Understanding the neural basis of the social difficulties in Autism Spectrum Disorder (ASD) is essential to develop effective interventions, test their efficacy and predict individual treatment outcome. Cognitive neuroscience offers a key level of analysis towards this endeavor by bridging the gap between molecular and cellular mechanisms to the behavioral phenotype.

            One hallmark symptom in Autism Spectrum Disorder (ASD) is disproportionate social impairments, including difficulty processing emotional information from faces. Despite an enormous body of research on this matter the results are often weak and conflicting. The neural basis for the affected behavior in ASD remains unclear. Here, we present a new approach focusing on naturalistic dynamic social contexts with the goal of finding robust, replicable neural correlates of difficulties in social information processing in ASD that generalize across individuals and stimuli.

            We investigated whether information in patterns of neural responses to emotional faces represent i) the valence of emotional facial expressions, ii) task of the subject and iii) the flexibility of neural patterns to task demands. In a first study, we measured blood-oxygen-level-dependent signal (BOLD) responses while neurotypical subjects watched short naturalistic movie clips of dynamic positive and negative facial expressions. For each movie clip, subjects were instructed to judge either the person's age or the valence of their emotional expression. Split-half multivoxel pattern analyses (MVPA) suggest that the task of the subject (attending to age vs emotion) could be robustly decoded from patterns in face-responsive regions (bilateral anterior and posterior superior temporal sulcus (STS), fusiform face area (FFA) and medial prefrontal cortex (MPFC)), independent of the stimulus. The property of the stimulus (valence of the facial emotion) was represented in STS and MPFC (replicating prior reports), although weaker than the task representation.

            In an ongoing follow up study, we investigate differences in patterns of neural responses in ASD versus neurotypical subjects. Preliminary results suggest aberrant task representation in MPFC in ASD. Additionally, a region in right STS showed reduced flexibility in neural activity to different task demands: while neurotypical subjects showed increased valence decoding when judging emotional valence versus age of facial expressions, this modulation was absent in ASD. I will discuss whether atypical cortical representation of social information in ASD might reflect reduced neural flexibility to optimally integrate task demands and perceptual input.