Measuring Contextualized Social Attention Using Eyetracking: A Promising Behavioral Biomarker of Autism Spectrum Disorders

Background: In the past decade, a number of studies have used eye tracking technology to quantify attention to social scenes. Taken together, these studies converge to demonstrate that individuals with ASD display increased attention to background objects and decreased attention to faces and people compared to controls (e.g., Bradshaw, Shic, & Chawarska, 2011; Klin, Jones, Schultz, Volkmar, & Cohen, 2002). More recently, research has focused on clarifying the mechanisms behind such diminished attention to the social world by examining how context impacts visual atypicalities in ASD.

Objectives: In this study, we extended this work by investigating the robustness of contextualized social attention as a behavioral biomarker of ASD in a large sample of children spanning all levels of functioning and a wide range of ages.

Methods:  162 children (N ASD = 106) were included in the study to examine the relationship between eyegaze patterns and phenotypic data. In order to look at group differences, a subset of 38 children with ASD was closely matched on age and IQ to 38 Typically Developing Controls (TDCs). Finally, another subset of 56 kids (N ASD = 32) who came at two different time points separated by a 9 week interval was analyzed to examine test-retest reliability. Social context was manipulated by having participants watch videos of children playing either jointly or in parallel. Eyegaze patterns were collected at a rate of 60 Hz using a Tobii X120 tracker. Total fixation duration to faces or objects relative to the fullscreen was calculated for each participant, and correlated with phenotypic information.

Results: Correlations on the entire sample demonstrated that Total Fixation Duration to faces was associated with independent measures of social skills (e.g., SRS, SCQ) but not with cognitive non-social scores (e.g., DAS).  Analyses on the matched groups revealed a Diagnosis x Stimulus-Type interaction (due to ASD children looking less at faces than the TDCs) and a strong Context x Diagnosis x Stimulus-Type interaction. This three-way interaction was due to group differences being significant in the joint condition but not in the parallel condition. We then looked at the test-retest reliability of this behavioral index (i.e., TFD to faces in the joint condition) using intraclass correlations and found it to be extremely reliable across our two time points.

Conclusions:  Our results revealed that 1) time spent looking at faces correlated with independent measures of social skills, 2) Interactive contexts are best for eliciting group differences, and 3) this behavioral index is stable across time. As such, we suggest that time looking at faces presented in an interactive context might constitute an interesting behavioral biomarker of autism.