Due to the broad etiological heterogeneity of autistic phenotypes, unitary theories of autism often fall short of describing the entirety, or even the majority, of the autism spectrum. Recent predictive coding hypotheses avoid this core issue by arguing autism is caused by altered parameters for updating mental models of the world, while leaving the details of those changes open. Attempts at building falsifiable predictive coding theories by specifying these details, such as the slow-updating and high-precision hypotheses of predictive coding in autism, have led to conflicting empirical results, implying they encounter the same dilemma as other unitary theories. Different prediction strategies may appear in different subsets of the heterogenous autistic population or be more relevant in specific tasks or contexts. In addition, there may be dissociations between prediction strategies and behavior in autism that lead to behavioral differences between autism and typical development without an altered underlying predictive mechanism. This project aims to assess multiple theories of predictive coding in a task-driven, non-social context and a task-free, social context. We use multiple measurement paradigms, including phenotypic characterization, eye-tracking, and functional neuroimaging, to obtain as complete a picture as possible of the entire cognitive mechanism, spanning multiple constructs across several domains of human functioning. This evidence will allow us to disentangle both individual-level heterogeneity in coding impairments from task- or context-level heterogeneity and dissociation between behavioral outcomes and neural correlates of predictive coding. Our work will contribute to our understanding of both the "how" and the "why" of temporal prediction by measuring how typical and atypical temporal prediction are encoded in the brain and variation in the links between temporal prediction and social behaviors across social behavioral phenotypes. Focusing on autism, which is typified by impairments in social function, allows us to determine how much temporal prediction is a direct factor in the ease with which individuals attain their desired social connectedness, or whether it is largely mediated through other cognitive constraints. More generally, this project will provide insight on what temporal prediction is for. In clinical knowledge and practice, this research will lead to improvements in our ability to precisely target interventions, particularly those involving structured sensory experiences, to specific patients by building on existing predictive coding capabilities to scaffold the development of social behaviors. It will also lead to assessments of these interventions' effects on cognitive mechanisms, rather than relying purely on behavioral or phenotypic outcomes. This will allow clinicians to more effectively capitalize on autistic individuals' existing skills to achieve their own social and relational goals. Such work is crucia...