Broad-Sense Heritability

Look at variation in a field and you cannot tell how much of it is the plants and how much is the ground they are standing on. Heritability is the partition: the share of observed variance attributable to genetic differences rather than environment. Broad-sense heritability counts all genetic variance — additive, dominance and epistatic together.

The number is a property of a population in an environment, not of a trait. The same trait can be highly heritable in a uniform glasshouse and barely heritable in a variable field, because the denominator changed. This is why heritability estimates from the literature transfer so poorly, and why they climb whenever a trial is better designed: reducing environmental variance raises heritability without altering a single gene.

Its use is triage. A heritability below about 0.2 says most of what you are looking at is noise, and selecting on individual plant performance will largely be selecting on soil. The standard responses are to replicate, to select on family means rather than individuals, and to test across more environments. Above 0.5, single-plant selection starts to work and early-generation selection becomes worthwhile.

Broad-sense is the right measure when the whole genotype is propagated intact — clonally propagated crops such as potato, sugarcane and fruit trees, and F₁ hybrids, where dominance effects are reproduced exactly in every plant sold. When parents are crossed and the offspring are the product, dominance and epistasis are broken up by recombination, and the narrow-sense measure is the honest one.