Genetics
Polygenic Inheritance
Why height is a bell curve, not on/off
Polygenic inheritance is when a single trait is controlled by many genes, each adding a small effect, so the trait varies continuously rather than falling into a few discrete classes. Because the phenotype is the running sum of dozens to thousands of tiny, independently inherited contributions — plus environmental variation — the population distribution is smooth and approximately normal: a bell curve. Human height, skin pigmentation, body mass, and blood pressure are textbook polygenic traits. Each individual locus still obeys Mendel's laws; the continuous spectrum simply emerges when you add many of them together.
- Also calledQuantitative / continuous-trait inheritance
- Genes per traitDozens to 10,000+ loci
- Distribution shapeContinuous, ~normal (bell curve)
- Height heritability≈ 0.80 (80%)
- Per-variant effectOften ≈ 1–2 mm of height
- Core assumptionEffects are additive (they sum)
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
From on/off switches to dimmer dials
Gregor Mendel got famously lucky. The seven pea traits he studied — purple versus white flowers, round versus wrinkled seeds, tall versus dwarf plants — each turned out to be controlled by a single gene with two cleanly dominant/recessive alleles. Cross two heterozygotes and you get a tidy 3:1 ratio you can literally count. Those are monogenic (single-gene) traits, and they behave like binary switches: you are either one thing or the other.
But step outside the pea patch and most of the traits you actually care about don't behave that way. There is no "tall allele" and "short allele" for humans that sorts everyone into two height bins. Instead, adult height runs in a continuous gradient from about 140 cm to 210 cm, with most people bunched in the middle and progressively fewer at the extremes. Skin color, body weight, blood pressure, fasting glucose, milk yield in cattle, kernel count in corn — all of these vary smoothly. They are polygenic: each is built from many genes acting together, where every gene is less a switch than a dimmer dial nudging the trait a little up or down.
The conceptual leap is simple but powerful. If one gene gives you a few discrete classes, then n genes — each with an "increasing" and a "decreasing" allele — give you up to 2n + 1 classes when their effects simply add up. Three genes already produce seven classes in a roughly bell-shaped pattern; ten genes produce twenty-one classes so finely spaced that, once you smear them with environmental noise, the steps vanish into a smooth curve. That smooth curve is the signature of polygenic, additive inheritance, and it is the bridge that reconciled Mendel's discrete particles of heredity with the continuous variation Darwin's cousin Francis Galton was measuring at the same time.
The additive model, gene by gene
The classic teaching model is the easiest place to build intuition. Imagine a trait — say, grain color in wheat — controlled by three independent loci: A, B, and C. At each locus there is an "uppercase" allele that adds one unit of pigment and a "lowercase" allele that adds zero. An individual's color is just the count of uppercase alleles across all three loci, from 0 (aabbcc, pure white) to 6 (AABBCC, deep red).
Now cross two triple-heterozygotes (AaBbCc × AaBbCc). Each parent passes a random uppercase or lowercase allele at each locus, so the offspring's pigment dose is the sum of six independent coin flips. The arithmetic gives a binomial distribution: only 1 in 64 offspring is pure white (0 uppercase) and only 1 in 64 is deepest red (6 uppercase), while 20 in 64 sit right in the middle with three uppercase alleles. Plot those frequencies and you get a discrete bell. There are still exact Mendelian ratios hiding inside — each locus segregates 1:2:1 on its own — but the visible result is a graded spectrum, not categories.
Three things turn that stepwise binomial into the smooth curves we measure in real populations:
- More loci. Height isn't three genes — a 2022 genome-wide association study of more than 5 million people mapped roughly 12,000 common variants contributing to it. Each adds only about a millimeter or two. With thousands of tiny steps, the staircase becomes a ramp.
- The Central Limit Theorem. Mathematically, the sum of many independent random contributions converges on a normal (Gaussian) distribution regardless of what each individual contribution looks like. Polygenic traits are exactly such sums, which is why the bell curve appears so reliably.
- Environmental variance. Nutrition, disease, hormones, and developmental noise shift each genotype's outcome up or down a little. This blurs the remaining discrete steps into one continuous distribution and is the reason genetically identical twins still differ in height by a centimeter or two.
