Neural networks underlying visual illusions: an activation likelihood estimation analysis

Visual illusory configurations have long been used to unveil the mechanisms underlying visual perception and the integration of contextual information. Neuroimaging studies employ different classes of optical illusions to identify which areas of the visual pathway are involved in visual perception and how these interact in feedback and feedforward loops to integrate specific visual features in a univocal percept. Here, we conducted a systematic review and meta-analysis of fMRI studies that contrasted activations in response to illusory contours, geometrical and motion illusions in order to reveal the cortical regions involved in the processing of illusory percepts, thus revealing the networks contributing to the integration of visual contextual information. Furthermore, we pooled the studies into two subgroups to reveal the areas of overlapping activation related to static visual illusions (e.g., Kanizsa figures, Müller Lyer, Ebbinghaus) and motion illusions (e.g., illusory self-motion, stereokinetic effect, Pinna illusion). The resulting networks represent the neural signatures of illusory processing and allow us to investigate the functional distribution of perceptual processing across the ventral and dorsal visual streams, observing the degree to which the two streams interact when building a visual percept. The Activation Likelihood Estimation (ALE) meta-analysis, conducted on 41 experiments taken from 19 studies emerging from the systematic review for a total of 243 foci, revealed a bilateral network of visual areas encompassing both ventral and dorsal visual regions, including the inferior and middle occipital cortices bilaterally and the right superior parietal gyrus. Furthermore, we used a meta-analytic connectivity modeling approach to explore the functional connectivity of the LOC, which resulted from the conjunction analysis as the only region shared between the static and motion illusion networks. Based on these results and previous connectivity accounts, we describe the networks of areas involved in the perception of contextual visual illusions, laying the foundation for a neural based classification system of illusions. Overall, these results describe a network of areas crucially involved in perceptual inference relying on feedback and feedforward interactions between areas of the ventral and dorsal visual pathways. The same network is proposed to be involved in hallucinogenic symptoms characteristic of schizophrenia and other disorders, with crucial implications in the use of illusions as biomarkers.