1. How does the brain compute visual information for navigational route planning?
The brain computes visual information for navigational route planning by processing several different scene features, such as 3D and semantic aspects. Identifying navigational affordances is a complex computational feat that requires localizing obstacles and finding ways around them, which necessitates 3D scene information. Semantic scene classification also benefits route planning as navigating typical basements, balconies, and garages require different procedures. To test the hypothesis, human electroencephalography (EEG) responses to indoor scene images were collected, capturing the temporal order of scene feature processing in the human brain during visual scene perception. Visual features were operationalized as activations of deep neural networks (DNNs) trained for 2D, 3D, and semantic tasks. Navigational features were captured using navigational affordance maps (NAM) constructed using human behavioral responses when planning exit routes in natural indoor scene images. Representational similarity analysis (RSA) was used to relate visual and navigational features to EEG data, revealing the temporal order in which these features are processed in the human brain. The results showed that navigational affordance representations emerged significantly later than visual features, supporting the view that the brain uses 2D, 3D, and semantic scene features to facilitate navigation planning.
read more
2. What is the temporal pattern of peak timings for unique variance explained by different DNN RDMs in the EEG data?
The temporal pattern of peak timings for unique variance explained by different DNN RDMs in the EEG data shows a hierarchy of scene feature processing. The highest uniquely explained variance by the 2D DNN RDM occurred first at 128.12+-3.56ms after stimulus onset, followed by 3D and semantic DNN RDMs peaking at 171.87+-30.79ms and 161.87+-10.45ms, respectively. The unique variance of the NAM RDM reached its peak significantly later than the other model RDMs at 296.25+-37.05ms after stimulus onset. This suggests a hierarchy of scene feature processing leading up to the representation of navigational affordances.
read more
3. Temporal dynamics of scene perception?
This study investigated the temporal dynamics of scene perception, focusing on the order of 2D, 3D, semantic features, and navigational affordances. The emergence of 2D, semantic, and 3D features preceded navigational affordance representations. Early emergence of low-level 2D features followed by high-level semantic features aligns with previous studies. 3D features, however, have received less attention in M/EEG studies. Our findings suggest that 3D features are processed in parallel with semantic features, and navigational affordance representation emerged later. Image complexity and task requirements may influence processing time. Future research could include dynamic scenes and additional computational models to further understand scene perception.
read more
4. How does the experimental paradigm engage participants in navigational affordance processing?
The experimental paradigm engages participants in explicit navigational affordance processing by asking them to imagine the directions of navigational paths relative to their viewpoint. Participants determine whether the paths lead to the left, center, or right while viewing images. Each image is presented for 200ms, followed by a randomly varying inter-trial interval between 600ms and 800ms. Image presentation trials are ordered in blocks of one to five trials, with catch trials included to ensure attentiveness and processing of spatial and navigational aspects. During catch trials, participants indicate whether an arrow on the screen points in the same or different direction as the previous navigational path. This design results in each image being repeated 75 times, providing a comprehensive engagement in navigational affordance processing.
read more