A modality-general bias towards meaning, not enhanced prior knowledge, is linked to hallucination-proneness in the general population
Abstract
Predictive processing theories propose that hallucinations arise from excessive influence of prior expectations on perception. Although these theories implicitly generalise across modalities, previous findings suggest potentially different mechanisms in vision and audition. In this study, we tested predictive processing accounts in the visual and auditory modalities in 177 healthy adults using well-matched prior knowledge tasks. Participants discriminated degraded meaningful targets from control stimuli before and after exposure to clear template stimuli, providing explicit prior knowledge. Signal detection analyses quantified pre-exposure response bias (criterion, C) and improvements in discrimination sensitivity (d′) following exposure. Hallucination-proneness was measured using the Cardiff Anomalous Perceptions Scale (CAPS). Prior knowledge improved discrimination equally across modalities, but individual differences in this benefit were unrelated to hallucination-proneness, with Bayesian analyses supporting the null hypothesis. Instead, hallucination-proneness was associated with a more liberal pre-exposure response bias in both modalities. When entered together in regression models, response bias was a stronger predictor of hallucination-proneness than prior knowledge benefit in both modalities. These findings suggest that hallucination-proneness is characterised by a modality-general bias towards meaningful interpretations of ambiguous sensory input, rather than enhanced directed use of prior knowledge.
Related
- Hallucination-proneness is associated with more extreme social evaluations of external voices
- HalluciGuard-eye: knowledge graph-enhanced retrieval-augmented generation for hallucination detection and mitigation in ophthalmic multimodal large language models
- Ultrasound-based breast nodule malignancy diagnosis using CNN enhanced by knowledge distillation
- Examining Physiological Hyperarousal in the Relationship Between Insomnia and Nightmare Proneness
- Consumption and knowledge of protein supplements in the Spanish adult population: A cross-sectional survey-based study.
BibTeX
@misc{hammers2026modalitygeneral,
title = {{A modality-general bias towards meaning, not enhanced prior knowledge, is linked to hallucination-proneness in the general population}},
author = {Freya Hammers and Xiaoran Wu and Ben Alderson-Day and Chelsi Clarke and Eamon Blake and Samuel Evans},
howpublished = {PsyArXiv},
year = {2026},
month = oct,
doi = {10.31234/osf.io/acwsy_v1},
url = {https://osf.io/preprints/psyarxiv/acwsy_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A modality-general bias towards meaning, not enhanced prior knowledge, is linked to hallucination-proneness in the general population
AU - Freya Hammers
AU - Xiaoran Wu
AU - Ben Alderson-Day
AU - Chelsi Clarke
AU - Eamon Blake
AU - Samuel Evans
PY - 2026
DA - 2026/10/07
PB - PsyArXiv
DO - 10.31234/osf.io/acwsy_v1
UR - https://osf.io/preprints/psyarxiv/acwsy_v1/
AB - Predictive processing theories propose that hallucinations arise from excessive influence of prior expectations on perception. Although these theories implicitly generalise across modalities, previous findings suggest potentially different mechanisms in vision and audition. In this study, we tested predictive processing accounts in the visual and auditory modalities in 177 healthy adults using well-matched prior knowledge tasks. Participants discriminated degraded meaningful targets from control stimuli before and after exposure to clear template stimuli, providing explicit prior knowledge. Signal detection analyses quantified pre-exposure response bias (criterion, C) and improvements in discrimination sensitivity (d′) following exposure. Hallucination-proneness was measured using the Cardiff Anomalous Perceptions Scale (CAPS). Prior knowledge improved discrimination equally across modalities, but individual differences in this benefit were unrelated to hallucination-proneness, with Bayesian analyses supporting the null hypothesis. Instead, hallucination-proneness was associated with a more liberal pre-exposure response bias in both modalities. When entered together in regression models, response bias was a stronger predictor of hallucination-proneness than prior knowledge benefit in both modalities. These findings suggest that hallucination-proneness is characterised by a modality-general bias towards meaningful interpretations of ambiguous sensory input, rather than enhanced directed use of prior knowledge.
ER -