Modelling the Percentage Looking Time to Social Scenes – Results from a Large Cohort with Autism Spectrum Disorder

Background: The social motivation account of Autism Spectrum Disorder (ASD) stresses the role of decreased social interest as an antecedent of socio-cognitive abnormalities in ASD (Chevallier, Kohls, Troiani et al., 2012). One strong candidate to possibly explain this account is reduced spontaneous social attention (e.g. to the human face and social-communicative cues). A number of studies report reduced attention to faces in children and adults with ASD, but others fail to replicate (Frazier, Strauss, Klingemier et al., 2017). Nonetheless, very few studies utilise large samples and a broad age range, and therefore unable to investigate socialattention at particular developmental stages, or the effects of varying degrees of social content in the stimulus.

Objectives: We report on the developmental profile of social attention and its sensitivity to context from childhood to adulthood in a large European cohort of individuals with ASD and controls.

Methods: We used eye-tracking data from the The EU-AIMS Longitudinal European Autism Project (Loth et al., 2017), which included 366 participants with ASD; 268 with TD; age range 6-30. Participants with ASD had an average IQ of 98.69 (SD = 19.48). Participants watched 6 static photographs depicting social interactions (Figure 1c). Trials with a percentage looking time (PLT) to the screen < 25% were excluded from analysis. Group differences in PLT were explored through 3 linear mixed models including the interaction between Areas of Interest (AOIs) 1) head, body and background people 2) face and hair 3) upper and lower face, group (ASD, TD), and age class (child, adolescent, adult), varying intercept per participant, and IQ as a covariate. We tested the linear contrasts of interest by subtracting the estimate of the TD group from the estimate of the ASD group within each age class.

Results: The groups significantly differed in terms of PLT to the head (difference = -2%, p = .01, d = 0.27) and to the face (difference = -2%, p = .01, d = 0.22) in the adolescent age class, and to the head (difference = -4%, p < .001, d = 0.57), the body (difference = 2%, p = .03, d = 0.26), the face (difference = -5%, p < .001, d = 0.55), and the upper face (difference = -4%, p < .001, d = 0.34) in the adult class - with participants with ASD looking on average less to the head/face and more to the body (Figure 1a). Examination of individual stimuli indicated differential sensitivity to group differences (Figure 1b). Analysis of temporal profiles of attention is ongoing.

Conclusions: We observed a tendency in the TD group to increased social attention with age. The ASD group, in contrast, maintained a flat profile. This suggests that reduced social attention is not an invariable feature of the ASD phenotype in childhood. On the other hand, social attention clearly deviates from typical development at later stages of development: the characterization of social attention indexed in this data may represent the consequences of living a life with ASD, with important consequences for intervention.