Heritability and its estimation

Concept and estimation

Quantitative genetics partitions phenotypic variance as V(P) = V(G) + V(E) + interactions. Broad-sense heritability, H² = V(G)/V(P), includes additive, dominance and epistatic variance. Narrow-sense heritability, h² = V(A)/V(P), retains additive variance and best predicts response to selection; R. A. Fisher supplied the variance framework.

Estimation uses resemblance among relatives:

• parent–offspring regression and sibling correlations estimate additive familial resemblance;
• classical twin studies compare monozygotic with dizygotic twins, often using h² ≈ 2(rMZ − rDZ) under restrictive assumptions;
• adoption studies separate resemblance to biological and rearing relatives;
• pedigrees and modern genomic relatedness methods estimate variance components in large samples.

Height is highly heritable in many well-nourished populations, yet mean stature can rise rapidly with nutrition. IQ estimates similarly vary by age, class and environment. Twin estimates assume comparable environments and may be inflated by gene–environment correlation, assortative mating or shared prenatal conditions. Population stratification can also distort genomic estimates.

Thus heritability is local to population, age, environment and measurement. It says nothing directly about between-population causes.