What Studying Autism Won't Teach You

There is a lot of research into autism. Some of it is primarily motivated by practical, real-life questions about how best to educate and accommodate autistic people. But other research is motivated by more scientific questions - what can autism tell us about the social brain?

Unfortunately, the answer is - not much.

Autism is a behaviourally defined condition. It is also a highly heterogeneous condition. What this means is that the category of autism lumps together a wide variety of kids, with very different underlying conditions, who share a set of behaviours.

This is a big problem for anyone trying to do scientific research on autism. Studying social skills in autism is semi-tautological - at best, all you'll learn is whether your test predicts real-life behaviour. Because autistic people are heterogeneous and defined by social skill impairment, any social skill test that is correlated with real-life impairment will show impairment in autism, by definition.

Case studies of autism would actually be more useful, because then the individual differences aren't smoothed and averaged out. But case studies also suffer from having no statistical power - no ability to tell coincidence from correlation. And we don't yet know the best way to subdivide autistics so we get more homogeneous groups to study.

Because autistics are pre-defined by social impairment, but have very different sets of impairments in other areas, studying them as a group will give you the impression that social skills are separable from these other skills, even if they aren't.

To illustrate, I'll discuss a few hypothetical cases.

Imagine one kid who has visual processing problems, and can't read facial expressions for the same reason he confuses a cat with a dog. If you test him alone, his impairment in facial expression recognition will be accompanied by severe impairments in every other visual skill, and he won't be impaired in understanding tone of voice, making it pretty obvious his impairment isn't primarily social.

Imagine another kid who has auditory processing problems. He struggles in even identifying speech as speech, much less understanding it. Obviously, he has a language delay, and this delay in language deprives him of the chance to 'look inside' another person's mind, and check his assumptions about what they're thinking and feeling. As a result, his social cognition is delayed, although his ability to read facial expressions is OK. Speech therapy addresses his language delays, teaching him to talk, but he doesn't get training in social skills except what he needs to cooperate with speech therapy.

Imagine a third kid, who is unable to feel embarrassed. He is impulsive and socially inappropriate, and because he's never embarrassed by his behaviour, he doesn't get why he shouldn't do it. The emotions he feels, he can recognize in others, but because he never feels embarrassed, he can't tell when others are feeling it.

Imagine a fourth kid, who has trouble shifting attention. Because many social cues are brief and fleeting, he'll miss them unless he happens to be paying attention to the right person at the right time. He also misses many other things, like the curb cut he tripped over because he wasn't watching where he was walking, or the little bird that spotted him walking past and flew away. On tests, his attention difficulties result in scattered performance - sometimes he gets it, sometimes he doesn't. And many skills are poor because he hasn't been paying attention to the right things at the right times to learn them.

Each kid has at least some social impairment, and could potentially be diagnosed as autistic. So let's say we average together their scores on various tests. This will obscure the one kid's visual impairment, because the other kids process visual stimuli just fine. It'll also obscure the other kid's auditory and language impairments. On a test of attention, the last kid's attention impairments will be obscured, because the other kids shift attention just fine. And if they even think to test embarrassment, they won't find much difference, because most of the kids get embarrassed fairly readily. The only areas where these kids' impairments overlap are in social areas, and so the average will show social impairment with no other impairments.

This apparent 'generalized social impairment without other impairments' is what a lot of autism research has found. And this has led to speculation that there may be some sort of 'social module' to the brain, some brain area or network that subserves all social skills, but isn't crucial to any other skill area. But as I've shown here, this apparent modularity could simply be an effect of averaging together results from many people with only one area of overlap in impairments.

What would really teach us about the social brain is to study groups pre-defined by biological, not behavioural measures. People with Fragile X Syndrome, Turner Syndrome, Williams Syndrome, Down Syndrome, etc. Or people with damage to brain regions thought to relate to social interaction, such as frontal lobes, amygdala, or anterior cingulate cortex.

There needs to be more research into the social skills of individuals with biologically-defined conditions. But what research has been done already suggests a very different view - social skills are not a unitary construct. There are face-specific impairments. There are emotion-specific impairments (like the never-embarrassed kid described above). There are specific impairments to social attention (or enhancements of it, in Williams Syndrome). Social skills are not a generalized skill area, but a collection of skills that work together in a certain context.

It's also important to consider the role of experience, and groups defined by certain kinds of experiences can be useful as well. Research into late-educated deaf people has highlighted the contribution that language skills make to social skills. Research into traumatized children has shown that emotional experiences can bias interpretation of social situations, so they're extra sensitive to signs of danger. And research into younger siblings has shown that a social partner who is just a bit ahead of you can accelerate your social development.

So if you want to understand the social brain, don't study autism. Study groups defined by something other than their behavioural traits.