What heritability is
Heritability is one number that answers a precise question: of all the variation in a trait within a population, what fraction is due to genetic differences? It is a ratio of variances — the genetic slice divided by the whole pie. If a trait's heritability is 0.7, then 70% of the *differences among individuals* in that group trace back to genetic differences, and 30% to everything else.
Broad-sense vs narrow-sense
There are two flavours. Broad-sense heritability (H²) counts all genetic variance — including the part where alleles interact in non-additive ways — as a fraction of total variance: H² = V_G / V_P. Narrow-sense heritability (h²) counts only the additive genetic variance V_A — the part that passes predictably from parent to offspring: h² = V_A / V_P.
Partition the genetic variance further:
V_G = V_A + V_D + V_I
| | |
additive dominance interaction(epistasis)
Broad-sense: H^2 = V_G / V_P (all genetics counts)
Narrow-sense: h^2 = V_A / V_P (only ADDITIVE counts)
Worked example:
V_A = 30, V_D = 6, V_I = 4, V_E = 10
V_G = 30 + 6 + 4 = 40
V_P = 40 + 10 = 50
H^2 = 40/50 = 0.80
h^2 = 30/50 = 0.60 <- predicts resemblance & breeding responseWhy care about the difference? Because narrow-sense h² is the one that predicts the future. Only additive effects pass reliably to the next generation, so h² tells a breeder how much a population will respond to selection, and tells us how much offspring will resemble their parents. Broad-sense H² is a fuller accounting of genetics but a worse predictor of inheritance across generations.
Three traps
- High heritability does not mean unchangeable. Eyesight is highly heritable, yet glasses fix it instantly. Heritability says nothing about whether an *environmental* intervention can shift the trait.
- It is tied to a population and an environment. If everyone eats the same diet, food can't create differences, so V_E shrinks and heritability rises — without any gene changing. Move to a place with uneven nutrition and the same trait's heritability drops.
- It says nothing about group averages. A high heritability *within* two groups tells you nothing about why the groups *differ* on average — that gap can be entirely environmental. This is one of the most abused errors in all of genetics.