Conflict, not category: What the pupil tracks in face pareidolia
Abstract
Illusory faces, the faces people see in inanimate objects, share the low-level image statistics of the objects they are made of while, nonetheless, being perceived as faces. Neural representations of illusory faces shift from face-like to object-like within roughly 200 ms, and observers explicitly categorise them as objects, yet the illusion remains compelling. We asked whether the pupil, a peripheral index of both light and arousal, is sensitive to this dual identity. Forty participants viewed sequences of human faces, illusory faces in objects, and matched non-face objects (100 images each) while performing an orthogonal task at fixation. Human faces and objects both evoked pupil constriction, whereas illusory faces produced dilation beginning at approximately 350 ms. Thus, rather than falling between faces and objects or resembling either category, illusory faces elicited a qualitatively different pupil response. This dissociation remained after controlling for luminance and emotional valence, despite luminance being the strongest determinant of pupil size. Critically, within the illusory images, the dilation scaled with independent measures of the strength of the face percept, including face-likeness and the probability of categorisation as a face, and did not diminish across repeated presentations. These findings are consistent with the pupil tracking the conflict between the object and face interpretations, rather than stimulus category per se. They suggest that although neural representations quickly settle on an object's veridical identity, the competing illusory face interpretation persists, revealing a physiological signature of the conflicting perceptual identities that define face pareidolia.
Related
- Bridging Sensory Extraction and Decision Dynamics: Face Pareidolia Emergent from Bottom-Up and Top-Down Interactions
- Emotional arousal predicts pupil responses following a single attention shift between bright and dark moving stimuli
- Pupil size and pupil change correlate with distinct cortical neurochemical systems across the adult lifespan
- Limitations of pupil diameter as a proxy for modes of locus coeruleus activity
- Open filament tracks drive Bacteroidota gliding motility
BibTeX
@misc{robinson2026conflict,
title = {{Conflict, not category: What the pupil tracks in face pareidolia}},
author = {Amanda Robinson and Jing Liang and Blake William Saurels and Jessica Taubert},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/wz3r7_v1},
url = {https://osf.io/preprints/psyarxiv/wz3r7_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Conflict, not category: What the pupil tracks in face pareidolia
AU - Amanda Robinson
AU - Jing Liang
AU - Blake William Saurels
AU - Jessica Taubert
PY - 2026
DA - 2026/09/04
PB - PsyArXiv
DO - 10.31234/osf.io/wz3r7_v1
UR - https://osf.io/preprints/psyarxiv/wz3r7_v1/
AB - Illusory faces, the faces people see in inanimate objects, share the low-level image statistics of the objects they are made of while, nonetheless, being perceived as faces. Neural representations of illusory faces shift from face-like to object-like within roughly 200 ms, and observers explicitly categorise them as objects, yet the illusion remains compelling. We asked whether the pupil, a peripheral index of both light and arousal, is sensitive to this dual identity. Forty participants viewed sequences of human faces, illusory faces in objects, and matched non-face objects (100 images each) while performing an orthogonal task at fixation. Human faces and objects both evoked pupil constriction, whereas illusory faces produced dilation beginning at approximately 350 ms. Thus, rather than falling between faces and objects or resembling either category, illusory faces elicited a qualitatively different pupil response. This dissociation remained after controlling for luminance and emotional valence, despite luminance being the strongest determinant of pupil size. Critically, within the illusory images, the dilation scaled with independent measures of the strength of the face percept, including face-likeness and the probability of categorisation as a face, and did not diminish across repeated presentations. These findings are consistent with the pupil tracking the conflict between the object and face interpretations, rather than stimulus category per se. They suggest that although neural representations quickly settle on an object's veridical identity, the competing illusory face interpretation persists, revealing a physiological signature of the conflicting perceptual identities that define face pareidolia.
ER -