Understanding Cognitive Behavioral Therapy for Psychosis Through the Predictive Coding Framework

By Julia M. Sheffield, Aaron P. Brinen, Brandee Feola, Stephan Heckers, and Philip R. Corlett in Biological Psychiatry: Global Open Science

---

Introduction

This article seeks to explain why Cognitive Behavioral Therapy for psychosis (CBTp) helps reduce persecutory delusions and how it produces change at a cognitive and neurobiological level. Although CBTp is recognized as an effective treatment, its cognitive and neurobiological mechanisms have not been well understood. The authors argue that the predictive coding framework offers a useful explanation. Predictive coding views the brain as a system that constantly generates expectations and updates them based on sensory information. When this process becomes disrupted, as seen in psychosis, neutral experiences may appear threatening and highly precise delusional beliefs can form to make sense of confusing sensory signals. The article examines how CBTp may support recovery by changing the way people gather and interpret the information that shapes their beliefs.

Research approach

The paper is a conceptual review that brings together research from CBTp, predictive processing, Bayesian learning theories, and cognitive neuroscience. It also uses David Marr’s three-level framework to describe how psychological processes operate at different layers. The authors argue that CBTp acts mainly at the level where beliefs are updated and interpreted rather than at the level of neural implementation or high-level therapeutic goals. Through this lens, CBTp can be understood as a method that adjusts the quality and meaning of sensory data that patients rely on when forming beliefs about their safety and their environment.

Discussion

The article explains that early versions of CBTp attempted to directly challenge delusional beliefs, which often led to resistance or reinforcement of the belief. Modern CBTp instead focuses on the psychological factors that maintain persecutory thinking, such as worry, negative self-beliefs, anomalous sensory experiences, sleep difficulties, reasoning biases, and safety behaviors like avoidance. Therapies such as the Feeling Safe Programme target these factors with structured exercises designed to generate new, meaningful experiences.

Classical models based on fear extinction do not fully explain psychosis. People with schizophrenia often show fear responses to neutral cues and struggle to learn that certain situations are safe. Their beliefs remain unusually stable and resistant to new information. Predictive coding offers a more complete framework. It proposes that disrupted prediction error signals create noisy or overly salient sensory experiences that seem significant even when they are harmless. In response, the brain forms strong beliefs to explain these sensations. Once established, these beliefs shape perception and are difficult to update.

CBTp allows to improve belief updating by changing the sensory information patients attend to. Reducing worry limits threatening mental images. Strengthening positive self-beliefs shifts interpretations of everyday events. Reappraising hallucinations reduces their influence on belief formation. Dropping safety behaviors provides opportunities for new, disconfirming experiences through sampling new sensory data. Over time, repeated exposure to safer interpretations allows alternative beliefs to become stronger while the persecutory belief loses its dominance.

Overall, the article argues that CBTp is effective not because it directly challenges delusions, but because it reshapes how people process and interpret sensory information, allowing new and safer beliefs to outweigh delusional beliefs.