Why Autistic Children Avoid Eye Contact: Two Competing Accounts

What the research found

Since Kanner’s early descriptions of autism, reduced eye contact has been read as a sign of social indifference, on the assumption that the eyes of other people simply hold less interest for an autistic child. The brain-based version of that assumption is precise. The amygdala theory of autism, set out by Simon Baron-Cohen and colleagues in 2000, holds that an underactive amygdala fails to flag eyes as salient. The amygdala is a small structure below the cortex that marks what is biologically, socially or emotionally important, and on this account it marks eyes too faintly to draw a child’s attention.

A competing hypothesis reverses the direction. Tanaka and Sung proposed in 2013 that the amygdala in autism responds to direct eye contact with an unpleasant degree of arousal, and that looking away is a strategy for bringing that arousal back down.

Nicole Stuart and colleagues at the University of Western Australia and the Telethon Kids Institute in Perth set out to arbitrate. Writing in the Journal of Autism and Developmental Disorders in 2023, they gathered studies that used functional MRI to measure brain activity in autistic and non-autistic participants while gaze to the eye region was tracked or experimentally directed. Eleven met their criteria, and eight supported eye avoidance.

The most direct evidence concerns timing. In a 2012 study by Kliemann and colleagues, a participant’s first fixation on a face was set to either the eyes or the mouth, and among autistic participants amygdala activity was positively correlated with how often they subsequently moved their gaze away from the eyes. This is a neural finding, recorded in the scanner rather than reported by participants, and it places the amygdala response before the avoidance.

Two studies compared directed gaze against free viewing, a design that permits something closer to a causal reading. Hadjikhani and colleagues in 2017 found equivalent amygdala activation in autistic and non-autistic participants viewing faces freely, and greater activation of the amygdala and superior colliculus in the autistic group once gaze was cued to the eyes. Tottenham and colleagues in 2014 found that the increase under cued gaze was largest in the participants who had looked at the eyes least when free to choose.

The three dissenting studies reported equivalence. Perlman in 2011, Dapretto in 2006 and Davies in 2011 each found amygdala activity in autistic participants comparable to that of controls under directed gaze. Across all eleven, none reported the reduced amygdala activity that the salience account requires.

What it means for you and your child

The two accounts predict different things about the same behaviour, and the amygdala is where they part company. An account built on diminished social interest predicts a muted response when a child looks at eyes; an account built on aversion predicts an elevated one. The elevated result was the common one here.

That difference is easily lost in practice. A child who looks away because faces do not interest him and a child who looks away because looking is uncomfortable present much the same behaviour to an observer, and only the second is describing something aversive. Autistic adults have long described the second. One respondent to a 2017 survey by Trevisan and colleagues said that eye contact “triggers a fight or flight response so strong that it overrides everything else,” and the neural evidence assembled here is consistent with that report.

The review is narrative rather than a meta-analysis. Its authors state inclusion criteria but describe no systematic search protocol and pool no statistics, so the result is a count of studies rather than a combined effect. Samples were predominantly male and predominantly people the original papers called higher functioning. No study measured arousal directly through heart rate or pupil size, which the authors name as the clearest gap: the arousal at the centre of the hypothesis has been inferred from brain activity and never recorded from the body.

The review also declines to make the account exclusive. Reduced reward from social stimuli and heightened arousal can operate together, and several researchers cited here propose that both contribute, in different proportions. Behavioural results are mixed too: an earlier review by Cuve and colleagues found that of sixteen eye-tracking studies, nine reported reduced fixation on the eyes and five reported no difference.

One finding runs the other way. Cued gaze brought activity in several cortical regions of the social brain closer to that of non-autistic participants, and one research group speculated that improving eye gaze might support perspective-taking. Only one of those studies compared cued gaze against free viewing, so that causal reading remains unestablished.

Eye contact is among the most closely policed autistic behaviours, and a target for increasing it still appears in behaviour plans. The review’s own finding is narrower than any such programme would need: directing gaze to the eyes raised amygdala activity in autistic participants for as long as the direction lasted, and no study tested whether repetition lowers that response over time. A rising count of seconds on a chart records compliance with a gaze target, which on these findings says nothing about whether the looking has grown any easier.