Do Abnormal Eye Movements Account for Impaired Social Cognition in Adults with ASD?

Background: Individuals with autism spectrum disorder (ASD) generally under-perform on a variety of social cognitive tasks relative to neurotypical (NT) controls. Individuals with ASD also exhibit differences in gaze to social images and videos. While there are correlations between atypical gaze and social cognitive abilities, to date, the causal nature of the relationship is unknown – in other words, does abnormal gaze result in abnormal social cognitive performance, or does altered social cognitive processing result in abnormal gaze?

Objectives: Using a novel “windowed viewing” experimental paradigm whereby individuals view stimuli through another person’s viewpoint, the present study investigated whether differences in social gaze in ASD underlies their social cognitive deficits.

Methods: Eye tracking data was collected from 5 high functioning ASD adults and 5 neurotypical (NT) adult “gaze donors” while they freely viewed an episode of The Office. New stimuli were created whereby for each video frame, the point of the donor’s fixation was used to create a window with a 3° visual angle around that point, thus approximating the foveal gaze pattern of each of the gaze donors.  107 NT undergraduates subsequently watched 20 video clips through ASD windows, 20 through NT windows, and 40 with full view (all without sound). After each clip, participants completed a multiple-choice task where they were asked to identify the complex emotion displayed by a particular character. All participants also completed the reading the mind in the eyes (RME) task (Baron-Cohen et al., 2001), which assesses the ability to use information from the eyes of faces to make social emotional judgments. The 5 ASD gaze donors were all in the bottom 10^thpercentile of RME performance, confirming their social cognitive deficits.

Results: Surprisingly, participants’ emotion recognition was equally accurate across the ASD-windowed (M=0.57[0.13]) and NT-windowed conditions (M=0.58[0.12]; t[106]=0.99, p=0.33). As expected, all participants performed significantly better in the full view condition (M=0.87[0.08]; t[106]=25.1, p<0.001). Emotion recognition accuracy was positively correlated with RME performance, suggesting both tasks assess related social cognitive abilities (r=0.46, p < 0.001). Typicality in the pattern of gaze by the donor was negatively correlated with participants’ emotion recognition performance (r=-0.68, p=0.03).  

Conclusions: These findings suggest that differences in gaze alone cannot account for difficulties ASD individuals have in recognizing complex social emotions. Not only do these results suggest that the point of foveation in both ASD and NT gaze patterns contains enough visual information for participants to perform equally well, but also that atypical patterns of gaze may actually facilitate performance. Thus, social deficits may not be caused by differences in selection of visual input, but rather the subsequent processing of this information. From these results we cannot conclude that differences in gaze alone cause deficits in social emotion recognition in high functioning adults with ASD, and suggest greater caution when interpreting the meaning of gaze differences in ASD.