Title : Dynamics of scene representations in the human brain revealed by 1 magnetoencephalography and deep neural networks 2 3

22. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not. ABSTRACT 23 24 Human scene recognition is a rapid multistep process evolving over time from single 25 scene image to spatial layout processing. We used multivariate pattern analyses on 26 magnetoencephalography (MEG) data to unravel the time course of this cortical process. 27 Following an early signal for lower-level visual analysis of single scenes at ~100ms, we 28 found a marker of real-world scene size, i.e. spatial layout processing, at ~250ms 29 indexing neural representations robust to changes in unrelated scene properties and 30 viewing conditions. For a quantitative explanation that captures the complexity of scene 31 recognition, we compared MEG data to a deep neural network model trained on scene 32 classification. Representations of scene size emerged intrinsically in the model, and 33 resolved emerging neural scene size representation. Together our data provide a first 34 description of an electrophysiological signal for layout processing in humans, and a novel 35 quantitative model of how spatial layout representations may emerge in the human brain. 36 37 38 39 40 41 KEY WORDS 42 43 Scene perception, spatial layout, magnetoencephalography, deep neural network, 44 representational similarity analysis 45 46. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not .