Social Visual Preferences in Children with ASD: Preliminary Results from a Novel Dynamic Eye Tracking Task

Background: Several eye-tracking studies have used social-nonsocial visual preference paradigms, i.e. side-by-side presentations of a social target versus a non-social distractor, to highlight atypical biases towards social information in individuals with autism spectrum disorder (ASD) (Pierce et al., 2011, 2015; Klin et al., 2009; Shaffer et al., 2017). In an ongoing study, we explore a modification of traditional social visual preference paradigms that incorporates uniform movement of targets and distractors, with the ultimate goal of increasing visual engagement to preference paradigms by participants while providing additional information for facilitating derivation of key eye tracking outcome measures.

Objectives:

• To examine the potential and limitations of a dynamic eye tracking social preference task in studying between-group differences in children with ASD and typically developing (TD) children.
• To examine relationships between social preferences in our task and clinically-relevant phenotypic variables in children with ASD.

Methods:

Twenty-nine children (N = 14 ASD, N = 15 TD) between the ages of 2 to 12 years watched three, 20 second trials. In each trial, two smaller videos were displayed: a social target video of a woman singing a song (Trials 1 and 3) or smiling (Trial 2), contrasted with distractor video of a nonsocial, dynamically evolving fractal. These videos began in the middle of the screen and moved horizontally in opposite directions, reversing towards the edge of the screen, and crossing over each other twice per trial. Videos were shown in the same order for every participant. Eye tracking was performed using an SR EyeLink 1000 Plus eye tracker.

Results:

An independent samples t-test showed that, across trials, the percentage of time TD children looked at the social stimuli (SocialLooking%) was greater than children with ASD (p =.042). However, when trials were examined individually, ASD and TD only significantly differed in Trial 1 (p = .046), not in Trials 2 and 3 (p = .082, 0.158). We subsequently performed a time course analysis, dividing comparable Trials 1 & 3 into four, equal time periods. A repeated measures ANOVA revealed TD children showed higher %SocialLooking only during time period one (p = .017). Phenotypically, relationships between ADOS and SocialLooking% in Trial 1 approached significance (r= -.480, p = .09). Age was negatively correlated with SocialLooking% in Trial 1 (r = .387, p = 0.035) and Trial 3 (p < .001).

Conclusions:

As predicted, children with ASD looked at the social stimuli less than TD children. Further exploration indicated these differences only occurred in the first portion of the first trial. It may be that TD and ASD differ only in initial, spontaneous preference for social stimuli, after which point familiarity effects wash away group differences. Results between social looking and autism symptoms are promising, but the current study may yet be underpowered to robustly detect these relationships. Relationship of age to social looking warrants caution, signalling the need to carefully consider developmental norms in this and potentially other preferential looking paradigms.