Atypical Scanning Patterns of Static Social Scenes in ASD: Results from the ABC-CT Feasibility Study

Background: Previous eye-tracking research has shown that children with ASD spend less time looking at faces and eyes than their typically developing (TD) peers (Hanley et al., 2013; Riby & Hancock, 2008, 2009; Sasson et al., 2007). When faces are shown in the context of complex social situations, these diagnosis group differences may be larger (Hanley et al., 2013; Speer et al., 2007).

Objectives: To examine face preference (%Face) of children with and without ASD when looking at static social photographs presented during eye tracking. These results were reviewed in conjunction with the preference for looking on screen at all (%Valid). Exploratory analyses were conducted to assess face preference differences across trials (to examine scene qualities that amplify group differences) and over time within trials (to investigate how long stimuli need to be shown for differences to be detected).

Methods: Eye-tracking data was collected from 51 children (N_ASD = 25, N_TD = 26) aged 4-11 using a 500 Hz EyeLink 1000 Plus. Forty-seven (N_ASD = 23, N_TD = 24; M_Age = 7.3 ± 2.2 years) remained after quality exclusions (trials: <50% onscreen looking or >2.5 degrees of calibration error; participants: <2 of 6 valid trials). Each static social image (from the EU-AIMS LEAP task battery) was presented for 20 seconds.

Results: Linear mixed models controlling for image and full scale IQ were used to examine group differences on %Valid and %Face. For %Valid, we observed a main effect of group (ASD<TD; p<.01, d=1.6). Exploratory time bin analyses (5s bins) revealed a main effect of time (%Valid decreasing over time; p<.001) and a group*time interaction (p<.01): ASD and TD groups had similar %Valid in the first five seconds, but between-group differences increased over time (6-10s: p<.05, 11-15s: p=.001, 16-20s: p<.001). For %Face, a group effect was observed in the expected direction (ASD<TD; p<.001, d=0.91). Exploratory analyses including a group*image interaction showed the ASD group looked less at faces than TD participants for all images except an image of a single person coloring (all other images involved multiple people). Further analyses examined how long images needed to be shown to observe between group differences. A group effect appeared for all presentation times examined, ranging from 0-10s (p<.001, d=0.82) to 0-1s (p<.01, d=.77), indicating pervasive differences in face looking. The strongest effect size was seen for 0-5s (p<.001, d=1.02), matching up to the time when no between-group differences in %Valid were observed.

Conclusions: Eye-tracking results from static social images revealed consistent and large differences in face preference between children with and without ASD. An exploratory content analysis suggested increased social complexity may increase group differences. An exploratory time analysis revealed face-looking differences as early as the first second of presentation, indicating that shorter trials may be sufficient for between group differentiation. Finally, the correspondence of the largest face-looking effect size in the first five seconds to a period where no between-group differences existed in overall looking potentially highlights the separability of social attention differences specific to ASD from general attention differences.