How to Predict Child Height — 4 Methods Compared, Accuracy & Best Age

Match the method to the child's age. Under 2 years, only the mid-parental target works: (mom + dad ± 5 in) / 2. From age 4, the Bayley-Pinneau multiplier — current height ÷ the fraction of adult height already attained — gives a ±2 in estimate. From age 4 with parents' data, the Khamis-Roche blend is slightly better (R² ≈ 0.93). The interactive matrix at the top of this guide shows exactly which method fits which window.

At home, the Khamis-Roche method is the best overall (R² ≈ 0.93, ±2.6 in) when both parents' heights are known. The chronological Bayley-Pinneau follows (R² ≈ 0.90, ±2 in) and needs only the child's data. A clinical bone-age X-ray plus Bayley-Pinneau beats everything (R² ≈ 0.98, ±1 in) but requires a radiograph ordered by a pediatric endocrinologist.

A defensible prediction opens at age 4 (Bayley-Pinneau, Khamis-Roche). Before 2, only the mid-parental genetic target applies. The doubling rule briefly works at its checkpoint (girls 18 months, boys 24 months). After 12 in girls and 14 in boys, puberty timing widens the error and a bone-age reading becomes the only way to tighten the estimate. At 16 (girls) / 18 (boys) growth plates close and no prediction is needed — the height is the height.

Estimation is possible at any age, but confidence changes sharply: birth to 2 years carries the mid-parental ±3.5 in band only; the doubling checkpoint at 18/24 months gives a ±3 in rough cut; ages 4–12 are the peak accuracy window (±2 in); puberty adds timing variance; growth-plate closure ends the question.

Certainty only arrives at skeletal maturity — around 16 for girls and 18 for boys, when growth plates fuse. Before that, every method is a probability statement: even the clinical gold standard (bone-age X-ray + Bayley-Pinneau) quotes a 95% interval of about ±1 in, not an exact figure.

The 6–10 year window. Growth is linear and pre-pubertal, the Bayley-Pinneau multipliers are most stable (R² ≈ 0.90), and the Khamis-Roche blend is at its best (R² ≈ 0.93). Before 4 the attainment fraction moves too fast; after 12 puberty timing injects variance.

The four home-usable methods are: (1) Bayley-Pinneau — divide current height by the attained-fraction for age and sex; (2) Khamis-Roche — a regression blend of the child's height and weight with the mid-parental target; (3) mid-parental height — (mom + dad ± 5 in) / 2; (4) the doubling rule — height × 2 at 18 months for girls or 24 months for boys. A fifth, the bone-age X-ray + Bayley-Pinneau, is the clinical gold standard but needs a radiograph.

It is a simplified mid-parental calculator: it averages the parents' heights with a sex adjustment and projects from the child's current percentile. That makes it a genetic-target tool (R² ≈ 0.55 band) rather than a growth-curve tool, so expect its output to sit near the mid-parental figure and to differ from Khamis-Roche or Bayley-Pinneau results by 2–4 in on the same child. It is a reasonable first look, not a final answer.

Partially — it is the weakest of the four methods (R² ≈ 0.40, ±3 in). It works best exactly at the checkpoint (24 months for boys, 18 for girls) and fails for early or late maturers, for preemies still in catch-up growth, and for any child whose length was measured lying down (a ~0.75 in systematic error). Use it as a party trick, then confirm with the multiplier methods from age 4.

The old folk rule says 'double the height at age 3', but the attainment data disagrees: a 3-year-old boy has reached about 54% of adult height (×1.85, not ×2.0) and a girl about 58% (×1.72). Doubling at 3 therefore overestimates by roughly 8% for boys and 16% for girls. The correct formal approach is the age-specific Bayley-Pinneau fraction — at age 3 the multipliers are about 1.85 (boys) and 1.72 (girls).

Use corrected age: corrected age = chronological age − weeks of prematurity, applied until 24 months. A baby born at 32 weeks (8 weeks early) seen at a chronological 12 months has a corrected age of 10 months, so any age-based multiplier must be read from the 10-month row. Doubling rules are unreliable for preemies in catch-up growth — rely on the mid-parental target first and the multiplier methods once corrected age passes 2 years.

Five ordinary reasons: (1) different underlying formulas — BabyCenter is a simplified mid-parental tool, WebMD a simplified Khamis-Roche, Mayo Clinic-style tools use Bayley-style lookup tables; (2) publication-vintage differences in the multiplier tables (1952/1959 Bayley, 1994 Khamis-Roche, CDC refreshes); (3) sex-adjusted vs unadjusted mid-parental variants; (4) unit rounding at input; (5) lying length vs standing height around ages 2–3. A spread under 2 in between two defensible methods is noise, not error.

