Traditionally, there have been two types of twin studies used to estimate heritability. The first is the classical twin study. In this, you take the phenotype correlation of MZ twins raised together and compare it to the correlation of DZ twins together. Since both types of twins shared the same home and same womb, the greater phenotypic similarity of MZ twins must be caused by their greater genetic similarity. And since MZ twins share twice as many segregating genes as DZ twins, we double the difference between the correlations to estimate heritability.

One criticism of this method is that even though both MZs and DZs share the same womb and home, the former are still more environmentally similar because of the unique experience of being an identical twin. If so, heritability will be overestimated.

The second type of study is taking the phenotypic correlation between MZ twins raised apart and using this as a proxy for heritability. The problem with this is even MZ twins raised apart shared the same womb and are often not raised far apart enough to prove much.

It occurred to me though that the perfect study would combine both methods and compare MZ twins raised apart with DZ twins raised apart. Now if the MZ twins are more alike, you can’t say it’s because of the unique experience of being an identical twin, because they don’t know their identical twin. And you can’t say it’s because they shared the same womb or were not raised far enough apart because the same applies to the DZ control group.

It’s absolutely BRILLIANT!!!!!!!!!!!!

To brilliant for me to be the first person to have thought of this.

Indeed I vaguely recalled the famous Minnesota study of twins reared apart (MISTRA) also having data on DZ twins raised apart. I recalled that MZ twins raised apart correlated about 0.75 on IQ and thus IQ had a heritability of 0.75. Well, I’ll just subtract the correlation of DZ twins raised apart from 0.75 and double the result (since MZ have twice as many common segregating genes) and I’d have myself a revised heritability estimate that was beyond reproach.

Sadly, I could not find the data anywhere.

And then I found this article by Jay Jospeph:

Bouchard and colleagues never published their full-sample DZA IQ correlations, even though they published full-sample DZA correlations for personality, “special mental abilities,” and most other MISTRA-studied psychological characteristics. To this day, they have prohibited independent researchers from inspecting the closely guarded MISTRA raw data. I show in my new article that the likely reason that they did not publish, share, or make available their full-sample DZA control group IQ data was that—based on the near full-sample DZA IQ correlations that were published in 2007 and 2012—the results would have revealed their failure to find a significantly higher MZA group versus DZA group mean correlation for any of the three IQ measures they used. An “important first step” requirement in the process of determining whether genetic factors influence IQ scores is finding that the MZA correlation is higher than the corresponding DZA correlation at a statistically significant level. The researchers bypassed this required step in their 1990 Science study, most likely because the hidden results failed to confirm their pre-existing belief that IQ was (strongly) influenced by genetic factors.9 Their strong genetic biases, it seems, led them to omit, bypass, and suppress their DZA IQ correlations in order to obtain the desired results.10

So it seems this excellent study was actually done, we’re just not allowed to see the results.