Objective Measurements of Social Visual Engagement Predict Parental Impressions of Child Behavior in a Clinical Sample

Poster Presentation
Friday, May 3, 2019: 5:30 PM-7:00 PM
Room: 710 (Palais des congres de Montreal)
T. A. Ponzo, A. I. Mendez, A. Klin, C. Klaiman, S. Shultz and W. Jones, Marcus Autism Center, Children's Healthcare of Atlanta and Emory University School of Medicine, Atlanta, GA

Parents have unique insights into their children’s development: their daily interactions provide experiences beyond lab or clinic that form impressions distinct from those of trained clinicians in a single appointment. Although clinician best estimate remains the gold standard for diagnosis of autism spectrum disorder (ASD), parents contribute critical insights that impact initial screening, symptom profile reports, and medical histories. Despite these strengths, parental impressions are inherently subjective, based on relatively small “samples” – i.e., parents’ own children rather than the many hundreds to which clinical professionals are exposed. Here, we measure the extent to which parent-report measures can be predicted by performance-based measures of social visual engagement collected via eye-tracking. These measures offer an assessment of the real-world validity of lab-based eye-tracking measures as well as the potential for objective metrics of child behavior.


To determine the extent to which parent-report measures of social-emotional functioning in children can be predicted by performance-based, objective measures of children’s social visual engagement collected via eye-tracking.


Participants were N=146 consecutive referrals (ages 1.1–4.75 years) to a community diagnostic clinic. Referrals were made based on concerns for ASD or other developmental delays. Parents completed the Ages & Stages Questionnaires: Social-Emotional, Second Edition (ASQ:SE-2), a questionnaire commonly used to identify early signs of social-emotional difficulties. Children viewed video scenes of age-matched peers engaging in naturalistic social interactions while eye-tracking data were collected. Percentage of fixation time to faces was calculated for each participant. We then used hierarchical logistic regression to predict parent responses on the ASQ:SE-2 from children’s eye-tracking measures.


Eye-tracking-based measures significantly predict ASQ:SE-2 parent responses. Items more relevant to social visual engagement (e.g., Does your child look at you when you talk to him?; Is your child interested in things around her, such as people, toys, and food?) are more strongly associated (r=0.406, p<0.001 and r=0.359, p<0.001, respectively), while items less relevant were unrelated (e.g., Does your child hurt himself on purpose?; Does your child try to hurt other children, adults, or animals?, both r<0.01, p>0.05). Hierarchical logistic regression yielded parameter estimates modeling (a) the odds of a child rarely or never looking at a parent’s face when talking (versus looking sometimes or most times) and (b) the odds of a child looking at least sometimes (versus most times) [Figure 1]. In both cases, t and p values indicate that eye-tracking-based measures significantly predict the odds of a child’s real-world behavior (t>2.6, p<0.01): for every 1% decrease in eye-tracking-based measures of face-looking, the model indicates a 15% increase in the odds that a child rarely or never looks at a parent’s face; likewise, every 1% decrease in face-looking indicates a 6% increase in the odds that a child looks only sometimes [Figure 2].


Objective measures of face-looking are significantly associated with parental impressions and predict specific aspects of a child’s social-emotional functioning. These results highlight the real-world relevance of lab-based eye-tracking measures and offer potential for capitalizing on measures that converge with the unique insights gleaned from parents.