Correlation means two things are statistically related; causation means one actually produces the other. That single distinction is the most important concept in psychological research, because confusing the two leads to wrong conclusions about human behavior, flawed treatments, and misleading headlines. Understanding why “related” does not equal “caused by” will change how you read every psychology study you encounter.
Correlation: A Relationship, Not an Explanation
A correlation exists when two variables move together in a predictable pattern. As one increases, the other increases (a positive correlation) or decreases (a negative correlation). Psychologists measure this with a number called a correlation coefficient, which ranges from −1 to +1. Zero means no relationship at all, while +1 or −1 means a perfect relationship.
In psychology, a coefficient of 0.7 or above is generally considered strong, 0.4 to 0.6 is moderate, and anything below 0.3 is weak. These cutoffs matter because many real-world psychological correlations are modest. A correlation of 0.3 between social media use and anxiety, for example, is real but leaves most of the variation in anxiety unexplained by social media alone.
The critical point: no matter how strong a correlation is, it tells you nothing about why the relationship exists. It simply confirms that a pattern is there.
Causation: One Thing Produces Another
Causation means that a change in one variable directly produces a change in another. Smoking causes an increased risk of lung cancer. Sleep deprivation causes impaired memory. In these cases, removing or changing the cause reliably changes the effect.
Psychologists have relied on a framework dating back to the philosopher John Stuart Mill to decide whether a causal claim is justified. Three criteria must be met. First, the cause has to come before the effect in time (temporal precedence). Second, the two variables must actually be related, meaning they covary. Third, and hardest to achieve, every plausible alternative explanation must be ruled out. No hidden third factor can account for the relationship. If any one of these three conditions is missing, the causal claim falls apart.
Why Correlation Gets Mistaken for Causation
The Third-Variable Problem
Two variables can appear tightly linked when, in reality, something else is driving both of them independently. The classic example: as air conditioner sales rise, so do drowning deaths. Air conditioners obviously don’t cause drowning. Summer heat is the hidden third variable, independently pushing up both numbers.
Psychology is full of subtler versions. Exercise correlates with happiness, but physical health may be the real driver, independently making people more likely to exercise and more likely to feel good. Nations that consume more chocolate per capita also win more Nobel prizes, but that likely reflects the fact that wealthy European countries both invest heavily in education and eat a lot of chocolate. The underlying variable is geography and wealth, not cocoa.
The Directionality Problem
Even when two psychological variables genuinely influence each other, correlation alone can’t tell you which direction the arrow points. If self-esteem and academic performance are correlated, does high self-esteem cause better grades, or do better grades build self-esteem? Both stories fit the same data equally well. A correlational study simply cannot distinguish between them without additional evidence about timing.
Coincidence at Scale
With enough variables, some will correlate by pure chance. The per-capita consumption of margarine in the United States tracked almost perfectly with the divorce rate in Maine over a ten-year span. That is a statistical coincidence, nothing more. When researchers or journalists scan hundreds of possible variable pairs, some meaningless matches are inevitable.
How Experiments Establish Causation
The tool that lets psychologists move from correlation to causation is the true experiment, and its defining feature is random assignment. Participants are randomly sorted into a treatment group (which receives the variable being tested) and a control group (which does not). Because the sorting is random, the two groups are, on average, equivalent in every way: personality, background, health, motivation, and any other factor you could name, including factors the researchers haven’t even thought of.
This is why random assignment is so powerful. It neutralizes all those confounding third variables at once. If the treatment group then shows a different outcome from the control group, the only systematic difference between them was the experimental treatment itself. That logic is what allows a direct cause-and-effect conclusion.
Without random assignment, you have a quasi-experiment. Participants end up in groups based on pre-existing characteristics rather than a coin flip, which means unknown differences between the groups may still explain the results. Quasi-experiments can suggest causation, but they can’t confirm it the way a true experiment can.
When Experiments Aren’t Possible
Many of the most important questions in psychology can’t be tested with a true experiment because doing so would be unethical. You cannot randomly assign children to abusive households to study the effects of childhood trauma. You cannot randomly assign people to develop depression to see whether it causes sleep disruption or the other way around. You cannot deliberately expose one group to poverty and shield another from it.
In these situations, psychologists rely on correlational designs, longitudinal tracking (following the same people over years to establish which variable changed first), and natural experiments where circumstances outside the lab create group differences. These approaches can strengthen a causal argument considerably, especially when the cause clearly precedes the effect in time and when researchers statistically control for known confounders. But they rarely close the door on alternative explanations as firmly as a randomized experiment does.
Real-World Consequences of the Confusion
Getting this wrong has real consequences. For years, doctors noticed that women taking hormone replacement therapy (HRT) seemed to have lower rates of heart disease. Some physicians began recommending HRT partly on this basis. But women who chose HRT were also more likely to come from higher socioeconomic groups with healthier diets and more exercise. When proper controlled trials were finally conducted, they revealed that HRT actually raised heart disease risk slightly. A correlation that looked protective was masking a treatment that was mildly harmful.
Psychology headlines create similar traps. A study finding that children who ate candy daily were more likely to be arrested for violent offenses later in life does not mean candy causes violence. Family environment, poverty, parenting style, and dozens of other variables could explain that link. Yet “daily candy leads to violent crime” is exactly the kind of headline that travels fast without the necessary caveats.
How to Spot the Difference
When you encounter a psychological claim, a few quick questions cut through the noise. Was the study an experiment with random assignment, or did it simply observe existing patterns? If it was observational, the results are correlational regardless of how large the sample was or how impressive the statistics looked. A statistically significant correlation still isn’t causation. Statistical significance tells you the pattern is unlikely to be due to chance. It says nothing about what caused the pattern.
Look at the language. Words like “linked to,” “associated with,” and “predicts” signal a correlation. Words like “causes,” “produces,” or “leads to” claim causation and should only appear when backed by experimental evidence or a very strong convergence of longitudinal, cross-cultural, and mechanistic data. If a news headline swaps “associated with” for “causes,” it has made a leap the science didn’t support.
Finally, ask yourself: could a third variable explain this? Could the direction run the other way? If either answer is yes and the study didn’t control for it through random assignment, the causal claim is premature. Thinking through these possibilities is not cynicism. It is exactly what psychologists themselves do when evaluating evidence, and it is the single most useful habit you can build as a consumer of psychological research.