Mismodeling the Social Self in Autism

by Jim Schnabel

March 10, 2008

One of the most striking signs of the autism spectrum disorders (ASD) is a difficulty in social interaction and communication. According to a recent brain-imaging study at the Baylor College of Medicine in Texas, the social troubles seen in ASD might be caused, at least in part, by a person’s failure to make a proper mental model of the self in a social environment.

The study, conducted in the laboratory of Baylor neuroscientist Read Montague, originated from some of his group’s previous work on human social behavior, reported in Science in 2005 and 2006. In that research, Montague and his colleagues took the novel approach of scanning two subjects simultaneously with functional magnetic resonance imaging (fMRI) while they played a trust-based investment game commonly used to study socioeconomic behavior.

The experiments soon had Montague and his postdocs scratching their heads. When the two game-playing subjects were scanned simultaneously, they showed distinctively patterned fMRI responses in a deep region of the brain called the cingulate cortex.

“It kept going on and going off,” remembers Montague. “But it didn’t correlate with the amount of money a subject would send, the amount of reward the subject would get, how many rounds had been played, whether the subject was male or female. I mean, we hunted that down. And we were very frustrated. And then we realized, the only thing left were the people.”

As a subject came to a decision about his own action in the game, his cingulate cortex tended to show one distinctive pattern. When the subject was confronted with evidence of the other party’s action in the game, his cingulate cortex lit up in the reverse pattern. But when the subject knowingly played against a computer that could not respond flexibly to his choices, so that the element of social interaction was removed, Montague found that the distinctive cingulate cortex responses disappeared.

Montague’s conclusion: The yin-yang response of the cingulate cortex reflected an internal modeling of “self” and “other” in a real, human social context. “We stumbled upon this response only because we were recording from two brains at once,” he says.

The apparent discovery of what Montague calls a “social agency” response in the cingulate cortex was interesting in its own right. But an obvious follow-up was to look at this region in people known to have social-interaction problems—for example, people with autism spectrum disorders. On Feb. 6, in the journal Neuron, Montague’s group reported on just such a study.

Montague and his colleagues recruited 18 relatively high-functioning male adolescents with ASD and asked them to play the investment game while being scanned by fMRI. A control group of IQ-matched adolescent boys was also scanned. Images of the cingulate cortexes of the boys with ASD, when compared to controls, appeared normal during the “other” phase when they were expected to be modeling their partners’ behavior.

But during the “self” modeling phase they showed markedly reduced responses. And the level of reduction was greater for those boys who had worse symptoms of ASD.

To help show that the cingulate response wasn’t limited to the social-exchange game, Montague arranged a second experiment: A large group of accomplished athletes was asked to watch a video of athletic activity, such as kicking a football. When asked to imagine performing that activity themselves, the subjects showed a similar “self” modeling pattern in their cingulate cortexes on fMRI scans.

Chris Frith, a social cognition expert at the University of Aarhus in Denmark, notes that Montague’s study is the first he has seen in which subjects with ASD undergo brain scans while participating in a real social interaction, as opposed to merely observing or thinking about such an interaction.

“High-functioning people with autism often know about how social interactions should go,” Frith says. “Their problem arises when [they are] directly engaged in social interactions with no time to think.”

To Frith, the key finding of the study was that the neural differences between controls and subjects with ASD “only appeared in the ‘self’ phase, i.e., when the subject had to decide what to do. There was no difference in the ‘other’ phase, when the subject learned what the other had done. This shows that the problem only applies to one aspect of the social exchange.”

Montague thinks that the diminished “self” response has to do with the model of one’s own intentions that one normally makes in a social context. “Individuals lacking a capacity to delineate their own intentions,” he wrote in his Neuron paper, “are likely unable to attribute the social goals or intentions of others”—which could explain why individuals with ASD have such difficulties socially.

Also noteworthy to Frith was the fact that the reduced “self”-related response seen in the adolescents with ASD was akin to the normal adolescents’ response when playing against a computer rather than a real person. People with ASD are often said to feel more at ease among computers and other inanimate objects than among people.

“I would like to know what is different about people’s attitudes and strategies when they are playing against a computer,” says Frith. “For example, do they believe that the computer won’t modify its style of play? Do they not care what the computer thinks of them? There are a number of interesting lines of research opened up by this result.”

“The fact that it was the ‘self’ region that was missing here is interesting,” agrees Lindsay Oberman, a postdoctoral researcher at Beth Israel Deaconess Medical Center in Boston who has co-authored several high-profile papers on social-cognition problems in autism.

Oberman cautions, however, that multiple brain regions show abnormalities in autism during tasks related to social behavior. “For example, when looking at faces, subjects with autism typically don’t activate the usual face-recognition region” in the visual cortex, she says. “And if we’re looking for the underlying mechanism by which these regions become dysfunctional in certain contexts, we need to get beyond just being able to pinpoint this brain region as impaired or that brain region as impaired.”

Montague argues that his work remains important. “If you’re ever going to move imaging-based neuroscience into psychopathology populations—people with ASD, people with mood disorders, people with mental illnesses of various sorts—you’re going to have to develop large databases of individual responses,” he says.

He plans further, simplified experiments to show that the cingulate cortex “self” response is a general one in ASD. “We’re hoping that this diminished ‘self’ response will still vary with the symptom severity when the ASD subjects look at pictures of themselves and others,” he says. “That way we could move into research with lower-functioning ASD kids and much younger-aged ASD kids. It would be really great if that worked out. So that’s what we’re up to next.”