Twin Studies

Thursday, April 3rd, 2014

In lieu of kidnapping twins, Dalton Conley explains, the way that researchers typically calculated how much a given trait — be that extraversion or earnings — was due to genetics was by comparing how alike identical twins were with respect to how alike (same sex) fraternal twins were: 

The logic is that the fraternal twins share half their genes on average and the identical twins share all of them, so the degree to which identical twins are more alike than their fraternal counterpart pairs reflects the genetic contribution to that trait.

If two-thirds of our kids’ chances in life were due to their family background, the field of behavioral genetics would have us believe that the vast lion’s share of that predictive power of family of origin was due to genetics.  According to these studies, about half of the variation in incomes or job situations was due to our genetic makeup.  And only about a sixth resulted from the household environment on which parents could exert some conscious influence.  The remaining one-third was a product of random events outside a family’s control: an inspiring teacher, a traumatic accident, or a lucky break at work.

I initially went into the field of genetics to prove these researchers wrong.  Genes couldn’t matter that much, I figured.  It just didn’t jive with what I saw around me: Siblings seemed so different from each other; I knew plenty of poor kids growing up that I could have imagined achieving great heights had they been reared in better circumstances; and, likewise, in my adulthood I had gotten to know plenty of folks who seemed to be of mediocre talent despite their huge paychecks.  Social environment had to count for more.  So I decided to go right after the geneticists’ core assumption.

That is, their nifty little calculation relies on one hugely problematic assumption known as the “equal environments assumption.”  Put in English, these researchers had to take as a given the notion that identical twins are not treated any more similarly to each other than fraternal twins are (and that identical twins don’t interact with each other more than fraternal twins do in ways that might affect the outcomes in question — i.e. that their mutual, reciprocal influence is no different than that of same gender fraternal twins).  Since in my own experience I often couldn’t even tell who was who in an identical twin set, it seemed obvious to me that identical twins were experiencing much more similar environments than fraternal twins were in ways that were not generalizable to us non-twins in the population, and thus the behavioral geneticists were inflating the effects of genes and correspondingly underestimating the impact of family environment.[2]

Determined to prove them wrong and save the day for social scientists, I thought of a trick that would have been unimaginable before the days of 23andme and the like: I would take the fraction of twins who thought they were identical when they were really fraternal (and vice versa) and run the same analysis on them.  If they thought they were fraternal twins their whole lives but the laboratory genetic test revealed they were actually identical, we could be sure that they weren’t raised with more similar environments because they had been (mistakenly) socialized as fraternal twins.  And ditto in reverse.  But when I ran these folks through the statistical models, the results didn’t refute the behavioral geneticists at all.  In fact, my models confirmed the high genetic heritability for everything from height to high school GPA to ADHD.

Leave a Reply