Heritability: splitting nature from environment
Because polygenic traits mix genes and environment, geneticists need a number that says how much of the population's variation is genetic. That number is heritability (h²) — the proportion of total phenotypic variance attributable to genetic variance in a given population and environment. It runs from 0 (all environmental) to 1 (all genetic).
Heritability is widely misunderstood, so two cautions are worth stating plainly. First, it is a property of a population, not an individual — saying height is 80% heritable does not mean 80% of your particular height came from genes. Second, it depends on the environment sampled. If everyone is fed identically, environmental variance shrinks and heritability rises; if environments vary wildly, the same genes yield a lower heritability. Heritability measures the relative spread of causes in a specific setting, not a fixed biological constant.
| Trait | Approx. heritability (h²) | Notes |
|---|---|---|
| Adult height | 0.80 | ~12,000 mapped variants; among the most heritable common traits |
| Body mass index (BMI) | 0.40–0.70 | Strong gene × food-environment interaction |
| Type 2 diabetes risk | 0.30–0.70 | Polygenic risk plus large lifestyle component |
| Systolic blood pressure | 0.30–0.50 | Hundreds of contributing loci |
| Skin pigmentation | ~0.80 | Largely driven by a handful of large-effect loci plus a polygenic tail |
| Schizophrenia liability | 0.60–0.80 | Thousands of small-effect common variants |
Polygenic vs. Mendelian: a side-by-side
The cleanest way to lock in the concept is to contrast it directly with the single-gene model. Notice that polygenic inheritance does not break Mendel's laws — every contributing locus still segregates and assorts exactly as Mendel described. The difference is purely in how many loci stack up and whether their effects are visible individually.
| Feature | Monogenic / Mendelian | Polygenic / quantitative |
|---|---|---|
| Genes per trait | One | Many (dozens to thousands) |
| Phenotype classes | Few, discrete (e.g., 3:1) | Continuous spectrum |
| Population distribution | Distinct peaks / categories | Smooth bell curve (~normal) |
| Effect of one allele | Large, often visible alone | Tiny — a few mm or % of trait |
| Role of environment | Usually minor | Often substantial |
| How it's studied | Pedigrees, simple crosses | Twin studies, GWAS, QTL mapping, heritability |
| Classic example | Cystic fibrosis; pea flower color | Height; skin color; blood pressure |
One subtlety: the boundary is a spectrum, not a wall. Skin pigmentation, for instance, is technically polygenic but is dominated by a few large-effect loci (such as SLC24A5) sitting on top of a long tail of small-effect variants. And some "Mendelian" diseases have polygenic modifiers that explain why two people with the same disease-causing mutation can have very different severity. Real genetics lives on the continuum between the two extremes.
Finding the genes: QTL mapping, GWAS, and polygenic scores
If a trait is built from thousands of nearly invisible nudges, how do you ever find the responsible genes? The answer is statistics at scale.
Quantitative trait loci (QTL) mapping was the first tool: cross strains that differ in a trait, then look for genetic markers whose inheritance correlates with the trait value in the offspring. Each correlated marker flags a chromosomal region — a QTL — that harbors one or more contributing genes. QTL mapping dominated crop and livestock genetics for decades and still underpins modern plant breeding.
Genome-wide association studies (GWAS) scaled the idea to whole populations. Genotype millions of people at millions of common variants, then test each variant for a statistical association with the trait. Because individual effects are minuscule, GWAS need enormous samples — the headline height study reached over 5 million participants to resolve its ~12,000 variants. The output is a catalog of variants and their effect sizes.
Summing those effects gives a polygenic score (often called a polygenic risk score for diseases): a single number estimating an individual's genetic predisposition by adding up their trait-increasing alleles, each weighted by its GWAS effect size. Polygenic scores are reshaping medicine — flagging people in the top few percent of genetic risk for coronary artery disease, breast cancer, or type 2 diabetes years before symptoms. But they carry hard limits: they are probabilistic shifts in odds, not verdicts; they explain only part of the heritability ("missing heritability" remains an open problem); and their accuracy drops sharply when applied to ancestries underrepresented in the original GWAS, an equity problem the field is still working to fix.
