Antônio Mello¹, Chandana Kodiweera¹, Daniel Stehr¹, Krzysztof Bujarski¹, Brad Duchaine¹; ¹Dartmouth College

A reference frame (RF) is the spatial coordinate system in which a visual stimulus is represented. Many theories of object recognition have proposed the existence of object-centered RFs in the human ventral visual pathway, but evidence for them is limited. Here, we investigate the RFs underlying the representation of high-level categories in a rare case of hemi-prosopometamorphopsia. Nagel is a 40-year-old man who sees distortions on the right side of faces and, less frequently, on the same side of hands. We presented Nagel with 432 photographs of faces and hands (216 of each) that varied in visual field presentation, the part of the stimulus visible, picture-plane orientation (0°, 90°, 180°, 270°), and viewpoint. He reported what he perceived for each image. Analyses using Fisher's exact test (Bonferroni-corrected) revealed no significant difference in the frequency of distortions across visual fields for either faces or hands (p = 1.00). Nagel also consistently perceived distortions on the right half-faces and right half-hands but not on their left counterparts (p = .025). Together, these findings demonstrate that his condition does not result from disruptions to representations encoded within retino-centered RFs. Moreover, distortions were almost exclusively confined to the right half of faces, regardless of orientation or viewpoint (i.e., distorting the same set of facial features), indicating disruptions to an orientation- and view-independent object-centered RF – specifically, a face-centered RF. In contrast, distortions affecting hands were primarily localized to the right-hand side of stimuli across the same conditions (i.e., distorting the right side of the stimulus, even when hands were rotated), suggesting disruptions to a stimulus-centered RF instead. Overall, these findings provide additional evidence for the existence of face-centered RFs in human face processing and indicate that the ventral visual system likely employs multiple RFs to encode high-level object representations.

Acknowledgements: We thank the Hitchcock Foundation for funding this research.