Keynotes | K&Ts | GACs | Talks | Posters | Search
Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms
Divisive normalization masquerades as predictive processing in visual cortex
Ningkai Wang1, Ralph Wientjens1,2,3, Jurjen Heij4,2, Gilles de Hollander5, Jan Theeuwes1,2,3,6, Tomas Knapen1; 1Vrije Universiteit Amsterdam, 2Royal Netherlands Academy of Arts and Sciences (KNAW), 3Institute Brain and Behaviour Amsterdam (iBBA), 4Spinoza Centre for Neuroimaging, 5University of Zurich, 6ISPA-Instituto Universitario
Presenter: Ningkai Wang
Neural responses in visual cortex are shaped by the temporal structure of sensory input: stimuli that conform to recent patterns are often suppressed, while surprising deviations elicit enhanced responses. These dynamics are typically attributed to predictive processing, in which the brain generates top-down expectations to minimize prediction error. However, the specific mechanisms underlying such modulations remain debated. Here, we combine ultra-high-field fMRI with a modified population receptive field (pRF) mapping paradigm to test whether temporal response modulations typically attributed to predictive processes can instead be explained by postdictive tuned normalization alone. We find a robust gradient across visual areas: early cortex exhibits response suppression, while higher areas show enhanced responses to unexpected stimuli. Crucially, this surprise-modulation gradient closely tracks spatial tuning, particularly pRF size and surround suppression strength. We show that these effects can be accounted for by feature-selective, delayed normalization without invoking top-down predictive mechanisms. Our findings offer a mechanistically grounded account of context-sensitive visual responses, reframing expectation effects as emergent properties of complex spatiotemporal tuning.
Topic Area: Computational Models of Vision & Visual Cortex