The Predictive Coding Theory of Autism

Predictive Coding is a neurocognitive theory that is grounded in Bayesian principles. According to this theory, the brain continuously generates and updates an internal model of the world¹. This model is used to anticipate incoming information (e.g., visual, auditory, verbal, and/or tactile), which is then compared with the sensory signals received. Mismatches between predictions and actual input, known as “prediction errors,” are used to update and refine future predictions. This allows the brain to generate expectations that can be generalised to new situations. Studies suggest that individuals on the autism spectrum have atypical predictive coding processing, which may help explain why they experience various sensory, perceptual, and social difficulties.

Predictive Coding in Neurotypical Individuals

The human brain has evolved to be highly efficient at processing information about the surrounding environment. Instead of reacting to every detail of incoming stimuli, it generates expectations about what is likely to happen next based on prior experience. When sensory input largely matches these expectations, neural processing is typically unconscious, smooth, and automatic. The brain can tolerate minor discrepancies between expectations and sensory input (e.g., the sound of a dog barking in the distance), allowing people to navigate complex and dynamic environments without being overwhelmed by constant prediction errors. However, when there is a significant mismatch between expectations (top-down predictions) and sensory input (bottom-up sensory signals), the unexpected information becomes more salient, recruiting additional attentional and cognitive resources while the brain updates its internal model (schema) to reduce the likelihood of future discrepancies. This neurocognitive mechanism enables efficient interaction with the environment by facilitating processing in areas such as:

• Perception: Filtering sights, sounds, and touch sensations to avoid becoming overwhelmed by minor or irrelevant details.
• Experiential Learning: Applying knowledge from past experiences (representations) to predict future situations and adapt behaviour accordingly.
• Social Interaction: Anticipating others’ actions, emotions, and intentions, allowing for effective communication and social interactions.

The balance between prior expectations and incoming stimuli is highly important. Prior expectations facilitate interpretation and the filtering of incoming information, while prediction errors help refine and update these expectations.

Atypical Predictive Coding in Autism

Research suggests that atypical predictive coding mechanisms in autism may contribute to some of the core difficulties experienced by those with the condition²:

1. Heightened Sensory Sensitivity: Sensory experiences may be perceived as unusually intense because the brain assigns excessive importance to minor prediction errors (e.g., the sound of a dog barking in the distance), thereby reducing reliance on prior expectations when interpreting incoming information. Visual stimuli, bright lights, sounds, or textures that neurotypical individuals readily filter out may be processed as highly salient by autistic individuals, potentially contributing to sensory overload and feelings of being overwhelmed.
2. Detail-Focused Cognition: With less input from prior expectations, the brain may focus more on peripheral sensory information. This can result in strong attention to detail and pattern-recognition, while making it harder to integrate information into broader contexts—such as grasping the overall meaning and nuances in situations, or seeing the “bigger picture”.
3. Difficulty with Change and Novelty: Unfamiliar or unexpected events may produce stronger prediction errors, causing anxiety or distress. This may help explain why individuals on the autism spectrum have a need or desire for predictability and routine.
4. Social Communication Challenges: Social interactions involve rapidly integrating and interpreting multiple cues such as facial expressions, gestures, vocal tones, semantic information and context. Atypical predictive coding can make it difficult for individuals with autism to form and update predictions about other peoples’ behaviours, which in turn may make it harder for them to predict what is likely to occur in social situations³. As a result, social interactions require greater conscious effort to anticipate and interpret others’ mental states, emotions, and intentions.
5. Cognitive Load and Fatigue: Because sensory input in new situations requires more deliberate conscious processing, with prediction errors occurring more frequently, novel social interactions or other new experiences can be overwhelming and mentally exhausting for individuals on the autism spectrum.

Implications for Intervention and Support

The Predictive Coding theory proposes that differences in generating and updating predictions can influence early learning, sensory processing, and social engagement in individuals on the autism spectrum. These insights highlight the importance of early interventions and supports that both stabilise expectations and promote adaptive learning. Interventions that provide structured and predictable environments and routines, in conjunction with gradual exposure to novel situations can help facilitate the individual’s ability to cope more effectively with new or novel situations when they occur⁴.

– The Autism Hub Team, 05/06/2026

References

1. Clark A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 181-204. doi:10.1017/S0140525X12000477
2. Van de Cruys, S., Evers, K., Van der Hallen, R., Van Eylen, L., Boets, B., de-Wit, L., & Wagemans J. (2014). Precise minds in uncertain worlds: Predictive coding in autism. Psychological Review, 121(4), 649-675. doi: 10.1037/a0037665
3. Keysers, C., Silani, G., & Gazzola, V. (2024). Predictive coding for the actions and emotions of others and its deficits in autism spectrum disorders. Neuroscience & Biobehavioral Reviews, 167:105877. doi: 10.1016/j.neubiorev.2024.105877
4. Haker, H., Schneebeli, M., & Stephan, K. E. (2016). Can Bayesian theories of autism spectrum disorder help improve clinical practice? Frontiers in Psychiatry, 7:107. doi: 10.3389/fpsyt.2016.00107

Author: autismhub