Abstract: Prediction is a fundamental brain function and a central mechanism by which individuals adapt to their environment. Predictive Coding Theories (PCT) is an essential family of theories of cortical function, providing a conceptual framework that allows us to explain how a brain constantly makes predictions based on prior belief about sensory input. Our behavior, perception, emotion, attention, and learning are all deeply impacted by the brain's prediction systems. The family of theories originated from the studies focusing on brain functions in typical developmental (TD) individuals and was subsequently introduced to understand the neural processes of psychosis, i.e., psychotic symptoms and psychotic disorders. In recent years, researchers have also applied PCT to explain the multi-domain deficits in perceptual-motor, cognitive-learning, and social-verbal domains in individuals with autism spectrum disorder (ASD), i.e., the Predictive Impairment in Autism (PIA). This theoretical hypothesis proposes that cognitive dysfunction in individuals with ASD may stem from an inability to build and update internal models to predict future events. As PCT consists of three inference models, including Bayesian inference, hierarchical inference, and active inference, many researchers have proposed different hypotheses based on different inference models. Pellicano and Burr (2012) proposed the Hypo-priors hypothesis according to the Bayesian inference model, which suggests that attenuated Bayesian priors may be responsible for the unique perceptual experience of individuals with ASD, leading them to perceive the world more accurately rather than modulated by prior experience. On the other hand, based on the hierarchical inference model, Vande Cruys et al. (2014) situated the core deficit of individuals with ASD in the high, inflexible precision of prediction errors (HIPPEA) from the perspective of environmental volatility. The reason why the hypothesis has aroused the interest of many researchers is that it could provide a deeper mechanistic account of ASD by integrating the multi-domain deficits with two core symptoms of social impairments and stereotyped behavior into a unified theoretical framework. This could contribute to the development of better therapies, accommodations, and even refined diagnostic criteria and tools. It also has the potential to link the study of ASD with research on the neuroscience of prediction, informing a new understanding of the neurobiological differences that lead to ASD and motivating the exploration of new targets for neuroactive medicines. Although the theory provides a credible and unified explanatory framework for ASD, empirical evidence does not consistently support it. In addition, there is inconclusive information about the underlying mechanism. It is unclear whether it results from an abnormality in bottom-up feed-forward connectivity due to alterations in the neuromodulatory systems, or an impairment in top-down feedback connectivity due to the dysfunction of predictive brain regions.Therefore, based on the theoretical overview of PCT, this paper systematically reviewed the evidence for and against the “Hypo-priors” and “HIPPEA” hypotheses in the three domains of perceptual-motor, cognitive-learning, and social-verbal, presenting the panorama of the current research situation in this field. Then, we explained how abnormal feed-forward and feedback connectivity may lead to predictive deficits in ASD. In addition, we clarified these concepts in the PCT, such as sensory input weight, prior belief weight, prediction error weight, prediction error accuracy, learning rate, sensory input accuracy, and prior belief accuracy. This paper clarified the relationship among these theoretical hypotheses, such as the PCT, Bayesian inference, hierarchical inference, active inference, PIA, Hypo-priors, HIPPEA, and abnormal feed-forward and feedback connectivity. Most importantly, this paper distinguished the abnormality of weight between sensory input and prior belief and the imbalance of precision between them. Although the Hypo-priors and HIPPEA are based on different models, they excessively impoverished the weight of sensory input. The abnormality of weight is due to the imbalance of precision. The latter is the potential mechanism of prediction deficits in ASD.

Key words: autism spectrum disorder, predictive coding, prior prediction, sensory input, feedforward/feedback connectivity