Two variables can rise or fall together without one causing the other. A hidden factor may affect both, or the direction of influence may run in reverse. Learn how researchers separate correlation from causation—and why a clear link can still tell only part of the story.

Multiple Choice

A researcher claims that a correlation between two variables implies causation. What is the correct critique?

The idea being tested is that correlation does not establish causation. Two variables can move together for several reasons other than one causing the other. A third factor could influence both and create a spurious relationship, or the direction of influence could run the other way (one variable affecting the other, or even bidirectional effects). To claim causation, you need evidence that changing one variable produces a change in the other while ruling out alternative explanations, typically through controlled experiments or strong causal inference methods. For example, temperature might drive both ice cream sales and outdoor activity levels, so they correlate without ice cream causing the other outcomes. This is why the proper critique is that correlation does not imply causation—you can have a relationship that’s due to a third variable or reverse causation.

A graph can make two things look tightly linked. Ice cream sales rise, and swimming accidents rise. The line on the chart climbs; the pattern feels persuasive. But does buying a frozen treat cause someone to swim—or get into trouble in the water? Not necessarily. A third factor, such as warm weather, may be driving both trends. This is the heart of a key idea in data literacy: correlation can reveal a relationship, but it doesn’t, by itself, show that one thing caused another.

What correlation tells us—and what it doesn’t

Correlation describes how two variables change in relation to each other. If they tend to rise together, they have a positive correlation. If one tends to rise as the other falls, they have a negative correlation. And if their changes don’t follow a clear pattern, they may have little or no correlation.

This information can be useful. A link between sleep and mood, for instance, might point researchers toward an important question. A connection between traffic volume and air pollution could help a city decide what to study next. Correlation is often where an investigation begins.

But a pattern is not a cause. Seeing that two variables move together doesn’t tell us why they do. One may affect the other, both may be influenced by something else, or the relationship may be a coincidence. A chart can show what happened; it can’t always explain what made it happen.

That distinction sounds simple, yet it’s easy to miss. When a relationship fits a familiar story, our minds tend to fill in the gaps. If people who carry umbrellas often arrive at work with wet coats, it might seem that umbrellas cause wet coats. More likely, rain causes both the umbrella use and the damp clothing. The story we tell about a pattern needs evidence beyond the pattern itself.

The hidden third factor

A third variable can create, strengthen, or hide a relationship between two others. Researchers sometimes call this a confounding variable. The name sounds technical, but the basic idea is everyday: something in the background may be shaping what we see.

Consider ice cream sales and swimming accidents. Both may increase during hot weather. If we ignore temperature, we might wrongly conclude that higher ice cream sales lead to more accidents. Yet warm days bring more people to pools and beaches, while also making ice cream more appealing. Temperature helps explain why the two numbers rise together.

Similar puzzles show up in health, education, business, and public policy. People who exercise more may report better health, but age, income, access to safe outdoor spaces, or existing health conditions could also matter. A company might find that customers who use a certain feature spend more. Perhaps the feature encourages spending—or perhaps the company’s most engaged customers are simply more likely to use it.

This doesn’t make correlations useless. It means they call for careful follow-up. Researchers ask what other factors could be involved, whether those factors were measured, and whether the observed link remains when they’re taken into account. A strong investigation doesn’t just admire the pattern; it checks what might be lurking behind it.

Sometimes the arrow points the other way

Another common trap is reverse causation. Two variables may be linked, but the assumed direction of influence may be wrong.

Suppose a study finds that people who feel stressed tend to drink more coffee. It’s tempting to say that coffee causes stress. That could be part of the story, but perhaps people under pressure reach for coffee to stay alert. Or both could be true: stress may lead to more coffee, and large amounts of caffeine may affect how some people feel. Real life can be less tidy than a single arrow in a textbook diagram.

Direction matters because a correlation alone doesn’t show which variable came first. Researchers need evidence about timing and about how changes unfold. If one event consistently occurs before another, that can help, but timing alone still doesn’t prove cause. A third factor might have set both in motion.

There can also be feedback. For example, low confidence may make someone avoid a challenging task, and avoiding it may make confidence fall further. In that case, influence runs in both directions over time. Reducing the relationship to “A causes B” would leave out much of the picture.

What stronger evidence looks like

To make a causal claim, researchers need to rule out plausible alternatives as well as they can. One approach is a controlled experiment. Researchers change one factor while keeping other conditions similar, then observe what happens. Random assignment can help spread differences among participants across groups, rather than allowing one group to have a built-in advantage.

Experiments aren’t always possible or ethical. No researcher can assign people to smoke for decades just to see what happens. In those cases, scientists may use observational data and causal inference methods. They can compare similar groups, track people over time, account for likely confounders, or study events that create conditions close to a natural experiment.

These methods can strengthen an argument, but they don’t turn uncertainty into certainty by magic. Results depend on data quality, the assumptions behind the analysis, and whether important variables were missed. A polished model can still give a misleading answer if it leaves out a crucial part of the story.

That’s why careful researchers use measured language. They may say a result is “associated with” an outcome, or that the evidence “suggests” a possible effect. Those phrases aren’t timid. They mark the difference between what the data show and what the researcher can reasonably conclude.

Read the claim as carefully as the chart

When you see a claim about two things being connected, ask a few plain questions. What exactly was measured? Could another factor affect both variables? Is the direction of influence clear, or could it run in reverse? Did the study track changes over time, compare suitable groups, or use an experiment?

It also helps to notice the size and setting of a study. A relationship observed in one group may not hold in another. A result based on a small sample may be less steady than one found across many settings. And a strong correlation can still have little practical importance; a relationship may be real but too small to matter much in everyday life.

These questions aren’t about dismissing every finding. They’re about giving evidence the right weight. Correlation can be a valuable clue—sometimes a very important one—but it is not the same as a demonstrated cause.

The ice cream example sticks because it makes the point visible: two trends can rise side by side while a third factor explains much of the connection. The same logic applies when the variables are less obvious and the stakes are higher. A careful reader looks beyond the line on the graph, considers hidden influences and asks what evidence would support a causal explanation. That habit keeps a compelling pattern from being mistaken for the whole story.