What can twin studies tell us about heritability?
A short primer on heritability and twin studies
According to Eric Turkheimer, the three laws of behavioural genetics are:
All traits are partly heritable.
The effect of genes on a trait outweighs that of family.
There is substantial variation in traits not accounted for by either genes or families1.
When we say, for the sake of argument, that IQ is 60% heritable, we do NOT mean that you inherit 60% of your score. Rather, heritability is the proportion of the population variance that correlates with genes. Take a large population that differs a lot in intelligence: some members possess genes that are linked to them being smart, some have genes associated with being dumb. Hence, heritability is the degree to which differences in traits across an entire population can be explained by genetics.
Also note, as Chomsky famously articulated, that correlation does not imply causation. This applies to genes too. For instance, in the era of slavery, the genes that cause an individual to be black of course predicted their chances of being a slave, yet it was not genes that caused slavery. In other words, heritability is not genetic determinism, although this should be self-evident from the definition of heritability I provided. If the environment changes, heritability estimates may also change. Epigenetics, whether a gene is expressed or not, is also influenced by environment. Therefore, for estimates of heritability to be reliable and informative, one must hold environment constant.
One way of estimating heritability is via twin studies, which operate according to two key methods:
Take a pair of identical twins separated at birth, whom share their genome and over half of their traits. The correlation between their traits yields an estimate of heritability.
Alternatively, measure the correlation in traits for identical twins raised together, and subtract that correlation for fraternal twins raised together. Fraternal twins share only half of the part of the genome that varies in the population, so by subtracting out the proportion of variation that can be attributed to environmental effects, you get the variation that is explained by differences in genes.
One gaping limitation in this approach however is that, crucially, the interaction between many genes in a genome predicts traits, not just one gene. In other words, traits are polygenic. You must impose assumptions over the functional form of these interaction effects. For example, is interaction between genes additive or multiplicative? Twin studies generally assume interaction is additive. Yet when this assumption is false, there is substantial mismeasurement. As identical twins share their entire genome, this captures all genetic interactions, so this is not an issue for the first method. Yet this is not the case for fraternal twins. As a result, this will tend to overestimate heritability under the second method, as the correlation in traits for fraternal twins conditional on a shared environment will be underestimated. Obviously, the superficially simple solution would be to forego the second method altogether, yet small sample sizes will limit the predictive power of the first method. There may simply not be enough identical twins separated at birth out there.
Whilst we are on the subject of interaction effects, we must account for assortative matching, which means that genes and environments are not independent. For instance, if I yield a genetic propensity towards extroversion, then I will likely seek out and associate with other extroverted peers, which increases my extroversion. Again, this inflates estimates of heritability. If interaction effects between genes and environments are indeed multiplicative when you assume additive, then part of the variance will be falsely measured as heritability. Likewise, people mate with traits more similar to them, so this will bias the genetic similarity of siblings upwards.
Another problem facing twin studies is that they assume that environmental effects are exogenous to whether the twins are identical or fraternal. In reality, identical twins are treated more similarly. This will therefore overstate heritability.
As such, twin studies tend to overestimate heritability. There are other methods such as GWAS, that often produce lower heritability estimates, which may be a subject for future posts. Nonetheless, heritability is real, even if its effects are exaggerated. Yes, genes do not determine our destiny, yet genes clearly matter for the sort of people we become, and hence our outcomes in life.
So peer effects explain much of the environmental effects on traits, yet families yield some influence over their children's environment, and assortative matching increases the link between genes and environment.