Eye-Tracking Measures of Social Monitoring in Children with Autism

Background: In previous research, we found that individuals with autism were significantly impaired in attributing social meaning to ambiguous visual information (the actions of animated geometric shapes). In another study, we measured visual scanning by individuals with autism while watching scenes of social interaction; their visual scanning showed evidence of altered visual salience and reduced social monitoring.  In the present study, we used scenes from the classic children’s film, The Red Balloon, to bring these two lines of research together. We defined a series of scenes in the film when a visual fixation to a particular location at a particular time would occur only as a consequence of a viewer’s attribution of intentionality to the film’s main characters, a boy and his friend, the red balloon. Because the agency of the balloon is dynamic during the film—at times an ordinary balloon, at times an animate character—appropriate social monitoring is an indication of social attribution.

Objectives: To study social monitoring as a measure of implicit social attribution in individuals with autism.

Methods: 60 children with autism spectrum disorders (ASD) and 28 age- and IQ-matched typically-developing (TD) participants watched scenes from The Red Balloon while eye-tracking data were collected. Social monitoring was measured during scenes of interaction between the balloon and human characters. Three types of episodes were examined and defined as interactions between the following: 1) the boy and balloon before the balloon becomes animate, 2) the boy and balloon after the balloon becomes animate, and 3) the boy, the balloon after it becomes animate, and a third character. Social monitoring was defined as looking from the social actor (the balloon) to the social responder (the boy or third character) in order to gather implicit social information about how the social responder is reacting to the balloon’s animacy. 

Results: Analyses reveal significant differences in amount of social monitoring between groups. During episodes requiring social monitoring (type 2 and 3), the TD group looked to the social responder for a significantly higher percentage of time than the group with ASD. The TD group also looked back and forth between the social actor and social responder significantly more than the group with ASD, thus showing more monitoring of the actions and reactions of the characters. For these episodes, the more a participant with ASD differed from the TD participants’ mean percent looking to the social responder, the greater the participant’s autism severity (assessed by the social and communication subscales of the ADOS).

Conclusions: Our data show that individuals with ASD appear to be most impaired when the actions of others are dependent upon the balloon as an animate, intentional being. The present study offers a useful and non-verbal method of assessing social monitoring which could be used in evaluating outcome from social skills groups for individuals with ASD. In addition, it has the potential to reflect the individual’s social profile pre- and post-social skills training.