News The latest developments in autism research.

Brain hiccup may explain some social problems in autism

by  /  20 February 2017
Fixed focus: Problems with taking another person’s perspective may underlie social difficulties in autism.

Martin Barraud / Getty Images

People with autism show an atypical pattern of brain activity when trying to adopt another person’s point of view, reports a new study1. The aberrant activity occurs in a brain region called the anterior cingulate cortex, and the degree of disruption tracks with social difficulties.

Activity in this region, which is located in the fold between the brain’s hemispheres, is known to be involved in ‘theory of mind’ — the ability to understand the thoughts and feelings of others2. Some people with autism have difficulty with theory of mind, says lead investigator Nicole Wenderoth, professor of neuronal control of movement at ETH Zurich in Switzerland. “They cannot step away from their own perspective.”

Wenderoth and her team explored whether blips in activity in the anterior cingulate cortex have anything to do with this problem.

They used functional magnetic resonance imaging to scan the brains of 16 people with autism and 20 controls as they played a computer game. In the game, players guessed which of two white doors contained a prize. The participants took turns choosing doors with two computer-generated players. They were told that one of these players was under the control of a person they had met before they entered the scanner.

Most of the time, the door containing the prize turned green when a player picked it. The door without the prize, by contrast, turned red. But on some trials, the colors were reversed, with red doors revealing a prize and green doors yielding nothing.

After each turn, the researchers asked the participants whether the outcome of a turn was expected — that is, a green door produced a reward and a red door did not — or unexpected. The participants with autism were less accurate than controls in identifying expected and unexpected outcomes, especially when they were assessing the other players’ turns. The results appeared 28 December in Brain.

Poor prediction:

The controls had lower levels of activity in the anterior cingulate cortex when they observed the turns of the player they believed to be controlled by a person than when evaluating their own or the computer’s actions. This decrease in activity was larger after an unexpected outcome than an expected one, suggesting that this brain region responds most to situations in which another person experiences an unexpected event.

At a group level, the participants with autism did not show this decrease in anterior cingulate cortex activity, even when the observed outcome was unexpected. But the individual responses varied: The smaller an individual’s decline in activity, the greater his social difficulties, as measured by the Autism Diagnostic Observation Schedule.

The findings suggest that, in people with autism, this brain region doesn’t respond in a typical way to a mismatch between what someone else expects to happen and what actually happens. The diminished response may make it difficult for individuals to take another person’s perspective, says Joshua Balsters, a postdoctoral fellow in Wenderoth’s lab.

The findings also mesh with the so-called ‘magical world’ theory, which states that people with autism have difficulty matching events with their causes, says Margaret Kjelgaard, associate professor of communication sciences and disorders at the MGH Institute of Health Professions in Boston, who was not involved in the study.

Wenderoth and her team found that, in controls, communication between the anterior cingulate cortex and a reward center of the brain tracks with the anterior cingulate cortex signal for unexpected outcomes. This relationship is absent in the autism group.

The researchers plan to study whether disruptions in the brain’s reward system underlie the difficulties that individuals with autism have in understanding another person’s expectation of receiving a reward.

  1. Balsters J.H. et al. Brain 140, 235-246 (2017) PubMed
  2. Apps M.A. et al. Neuroimage 64, 1-9 (2013) PubMed
  • Zoran Bekric

    This trial seems to turn on the use of the word “expected”.

    If the prize was behind the green door “most of the time”, but “on some trials” it was behind the red door, what was the actual ratio? If the prize was behind the red door 25% of the time, then after three green doors, it would be rational to “expect” it to be behind the red door. So, if the prize is does turn out to be behind the red door, then it appears where it was “expected”.

    Perhaps the controls were over-generalising, and always “expected” the prize to be behind the green door, because that’s what happened most of the time. As such predicting the “expectations” of the controls would be difficult because not only would one have to analyze the frequency of how often the prize appeared behind which colored door, but also model the errors the other party may have made in their analysis of that frequency.

    While failing to model errors such as over-generalisation is a legitimate failure of theory of mind, it is still a fairly complex task.

    So, a couple of questions:

    I) Were there any differences between the frequency models developed by the controls and the subjects with autism?

    II) How well did each group — controls and test subjects — predict the other models developed by other participants? Did they just assume that everyone would have derived the same frequency model as they did? Or did they anticipate that other people might develop different frequency models and, if so, how accurately did they predict what those models were?

    • Joshua Balsters

      Thank you very much for your comment.

      First, I’d like to clarify one feature of the task. At the start of each trial both doors were white and the participant would choose a door (or watch a door being chosen by the person outside the MRI scanner or the computer). After a door was selected it would change to green or red indicating to all agents there probably was, or was not, a prize behind that door. Participants couldn’t choose based on the colour of the door, the colour change only occurred after a door was selected. The rule, or frequency model, was the same for all agents and participants, 66% of trials were green-win/red-neutral. Participants were also explicitly told before the experiment that a green door meant they would probably win and red door meant they probably would not.

      You are correct that our analysis assumes “over-generalisation” and our age and IQ matched typically developing group correctly indicated whether the outcome was expected (i.e. green-win/red-neutral) or unexpected (i.e. green-neutral/red-win) for around 95% of all trials. Crucially, there was not difference in performance when making these judgments for yourself, another person, or a computer. ASD individuals showed generally lower accuracy on the task (less “over-generalisation” is in deed one potential explanation for this finding), however, ASD individuals were significantly better at indicating whether the trial was expected or unexpected for themselves (around 90% correct) compared to the other player or the computer (around 80% correct). These behavioural results can be seen in Fig 1d,e of the paper. This behavioural result was very important because it highlighted that individuals with ASD had greater difficulty monitoring the expectations and outcomes for other Agents than they do for themselves.


Log in to your Spectrum Wiki account

Email Address:



Request your Spectrum Wiki account

Spectrum Wiki is a community of researchers affiliated with an academic or research institutions. To be considered for participation, please fill out this form and a member of our team will respond to your request.


Email Address:

Title and Lab:

Area of Expertise: