A situation shows causation when a change in one thing directly produces a change in another, not just when two things happen to move together. The difference matters because plenty of situations look causal on the surface but fall apart under scrutiny. To spot genuine causation, you need to check for three ingredients: the cause came before the effect, there’s a direct mechanism linking them, and no hidden third factor is actually driving both.
Causation vs. Correlation
Correlation means two variables move in sync. When one goes up, the other tends to go up (or down). That’s all it tells you. It says nothing about which one is driving the other, or whether something else entirely is pulling the strings. Causation is a stronger claim: a change in variable X directly produces a change in variable Y.
A classic example separates the two instantly. Ice cream sales and drowning deaths both spike during the same months. The correlation between them is real and measurable. But eating ice cream doesn’t cause drowning. Hot weather is the hidden third factor. Summer heat drives people to buy ice cream and to swim, and more swimmers means more drownings. Remove the warm weather from the picture, and the correlation between ice cream and drowning collapses. That hidden driver is called a confounding variable, and it’s the single most common reason people mistake correlation for causation.
What a Causal Situation Looks Like
A situation genuinely shows causation when it passes a simple mental test called the counterfactual: if the cause had not happened, the effect would not have occurred. The philosopher David Hume put it this way in the 18th century: “We may define a cause to be an object followed by another, where, if the first object had not been, the second never had existed.”
Imagine someone takes a medication and their infection clears up. To know if the medication caused the recovery, you’d need to ask: would this person still have the infection if they hadn’t taken it? If the answer is yes, the medication was the cause. In real life you can’t rewind time and test the same person both ways, so researchers use experiments to approximate that test, which is where controlled trials come in.
Here’s a quick way to evaluate any situation you’re given. Three conditions must all be true for causation:
- Time order: The cause happened before the effect.
- A plausible mechanism: There’s a logical, physical explanation for how X could produce Y.
- No confounders: You’ve ruled out other variables that could explain the link.
Examples That Show Causation
Smoking and lung cancer is the textbook case. By the early 1960s, 25 case-control studies showed that smokers were roughly 5.4 times more likely to develop lung cancer than nonsmokers. Seven long-term cohort studies put the mortality ratio even higher, at about 10.8. But the evidence didn’t stop at correlation. Researchers demonstrated a dose-response relationship: people who smoked two packs a day had higher cancer rates than those who smoked one pack. Pathology studies found that heavier smokers had more precancerous cells lining their airways. And the mechanism was biologically straightforward, since the lung sits directly in the path of inhaled smoke. By 1964, the U.S. Surgeon General’s report declared cigarette smoking a cause of lung cancer, not merely a correlate.
A simpler, everyday example: flipping a light switch and the light turning on. The switch happened first (time order), you understand the electrical circuit that connects them (mechanism), and nothing else in the room changed at that exact moment (no confounders). That’s causation.
Contrast that with a situation where students who eat breakfast score higher on exams. That correlation could reflect causation, or it could reflect the fact that students from more stable households are both more likely to eat breakfast and more likely to have study support at home. Without controlling for family income, parental involvement, sleep quality, and other confounders, the breakfast-to-test-score link remains a correlation, not a proven cause.
How Experiments Prove Causation
The gold standard for establishing causation is a randomized controlled trial. Participants are randomly split into two groups: one receives the treatment, the other receives a placebo or standard care. Randomization is what makes it powerful. By assigning people at random, the groups end up balanced on every characteristic, including ones the researchers didn’t think to measure, like genetics or lifestyle habits. Any difference in outcomes can then be attributed to the treatment itself rather than to some lurking confounder.
This is why observational studies (surveys, health records, population data) can identify strong correlations but rarely prove causation on their own. In an observational study, people self-select into groups. Smokers differ from nonsmokers in dozens of ways beyond their tobacco use, including diet, exercise, income, and stress levels. A well-designed experiment strips those differences away. When that’s not possible for ethical or practical reasons, researchers rely on converging lines of evidence: time order, dose-response patterns, biological plausibility, consistency across different populations, and the removal test (does the effect disappear when the suspected cause is removed?).
Common Traps That Mimic Causation
Confounding variables are the most frequent trap, but two others catch people off guard. The first is reverse causation, where the effect is mistaken for the cause. In studies of body weight and mortality, researchers noticed that being underweight seemed to predict earlier death. But in many cases, people were underweight because they were already sick, and the illness caused the weight loss, not the other way around. The low weight was a symptom, not a cause.
The second trap is coincidence on a large enough scale. With thousands of variables tracked across millions of people, some pairs will correlate purely by chance. The divorce rate in Maine has famously tracked the per-capita consumption of margarine for years. No one seriously proposes a mechanism connecting the two. When evaluating whether a situation shows causation, always ask: is there a plausible physical pathway connecting these two things? If not, the correlation is likely noise.
A Checklist for Identifying Causation
When you encounter a situation on a test, in the news, or in a research claim, run through these questions:
- Did X happen before Y? If the timeline is unclear or reversed, causation isn’t established.
- Is there a dose-response pattern? More of X should produce more (or less) of Y. Smokers who smoke more get lung cancer at higher rates. That gradient is strong evidence.
- Is there a mechanism? Can you explain, even simply, how X physically leads to Y?
- Were other explanations ruled out? Was the study randomized? Were confounders like age, income, or preexisting conditions controlled for?
- Is the finding consistent? Does it hold up across different groups, time periods, and study designs?
A situation that checks all five boxes is showing causation. One that only shows two things rising or falling together, without controlling for alternatives or establishing a mechanism, is showing correlation. The distinction is not just academic. It determines whether a medical treatment works, whether a policy will have its intended effect, and whether the headline you just read actually means what it claims.