# What does a correlation coefficient tell you?

> Written with AI · [Psychology: research methods and statistics](https://lessoons.com/en/psychology-research-methods) · Grade 10 · 2026-11-13 · https://lessoons.com/en/psychology-research-methods/what-does-a-correlation-coefficient-tell-you

A correlation coefficient tells you how strongly two variables are related and in which direction, in a single number from −1 to +1. The sign gives the direction: plus means they rise together, minus means one falls as the other rises. The size gives the strength: near 0 there is no straight-line relation, near 1 or −1 a very strong one. It never shows cause.

## The topic explained

### One number for a relationship

A correlation coefficient, written r, sums up how two measured variables go together. The most common one, Pearson's r, was developed by Karl Pearson from an idea Francis Galton introduced in the 1880s. It always lies between −1 and +1, and it answers two separate questions. Its sign gives the direction: a positive correlation means that higher scores on one variable go with higher scores on the other, a negative correlation means that higher scores on one go with lower scores on the other. Its size, ignoring the sign, gives the strength: at +1 or −1 every point lies exactly on a straight line, and at 0 there is no straight-line relation at all.

Every correlation starts from a scatter graph, one point per person. In the invented data below, ten students who studied for more hours tended to score higher, and the points lie close to a rising line, so r is high and positive.

### How big is big? A real example from Israel

A rule of thumb from the statistician Jacob Cohen (1988) reads the size of r, whatever its sign: below 0.10 negligible, 0.10 to 0.29 low, 0.30 to 0.49 moderate, and 0.50 and above high. Israel's National Institute for Testing and Evaluation uses it to judge the psychometric entrance exam. Among 100,863 students in all Israeli universities, from the 2005/6 to the 2009/10 cohorts, the exam's general score correlated 0.43 with first-year grades, the bagrut average 0.36, and the two combined 0.46: moderate correlations, and the combination predicts best.

A frequent misreading is that r = 0.43 means the exam predicts 43 percent of a grade. The square of r, here about 0.18, is closer to that idea: roughly 18 percent of the differences in grades go with differences in the score. Moderate correlations are normal in psychology, because behavior has many causes.

### Negative is not weak

Students often treat a negative correlation as a weak or bad one. The sign says only which way the points slope. In the invented data here, students with more screen time late at night slept less, and the points fall steeply along a line, so r is close to −1: a strong relationship. A coefficient of −0.92 is just as strong as one of +0.92, and much stronger than +0.30. In an exam answer, describe a coefficient with both of its parts, for example "a strong negative correlation", and then say in words what it means for the two variables: the more of one, the less of the other.

### What r cannot tell you

Pearson's r measures only straight-line relations. Points that rise and then fall in an arch can be perfectly related and still give r = 0, so always look at the scatter graph first. One extreme point, an outlier, can pull r up or down a long way. A narrow range lowers r too: the institute computes its coefficients only for students who were admitted, whose scores are bunched together, and corrects them for this range restriction. Above all, r is still a correlation, so even a strong one does not show which variable, if either, causes the other.

## Practice: reading correlation coefficients

For 10th grade psychology students. Answer each question from the page, then check your work against the answers.

1. **What are the smallest and largest values a correlation coefficient can take, and what does each extreme mean?** — From −1 to +1. At +1 all points lie on a rising straight line; at −1 all points lie on a falling straight line; at 0 there is no straight-line relation.
2. **Fill in the direction and strength of each coefficient, using Cohen's rule of thumb.** — +0.85: positive, high. −0.62: negative, high. +0.08: positive, negligible. −0.35: negative, moderate.
3. **Look at the scatter graph. Is the correlation positive or negative, and roughly how strong?** — Negative: more days absent go with lower grades. The points fall clearly with some spread, so it is strong; the data give r = −0.85.
4. **Which is the stronger correlation, −0.70 or +0.40? Explain.** — −0.70 is stronger, because strength is the size of r ignoring the sign: 0.70 is larger than 0.40. The minus only says that one variable falls as the other rises.
5. **The psychometric exam's general score correlated 0.43 with first-year grades. A friend says: "So it predicts 43 percent of your grade." Correct the friend.** — 0.43 is a moderate correlation, not a percentage. Its square, about 0.18, means roughly 18 percent of the differences in grades go with differences in the score; much of a student's grade depends on other things.
6. **These nine points are clearly related, yet r is 0. Explain why.** — Pearson's r measures only straight-line relations. Here Y rises and then falls in an arch, so the rising half and the falling half cancel out and r = 0, although the points follow an exact curve. This is why you check the scatter graph first.
7. **A study finds r = 0.65 between the number of friends teenagers report and their happiness. Write two conclusions the study supports and one it does not.** — It supports that the two are positively related and that the relationship is strong by Cohen's guide. It does not show that friends cause happiness: happier teenagers may make more friends, or a third variable, such as an outgoing personality, may lead to both.

## Common questions

### What is a good correlation coefficient in psychology?

It depends on the question, but by Cohen's rule of thumb 0.10 is low, 0.30 moderate and 0.50 high. Many real findings in psychology are moderate, because behavior has many causes, so 0.30 to 0.50 is often an important result.

### What is the difference between correlation and causation?

A correlation shows that two variables go together; causation means one changes the other. A correlation can come from a third variable or run in the reverse direction, and only an experiment that changes one variable can show cause.

### Do I need to calculate r by hand?

Usually not at this level. You are expected to read r from a table or a computer printout, to describe its direction and strength, and to sketch or read a scatter graph. The calculation compares how far each pair of scores lies from the two means.

### What is r squared?

The square of the correlation coefficient. It estimates the share of the differences in one variable that go with differences in the other. For r = 0.5, r squared is 0.25, about a quarter.

### Can a correlation be higher than 1?

No. Pearson's r always lies between −1 and +1. A value outside that range means a calculation or typing error.
