In most twin studies, the twins themselves (or their parents) report “twin type” – i.e., whether the twins are identical (monozygotic/MZ) or fraternal (dizygotic/DZ). This is reasonably accurate, but falls far short of DNA testing. Most behavioral genetic studies rely on moderately flawed data.
You might think that this fact undermines the whole literature. But in fact, researchers have turned this weakness into a strength. Here’s how:
1. Twin studies normally assume that MZ twins are treated no more similarly than DZ twins. Many critics suggest that MZ twins might be more alike partly (or entirely!) because their parents treat them more alike.
If all parents correctly identified their twins’ zygosity, it would be very hard to answer this challenge. But the existence of parental error allows a striking test: If parents treat twins they believe to be identical more similarly than twins they believe to be fraternal, then misidentified DZ twins should be more alike than correctly identified DZ twins, and misidentified MZ twins should be less alike than correctly identified MZ twins. As Pinker explains in The Blank Slate:
A neat test is provided by identical twins who are mistakenly thought to be fraternal until a genetic test proves otherwise. If it is expectations that make twins alike, these twins should not be alike; if it is the genes, they should be. In fact the twins are as alike as when the parents know they are identical. [Note: Here’s Pinker’s cite.]
2. So heritability estimates are not artificially inflated by environmental similarity. But wait – there’s more! If a modest fraction of so-called MZ twins are actually DZ, and a modest fraction of so-called DZ twins are actually MZ, then standard heritability estimates are artificially deflated.
Here’s why: If 10% of all twins types are misidentified, then standard estimates don’t show the effect of sharing 50% of your genes versus 100%; they only show the effect of sharing 55% of your genes versus 95%. That means that the true effect of sharing 50 percentage-points of additional genetic material is actually 1.25 times larger than standard estimates suggest!
No empirical research is perfect. People who don’t like the results often revel in this fact: “No one can prove that I’m wrong!” Wiser heads know, however, that many of these errors are not a roadblock to intellectual progress, but a path to deeper knowledge.
READER COMMENTS
BGC
Mar 18 2008 at 1:09am
Right – in epidemiology I was taught that these ‘misclassification errors’ always have the effect of reducing measured differences.
And twin studies under-estimate genetic effects for another reason too – there are genetic mutations which accmulate during the lifespan which make monozygotic twins diverge genetically the older they get. So a trait could be genetically determined, but differ in identical twins.
One lesson is that genes are even more powerful than we already know, another lesson is that the residual differences left-over after measured genetic effects have been subtracted cannot be ascribed to the *environment* since some of the unexplained variation is merely due to misclassiciation and other types of experimental error and technical limitations.
Comments are closed.