, in a predictable direction. OpenStax's Psychology 2e puts it directly: when two variables are correlated, it simply means that as one variable changes, so does the other. Causation is a stronger claim: it says one thing makes another happen, that changes in one variable cause changes in another. Noba's Research Designs module makes the same distinction in method terms: correlational research measures variables as they naturally occur and computes how they go together, while experiments actively change one variable and watch for changes in another. The two-word summary researchers repeat is simple: correlation does not mean causation. The rest of this lesson is about what that phrase means and why everyone, including experts, has to keep saying it.
Why the mind leaps from link to cause
Humans are pattern-seeking animals, and that is usually a feature. Spotting that dark clouds mean rain, or that a certain rustle means danger, kept our ancestors alive. The same machinery runs in modern life, but it overproduces: the brain treats a few coincidences as a reliable rule and then dresses the rule up as a cause. OpenStax flags exactly this as a standard topic in analyzing findings: our tendency to look for relationships between variables that do not really exist. Here is an original example. A fan notices that his team has won the last three games he watched while wearing his lucky socks, and he now refuses to watch without them. His sock-wearing and his team's winning are, in his memory, correlated, and he has quietly decided the socks cause the wins. Nothing about the socks explains the games; the pattern is a product of attention, not of the socks. That is the engine behind most bad causal headlines: a real or imagined link, a mind that loves patterns, and a story that turns goes-with into causes.
The third-variable problem
Sometimes a correlation is real, meaningful, and still not causal, because a hidden factor drives both variables. Researchers call that hidden factor a third variable, or a confounding variable. Consider a fictional dataset from a seaside town. Across a summer, the town records daily sunshine hours, sunglasses sold at the boardwalk, and first-aid visits for minor beach injuries. All three rise together. A careless headline could claim that buying sunglasses causes beach injuries. The real driver is simpler: sunny days draw bigger crowds to the beach, and bigger crowds mean more sunglasses sold and more scrapes and sunburns treated. The sunshine, not the sunglasses, is the third variable. OpenStax describes the same structure: while variables are sometimes correlated because one does cause the other, it could also be that some other factor is actually causing the systematic movement in both variables of interest. Whenever you meet a link, ask first: what else moves with both of these?
Reverse causation
The second trap is the direction of the arrow. A correlation says two things move together; it does not say which one is pulling. Consider a fictional workplace study: employees who drink more coffee report worse sleep, and a wellness newsletter concludes that coffee ruins sleep. That may be true, but the data alone cannot rule out the opposite reading: people who sleep badly arrive tired and reach for more coffee. The arrow may point from poor sleep to coffee, not from coffee to poor sleep. This is why the OpenIntro statistics textbook notes that correlation does not specify which variable is explanatory and which is the outcome. The same two variables, the same correlation, two completely different causal stories. Before accepting any headline, ask the direction question: which one could plausibly make the other happen, and could the story work in reverse?
Experiments: the method that can test cause
If correlation can only raise questions, what answers them? Experiments, the subject of a sibling lesson. In an experiment A research method in which the investigator actively changes one variable while controlling conditions, in order to test cause and effect. Full entry →, the researcher actively changes one variable, the suspected cause, while controlling the conditions around it, then measures whether the other variable changes in response. Noba's module states the logic plainly: experiments allow researchers to make causal inferences. OpenStax agrees: correlational research can find a relationship between two variables, but the only way a researcher can claim that the relationship is cause and effect is to perform an experiment. The honest note is equally important. Experiments are powerful but imperfect: they often take place in artificial settings, and many questions cannot be tested experimentally for ethical or practical reasons. So the honest framing for this whole topic is: correlation is a question, and experiments are the best available answer, a good answer that still deserves more studies, more checks, and more questions.
Eli explains
The same idea, in plain words
Explain it like I’m 10
Correlation and causation are two different claims about the same pair of things. Correlation says: these two move together. Causation says: this one makes that one move. The first is a description; the second is a story about a mechanism, and a description never proves the story. Two things can move together because one drives the other, because the other drives the one, because a third thing drives both, or because your mind simply noticed the times they matched and ignored the times they did not. That is why researchers keep the phrase correlation does not mean causation close at hand: not because links are unimportant, they are the starting point of every investigation, but because a link alone does not tell you which story is true.
Picture it like this
Think of a correlation as two friends who always arrive at the same party together. Seeing them walk in side by side tells you they are connected, and you can predict that if one shows up, the other probably will too. But it does not tell you who invited whom, or whether a third friend organized the whole evening and drove them both.
