Heterosis – the Gap Between an Offspring and the Mean of Its Parents
Heterosis is a calculation, not a property. Take two parent plants, measure the same trait on both, average it, and compare that average with the offspring. If the offspring comes out above it, the difference is the heterosis. The most useful thing to know about that number is how little it explains on its own: it says nothing about where the advantage comes from, nor whether it can be repeated. It only becomes readable once you know which parents it was calculated against.
The case that shows it
In cannabis the calculation has been run both ways on the same material. Five crosses made from four closely mated lines were set against two different reference points: once against their immediate parents, once against the open-pollinated source accessions those lines had originally come from. Open-pollinated here means pollinated without control, the state before any breeding work.
Measured against the parents, all crosses but one came out larger, and seed yield ran between 3.9 and 155 % above the average of the two parent lines. Measured against the source accessions, only two of five exceeded both ancestors in total plant mass.
Same material, same plants, two results that look contradictory. They are not: the advantage over the parent lines largely consists of undoing damage the breeding itself created.
Where that damage comes from
Mating a line closely over several generations drives it towards true breeding: at more and more places in the genome the plant carries two identical rather than two different variants. That is the point of the exercise, because only then does a line pass its traits on unchanged. It comes at a price, since harmful variants end up in double copy too and then take effect instead of being masked by a healthy second copy. The line loses vigour, and that loss is called inbreeding depression.
Cross two such lines back together and the loss largely disappears. Exactly that recovery is baked into the calculation against the parent lines, and it is not a gain over the source material but a return to it.
The four components and what the distinction is for
Separating this cleanly is only possible at the level of whole populations, never on a single cross. Take two separated groups of a species, mate them at random, and hold the first daughter generation against the average of both groups: that gap is called panmictic mid-parent heterosis, panmictic being the technical word for this random mating. Mate the daughter generation at random again and half the effect remains.
Baseline heterosis is the share that recovers what was lost while both parent groups were closely mated – the undoing of inbreeding depression. Inbred mid-parent heterosis is the sum of both, and therefore the gap between the cross and the average of all closely mated parents. That is the quantity measured in the cannabis case above.
The common shorthand that heterosis is the opposite of inbreeding depression applies to only one of these four: baseline heterosis. The others can arise even where no single variant masks another, because it is enough that variants at different places in the genome act together. Inbreeding depression cannot arise that way, and the whole distinction rests on that difference.
When the sign flips
The gap can also fall on the other side. It is then called outbreeding depression, and it arises not from a missing variant but from matched combinations of variants being pulled apart in the cross. Such combinations have been selected together over a long time within one origin and fit one another; a distant cross takes them apart.
That gives a rule of thumb with a built-in limit: heterosis rises with the genetic distance between parents, but only up to the point where this loss takes over.
The sign can also follow from the chosen measurement. Positive heterosis for “days to flowering” and negative heterosis for “speed of development” are the same observation written down twice.
Not every trait, and not evenly
Within one and the same cross a trait may show heterosis and the next may not, and neither predicts the other. Crosses between the rice subspecies indica and japonica grow more vigorously and are less fertile at the same time. Maize crosses show more heterosis under poor conditions than under good ones, even though absolute performance is lower there.
Composite traits can show heterosis while their components sit at the parental average. Among the breeding pools of oil palm, one carries few heavy fruit bunches and the other many light ones; their crosses come out 25 % above the parental average for fruit output, while bunch number and bunch weight each sit exactly on it.
Two familiar reference points do not belong in this calculation. Better-parent heterosis measures against the stronger of the two parents, commercial heterosis against a market variety. Both are statements of performance rather than measures of heterosis, because their reference point is not the average of the parents.