In psychology, correlation is a statistical measure of how two variables relate to each other. It tells researchers whether two things tend to move together, like stress and sleep quality, or height and weight, and how strongly that relationship holds. Correlation is one of the most widely used tools in psychological research, but it comes with a critical limitation: it cannot tell you whether one variable actually causes the other to change.
How Correlation Is Measured
Psychologists quantify correlation using a number called the correlation coefficient, represented by the letter “r.” This value falls on a scale from -1.0 to +1.0. A value of +1.0 means two variables move together in perfect lockstep: as one rises, the other rises by a perfectly predictable amount. A value of -1.0 means they move in perfectly opposite directions. A value of 0 means there is no relationship at all between the two variables.
In practice, perfect correlations of +1.0 or -1.0 almost never appear in psychological research. Human behavior is messy, and dozens of factors influence any given outcome. Instead, researchers work with values somewhere between those extremes and interpret how strong the relationship is. The widely used benchmarks proposed by the statistician Jacob Cohen classify an r of .10 as a small correlation, .30 as medium, and .50 as large. So if a study finds a correlation of r = .35 between loneliness and social media use, that would be considered a medium-strength relationship.
Positive vs. Negative Correlations
A positive correlation means both variables increase together. Height and weight are a classic example: taller people generally weigh more. Age and the number of wrinkles on your skin is another. In psychology, you might see a positive correlation between the number of hours spent studying and exam performance, or between exercise frequency and self-reported mood.
A negative correlation means that as one variable goes up, the other tends to go down. For example, researchers consistently find a negative correlation between stress levels and sleep quality. The more stressed someone feels, the less restful their sleep tends to be. Similarly, there is a negative relationship between the amount of time children spend reading and the number of behavioral problems reported by their teachers. The sign of the correlation tells you the direction, not the strength. An r of -.60 is just as strong as an r of +.60; it simply describes a relationship moving in the opposite direction.
Reading a Scatterplot
The easiest way to visualize a correlation is with a scatterplot, a graph where each dot represents one person (or one data point) plotted according to their scores on two variables. If the dots cluster tightly around an upward-sloping line, you’re looking at a strong positive correlation. If they cluster around a downward slope, it’s a strong negative one. The more the dots scatter away from any clear line, the weaker the relationship. A scatterplot that looks like a random cloud of dots suggests little to no correlation between the variables.
Why Correlation Does Not Mean Causation
This is the single most important concept to understand about correlation in psychology, and it trips up even experienced readers of research. Just because two variables are correlated does not mean one causes the other. There are two specific reasons this matters.
The Directionality Problem
When two variables are correlated, it’s often impossible to know which one is driving the other. Vitamin D levels and depression are a good example: studies show they are correlated, but researchers cannot easily determine whether low vitamin D leads to depression or whether being depressed causes people to eat less nutritious food and spend less time outdoors, reducing their vitamin D. The same issue appears with physical activity and self-esteem. Does exercising make you feel better about yourself, or do people with higher self-esteem tend to exercise more? Both directions are plausible, and a correlation alone cannot settle the question.
The Third Variable Problem
Sometimes two variables appear related only because a hidden third factor is influencing both of them. The textbook example: ice cream sales and violent crime rates are positively correlated. But eating ice cream does not cause crime. Hot weather is the third variable. People buy more ice cream when it’s hot outside, and they also spend more time outdoors where interpersonal conflicts are more likely.
This plays out in more subtle ways in real research. One study found that children who regularly eat cereal tend to have healthier body weights, a finding that cereal companies were eager to publicize. But the correlation likely reflects a third variable: children who eat cereal for breakfast may also have more structured eating habits, more physical activity, or more involved parents overseeing their overall diet. The cereal itself may have little to do with it.
Another example involves violent video games and aggression in children. These two variables are correlated, but the quality of parental attention could be a confounding factor. Children who receive less parental supervision may both play more violent video games and exhibit more aggressive behaviors, without one directly causing the other.
How Psychologists Use Correlation Beyond Research Studies
Correlation isn’t just for studying relationships between behaviors. Psychologists also rely on it to build and evaluate their measurement tools, like personality questionnaires, IQ tests, and clinical screening instruments.
One common application is called test-retest reliability. To check whether a psychological test gives consistent results, researchers administer it to the same group of people on two separate occasions and then calculate the correlation between the two sets of scores. If the test is measuring something stable, like a personality trait, the scores from both sessions should be highly correlated. In psychometrics, the field dedicated to psychological measurement, a correlation of about .70 is considered the minimum threshold for acceptable reliability. Values above .80 indicate good reliability, and above .90 is considered excellent.
However, a high test-retest correlation doesn’t automatically mean a test is flawless. The score can be inflated by systematic patterns in how people respond, like remembering their previous answers, or by factors unrelated to what the test is actually trying to measure. That’s why psychologists typically look at several types of evidence together rather than relying on a single correlation to judge a test’s quality.
Why Correlational Research Still Matters
Given its limitations, you might wonder why psychologists bother with correlation at all. The answer is practical. Many of the questions psychologists care about cannot be studied with experiments. You can’t randomly assign people to experience childhood trauma, or to develop a specific personality type, or to live in poverty for ten years and then measure the effects. In these cases, correlational research is the only ethical option.
Correlation also serves as a starting point. When researchers discover that two variables are related, that finding generates hypotheses that can be tested with more controlled methods. A correlation between sleep deprivation and anxiety, for instance, might lead to an experiment where one group of volunteers has their sleep restricted under controlled conditions while another group sleeps normally, with both groups then assessed for anxiety symptoms. The correlation identifies the pattern; the experiment tests the mechanism.
Understanding what correlation means, and especially what it doesn’t mean, is one of the most practical skills you can take from a psychology course. It changes how you read headlines, evaluate health claims, and interpret the studies that shape public conversation about human behavior.