Regression to the Mean Fallacy

The regression-to-the-mean fallacy occurs when a return from an unusually high or low result toward a typical level is credited to a cause without enough evidence. Regression to the mean is a statistical pattern, not a force that pushes results toward average.

A clinic tests a person on a particularly difficult day and gets an unusually high symptom score. A later, lower score may reflect treatment, ordinary fluctuation, or both. Repeated measures and a suitable comparison help clarify what changed.

A measurement often combines a person's or system's usual level with temporary influences and random variation. If a result is unusually extreme partly because of those temporary factors, the next measurement is likely to be less extreme on average.

For example, a student who scores far below their usual level on one exam may do better next time even if no intervention occurs. A tutoring program could also help, but the change alone cannot show how much improvement came from tutoring.

The same issue appears when people are selected for help after an unusually poor result. A later improvement may reflect the support, natural variation, or both, so evaluating an intervention needs a suitable comparison.

The fallacy can also work in the other direction. After an exceptional performance, a person may do less well next time and observers may blame a new problem, even though the first result was partly an unusually high fluctuation.

A salesperson has an exceptional month, then returns to their usual sales level. A manager may assume the salesperson lost motivation, but the first month may have been unusually favorable. A longer record can distinguish a lasting change from a temporary peak.

A single before-and-after comparison is often difficult to interpret. Repeated measurements, comparison groups, and attention to how participants were selected can help distinguish a treatment effect from expected variation.

When a result is extreme, ask how stable the measure is and how the person or system usually performs. Record multiple observations where practical instead of treating one point as a permanent baseline.

If evaluating a program, compare outcomes with a similar group or another credible design. Consider what would likely have happened without the intervention, and be cautious about attributing all change to the new action.

Avoid reacting to one unusually good or bad result as if it reveals a lasting shift. Follow up over time and look for a pattern that persists beyond the most extreme observation.

Recognizing regression to the mean does not dismiss real improvement or decline. It helps separate the change associated with an intervention from the change that may have occurred anyway.

Psychological perspective

Extreme measurements are more likely than ordinary ones to include an unusually favorable or unfavorable temporary influence. On a later measurement, that influence may not recur, making the result look more typical even when the underlying process has not changed.

This statistical concept is not a claim that interventions never work or that every change is random. It warns against causal conclusions from selected extremes and a single follow-up, where several explanations may fit the same result.

Thinking clearly

Check the usual range and reliability of the measure before interpreting an extreme result. Gather repeated observations when possible, and note how the person or cases were selected.

For an intervention, use a credible comparison and consider what would have happened without it. Separate evidence that a result changed from evidence explaining why it changed.

Follow outcomes beyond one unusually high or low point. Look for a sustained pattern and report uncertainty rather than giving a program or person credit for every movement toward the typical level.

Recognizing regression to the mean does not dismiss real improvement or decline.

What baseline or comparison would help separate an intervention's effect from ordinary variation?