Regression to the Mean Between Generations Explained

Regression to the mean between generations is the pattern in which unusually tall or short parents tend to have children who are still taller or shorter than average, but by less than the parents were. Extreme parents produce less extreme children, not equally extreme ones.

The idea comes from one study of family heights. It is also where the statistical word "regression" comes from, so it appears in textbooks, in genetics, and in any field that compares one measurement with a later one. If you have ever wondered why a record-setting quarter is usually followed by an ordinary one, the explanation starts here.

In 1886 Francis Galton published "Regression towards Mediocrity in Hereditary Stature" in the Journal of the Anthropological Institute. He had heights for 930 adult children from 205 sets of parents.

To compare mothers with fathers, he multiplied each mother's height by 1.08 to put it on a male scale. He then averaged the two parents into a single figure he called the "mid-parent." He set the population average, which he called the "level of mediocrity," at 68¼ inches without shoes. Mediocrity here only meant ordinary.

His central finding, in his own words: "the height-deviate of the offspring is, on the average, two-thirds of the height-deviate of its mid-parentage." A deviate is how far someone sits from the average. If the parents average 3 inches above it, their children average about 2 inches above it.

Galton explained this as a "double source" of inheritance. A child's heritage, he wrote, "comes partly from a remote and numerous ancestry, who are on the whole like any other sample of past population, and therefore mediocre, and it comes partly only from the person of the parent." He carried the same logic further out. From a single parent he estimated a child inherits one-third of a deviation, and a grandchild one-ninth.

The modern reading is simpler. Height is shaped partly by things passed down and partly by things that are not. A very tall parent is usually tall for both reasons. The child inherits the first share but not the second, so on average the child lands nearer the middle. The same arithmetic sits behind every case of regression to the mean, with or without genes.

No, and Galton himself changed his mind on this point. He first called the effect "reversion," which suggested a biological pull back toward ancestral type. According to a history of the term in The American Statistician, further work on family data showed the effect runs both ways. Start with very tall children and look back, and their parents are also closer to average than the children are. That is why he switched to the word "regression."

A force inside the body cannot make parents regress toward their children. Imperfect correlation can. When two measurements are linked but not perfectly, the extreme cases on either side will tend to sit less extremely on the other. A modern re-analysis of Galton's data notes that regressing mid-parent on child gives a slope of about one-third, not the reverse of two-thirds. This lopsided result is what correlated pairs produce, and it is not a biological mechanism.

The same re-analysis notes that the spread of heights did not shrink over generations. If regression were a squeezing force, everyone would converge on one height. That does not happen because each generation produces its own new extremes, from the parts of height that are not inherited.

The secular trend is the second common mix-up. It describes whole populations getting taller over time. CDC data from NHANES put US men aged 20 to 74 at just over 5 feet 8 inches in 1960 and 5 feet 9.5 inches in 2002. Over the same period, women rose from slightly over 5 feet 3 inches to 5 feet 4 inches. European populations have gained an estimated 10 to 30 mm a decade since the 1800s.

That rise is not genetic drift or statistical reversion. A twin study of 143,390 pairs found the heritability of height stayed in a stable range across birth cohorts from 1886 to 1994, while average height rose sharply. The trend can also reverse. A 2025 NHANES analysis of 44,322 US-born adults found average height falling in people born from around the early 1980s, by between 0.68 cm and 1.97 cm depending on the group. The authors tie the fall to nutrition and health conditions in childhood.

So both effects can show up in the same family. A tall father's son may regress toward the average, while the average itself moves.

Galton's two-thirds is a statement about averages across many families. It does not predict any one child. Plenty of tall parents have children taller than themselves. The rule only says that, averaged over many such families, the children will sit closer to the middle.

The check carries over to work. When you pick out a group because it was extreme, such as top reps, worst-performing stores, or the users who churned fastest, expect the next measurement of that group to be less extreme, even if you do nothing. Before crediting a fix or blaming a decline, ask how far the group would have drifted back anyway. A worked version with real numbers shows how to size that drift.