Blog / Why an Extreme First Self-Test Score Can Move Toward the Mean
Why an Extreme First Self-Test Score Can Move Toward the Mean
Regression to the mean is a selection mechanism, not proof that a person changed or that an intervention worked.
October 13, 2026 · 5 min read · By Cepingku Editorial Team
Ready to start?
Start the Big Five Personality Test →An unusually high or low first self-test score can be followed by a less extreme score even when the underlying attribute did not change. That pattern does not prove the first score was wrong, and it does not prove that a course, coaching session, or intervention between attempts worked. Barnett, van der Pols, and Dobson describe regression to the mean as a repeated-measurement phenomenon produced by non-systematic random error or fluctuation around a true mean. When observations are selected because the baseline value is extreme, the selected group is especially likely to include fluctuation in that direction.
A fictional model shows the selection mechanism
Assume, only for teaching, a fictional cohort in which every person’s stable level is 50 and an observed score equals stable level plus random fluctuation. One person’s initial score of 70 can be written as 70 = 50 + 20. If there is no true change, no practice effect, no scoring change, and the next fluctuation is independent of the first and has expected value zero, the conditional expected retest score is 50 + 0 = 50. This is not a prediction that this individual must score 50. The next value can still be higher or lower. It shows the selection mechanism in that homogeneous fictional cohort: a group selected for first scores like 70 contains positive random fluctuations that need not recur at retest. A real selected group has heterogeneous, unknown stable levels; under the relevant random-error and selection assumptions, its expected retest is merely less extreme, rather than every score returning to 50.
The mirror case applies after an unusually low first score. Selecting the lowest baseline scores preferentially selects negative fluctuation; a later group average can rise toward the middle without an intervention. Barnett and colleagues note that the phenomenon becomes more visible with greater measurement error and when follow-up is examined only in a sub-sample selected with a baseline value. Their paper explicitly discusses psychometric measures as well as other repeated measures.
Why one selected group before and after is insufficient
Suppose only people in the lowest 10 percent on a first assessment receive a study programme. If their average rises later, the programme may have helped, but regression to the mean is also a live explanation because the enrolment rule selected unusually low observations. Real change, practice, fatigue, changed circumstances, and changed testing conditions may also contribute. The same logic applies in reverse when a high-baseline group later falls. A before-and-after average alone cannot separate these causes.
A design can make the claim more testable. Barnett and colleagues describe random allocation to comparison groups: both groups should then be similarly affected by regression to the mean, so the difference between their mean changes is more informative than either group’s change alone. They also discuss selecting participants from two or more baseline measurements, using their mean or a later baseline to reduce the influence of one unusual reading. Neither step makes every alternative explanation disappear, but each addresses the selection mechanism directly.
Use a retest as evidence with its conditions attached
For a personal retest, record the version, language, setting, elapsed time, and whether similar items were practised. Treat an extreme first result as a reason to check conditions and repeat carefully, not as proof of change. The Big Five Personality Test organises one set of self-reported responses across five dimensions. This article does not use another instrument’s error, norms, or validity to make claims about this site’s test, and regression to the mean is not an individual forecast.
Want to see where you land?
Take the Big Five Personality Test and get your results in minutes.
Take the Big Five Personality Test →