, , , , , , , ,

Using twin studies, scientists divide phenotypic variation into three categories: DNA variation, shared environmental variation, and unshared environmental variation. Shared environment are all the experiences MZ twins reared together have in common (same upbringing, same schools, same womb) while unshared environment are all the experiences they don’t share (position within the womb, getting hit on the head, having an inspiring teacher).

The best estimate using massive datasets suggest that within Western democracies, DNA explains 41% of IQ variation at age 9, 55% at age 12, 66% at age 17, and 74% in adulthood. By contrast shared environment explains 33% at age 9, 18% at age 12, 16% at age 17, and 10% in adulthood (Bouchard 2013, figure 2). That leaves unshared environment explaining 26% of the variation at age 9, 27% at age 12, 19% at 17, and 16% in adulthood.

You don’t have to believe these associations are causal, but they are real. They’ve been more or less replicated using studies comparing (1) MZ twins with DZ twins, (2) MZ twins raised apart, (3) unrelated people reared in the same home. Although all of these methods depend of different assumptions, they all converge on the same conclusion: the predictive power of DNA skyrockets from childhood to adulthood while the predictive power of shared environment plummets. The same pattern (known as the Wilson effect) has also been observed for other phenotypes and in other species.

But why? Shouldn’t environment get more important as we age since experience has increasing time to accumulate? One theory is that more and more genes become active as we age. A more popular theory is that we select environments that maximize our genotype, so environment becomes just a magnifier of genes, not a causal force in its own right. So genetically smart people will stay in school and genetically strong people will lift weights and take steroids etc. People invest in where they’re more likely to be rewarded.

But here’s where things get really interesting. The Wilson effect behaves differently on different types of IQ tests. In his book Does your Family make you smarter? James Flynn notes that cognitive inequality increases from childhood to later adulthood (because good genes cause good environments and bad genes cause bad environments, the smart get smarter and the dumb get dumber, relative to the average person their age) but this pattern is much more pronounced on some tests than others.

Flynn describes three types of tests:

  • Type 1: Tests that show large family effects (shared environment) that decay slowly. This include tests involving vocabulary (define “rudimentary”), general knowledge (How old is the Earth?) verbal abstraction (how are a brain and a computer the same?) and social comprehension (why do you need a passport to travel?)
  • Type 2: Tests that show small family effects that decay fast. These include spatial manipulation (use these two triangles to make a square) and noticing incongruities (what’s missing or absurd in a picture of a common object or scene).
  • Type 3: Tests that show that large family effects that decay fast. These tests include clerical speed and arithmetic.

Flynn argues that type 1 tests involve skills that children learn from observing their parents talk, hence the large family effect. By contrast he says of type 2 tests:

Aside from the occasional jigsaw puzzle, they have no part in everyday life. Children never see their parents performing these cognitive tasks as part of normal behavior. Family effects are weak, even among preschoolers. Since these subtests match environment with genetic potential so young, they would be an ideal measure (for, say, 5-year-olds) of genes for intelligence.

From pages 53-54 of Does Your Family Make You Smarter? by James Flynn

In other words, Type 2 tests measure “novel problem solving”, while type 1 tests measure acquired abilities. A more provocative interpretation is type 2 tests measure real intelligence, while type 1 just measure knowledge and experience. This is the age-old distinction between aptitude tests vs achievement tests, culture fair vs culture loaded, fluid vs crystallized.

And yet Flynn largely rejects Cattell-Horn-Carroll’s theory that fluid ability (novel problem solving) is invested to acquire crystallized ability (accumulated knowledge) writing:

…fluid skill is just as heavily influenced by family environment as the most malleable crystallized skill (vocabulary) and therefore, neither skill deserves to be called an investment and the other a dividend.

From page 132 of Does Your Family Make You Smarter? by James Flynn

Flynn of course is referring to the greatest irony in the history of psychometrics and the biggest mistake of Arthur Jensen’s career: the Raven Progressive Matrices (long worshiped by Jensen and Jensenistas as the most culture fair measure of pure intelligence ever invented) is a type 1 test!

Which of the 8 choices completes the above pattern? Image from  from Carpenter, P., Just, M., & Shell, P. (1990, July)

But let’s not throw the baby out with the bath water. There’s no need to abandon CHC investment theory just because a major test got mischaracterized. But at the same time, it doesn’t feel right to reclassify the Raven as a crystallized test, Research is needed to understand why the Raven is so culturally sensitive when it superficially looks like a measure of novel problem solving. Is it measuring some kind of implicit crystallized knowledge we’re not conscious of like being familiar with patterns, columns and rows and reasoning through the process of elimination, or are the family effects on the non-cognitive part of the test (having the motivation to persist and concentrate on such an abstract task). Flynn argues that the brain is like a muscle, but if so, the Raven is an exercise most have never done before, so why isn’t it a type 2 test?

Flynn might argue that if your family helped you with abstract problems in algebra or had philosophical discussions about hypothetical concepts, you’ve been exercising for the Raven all your life, but this seems like a bit of a stretch. All the research shows that cognitive training has narrow transfer (i.e. practicing chess will only make you slightly better at checkers, and not at all better at scrabble) though perhaps the Raven’s uniquely abstract (general) nature allows it to slightly buck this trend.