Divergent Patterns of Speaker-Intention Discrimination and Emotional Valence Ratings in Autistic and Non-Autistic Adults

Autistic people often experience challenges in interpreting nonliteral language, such as sarcasm, teasing, and prosocial lies, which rely on speaker-intention discriminations. Prior research on speaker-intention discriminations in autism has often employed static stimuli that lack ecological validity, did not capture multimodal communication cues, and primarily focused on children. The present study examined how autistic and non-autistic adults infer speaker intentions and emotional valence in dynamic, naturalistic videos. Participants (N = 73) viewed 50 brief video vignettes from the Relational Inference in Social Communication database, depicting literal, sarcastic, teasing, or prosocial lie scenarios. After each video, they identified the speaker’s communicative intent and rated the emotional response of the addressee. Participants also completed two measures of autistic characteristics (Autism Spectrum Quotient; Comprehensive Autistic Trait Inventory). Autistic participants showed lower speaker-intention discrimination than non-autistic participants, whereas valence ratings differed modestly by group but followed broadly similar patterns across intentions. Autistic traits predicted lower speaker-intention discrimination in non-autistic adults but were unrelated (Comprehensive Autistic Trait Inventory) or positively associated (Autism Quotient) in autistic adults. These findings highlight both shared and differing aspects of speaker-intention discriminations and happiness rating across groups and support the value of combining categorical and trait-based approaches to better understand individual differences in social communication.