No evidence for a generalised predictive processing deficit in psychosis: A meta-analysis

Abstract Background The influential predictive processing theory of psychosis posits that symptoms arise from an imbalanced weighting of predictions (priors) and sensory evidence, yet studies present conflicting results. This is particularly problematic as findings from single tasks with modest sample sizes are frequently used to advance a theory for a generalized altered reliance on priors in psychosis.

Methods: This study presents a random-effects, multi-level meta-analysis (PROSPERO CRD42024574379) evaluating evidence for aberrant reliance on priors in psychosis across perceptual tasks. The search identified articles in Embase, MEDLINE, APA PsycINFO, and APA PsycArticles published between 1st January 2005—31st October 2024, with risk of bias assessed using the Newcastle-Ottawa Scale. Included articles (34 results from 27 studies) compared adults with schizophrenia-spectrum psychosis (SZ; n = 904) to healthy controls (n = 1,039) on behavioral measures representing reliance on priors.

Results: Results provided no evidence for atypical reliance on priors in psychosis (g = .03, 95% CI [−0.27, 0.34]; p = .818) or associations with delusions (6 results; SZ = 183; r = −.16, 95% CI [−0.51, 0.19]; p = .293) or hallucinations (10 results; SZ = 370; r = .04, 95% CI [−0.28, 0.36]; p = .780). In contrast with the theory that psychosis differentially affects priors at different levels of the cognitive hierarchy, a sub-group analysis indicated that a two-level hierarchical model of priors did not reconcile conflicting results (F (1,32) = 0.1, p = .758).

Conclusions: These findings do not suggest that psychosis is associated with a generalized predictive processing deficit spanning multiple aspects of perception.