Problems with Simon Baron-Cohen's Thesis

I see that Simon Baron-Cohen has a piece in Seed about his theory of autism. I am really skeptical of many of his arguments related to autism, so I thought I would discuss a couple of them. Here is his core argument:

So what has all of this got to do with autism? We know that autism runs in families, and that if a child with autism is a twin, the chances of the other twin also having autism is much higher if the twins are identical. This tells us that genes are likely to be an important part of the explanation, and that one should look at the parents of children with autism for clues. Furthermore, our studies have uncovered four findings that implicate assortative mating in autism. First, both parents of children with autism are likely to be super-fast on attention tasks, in which the aim is to spot a detail as quickly as possible. Second, both parents have an increased likelihood of having had a father who worked in the field of engineering. Third, both parents are more likely to have elevated scores on subtle measures of autistic traits. And fourth, both parents show a trend toward a more male pattern of brain activity when measured using MRI.

The chances of both parents displaying these similarities are vanishingly small. Something must be causing two such individuals to be attracted to one another. I propose that "something" is strong systemizing--the drive to analyze the details of a system in order to understand how it works.

All human brains have a systemizing mechanism that is set at different levels. The extreme behavior of children with autism (whose systemizing mechanism may be set too high) may seem like a far cry from the more moderate behavior seen in their parents and grandparents, but this new theory proposes that across the generations these are only differences of degree. Think of a child with autism, who appears lost in his or her own world, totally focused on lining up Lego bricks into colorful patterns for hours at a time. This is hypersystemizing. Then think of the parents we talked about at the outset: a fascination with weather reports is one example of systemizing; a fascination with model trains is another, albeit milder, example.

Evidence from parents in the general population suggests there is assortative mating for systemizing, such that people who are attracted to systems are more likely to have a partner who shares this characteristic. Combined with the fact that both parents of children with autism are likely to be very detail-oriented, highly analytic, and to have a father who worked in a field requiring good systemizing skills, this suggests that the genes involved in systemizing may be linked to the genes that cause autism.

Although these genes remain to be identified, the assortative mating theory throws up some testable predictions: First, autism should be more common in families where both parents are strong systemizers. For example, some media reports have claimed that autism is more prevalent in areas like Silicon Valley, but we need well-controlled tests to see if this is true. Second, since the drive to systemize is stronger in males than in females, it should be the case that both mothers and fathers of children with autism are more likely to have strongly "male" interests and behaviors. Finally, if systemizing is linked in part to prenatal testosterone levels (which studies from our lab suggest may be the case), then mothers of children with autism may be more likely to have testosterone-linked medical conditions. Again, a highly testable prediction.

First, I do not dispute the idea that autism is genetically heritable. Overwhelming evidence suggests that this is the case. What I dispute is this idea that assortative mating between two people with "systematizing" brains is causative. I have a several reasons for this:

• 1) While autism is strongly heritable (80%), the tendency towards systematizing traits is not. I am not sure what data Baron-Cohen is using, but the data I have read shows that the heritability for specific mental abilities such as perceptual speed (like the test he gave to the two parents) is only about 40%. Likewise, personality and interests are not particularly heritable either. Estimates for personality suggest that it is about 35% heritable, and the same amount for estimates for occupational interest.

  If indeed there are genes for systematizing brains -- I have my doubts -- and we believe these genes are causative -- I have my doubts -- would we not expect them to be at least as heritable as autism is?

• 2) This business about strongly male interests is a distinction without a difference. Surely, girls are interested in different things on average than boys, but I have presented evidence repeatedly that this is not the result of a genetic/biological difference between boys and girls but rather a sociocultural difference. If we believe that these behaviors are socioculturally derived and we think that they are associated with autism, wouldn't we then have to think that autism is sociocultural? The evidence shows that autism is not sociocultural, so I have to conclude that the association with the "male" interests is an epiphenomenon.
• 3) I am willing to accept this idea that prenatal testosterone has a role in the etiology of autism, but you can't say that it is prenatal and that is also genetic. I recognize that environmental and genetic effects are often additive, but in this case we have a disease that is primarily genetic. I feel like he is trying to have it both ways, asserting that it is an environmental effect that causes the systematizing and the autism and then saying that it is a genetic effect.
• 4) While it is unfair to expect this, I would note that he has in no way explained how extreme systematizing brains result in all of the other clinical findings in autism such as social withdrawl and developmental delay. The rejoinder could be that extreme systematizers often neglect the social aspects of development -- that the geeky engineer is often socially awkward and autism is the logical extension of that. However, this explanation smacks of over-generalization. People who are systematic are sometimes socially inept, but they are not always socially inept. Thus, the explanation of the clinical expression of the disease in this hypothesis smacks of hand-waving.

In short, I don't buy Baron-Cohen's thesis because it resembles of so many other theories of late trying to explain the increase in autism in terms of correlation rather than causation -- first it was mercury in inoculations, then television, now that the engineers are breeding. I am certain that I could find a variety of traits that are common among the parents of autistic children, but that does not mean those traits are heritable or that they caused the disease.