Autism spectrum disorder is a neurodevelopmental condition involving social-communicative difficulties. Autistic individuals and their parents exhibit unique social attention patterns with links to social communication, implicating it as a potential neurocognitive marker for autism. Brain responses related to social skills can change in response to experimental manipulation, offering exciting prospects for altering social attention. Several randomized clinical trials (RCTs) have demonstrated enhanced social competence in autistic youth following social skills training (SST). Yet, no study has examined how visual/social attention may change/predict response to SST in autistic individuals. Developing a comprehensive social attention metric from eye-tracking variables during a social communication task within a SST RCT in autism could identify novel neurocognitive outcomes for social learning, with genetic and clinical significance.

Participants (N=35, 94% male, 47% reported non-Caucasian, 32% reported Hispanic, verbally fluent, 7-11-years) were randomly assigned to 1) treatment (intranasal oxytocin + SST group) or 2) control (facilitated play) for 12 weekly sessions. Following prior work (Nayar et al., 2022), visual attention characterizing where and how participants looked towards social/non-social information during a social-emotional narrative task was assessed using dynamic analytic methods (e.g., perseverative/regressive refixations, scene exploration), reduced via a principal component analysis (PCA). Narrations were transcribed and computationally-derived linguistic markers of core narrative features were extracted (e.g., word count, causality, cognitive/affective words). Data were collected at baseline and week 12 (treatment endpoint); with eye-tracking variables also available at follow-up (week 16). Linear mixed-effects models captured change over time between groups. Links between visual attention, social-behavioral/cognitive treatment outcomes per prior work (Soorya et al., 2015), and narrative variables were assessed using Pearson correlations.

Compared to controls, the treatment group demonstrated decreased social attention (PCA score) following treatment (p<.01), driven by reduced perseverative/regressive social refixations (ps<.05) particularly at maintenance (ps<.05). In contrast, increased exploration of social and non-social information was found in the treatment versus control group (p=.08). Similarly for narrations, only the treatment group produced longer narratives at endpoint relative to baseline (p<.01). An overall time effect revealed increased use of causal statements over time across groups (p<.05). Finally, although no significant correlations emerged between gaze and narrative metrics, both groups showed a positive association between visual exploration and social-cognitive outcome at maintenance (p<.01).

This study provides initial evidence supporting the malleability of social-communicative mechanisms post-SST in autism. The treatment group exhibited reduced perseverative/regressive social refixations, coupled with increased exploration across social/non-social stimuli and longer narratives (indicative of enriched storytelling). Notably pronounced during the maintenance phase, findings suggest 1) delayed augmentative effects of oxytocin on visual attention (Soorya et al., 2023) and 2) a potential shift from local to global perceptual strategy underlying treatment-induced changes in visual attention and narration. Associations between better social-cognitive outcomes and greater scene exploration further underscores the treatment’s potential to enhance global perceptual processes. Overall, this work highlights the utility of visual attention and narrative features as indicators for tracking SST outcomes in autism, prompting future investigations to disentangle local/global processing's role in downstream social-cognitive changes.