Why it matters
- Medicine. Most common diseases — heart disease, diabetes, hypertension, major psychiatric disorders — are polygenic. Polygenic risk scores enable earlier, stratified screening.
- Agriculture. Yield, drought tolerance, and growth rate are quantitative traits; genomic selection on polygenic scores now drives crop and livestock improvement faster than phenotype-only breeding.
- Evolution. Polygenic adaptation lets populations shift gradually by changing allele frequencies at many loci at once, rather than waiting for one large-effect mutation.
- Resolving a historic debate. The Fisher 1918 "infinitesimal model" showed that many small Mendelian genes produce continuous variation — reconciling Mendelians and biometricians and founding quantitative genetics.
- Understanding human variation. It explains why traits like intelligence-test scores or height are continuous, familial, and yet have no single "gene for" them.
Common misconceptions
- "Polygenic traits ignore Mendel's laws." Each locus still segregates 1:2:1; the bell curve is just the sum of many such loci.
- "High heritability means environment doesn't matter." Heritability is population- and environment-specific; a highly heritable trait can still be very responsive to changed conditions.
- "A polygenic score predicts my outcome." It shifts probabilities across a group; it does not determine any individual's fate.
- "More genes means more dominant alleles win." The classic model is additive — alleles sum; dominance and epistasis are layered on top, not the foundation.
- "Polygenic equals pleiotropy." Polygenic = many genes → one trait. Pleiotropy = one gene → many traits. Opposite directions.
Frequently asked questions
What is polygenic inheritance?
Polygenic inheritance is when a single trait is controlled by many genes at different loci, each adding a small effect. Instead of a few discrete categories (tall vs short), the trait varies continuously across a smooth range. Because the phenotype is the sum of many small independent contributions plus environment, the population distribution is approximately a bell curve. Height, skin color, body weight, and blood pressure are classic polygenic traits.
Why does polygenic inheritance produce a bell curve?
Each locus is inherited independently and adds roughly the same small increment. Getting an extreme phenotype requires inheriting many "increasing" alleles at once, which is rare; getting an intermediate value can happen many different ways, which is common. Summing many independent random contributions converges on a normal distribution by the Central Limit Theorem. Add environmental variation and the peak smooths further — so most individuals cluster near the average and few sit at the extremes.
How is polygenic inheritance different from Mendelian inheritance?
Mendelian (monogenic) traits are controlled by one gene and fall into a few discrete classes with clean ratios — like pea flower color (purple vs white) or cystic fibrosis (affected vs carrier vs unaffected). Polygenic traits are controlled by many genes and vary continuously, so you can't sort people into a handful of categories. Each individual polygenic locus still obeys Mendel's laws; the continuous distribution simply emerges from adding many of them together.
How many genes affect human height?
A 2022 GWAS of over 5 million people mapped roughly 12,000 common genetic variants that together explain most of the heritable variation in adult height. Height is about 80% heritable, yet each individual variant changes height by only about a millimeter or two. No single "height gene" exists — the trait is the cumulative sum of thousands of tiny effects spread across nearly every chromosome.
What is a polygenic score?
A polygenic score (or polygenic risk score) sums up a person's trait-increasing alleles across thousands or millions of variants, each weighted by its measured effect size. It produces a single number estimating genetic predisposition for a trait or disease — height, coronary artery disease, type 2 diabetes, schizophrenia. Scores are probabilistic, not deterministic: they shift the odds across a population but cannot tell any one individual their exact outcome, and accuracy drops sharply across ancestries not represented in the training data.
Is polygenic the same as pleiotropy or epistasis?
No, and they're easy to confuse. Polygenic means many genes affect one trait. Pleiotropy is the reverse: one gene affects many traits. Epistasis is when genes interact non-additively, so one locus masks or modifies another. Classic polygenic models assume additivity (effects simply sum), but real traits include some epistasis and gene-environment interaction layered on top of the dominant additive signal.