Adult height predictor

A 1970 formula that is still the standard of care

The way to predict a child's adult height has not changed in more than fifty years. You add the parents' heights together, add 13 cm if the child is a boy or take 13 cm off if she is a girl, and halve the result. James Tanner published it in the Archives of Disease in Childhood in 1970 and paediatricians still use it today.

What has changed is how much is known about how well it works. And the headline finding, from a 2024 study that measured 303 adults from 23 unusually large families, is one that height calculators almost never print: mid-parental height explains 36% of the variation in how tall children end up.

Thirty-six per cent

That figure deserves a moment. It does not mean the formula is wrong. It means that nearly two thirds of what decides your child's adult height is not captured by their parents' mid-point.

Height is highly heritable — the same study put heritability at 74% — but heritable and predictable are different things. Hundreds of genetic variants contribute, and a child inherits a particular half of each parent's set rather than an average of both. Two siblings with the same parents routinely differ by several inches, and nothing in the formula distinguishes them.

The error bars, and why there are three of them

Tanner's own estimate was that a child's adult height should fall within ±8.5 cm of the prediction, which covers the 3rd to the 97th centile. He later widened it to ±9 cm for girls and ±10 cm for boys. A 95% interval expressed as ±1.64 standard deviation scores works out at ±10.6 cm for girls and ±11.7 cm for boys under the CDC standards.

The 2024 study measured it directly rather than deriving it, and found the standard deviation around the prediction to be 4.7 cm for sons and 4.4 cm for daughters. That is the number this calculator uses to tell you how unusual a given miss would be — about a one in sixty chance of being 10 cm or more below the prediction for a boy, and roughly one in ninety for a girl.

Whichever version you take, the door is wide. A prediction of 170 cm honestly means somewhere between about 160 and 180.

Two things the formula gets slightly wrong

The first is the 13 cm. It assumes the gap between men and women is constant at every height, and it is not. In the CDC charts, men at the 3rd centile are 12.2 cm taller than women at the 3rd centile, while at the 97th the gap is 14.7 cm. Multiplying a woman's height by 1.08 fits the data better — which is, as it happens, exactly what Galton did in 1886.

The second is regression to the mean, and it is the more consequential of the two. Very tall parents tend to have children shorter than the mid-parental prediction; very short parents tend to have taller ones. Galton described this in the paper that gave statistics the word regression, and it is still not corrected for in routine clinical practice. The practical consequence is that this prediction is at its least reliable exactly where people are most curious about it: at the extremes.

The correction you can switch on

There is one bias with an easy fix. Adults start losing height from around thirty, so a parent measured today is shorter than the parent whose genes were passed on. In the 2024 cohort, whose parents averaged 62 years old, children ended up on average 2.7 cm taller than the uncorrected prediction. Correcting for parental age shrank that systematic error to 0.06 cm.

Tanner himself suggested adding 1.5 cm for parents aged 45 to 55 and 3 cm for parents over 55. Current paediatric endocrinology guidelines apply no correction at all, which is why it is off by default here rather than applied silently. If the parents are in their thirties it makes no difference and should stay off.

How clinicians actually use this

Not on its own. A paediatrician calculates the target height from the parents and compares it with the child's projected height — their current height plotted on a growth chart and extended to adulthood along the same centile, or estimated from a bone age X-ray. It is the gap between the two that matters. A child tracking well below their target height is the signal that prompts investigation, not the target height itself.

That comparison needs a measured child and a growth chart, which is exactly what a calculator cannot supply. If you are worried about how your child is growing, a GP or health visitor with the red book is the right place to take it, and they will want the measurements rather than a prediction.

What is missing entirely

Nutrition. Chronic illness. Hormonal conditions. Being born very preterm. Any of these can move adult height by more than the whole error bar, and none of them appear in a formula that takes two numbers as input.

The 2024 authors add one more caution about their own figures: 80% of their participants were of Ashkenazi Jewish ancestry, and whether the measured standard deviations transfer to other populations has not been confirmed.

None of which makes it less fun. It is a real formula from a real paper with real error bars, and knowing how wide those bars are is the difference between a party trick and a small piece of understanding.