Understanding why some patients continue to experience delusions long after an acute psychotic episode — while others recover more fully — is one of the most pressing unanswered questions in schizophrenia care. A new computational approach may offer a measurable cognitive marker that tracks not just the presence of delusions, but their trajectory over time, potentially reshaping how clinicians monitor recovery.
This longitudinal study followed 75 participants with schizophrenia-spectrum disorder (SSD), recruited shortly after discharge from inpatient psychiatric hospitalization, alongside 71 non-clinical comparison participants over six months with assessments at six timepoints. Using a Hierarchical Gaussian Filter applied to a three-option probabilistic reversal learning task, researchers estimated each participant's volatility prior — a computational parameter reflecting how strongly an individual expects their environment to be unpredictable and rapidly changing. At baseline, SSD participants showed markedly elevated volatility priors relative to controls, alongside significantly higher paranoia and delusional ideation scores. Over six months, all three measures declined, yet none fully normalized in the SSD group. Crucially, changes in volatility priors tracked longitudinally with changes in delusion severity, suggesting a dynamic, state-like relationship rather than a fixed trait.
This finding matters because it repositions the volatility prior from a theoretical curiosity in computational psychiatry to a potential clinical monitoring tool. The Bayesian brain framework has long proposed that psychosis may reflect pathological inference — the brain over-updating its internal models in response to environmental signals. This study provides rare longitudinal evidence supporting that framing in a clinical cohort. Key limitations include the relatively modest sample size and the absence of a randomized intervention, meaning causal direction cannot be firmly established. Whether volatility priors could eventually serve as treatment-response biomarkers — helping clinicians distinguish patients likely to relapse from those on stable recovery trajectories — remains an open question. This is incremental but meaningfully directional evidence.