Where the picture stops working
The party analogy breaks down because people can simply ask the friends how they got there; variables cannot answer questions. A correlation also never reveals whether the connection is causal, coincidental, or shared, because the data are silent on the story. And unlike friends, who either show up together or do not, correlations are matters of degree: two variables can be weakly or strongly linked, and even a strong link says nothing about cause.
Worked example
A campus newspaper runs the headline Breakfast Boosts Exam Scores after a survey finds that students who eat breakfast average higher marks on morning exams. A careful reader applies this lesson. First, is the claim correlation or causation? The survey shows a link, nothing more. Second, could a third variable drive both? Students who eat breakfast may also sleep more regularly, arrive on time, and keep steadier routines, any of which could explain the scores. Third, does the arrow have to point from breakfast to grades? It could point the other way: students who are already organized and doing well may be the ones with time for breakfast. The headline is a question, not an answer. To test it, a researcher would need an experiment, randomly assigning students to eat or skip breakfast before the same exam and comparing results. Until that happens, the honest summary is: breakfast goes with better scores; whether it causes them is still open.
Key takeaway
Correlation shows that things move together; causation means one thing makes another happen. A link raises a question, a well-run experiment is the best way to answer it, and even then good science keeps asking.
Quick check
3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.
Why do researchers say that correlational findings cannot support cause-and-effect claims on their own?
Choose an answer, then check it.
In a fictional seaside town, daily sunshine hours, sunglasses sold at the boardwalk, and first-aid visits for beach injuries all rise together across the summer. A headline claims that buying sunglasses causes beach injuries. Which explanation best fits the data?
Choose an answer, then check it.
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related
You’ll learn to
- Define correlation and causation, and state the difference between a link and a cause-and-effect relationship.
- Explain why the human tendency to see patterns makes correlational findings easy to over-read as causal.
- Identify the third-variable problem and reverse causation as two ways a real correlation can exist without causation.
- Apply the correlation-versus-causation distinction to evaluate a research headline or finding.
- Explain why experiments are the method that can test a cause, and why even good experiments leave questions open.
Common mistakes
A strong correlation proves causation.
Strength describes how reliably the variables move together; it says nothing about why. Even a near-perfect link can be driven by a third variable, so strength is never proof of cause.
If a finding is not causal, it is worthless.
Correlational evidence is how research discovers leads. A link raises the question that an experiment can then answer, so correlation is valuable even when it cannot settle causation.
The direction of the arrow is obvious from the data.
Correlational data do not show direction. Reverse causation is always a live possibility, so the arrow has to be argued for, not assumed.
One good experiment settles it forever.
Experiments are the best tool for causal claims, but they are often run in artificial settings, and a single study still needs replication and converging evidence before it becomes a confident conclusion.
Easily confused
Correlation vs. Causation
Correlation says two variables move together; causation says one makes the other happen. A link can be real without any mechanism connecting the two.
Third-variable problem vs. Reverse causation
With a third variable, a hidden factor drives both observed variables; with reverse causation, the two variables are truly connected but the causal arrow points opposite to the headline's story.
Correlational study vs. Experiment
A correlational study observes and measures variables as they naturally occur; an experiment manipulates one variable under controlled conditions, which is what allows a causal test.
Key vocabulary
- correlation
- A statistical relationship in which two variables tend to change together; the link may or may not be causal.
- causation
- A cause-and-effect relationship in which changes in one variable produce changes in another.
- third variable (confounding variable)
- A hidden factor that influences both of the variables being studied, creating the appearance of a direct link between them.
- reverse causation
- A situation in which the causal arrow runs opposite to the assumed direction, so the outcome actually drives the cause.
- illusory correlation
- A perceived relationship between variables that does not really exist, produced by the mind's tendency to notice matches and overlook mismatches.
- experiment
- A research method in which the investigator actively changes one variable while controlling conditions, in order to test cause and effect.
- variable
- Any measurable characteristic that can take different values, such as hours of sleep, exam scores, or coffee consumption.
Sources & references
- Research Designs — Noba Project (Christie Napa Scollon, Singapore Management University)
- 2.2 Approaches to Research — Psychology 2e — OpenStax / Rice University
- Psychology 2e, Section 2.3: Analyzing Findings — OpenStax, Rice University
- OpenIntro — Introduction to Modern Statistics (2e), Section 7.1: Fitting a line, residuals, and correlation — OpenIntro (openintro.org)
EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.
Researched 2026-08-22
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.