Twin studies are studies conducted on human twins, usually involving identical (monozygotic, or MZ) twins in some way. These studies typically aim to infer how much of the variation in a human trait is due to genetic factors and how much is due to the environment. Most commonly, these studies use the classical twin method in which similarity (correlation) is calculated between identical twins for a given trait, and compared to the similarity between fraternal (dizygotic, or DZ) twins for the same trait. These two correlation coefficients are often then entered into Falconer's formula to generate a heritability estimate. (Falconer's formula states that the narrow-sense heritability, which is the percent of variation due to additive genetic factors, is equal to double the difference between the identical and fraternal twin correlations.) Falconer's formula is now regarded as having flawed assumptions. But there are other types of twin studies as well, notably the design involving identical twins reared apart (the idea is that they should have the same genes, but be raised in different environments). Perhaps the most famous of these is the "landmark" Minnesota Study of Twins Reared Apart that began in 1979. Of course, there are problems with these types of studies as well, including that "reared apart" is not necessarily the same as "reared in significantly different environments".
Limitations of twin studies
One fundamental limitation of twin studies is that, being observational studies, they cannot account for differences which do not exist (or do not exist in significant numbers of people) in the population, or in the sample of the population in the study. This includes both differences in effects and differences in causes (both environmental and genetic). Thus, even if a heritability estimate from a twin study indicates that a property is extremely heritable, it would be incorrect to conclude that no possible environmental intervention could affect it very much. We know that deliberate interventions like administering human growth hormone to short children can affect their height, but in a very poor population without much access to modern medicine, this environmental intervention might not be present and thus would not "show up" in a heritability estimate from a twin study performed on that population. Moreover, just because a characteristic is highly heritable now, it might not be in 100 years time, if a new medical intervention or technology is invented that can affect that characteristic, and that medical intervention or technology becomes widely used.
Also, on the other side of the nature versus nurture debate, it would clearly incorrect to take a heritability estimate from a twin study done on an isolated tribe as evidence supporting the (incorrect) belief that the ability to digest milk is highly non-heritable in all human beings, because it could just be that the gene for lactose tolerance is absent, or exists only in very low levels, in that population. This contrived example illustrates the danger in simplistic extrapolation of heritability findings from one human population to another very different population.
- How to calculate heritability, Cureffi.org
- Heritability in Inflammatory Bowel Disease: From the First Twin Study to Genome-Wide Association Studies
- Twins Reared Apart, Psychology Today
- Sources of human psychological differences: the Minnesota Study of Twins Reared Apart, Science, Bouchard et al., October 12, 1990, 250(4978):223-8
- Studies of Reared-Apart (Separated) Twins: Facts and Fallacies, Jay Joseph, Mad in America blog, December 15, 2014