POSTER DETAILS
Contextual representations of natural scenes in monkey and mouse V1 neurons
Paolo Papale, Koen Seignette, Feng Wang, Andrew T. Morgan, Xing Chen, Amparo Gilhuis, Maaike Van der Aa, Paul Neering, Chris Van der Togt, Lucy S. Petro, Lars Muckli, Pieter R. Roelfsema, Matthew W. Self, Christiaan N. Levelt
Date / Location: Monday, 11 July 2022 / S03-481
Add this poster event to your π
Add this poster event to your π
Neuronal activity in the primary visual cortex (V1) is governed by feedforward information processing in each neuronβs receptive field (RF) and by contextual information in surrounding regions. Ongoing feedforward responses are rapidly affected by contextual inputs, so it is challenging to dissociate their functional roles. We used full and occluded views of natural scenes to reveal the impact of contextual mechanisms on V1 representations in the absence of information in the RF, as well as the effect of training on the selectivity and response dynamics of V1 neurons. We presented partially occluded and full (non-occluded) natural scenes while recording V1 activity from monkeys using electrophysiology (24 images per condition; 211 sites in two macaques) and from mice using chronic widefield and two-photon calcium imaging (6 images per condition; ~900 neurons in six animals). In mice, V1 activity was recorded both before and after perceptual training to detect 4 full natural scenes. We successfully decoded stimulus identity from V1 activity in both species without visual stimulation in neuronal RFs. In monkeys, decoding was significant in the sustained phase of the response, from ~75ms after stimulus onset. In mice, we found significant decoding both before and after perceptual training (all ps < 0.05, permutation test), and training mice on fullscreen images weakened V1 responses to these images but strengthened responses to occluded images. Our results support predictive coding theories of neuronal activity and lay the groundwork for investigating the precise contextual features that drive and modify V1 responses.