A correlation coefficient is a number between -1 and +1 that measures the strength and direction of a relationship between two variables. In psychology, it’s one of the most common statistical tools, used to answer questions like whether sleep affects grades, whether stress relates to health outcomes, or whether age predicts certain cognitive changes. The closer the number is to +1 or -1, the stronger the relationship. A value of 0 means no relationship at all.
How the Number Works
The correlation coefficient is represented by the letter r. Its sign tells you the direction of the relationship, and its distance from zero tells you the strength.
A positive correlation means both variables move in the same direction. As one increases, the other tends to increase too. Height and weight are a straightforward example: taller people generally weigh more. In psychology, a classic positive correlation is the relationship between hours of sleep and academic performance, where more sleep tends to accompany higher grades.
A negative correlation means the variables move in opposite directions. As one goes up, the other tends to go down. Researchers at the University of Minnesota found a negative correlation of r = -0.29 between the number of days students slept fewer than five hours and their GPA. More sleep-deprived nights, lower grades.
A correlation of exactly +1 or -1 is called a perfect correlation, meaning every data point falls precisely on a straight line. That almost never happens in real psychological data, where human behavior is messy and influenced by many factors at once.
Small, Medium, and Large Correlations
Not every correlation is equally meaningful. In 1988, the psychologist Jacob Cohen proposed benchmarks that researchers still use today to interpret effect sizes:
- Small: r = .10
- Medium: r = .30
- Large: r = .50
These apply to both positive and negative values, so r = -0.50 is just as strong as r = +0.50. The sign only tells you the direction, not the strength. Cohen himself noted these benchmarks should be used as rough guides when you don’t have more specific context. In some areas of psychology, a correlation of .30 might be considered impressively strong because human behavior is shaped by so many overlapping influences. In others, .30 might be unremarkable.
For perspective, many well-established findings in psychology produce correlations in the .20 to .40 range. A correlation doesn’t need to be large to be important, especially when it applies to millions of people or informs real decisions about health, education, or policy.
Types of Correlation Coefficients
The most widely used version is the Pearson correlation coefficient, sometimes written as Pearson’s r. It measures the linear relationship between two continuous variables and works best when the data follows a roughly normal (bell-curve) distribution. If you’re looking at the relationship between hours studied and exam scores, and both variables are measured on a continuous numerical scale, Pearson’s r is the standard choice.
When data doesn’t follow a normal distribution, or when one or both variables are ranked rather than measured on a precise numerical scale, researchers use Spearman’s correlation (sometimes called Spearman’s rho). This version captures any consistent upward or downward trend between variables, not just perfectly straight-line relationships. If you asked people to rank their stress level from 1 to 10 and then ranked their sleep quality, Spearman’s would be the better fit. Both coefficients range from -1 to +1 and are interpreted the same way.
What a Scatterplot Reveals
The easiest way to visualize a correlation is with a scatterplot, where each dot represents one person (or one observation) plotted according to their values on two variables. A strong positive correlation looks like a cloud of dots tilting upward from left to right. A strong negative correlation tilts downward. When r is close to zero, the dots look like a shapeless blob with no clear direction.
Scatterplots are especially useful for catching something the correlation coefficient can miss: a curvilinear relationship, where the pattern between two variables isn’t a straight line. The relationship between anxiety and performance is a well-known example. A little anxiety can improve focus and performance, but too much anxiety makes performance collapse. Plotted on a scatterplot, this creates an inverted U-shape. The Pearson correlation for this kind of relationship could come out near zero, even though a clear and meaningful pattern exists. Always looking at the scatterplot alongside the number helps you avoid being misled.
Why Correlation Does Not Equal Causation
This is the single most important thing to understand about correlations in psychology. When two variables are correlated, there are three possible explanations, and the correlation alone cannot tell you which one is correct.
First, variable A might cause changes in variable B. Second, variable B might cause changes in variable A. Third, some other variable, C, might be causing both A and B to move together. That third option is called the third-variable problem, and it’s responsible for some of the most misleading statistics you’ll encounter.
A famous example: ice cream sales and crime rates are positively correlated. But ice cream doesn’t cause crime, and crime doesn’t cause ice cream sales. The third variable is temperature. Hot weather independently drives both. In psychology, if someone reports that tattoo coverage correlates with income, the explanation could be peer influence, career self-selection, or dozens of other factors shaping both variables at once.
Correlational research is still genuinely valuable. It identifies relationships that can guide further investigation, it works in situations where experiments would be unethical (you can’t randomly assign people to experience trauma), and it can reveal patterns across large populations. But it tells you that a relationship exists, not why.
Statistical Significance and Sample Size
When you see a correlation reported in a psychology study, it usually comes with a p-value, which indicates how likely it is that the result occurred by chance. A p-value below .05 is traditionally considered statistically significant, meaning there’s less than a 5% probability that the correlation is a fluke of random sampling.
Here’s the catch: sample size heavily influences whether a correlation reaches statistical significance. With a very large sample, even a tiny correlation of r = .05 can be statistically significant, yet it may have no practical importance whatsoever. Conversely, a moderate correlation in a small sample might not reach significance simply because there aren’t enough data points. The p-value tells you whether the pattern is likely real, but the size of the correlation coefficient tells you whether it actually matters. Both pieces of information are needed to draw useful conclusions.
Limits of the Correlation Coefficient
Beyond the causation problem and the curvilinear blind spot, the correlation coefficient has a few other limitations worth knowing about. It’s sensitive to outliers: a single extreme data point can dramatically inflate or deflate the value of r. It only captures the relationship between two variables at a time, which is a serious constraint in psychology, where behavior is almost always shaped by multiple factors simultaneously. And it assumes that the relationship between variables is consistent across the entire range of data, which isn’t always the case. The link between income and happiness, for instance, is stronger at lower income levels and flattens out at higher ones.
Despite these limitations, the correlation coefficient remains one of the most useful and frequently reported statistics in psychology. It gives you a single, interpretable number that summarizes how two aspects of human behavior or experience relate to each other, and it forms the foundation for more advanced techniques like regression and factor analysis that build on the same basic logic.