Purchase this article with an account.
Andrew Morgan, Lucy Petro, Luca Vizioli, Lars Muckli; Retinotopically occluded subsections of early visual cortex contain contextual information about individual scenes, category and depth.. Journal of Vision 2015;15(12):516. doi: https://doi.org/10.1167/15.12.516.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Introduction Activity in early visual cortex contains information about context, even in the absence of meaningful feedforward input (Smith & Muckli, 2010; Vetter, et al. 2014). To explain these contextual effects, previous work suggests that other brain areas predict sensory input, and that predictions are transmitted to early visual areas via dynamic and diffuse feedback. Here, we aimed to enhance our knowledge regarding the level of detail in feedback representations, which may depict scene-specific features (Naselaris, et al. 2014), or conversely, may be categorical, which has been observed when context was provided by auditory stimulation (Vetter, et al. 2014). To adjudicate these possibilities, we examined whether activity patterns in occluded portions of early visual cortex are sufficient for discriminating individual scenes, or if the absence of feedforward input reveals categorical scene groupings. Methods We blocked feedforward input to subsections of retinotopic visual cortex (Smith and Muckli, 2010) while participants viewed 24 real-world scenes. Scenes spanned multiple categories and spatial depths, two higher-level characteristics previously studied in early visual cortex (Walther, et al. 2009; Kravitz, et al. 2011). This allowed us to examine possible groupings in feedback representations. We investigated the information content of response patterns in occluded subsections of V1 and V2 using functional magnetic resonance imaging and multi-voxel pattern analyses. Results & Conclusions Category, depth, and individual scenes can all be decoded from occluded portions of early visual cortex. Scene decoding errors were uniformly distributed amongst scenes rather than being concentrated within category- or depth-based groupings. These results indicate that non-feedforward processing in early visual cortex is specific to individual scenes and retains high-level structure when context is generated in the visual domain, as would be expected by filling-in of visual information by feedback.
Meeting abstract presented at VSS 2015
This PDF is available to Subscribers Only