A student who scores far above their usual mark on one exam often slips a little on the next, while a student who bombs an exam often climbs a little afterward. An employee who is praised may ease off the following month; one who is scolded may look better the next month. Many people credit praise or criticism, but often no special cause is needed—it is a statistical pattern. That pattern is regression toward the mean.
In plain terms, regression toward the mean means that an extreme high or low reading, when measured again, tends to come out closer to the average. Every measurement mixes true ability or nature with that day’s luck and error, and the more extreme the value, the more likely luck was leaning hard in one direction.
In this article, we will look in turn at what regression toward the mean means and Galton’s discovery, why extremes drift back toward the average, easy everyday examples, and how not to mistake regression for a cause.
How Regression Toward the Mean Changes the Way We Read Numbers
What It Means and Galton’s Discovery
The person who named the phenomenon was the British scientist Francis Galton. After experiments with sweet-pea seeds, Galton in 1886 compared the heights of about 205 sets of parents and about 930 adult children with their parents' heights. Children of very tall parents were usually less tall than their parents; children of very short parents were usually less short. The children’s departure from the average height was about two-thirds of the parents’ departure.
Galton called this “regression towards mediocrity,” and the word “regression” in today’s regression analysis comes from that usage. The important point is that it does not mean people’s heights grow more alike generation by generation. Groups that start at an extreme move closer to the mean when measured again, but the spread of the overall distribution stays in place.
- Extreme Value: a measurement far from the average—the starting point where regression shows most clearly.
- Correlation Coefficient: how closely two measurements move together; the farther it is below 1, the stronger the regression.
- Measurement Error: the random wobble that makes the same target read a little differently when measured again.
Why Extremes Drift Back Toward the Mean
Regression toward the mean is not a mysterious force; it comes from the structure of measurement. It helps to think of any result as “true level” plus “chance at that moment.”
(1) Results Are Skill Plus Luck
Someone who posts a top score likely had good skill—and good luck that day. Luck is redrawn on the next measurement, so the same streak of fortune rarely repeats.
(2) Weaker Correlation Means a Bigger Pull Back
Heights, where parent–child correlation is fairly strong, pull back only a little; daily sales or game scores, where chance plays a larger role, pull back a lot. If two measurements moved in perfect lockstep, regression would vanish.
(3) It Is a Group Tendency, Not Individual Destiny
It does not mean every top student must fall on the next exam. It is a probabilistic tendency for the average of a group picked at an extreme to move toward the overall mean on the next measurement. Looking backward works the same way: today’s top scorers were often closer to the mean on the previous exam. Regression appears whenever you select on an extreme, regardless of the direction of time.
Easy Everyday Examples
Psychologist Daniel Kahneman heard a famous pushback from Israeli Air Force flight instructors: praise a fine flight and the next one is worse; scold a poor flight and the next one improves. Kahneman realized this was not the power of praise and punishment but regression toward the mean. After a good flight, a drop is naturally likely; after a bad one, a rise is naturally likely—so scolding looked effective.
Familiar Scenes of Regression
- Sophomore Slump: when a standout debut season is followed by a plainer next season, regression often carries much of the explanation.
- Treatment at the Peak: people seek care when symptoms are worst, so almost any treatment can look as if it helped afterward.
- Profit Convergence: in 1933 Horace Secrist treated the fade of high-performing firms toward the ordinary as a law of business, but the statistician Harold Hotelling pointed out that regression alone could explain it.
How Not to Mistake Regression for a Cause
Regression toward the mean is dangerous because it makes us credit our own actions for changes that would have happened anyway. That mistake is called the regression fallacy.
(1) Keep a Comparison Group in View
When you evaluate a new teaching method or treatment, measure a control group that did nothing under the same conditions. If both groups improve equally, the difference is likely regression. One reason clinical trials use a placebo arm is precisely this.
(2) Measure More Than Once
Do not select and evaluate on the same single measurement. Judging from the average of several measurements shrinks the share that belongs to chance.
(3) Delay Interpretation After Extremes
Changes that follow a record high or low should first be suspected as regression; only after enough time has passed should you dig into causes.
Regression toward the mean is the statistical pattern in which chance mixed into an extreme result fades on the next measurement and the value drifts toward the average. The idea that began with Galton’s height study cools the eye we bring to praise and blame, treatment effects, and company results.
Before you hunt for a cause behind the change that follows a striking number, first ask how much of it is chance returning—that is the first step toward applying regression toward the mean wisely in everyday life.
