Heritability

Based on Wikipedia: Heritability

Here's a number that has caused more confusion than almost any other in modern science: 0.6. That's roughly the heritability of personality traits. And if you're like most people, you probably think it means sixty percent of your personality comes from your genes and forty percent from your environment.

You'd be wrong. And understanding why you're wrong turns out to be surprisingly important.

What Heritability Actually Measures

Heritability is a statistical concept that asks a deceptively simple question: In a particular population, how much of the variation we see in a trait is associated with genetic differences between individuals?

Notice those italicized words. They're doing a lot of work.

Heritability doesn't tell you how much of your height, intelligence, or tendency toward anxiety was determined by your genes. It tells you something different: if we gathered everyone in a population and measured how much they differ from each other on some trait, what fraction of that difference correlates with genetic variation?

This distinction matters enormously. Consider height. The heritability of height in Western populations is about 0.8, meaning eighty percent of the variation in height among people correlates with genetic differences. But this doesn't mean eighty percent of your height comes from genes. If it did, a person who is six feet tall would have about fourteen inches of "environmental height" tacked on—which makes no sense at all.

Your actual height is one hundred percent dependent on both your genes and your environment. You need both. Without genes, there's no blueprint. Without food, water, and a functioning metabolism, that blueprint never gets built.

The Population Problem

Here's where things get even more counterintuitive. Heritability can change without any genetic change occurring whatsoever.

Imagine two scenarios. In the first, everyone in a country has access to excellent nutrition, healthcare, and education. Environmental differences between people are minimal. In this world, most of the variation you see in traits like height or cognitive ability will track genetic differences, simply because there's not much environmental variation to explain anything. Heritability will be high.

Now imagine the same population, same genes, but with massive inequality. Some people are malnourished, others well-fed. Some receive excellent education, others none. Now environmental differences explain much more of the variation between people. Heritability drops.

The genes haven't changed. The heritability has.

This is why heritability estimates are always specific to a particular population in a particular environment at a particular time. A heritability estimate from rural Guatemala cannot be applied to suburban Denmark. One from the 1950s may not hold today.

The Twin Studies Revolution

So how do scientists actually estimate heritability? The most famous method involves twins.

Identical twins share essentially all of their DNA. Fraternal twins share about half, just like regular siblings. If a trait is highly heritable, identical twins should be much more similar to each other than fraternal twins, even when raised in the same household.

The ultimate test comes from identical twins separated at birth and raised in different families. These individuals have identical genotypes but different environments. By comparing how similar they remain despite their different upbringings, researchers can tease apart genetic and environmental contributions to variation.

Such studies have produced remarkable findings. Identical twins raised apart often share eerily specific habits, preferences, and quirks—far more than chance would predict. But these studies also have limitations. Separated twins still shared a prenatal environment for nine months. And "different households" doesn't necessarily mean different socioeconomic conditions or cultures; adoption agencies often place children in similar environments.

Narrow Sense Versus Broad Sense

Scientists distinguish between two types of heritability, and the difference matters for anyone interested in breeding animals, growing crops, or understanding human traits.

Broad-sense heritability captures all genetic contributions to variation. This includes not just the straightforward additive effects of individual genes, but also dominance effects (where one version of a gene masks another) and epistatic effects (where genes interact with each other in complex ways).

Narrow-sense heritability is more restrictive. It only counts the additive effects—the portion of genetic variation where you can essentially add up the contributions of individual gene variants to predict the trait.

Why does this matter? Because narrow-sense heritability is what responds to selection. When breeders choose the tallest plants or the fastest horses to reproduce, they're selecting based on the offspring's expected trait value. That expectation depends on the additive effects passed from parent to child. A parent passes exactly one copy of each gene to each offspring, so only the average effect of those individual gene variants—the additive component—predictably carries forward.

This is why narrow-sense heritability tells you how quickly a population will respond to selective breeding. High narrow-sense heritability means selecting for a trait will produce substantial change in the next generation. Low narrow-sense heritability means selection will be frustratingly slow, even if broad-sense heritability is high.

The Missing Heritability Problem

In the early days of genome-wide association studies—massive projects that scan hundreds of thousands of genetic variants across thousands of people—researchers expected to find the specific genes responsible for heritable traits. Twin studies had shown that traits like height and intelligence were substantially heritable. The genes had to be somewhere.

They found some. But not nearly enough.

For height, which has a heritability around 0.8, early genome-wide association studies could only account for about five percent of the variation by identifying specific genetic variants. The other seventy-five percent became known as "missing heritability."

Where did it go? Several explanations have been proposed. Perhaps many thousands of genetic variants each contribute tiny effects, too small to detect individually but adding up collectively. Perhaps rare variants, different in each family, play important roles but don't show up in population-wide studies. Perhaps gene-gene interactions and gene-environment interactions are more important than the additive models assume.

This last possibility is particularly interesting because most heritability estimates assume additivity—that genetic effects simply add up without complex interactions. If this assumption is wrong, the traditional heritability estimates themselves might be misleading.

Canalization and Plasticity

Genes don't operate in a vacuum. They interact with environments in ways that can make heritability estimates slippery.

Some traits are canalized, meaning the genes ensure the trait develops the same way across a wide range of environments. Humans develop five fingers on each hand with remarkable consistency, whether they grow up in tropical rainforests or arctic tundra, whether well-nourished or malnourished. The genetic program is robust against environmental variation. For such traits, heritability might be low simply because there's little variation to explain—almost everyone has the same outcome.

