Heritability

The concept of heritability comes to us from population genetics. The idea was developed but not named by the geneticists Sewall Wright and Ronald Fisher. Both were theorists and applied their ideas to both agricultural and human psychological traits (i.e., intelligence). Wright developed the method of path analysis, which is now widely used in behavior genetics to compute heritability coefficients. These methods are very flexible and have been modified to incorporate modern molecular genetic findings. This entry outlines the purpose of the heritability formula, describes its various components, provides some examples of how the heritability of a trait is computed, and discusses some of the limitations of the concept (what it does not tell us).

The heritability coefficient indexes genetic influence on a quantitative trait (i.e., height, weight, intelligence) for a specific population. It tells us how much of the variance (a measure of the extent of differences between people) in a population is due to genetic factors. It is a population statistic (describes a population, just as the mean of a trait is descriptive of a population) and does not apply directly to individuals. It is useful to distinguish between “broad heritability” and “narrow heritability.”

Broad heritability tells us the proportion of total phenotypic (measurable) variance (σ2P) in a trait that is due to all genetic influences (σ2G). It can vary from sample to sample, environment to environment, and age to age. Thus the formula:

This formula should be read as broad heritability = all genetic variance/phenotypic variance. The term h2 is the symbol for heritability and is not to be interpreted to mean the value is to be squared.

We know from Mendelian theory that genes act in a number of ways. Many act additively: The more genes positive for the trait the more the trait expresses itself (height is a good example). The idea is that these additive genes, each of small effect, sum up to influence a trait. In addition, some genes dominate other genes (Mendelian dominance). This is a form of interaction (nonadditivity) between two genes at the same place (locus) on the paired human chromosomes. Another form of nonadditivity is epistasis, the interaction between genes at different loci on the chromosomes. Additive genes create similarity between relatives in a straightforward manner. Nonadditive effects work in a much more complex way and their effects are difficult to estimate with precision in human studies.

Narrow heritability includes only the proportion of variance because of genes that act in an additive manner. This is called the additive genetic variance (σ2A). Thus the formula:

As in the previous example, this formula should be read as narrow heritability = additive genetic variance/phenotypic variance. Again the term h2 is the symbol for heritability and is not to be interpreted to mean the value is to be squared.

Environmental sources of variance are, of course, conceptually important. One such source is called shared environmental variance or common [Page 390]environmental variance (σ2C). This environmental influence is what has been thought to make twins and siblings similar (similar child rearing, same home, parents with the same educational background, etc.). For a long time psychologists thought that common environment was an extremely important source of twin and sibling similarity, but behavior genetic studies have shown that it is of minor importance for many traits measured in adulthood. For a number of traits (i.e., intelligence and social attitudes), shared environment is very important in the early years but much less so in adulthood. We can estimate the importance of shared environment in a number of ways, but two are particularly simple. Both use correlation coefficients. The correlation coefficient used in twin and adoption studies can run from zero to one and can, in the example below, be interpreted as a percentage of variance. In most other contexts, a correlation coefficient must be squared to interpret the results as a percentage. This point is widely misunderstood. Consider monozygotic twins reared together (MZT), often incorrectly called identical twins. They have all their genes in common. Consequently, the correlation between them on a trait such as intelligence reflects all the effects of genes (additive and nonadditive) plus the similarity because of being reared in the same family (shared environment). Monozygotic twins reared apart (MZA), if they were placed in adoptive homes randomly, would be similar only because they had all their genes in common (additive and nonadditive effects). Thus the MZA correlation would directly estimate the broad heritability. The difference between the two correlations (MZT – MZA) would estimate shared environmental influence. The IQ correlation between MZT twins, in adulthood, is about .86. The IQ correlation for MZA twins (almost all adults) is .74 (weighted mean of five studies). The difference suggests a shared environmental influence of .12. The second method of estimating shared environmental influence is to examine the similarity of unrelated individuals reared in the same home (URT). For childhood data the URT correlation is .26 (weighted mean of 14 studies). In adulthood the URT correlation is .04 (weighted mean of five studies). This method suggests a shared environmental influence of .04. Other methods yield estimates between these two figures and also show a strong age effect, with genetic influence increasing with age.

...