Heritability

Heritability is the proportion of phenotypic variance in a population that is attributable to genetic variance. It is not — and this cannot be emphasized strongly enough — a measure of how "genetic" an individual trait is, nor a statement about the immutability of a phenotype, nor a ratio of nature to nurture in any organism's development. It is a population statistic, invented by Ronald Fisher in 1918 as part of the mathematical synthesis of Mendelian genetics and biometric statistics, and its continued misuse in scientific and public discourse is one of the most persistent epistemological failures in modern biology.

The Mathematical Definition

In its simplest form, the heritability \(h^2\) of a trait is defined as:

\[h^2 = \frac{V_G}{V_P}\]

where \(V_G\) is the genetic variance in the population and \(V_P\) is the total phenotypic variance. The denominator \(V_P\) can be decomposed further:

\[V_P = V_G + V_E + V_{GE} + V_e\]

where \(V_E\) is environmental variance, \(V_{GE}\) is gene-environment interaction variance, and \(V_e\) is measurement error. Heritability thus depends on the relative magnitudes of all variance components in a particular population under particular environmental conditions. It is not a property of the trait. It is a property of the population-environment system.

This is why heritability estimates for the same trait can vary dramatically across populations. Height heritability is near 0.9 in affluent Western populations where nutrition is uniform and abundant, but much lower in populations where nutritional variation is a major source of height differences. The trait — height — has not changed. The environment has, and with it, the variance partitioning.

What Heritability Does and Does Not Mean

Heritability answers a specific question: "Given the range of genetic and environmental variation present in this population, what fraction of the observed differences between individuals is associated with genetic differences?" It does not answer, and cannot answer, the following questions:

• Can this trait be changed? High heritability does not imply immutability. Eyeglasses correct highly heritable myopia. Vaccines prevent highly heritable susceptibility to infectious disease.
• Are group differences genetic? Heritability within a population says nothing about the causes of differences between populations. If Group A and Group B experience different environments, the genetic variance within each group is irrelevant to explaining the between-group gap.
• How much does genetics matter for this individual? Heritability is a population parameter. It cannot be applied to individuals any more than the mean height of a population tells you the height of a specific person.

The persistence of these confusions — despite a century of clarification by quantitative geneticists — suggests that heritability is not merely a difficult concept. It is a concept that serves ideological functions. The Nature vs Nurture debate, which heritability was supposed to resolve mathematically, instead became a venue for laundering political commitments through statistical language.

Heritability and the Architecture of Causation

The deeper problem with heritability is that it presupposes a linear, additive decomposition of causes that rarely matches biological reality. Gene-Environment Interaction means that the effect of a genetic variant depends on environmental context. Gene-environment correlation means that genotypes are non-randomly distributed across environments. Epigenetic modification means that environmental exposures can alter gene expression across generations. Each of these phenomena violates the assumptions of the classical heritability model, yet each is now known to be widespread.

The heritability framework treats these complexities as statistical nuisances — terms to be estimated and controlled — rather than as signals that the underlying causal architecture is not decomposable into independent genetic and environmental components. When model selection in genetics consistently favors models with large interaction terms, the appropriate response is not to report heritability "corrected" for interaction. The appropriate response is to abandon the variance-partitioning framework for one that captures the dynamic, contingent, and feedback-laden nature of development.

The Social Life of a Statistic

Heritability estimates have a peculiar social trajectory. They are produced in the technical literature with extensive caveats, then migrate into textbooks with fewer caveats, then into popular science with the caveats removed entirely, and finally into policy discourse as settled facts about human nature. At each stage, the statistical uncertainty accumulates interpretive certainty.

The high heritability of intelligence (estimates of 0.5–0.8 in adult populations) is the most politically charged case. The data support the claim that genetic differences explain substantial variance in psychometric test performance within studied populations. They do not support the claims routinely attached to this finding: that intelligence is fixed, that educational interventions are futile, that group differences are genetic, or that social inequality reflects natural inequality. Each of these inferences is an additional step requiring additional evidence, and each has been contested on empirical grounds.

The correct interpretation of intelligence heritability is narrower and more interesting than the political deployments suggest. What the heritability estimate actually measures is the predictability of adult test performance from genetic similarity, given the current distribution of environments. In a world with radically different educational or nutritional distributions, the same genetic architecture would produce different heritabilities — and possibly different phenotypic outcomes.

The concept of heritability survives not because it captures biological reality but because it is computable. It is a variance component in an ANOVA table, and ANOVA tables are what we know how to make. The deeper developmental reality — genes and environments as inseparable processes operating through time, not as independent variables to be partitioned — is harder to model and impossible to reduce to a single number. The persistence of heritability as the dominant framework for thinking about genetic influence is therefore not a scientific achievement. It is a methodological inertia that protects simplicity at the cost of truth.