Patterns of Visual Engagement Differ As a Function of Cognitive Profile in School-Aged Children with ASD

Background:   Heterogeneity in autism spectrum disorder (ASD) is an obstacle to advancements in identifying and treating causes of the disorder.  Measures capturing the core underlying features of ASD, such as reduced interest in socially adaptive stimuli, may provide a means for parsing phenotypic heterogeneity in ASD.  For example, previous research has revealed that the social adaptive value of where children looked when viewing social scenes differed significantly based on IQ profile (Rice et al., 2012). To further investigate how different patterns of visual engagement are related to cognitive functioning, the present study uses a novel approach for quantifying not only where a child is looking but also their level of engagementwith scene content.

Objectives:   Examine whether visual engagementwith scene content differs between subgroups of ASD characterized by different cognitive profiles. 

Methods:  School-age children with ASD (n=174) watched age-appropriate, socially relevant videos while eye-tracking data were collected. The ASD sample was divided into four subgroups as in Rice et al. (2012): Participants with a verbal IQ (VIQ) advantage (VIQ-NVIQ>12), a non-verbal IQ (NVIQ) advantage (NVIQ-VIQ>12), an even IQ profile and higher full-scale IQ (FSIQ), and an even IQ profile and lower FSIQ (see Table 1). Viewer engagement was quantified by measuring patterns of eye-blink inhibition, a method that capitalizes on the finding that people unconsciously adjust the timing of eye-blinks to minimize the likelihood of missing critical information (Shultz et al. 2011). Probabilistically, people are least likely to blink when highly engaged with what they are viewing and most likely to blink when less engaged. Permutation testing identified periods of statistically significant blink inhibition (indicating moments when children were highly engaged) and statistically significant increased blinking (moments when children were less engaged) for each subgroup separately. Percentage of visual fixation time on eyes, mouth, body, and object regions were calculated for each child.

Results:  Comparisons of visual fixation over the entire viewing session revealed that VIQ and high FSIQ subgroups looked more at mouths compared with the NVIQ subgroup (p<.01) (Figure 1A). No other differences were found. In contrast to the relatively similar patterns of visual fixation between subgroups, patterns of eye-blinking revealed striking differences in whensubgroups were highly engaged with scene content. Only 1.24% of highly engaging movie frames were perceived as engaging to all 4 subgroups. By contrast, 75.76% of highly engaging movie frames were perceived as engaging by only one subgroup (Figure 1B). These data suggest that subgroups likely engage with and experience these movies in very different ways. Ongoing analyses are aimed at further investigating the type of content that is perceived as engaging to each ASD subgroup.

Conclusions:  This study identified patterns of visual engagement in one of the largest eye-tracking samples of school-age children with ASD. Our findings demonstrate that patterns of engagement with social content are influenced by the cognitive profile of children with ASD. These measures provide a promising means for parsing heterogeneity in ASD and represent an important step towards developing interventions tailored to an individual’s specific learning style.