The textbook exercise (often numbered 2.1 in child-development coursework) usually applies the doubling rule or a linear percentile extrapolation. Those are teaching simplifications: the doubling rule carries R² ≈ 0.40 and percentile extrapolation assumes the channel never shifts. The published methods this guide compares — Bayley-Pinneau, Khamis-Roche, mid-parental — are the adult versions of the same idea, with measured error bars attached.

Yes — an age-matched child height predictor is exactly what the Bayley-Pinneau attainment table provides: for every age and sex, divide current height by the attained fraction of adult height. That fraction climbs from roughly 47% at age 2 to 84–88% by age 12, so the prediction tightens as the child grows. Any age from 2 to 16 can be read off the multiplier table above — that age-keyed lookup is the predictor in its published form.

Only as a rough sanity check, never as a forecast — so no, arms alone cannot determine a child's adult height with usable precision. In adults arm span approximately equals standing height (the span-minus-height difference is close to zero for most European-origin populations and slightly positive for many African and South Asian populations), but in children the wingspan lags stature until the pubertal spurt and only catches up near skeletal maturity, so multiplying a child's arm span by a fixed constant stacks age-dependent error on top of normal variation. Trying to predict a child's height by legs faces the mirror problem: the leg-length to sitting-height ratio shifts throughout childhood, and a long-legged early maturer can finish average once the torso catches up; a child height predictor based on leg length typically misses by ±3–4 in, clearly worse than the ±2 in Bayley-Pinneau estimate from total height and age. If body proportion itself looks unusual — very long limbs or unusually short legs relative to the trunk — that is a pediatrician's assessment, usually with a bone-age X-ray, rather than a limb-ratio shortcut.

You cannot choose your child's height, but you can influence where inside the genetic range they finish. Adult height is roughly 80% heritable, so the ceiling and the floor are set by the polygenic draw from both parents — no supplement, stretching device, hanging bar, or sport has ever been shown to add height beyond that genetic envelope, and growth-hormone treatment is reserved for diagnosed medical conditions rather than short-but-healthy children. What the remaining ~20% decides is whether the child reaches the top or the middle of their potential: adequate protein and calories through infancy and the preschool growth window (the periods most sensitive to nutritional stunting), deep sleep when growth hormone pulses most strongly, regular weight-bearing exercise, and prompt treatment of chronic illness. Realistically those factors are worth a few centimeters within the family's ±3.5 in target band; expecting them to produce a 6′4″ child from two average-height parents is not. Worry about the controllable inputs, and treat the genetic corridor as the boundary of the possible.

Menarche arrives relatively late in the pubertal sequence — after the fastest growth has already happened — but the first period does not end growth. After it begins, a girl typically gains another about 2–3 inches (roughly 5–7.5 cm), most of it in the first 12–18 months, and the growth plates usually fuse completely about 2–2.5 years after menarche, which is when final height is settled. Two consequences for a girl's height predictor around her period: estimates made before menarche stay on the standard chronological or bone-age fractions, while a prediction run after the first period should use the post-menarche attainment fraction (she is already at roughly 96–97% of adult height), which is why post-period home predictions carry the narrowest error bands. Early menstruators — first period before roughly age 11 — have less total growing time and often finish slightly shorter than the pre-period projection suggested; late menstruators tend to come in on the high side. When the remaining inches genuinely matter, a bone-age X-ray after menarche reads growth-plate status directly and beats any website estimate.

Weakly at birth, strongly from age 2–3 onward — and the distinction is the key fact in the child growth to adult height association. The correlation between a child's height at birth and adult height is only modest (roughly r ≈ 0.2–0.3), because newborn length reflects the prenatal environment — maternal nutrition, placental supply, and gestation length — at least as much as the child's own growth genetics; a long birth length does not reliably correlate with adult height, and catch-up plus catch-down growth across the first two years shuffles the ranking substantially. The association tightens once a percentile channel establishes itself: from about age 2–3, height percentile as a child correlates with adult height increasingly well (around r ≈ 0.7 and higher by school age), and a child who holds one channel for years is precisely the case the Bayley-Pinneau method is built for. The actionable signal is persistence rather than any single measurement — a stable 60th-percentile channel from age 3 to 10 is genuinely indicative of adult height near the 60th percentile, while crossing two channels up or down is what breaks the association and warrants a growth-chart review. That asymmetry — weak birth-length correlation, strong childhood-channel persistence — is why no serious method opens before age 2 and why the peak prediction window is ages 4–12.