Predictive Processing In Hallucination Proneness: Perceptual vs. Cognitive Expectations - Validation Study
Year: 2026
Type: preprint
Abstract: This study constitutes the second part of a "discovery–validation" approach as a response to replication crisis. It comprises an independent dataset with a study procedure identical to our discovery study preregistered in June 2025 (doi.org/10.17605/OSF.IO/QCTUJ). The discovery study was analyzed prior to preregistering the validation study to inform a refinement of our primary analyses. Sample size estimation for this study is based on results of the mentioned previous study (discovery sample). All hypotheses remain identical. Instead of referring the reader to the mentioned preregistration, we here repeat the study description: Theoretical Background: Bayesian Predictive Processing (BPP) provides a promising framework for explaining hallucinatory experiences. According to BPP, conscious perception arises from the probabilistic integration of bottom-up sensory input with top-down prior expectations (priors). Hallucinations may result from an aberrant weighting of these priors during perceptual inference. However, it remains unclear whether individuals prone to hallucinations exhibit strengthened (e.g., Powers et al. 2017; Teufel et al. 2015) or weakened (e.g., Eckert et al., 2023; Weilnhammer et al., 2020) priors. A hierarchical extension of the BPP framework proposes that this imbalance may not be uniform across levels of representation. Specifically, perceptual (implicit) priors may be attenuated, whereas cognitive (explicit) priors may be exaggerated in individuals with hallucinatory tendencies (Corlett et al., 2019; Sterzer et al., 2018). The goal of the study is to empirically test his hypothesis by separately assessing the influence of perceptual and cognitive priors in a healthy sample varying in hallucination proneness. This will be achieved through two experimental paradigms designed to isolate perceptual versus congitive prior contributions to ambiguous stimulus interpretation. Study: Perceptual priors are operationalized through a serial dependence paradigm, where previous choices inform current decisions under high perceptual uncertainty. Participants will discriminate between house and face stimuli embedded in white noise at four evidence levels, presented for 500 ms. Stimuli will be shown in ten blocks of 40 trials. Cognitive priors will be operationalized by a learning paradigm, in which participants will be presented with two-tone (so-called "mooney") images of people and animals before and after a learning phase displaying the respective grayscale images, providing the cognitive priors. The degraded images will be shown for 800 ms and grayscale images for 1500 ms. The experiment consists of seven blocks with 40 trials each. In both experiments, participants will respond to stimuli by indicating their choice via key press on a response screen that follows each stimulus presentation. The response mapping will be orthogonalized to avoid systematic biases. Participants have a two-second response window before the trial is counted as a timeout. The interval between the response screen and the next trial will be jittered between 750 and 1500 ms (average 1000 ms). All categories (and difficulty levels in experiment 2) are displayed equally often per block. Hypotheses: We expect that individuals with higher proneness to hallucinations will show weaker weighting of perceptual priors (indicated by a smaller bias towards the previous decision in the current trial) and stronger weighting of cognitive priors (indicated by a larger increase in recognition of two-tone images after the learning phase). Results will help integrate conflicting findings on the role of priors in hallucinations using a theory-driven approach. Investigating a non-clinical population is intended to contribute to a better understanding of the development of psychotic symptoms.
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7136281166