Predictive Processing: A New Lens on Psychosis That Unites Brain and Behavior
In recent years, the concept of predictive processing has surged to the forefront of cognitive neuroscience, promising to unify a wide range of brain functions under a single computational framework. This paradigm suggests that the brain does not passively receive sensory input but actively generates predictions about incoming data, constantly comparing these forecasts to actual sensory signals. Such a dynamic system enables efficient perception, learning, and action. However, its implications stretch far beyond normal cognition — opening new avenues to understand the perplexing phenomena of psychosis.
A groundbreaking narrative review by Goodwin, Diederen, Hird, and collaborators, soon to be published in Nature Mental Health, meticulously revisits and extends the seminal work of Sterzer and colleagues. It evaluates predictive processing as a potent model for elucidating the neurocomputational mechanics underpinning psychotic experiences, from mild manifestations in non-clinical populations to severe, chronic manifestations in schizophrenia. Through this examination, the authors aim to reconcile competing hypotheses about the origins of psychosis — whether disruptions arise predominantly from aberrant top-down expectations or bottom-up sensory abnormalities.
The review stresses the importance of priors and sensory likelihoods — two key elements of predictive processing theory — in understanding the disorder. Priors represent the brain’s pre-established beliefs or expectations, while likelihoods correspond to sensory evidence. Psychosis, it is posited, may emerge from either overly precise priors that overwhelm noisy sensory signals (a top-down disruption) or from diminished priors coupled with excessively precise sensory inputs (a bottom-up disruption). Reconciling these perspectives could demystify the variability observed across psychosis stages and presentations.
Psychotic phenomena manifest along a continuum. Non-clinical psychotic experiences, such as mild hallucinations or delusional thoughts that do not severely impair functioning, provide a window into early, subthreshold mechanisms of perceptual inference gone awry. Intriguingly, alterations in the balance between priors and sensory fidelity may underlie these benign symptoms, indicating that predictive processing dysregulation can exist even outside clinical diagnosis.
As psychosis progresses toward more severe clinical states — including those identified as high-risk or undergoing first-episode psychosis (FEP) — the interplay of predictive disruptions becomes more pronounced. Individuals in these stages show more robust deviations in how brain networks generate and update predictions, affecting perception and cognition. The review highlights experimental evidence showing that aberrant precision weighting of priors and sensory inputs correlates with symptom intensification and functional decline.
In established schizophrenia, the model continues to hold explanatory power. Aberrant predictive coding manifests as distorted perceptual experiences — hallucinations and delusions — and cognitive disturbances. The review carefully synthesizes recent neuroimaging and behavioral studies linking these symptoms to quantifiable shifts in predictive hierarchy function. Notably, it underscores that these disturbances vary within and across patients, supporting a transdiagnostic approach rather than a monolithic disease model.
A major strength of this review lies in its ambition to bridge the gap between top-down and bottom-up views. Historically, debates centered on whether psychosis arises because strong, inaccurate priors swamp the brain’s interpretation of sensory data, or conversely, because overly noisy priors fail to constrain hyper-salient sensory influx. By parsing the nuanced dynamics among priors and likelihood precision across psychosis stages, Goodwin and colleagues provide a synthesized framework that embraces both sides — illustrating that psychosis may manifest differently across individuals and illness phases, shaped by varied combinations of predictive disturbances.
Beyond framing psychosis as a disorder of predictive processing, the review explores the promise of this framework in clinical translation. Predictive processing metrics might serve as biomarkers for earlier detection, risk stratification, and treatment efficacy evaluation. For instance, deficits in sensory attenuation or abnormal belief updating patterns detected via computational behavioral paradigms could identify at-risk individuals long before overt symptoms emerge, facilitating preventive interventions.
Therapeutic potentials abound as well. Understanding the computational origins of psychotic symptoms opens the door to new interventions aimed at recalibrating predictive coding mechanisms. Cognitive remediation therapies could be tailored to modulate the precision of priors or sensory evidence weighting. Furthermore, neuromodulatory techniques such as transcranial magnetic stimulation might precisely target brain circuits involved in aberrant predictions.
The authors emphasize, however, that the field is still in its early days regarding clinical application. They call for standardized experimental paradigms that rigorously quantify predictive parameters across diverse populations and longitudinal designs that monitor how predictive disruptions evolve over the course of illness. Such rigor is essential to move predictive processing from theoretical promise to clinical reality.
This comprehensive review reinvigorates the vision of predictive processing as a unifying lingua franca for psychiatric neuroscience. By aligning computational complexity with clinical phenomena, it pushes the boundaries of how we understand psychosis — not simply as a constellation of incoherent symptoms, but as the direct consequence of fundamental disruptions in brain function. This framework invites a paradigm shift: from symptom-based diagnoses to mechanistic, model-driven approaches.
In doing so, the review helps demystify why psychosis appears so heterogeneous and resistant to conventional treatment. If disrupted predictive coding lies at its heart, resolving psychosis will require nuanced interventions that restore balanced brain inference, individualized to the profile of disruption. This promises a future where precise, mechanism-informed therapies replace trial-and-error approaches, improving outcomes markedly.
Moreover, the framework transcends psychosis, offering insight into a range of psychiatric conditions characterized by altered perception and cognition — from mood and anxiety disorders to autism spectrum conditions. The authors underscore predictive processing’s potential as a transdiagnostic scaffold, deepening our grasp of mental illness complexity.
Ultimately, this narrative marks a decisive step in harmonizing decades of behavioral, neurobiological, and computational research. It demonstrates the power of predictive processing to unify disparate empirical findings into coherent neurocomputational models, empowering researchers and clinicians to think beyond traditional boundaries.
As good science should, this review leaves readers with both answers and questions, charting a path for future inquiry. By integrating rigorous theoretical perspectives with cutting-edge experimental data, it inspires renewed efforts to unravel the neurocomputational roots of psychosis and leverage these insights for transformative clinical gains.
With mental health challenges mounting worldwide, such innovative approaches could not be more timely. The fusion of computational neuroscience and psychiatry embodied here offers hope for profound advancements in our understanding, diagnosis, and treatment of psychotic disorders. The predictive brain, it seems, may hold the key to unlocking the mysteries of psychosis — and opening a new era in psychiatric care.
Subject of Research: Predictive processing and its role in the neurocomputational mechanisms underlying psychosis across different stages of illness.
Article Title: Predictive processing accounts of psychosis: bottom-up or top-down disruptions.
Article References:
Goodwin, I., Diederen, K.M.J., Hird, E.J. et al. Predictive processing accounts of psychosis: bottom-up or top-down disruptions. Nat. Mental Health (2026). https://doi.org/10.1038/s44220-025-00558-5
Image Credits: AI Generated