Other traits show phenotypic plasticity—the same genes produce different outcomes depending on the environment. Water fleas develop defensive spines when chemical signals indicate predators are nearby, but remain spine-free in predator-free water. Same genes, different environments, different bodies. For highly plastic traits, heritability depends enormously on which environments are present in the population being studied.

The Behavioral Genetics Controversy

Nowhere has heritability been more contentious than in studies of human behavior and cognition.

Research consistently finds substantial heritability for traits like intelligence, personality dimensions, and susceptibility to mental disorders. These findings have been replicated across many populations and using multiple methods. The science is solid.

But the interpretation is where people go astray.

A high heritability for intelligence does not mean that education is pointless. It doesn't mean that group differences in test scores are genetic. It doesn't mean individuals are destined by their DNA. Remember: heritability describes variation within a population under current conditions. Change those conditions, and heritability changes too.

The classic analogy involves plants. Take genetically identical seeds and grow some in nutrient-rich soil, others in nutrient-poor soil. The plants in poor soil will be stunted. Within each group, all variation is environmental (since the genetics are identical, heritability within each group is zero). But the difference between groups is entirely environmental too—give the stunted plants better soil, and they'll grow tall.

High heritability within groups tells you nothing about what causes differences between groups. This logical error has been the source of enormous mischief in the history of behavioral genetics.

The Two Schools

The mathematical foundations of heritability come from two different traditions that developed in the early twentieth century.

One school traces to Sewall Wright at the University of Chicago, who developed path analysis—a method for decomposing correlations between relatives into genetic and environmental components. If parents and offspring correlate on a trait, how much of that correlation flows through shared genes versus shared environment? Wright's approach was later extended by geneticists at Iowa State University and elsewhere.

The other school began with the legendary statistician Ronald Fisher, who developed the analysis of variance—a technique for partitioning variation into different sources. This approach was elaborated at the University of Edinburgh and North Carolina State University, focusing on agricultural applications like livestock breeding.

Today, these approaches have largely merged into sophisticated statistical methods that can estimate heritability from complex pedigrees, from comparisons of twins and siblings, or even from patterns of genetic similarity measured directly from DNA. Each method has assumptions, each has limitations, but they generally converge on similar answers for well-studied traits.

What High Heritability Doesn't Mean

Let's be explicit about the misconceptions, because they're so common and so consequential.

High heritability does not mean a trait is unchangeable. Height is highly heritable, yet average height has increased dramatically over the past century due to improved nutrition. Genes set possibilities; environments determine which possibilities are realized.

High heritability does not mean a trait is unaffected by environment. A trait can be highly heritable and yet completely dependent on environmental conditions for its expression. Phenylketonuria, a metabolic disorder, is entirely genetic—heritability of one hundred percent. But its devastating effects can be prevented entirely by an environmental intervention: a diet low in the amino acid phenylalanine.

High heritability does not mean the trait is "genetic" in any simple sense. It's a statement about variation in a population, not about the causal architecture of development. A trait could require elaborate environmental inputs to develop and still show high heritability if everyone in the population receives similar inputs.

High heritability in one population does not imply high heritability in another. Change the environmental variation, and heritability changes.

The Threshold Model

Not all traits are continuous like height. Some are binary: you either have the condition or you don't. You either develop schizophrenia or remain healthy. You have the extra toe or you don't.

For such traits, geneticists use the liability threshold model. The idea is that an underlying continuous variable—liability—is influenced by many genetic and environmental factors. When liability exceeds a certain threshold, the discrete trait appears.

This model allows researchers to estimate heritability for conditions that don't vary continuously. If identical twins are more likely than fraternal twins to share a diagnosis, we can work backward to estimate how heritable the underlying liability is. This approach has been applied to conditions ranging from autism to diabetes to addiction.

Modern Genomic Approaches

The explosion of genetic data has transformed heritability research. Instead of relying on twins and family structures, researchers can now estimate genetic similarity directly from DNA sequences.

One approach compares unrelated individuals who happen to be slightly more or less genetically similar by chance. Even among strangers, some pairs share more genetic variants than others. If these slightly-more-similar pairs also tend to be more similar on a trait, that provides evidence for heritability.

These genomic methods have largely confirmed the heritability estimates from twin studies while also revealing that the genetic effects are spread across vast numbers of variants, each contributing a tiny amount. For most complex traits, there is no single gene "for" the trait—instead, thousands of genetic differences each nudge the outcome slightly in one direction or another.

Why It Matters

Understanding heritability correctly matters for science, for medicine, and for society.

In medicine, knowing that a condition is highly heritable doesn't tell doctors to give up on treatment—it tells them to look for biological mechanisms that might be targeted with drugs or other interventions. The heritability of Type 1 diabetes is quite high, but that hasn't stopped researchers from developing insulin therapy and working toward cures.

In education, knowing that academic achievement is partly heritable doesn't mean some children are destined to fail. It suggests that different children may need different approaches to reach their potential. One-size-fits-all education is a bad fit for a population with real variation.

In public policy, heritability estimates are often misused to argue that interventions are pointless. But this gets the logic backward. High heritability in an unequal society might actually indicate that equalizing environments could make a substantial difference. If environmental variation were reduced, more of the remaining variation would be genetic—heritability would rise—but everyone might be better off.

The number itself is neutral. What we do with it depends on understanding what it actually means.