At first glance, it might sound strange to suggest that an IQ test given at age seven is more accurate than IQ test taken when you’re an adult and your brain is fully developed.  Indeed one of the most stunning discoveries of behavioral genetics is the doubling (yes doubling) of IQ’s heritability from childhood to older adulthood.  Of course the majority of these studies estimate heritability by comparing the correlation of MZ twins raised together with DZ twins raised together; a method that is vulnerable to criticism because of the equal environment assumption. But we have corroborating evidence (i.e. the zero correlation between the IQs of unrelated people raised together, the vanishing of IQ gains caused by adoption) that IQ does indeed become more genetic with age, at least within a given developed country.

But the rising heritability of IQ might be misleading.  We know from the Flynn effect that shared environment effects like schooling and socioeconomic status have HUGE effects on IQ (though perhaps not intelligence) but as scholars Dickens and Flynn brilliantly noted, within generations, these cultural effects get counted as genetic effects because people with certain genes spend more time in culturally stimulating environments.

So when we say IQ is 80% genetic by late adulthood, we’re not necessarily saying genes DIRECTLY cause IQ.  Rather genes cause you to engage in IQ increasing behavior. It’s a bit like saying weight is highly genetic, when what we really mean is diet and exercise choices are genetic (which cause weight). So adult IQ is not necessarily measuring who is the most biologically clever.  It could merely be measuring who got a PhD in English literature, thus artificially boosting their score on vocabulary beyond what it should have been.

In order to truly measure REAL intelligence (if you believe in such a construct), you need to return to childhood, where people haven’t had as much time to coach themselves for IQ type tests.  Of course children who come from rich and educated homes have an unfair advantage on IQ tests but because the gene-environment covariance is much smaller in kids than in adults, the effects of shared environment can be separated from the effects of genes by statistically adjusting for social class.

How do you adjust childhood IQ for social class?  Let’s say you have a seven year-old child from an upper class family who scores 120 on an IQ test (white norms) This implies he’s smarter than 90% of white Americans his age.  However most white kids his age were not raised in upper class homes.  Now you could compared him only to other white kids from his class, but that’s misleading because they’re a genetically elite sample (since they inherited not only their social class from their parents, but the genes that caused it).  What you need to do is compare him to a genetically random sample of white kids raised in upper class homes, and the closest we have to that are adoption studies.

After correcting for old norms and converting to white norms, the sixteen white adopted kids in the Minnesota transracial adoption study had an age seven IQ of 109 with an SD of 11.3, compared to the general U.S. white population which by definition has a mean IQ of 100 with an SD of 15.  What this shows is that if every white seven year-old in America was raised by the upper-class, the mean IQ would rise and the variability would shrink.  This makes sense because we’ve removed a huge chunk of the variance from IQ (cultural effects) leaving hopefully only direct biological effects (genes and biological environment).

So a white upper class seven-year old with an IQ of 120 is 0.97 SD above the mean if all white kids were raised in his social class.  Thus his social class adjusted IQ would be 115.

If being raised in an upper class home causes the  mean IQ to increase by 9 points (and the SD to drop to 11.3) we might guess that being raised in a lower class home causes the mean IQ to drop by 9 points.  So if all white kids were raised by low income high school dropouts, the IQ distribution at age seven might be 91 (SD 11.3).

So a seven-year-old from a low class family with an IQ of 120 is an astonishing 2.57 SD above the mean if all white kids were raised in such homes.  His social class adjusted IQ would be an incredible 138!

So here we see two kids with an identical IQ score, but by adjusting for social class, we find that one kid is actually 23 IQ points smarter than the other.  The kid from the lower class will probably get a good education and enter the upper class when he’s older, and by 40, his measured IQ might become as high as his age seven adjusted IQ.  By contrast the upper class kid might go down a bit in social class, causing his measured IQ at age 40 to also match his adjusted IQ at age seven.

On the other hand, if the upper class kid is especially curious, and the lower class kid has no interest in learning, the opposite might occur.  This is why IQ tests when measured in childhood, if properly adjusted, might be far better measures of real intelligence than measurements taken after kids are old enough to shape their own environments in ways we can’t adjust for.