Circular inference in visual cortical hierarchies in schizophrenia

In schizophrenia, disrupted perceptual processes contribute to positive symptoms. However, it remains unclear how top-down priors and bottom-up likelihoods are represented across the hierarchy of visual cortices. To identify these processes within cortical hierarchies, the circular inference model offers a framework to account for the biased integration of priors and likelihoods that may underlie hallucinations and delusions.

Twenty-one participants, including 10 schizophrenia patients and 11 healthy controls, conducted a perceptual decision-making task while undergoing functional MRI (fMRI) scanning. Participants first received probabilistic hints about forthcoming objects, then viewed a blurred image, and finally answered the object identity and confidence using a slider. Subjects' performances were analyzed based on Bayesian perception models, and the variables representing top-down prior and bottom-up likelihood were used for fMRI data analysis.

Region of interest analysis revealed significantly enhanced top-down-related activity throughout the visual cortex in schizophrenia, compared with healthy controls (F(1,133) = 8.82, p = 0.003, ${\eta }^2$  = 0.059). In addition, exploratory ROI analyses found increased bottom-up-related activity in the early visual areas in individuals with higher delusion scores (V1: r = 0.41, p = 0.059, V2: r = 0.47, p = 0.030).

Using model-based fMRI analysis, these findings provide evidence for enhanced top-down activity across the visual hierarchy in schizophrenia. Exploratory analyses also suggested a potential association between bottom-up hypersensitivity and delusional ideation; however, this association did not survive multiple-comparison correction. Our results are consistent with altered circular information flow in the visual cortex and highlight the value of model-based neuroimaging for uncovering latent abnormalities in perceptual processing.