Personality & Individual Differences · Simplified
The Same Genes, A Different Heritability
Heritability is a fraction. This page leaves the top of the fraction alone and changes the bottom.
The one sentence to take away
Heritability is the proportion of the variation between people in a particular population, living in the range of environments that population happens to contain, that is associated with genetic variation. It is not the proportion of any individual's trait caused by their genes, no such proportion exists, and it says nothing about whether the trait can be changed.
Step 1 of 2
First, the definition
A study reports that a trait has a heritability of 0.50. What does that mean?
Step 2 of 2
One population, one dial
Below is a fictional population. The amount of genetic variation in it is fixed and does not change at any point on this page. Neither does the amount that families differ from one another. The only thing the slider changes is how much people's individual circumstances differ, the things that make two children in the same family turn out unlike each other.
Key terms
- Heritability
- The share of the variation in a trait, in one population at one time, associated with genetic variation. A property of that population, not of any person in it.
- Shared and non-shared environment
- The part of the environment that makes members of the same family alike, and the part that makes them different. The second is usually the larger of the two, and it is not upbringing in the ordinary sense.
- Falconer's formulae
- The arithmetic that turns two correlations into three variance components: double the gap between the identical and fraternal correlations for the genetic share, and so on.
The bar is all the variation there is between people in this population, in the units the slider uses. The genetic segment is the same length at every setting. Heritability is that segment as a share of the whole bar, so the bar getting longer is the only thing that has to happen for heritability to fall.
| Quantity | What the study sees | Estimated by the formulae | The truth in the model |
|---|
What this shows
A number about a population, at a time, in a range of environments
Key idea: Not one gene changed anywhere on this page. The amount of genetic variation stayed at exactly 1.00 from beginning to end, and the heritability of the trait ran from 0.19 to 0.67 all the same, because heritability is a ratio and the slider was moving the denominator. Three things follow, and they are the three things heritability is most often taken to mean. It cannot be the proportion of an individual's trait caused by their genes, because it changed without any individual changing. It cannot say whether a trait can be altered, because the whole demonstration consisted of altering the environment, and doing so raised heritability rather than lowering it. And it does not travel: a figure from one population, at one time, across the range of environments that population happened to contain, is a description of that and of nothing else.
The model here is the simple one the formulae assume, and it is simple in ways real data are not. There is no sampling error at all: the correlations shown are the ones the model implies exactly, whereas a real study estimates them from a finite number of pairs and reports them with a confidence interval, and heritability estimates from real twin studies are considerably less precise than the two decimal places here suggest. The formulae also rest on assumptions that this model satisfies and reality may not, chiefly that identical and fraternal pairs share their environments to the same degree, and that partners do not resemble each other on the trait; when those fail the estimates go wrong in specific directions. There is no gene-environment interaction and no correlation between genes and environment, both of which are real and both of which this decomposition simply assigns elsewhere. And one thing that does not need qualifying: none of this bears on anybody as an individual, and heritability was never a statement about a person.
The longer version lets you set all three variance components separately, generates fictional twin and sibling pairs with sampling error, and offers three assumption violations whose direction of bias you can watch. It is at Twin-Study Simulator in the main collection.