Representation of navigational affordances and ego-motion in the occipital place area

Frederik S. Kamps, Emily M. Chen, Nancy Kanwisher, Rebecca Saxe

TL;DR

This study investigates how the human brain represents information crucial for navigation, specifically "navigational affordances" (where you can go) and "ego-motion" (how you are moving). Using dynamic first-person video stimuli, researchers found that the Occipital Place Area (OPA) in the brain is uniquely tuned to both types of information, and even their conflicts, suggesting it plays a key role in real-time navigation planning and guidance.

What Problem Does This Paper Address?

Navigating complex environments is a fundamental human ability, heavily reliant on visual input. While we know a network of brain regions, including the Occipital Place Area (OPA), Parahippocampal Place Area (PPA), and Medial Place Area (MPA), is involved in processing visual scenes, their specific roles in navigation remain unclear. Previous research often used static images, which don't capture the dynamic, first-person experience of moving through a scene. This leaves a gap in understanding how the brain processes the crucial information needed for real-time navigation: where can I go (affordances) and how am I moving (ego-motion)? This paper aims to bridge this gap by using dynamic stimuli to pinpoint the OPA's specific contributions to these navigationally relevant visual representations.

Methods

The researchers employed a region-of-interest (ROI) approach, focusing on the OPA, PPA, MPA, and Early Visual Cortex (EVC) as a control. Participants viewed dynamic, first-person video clips depicting various scenarios involving navigational affordances (presence and location of open doorways) and ego-motion (forward, backward, left turn, right turn).

Key methods included:

  • fMRI: Functional Magnetic Resonance Imaging was used to measure brain activity.
  • Dynamic Stimuli: Unlike previous studies using static images, this research utilized 3-second video clips simulating first-person movement through scenes. This is crucial for capturing the experience of ego-motion.
  • Multivariate Pattern Analysis (MVPA): This technique was used to analyze the patterns of brain activity across voxels within each ROI. It allowed the researchers to determine if different conditions (e.g., left doorway vs. right doorway) elicited distinct neural patterns, indicating representational differences.
  • Univariate Analysis: This standard fMRI analysis measured the average signal change in an ROI for different conditions, providing a more direct measure of activation strength.
  • Contralateral Bias Testing: The study specifically investigated whether OPA responses were stronger for stimuli presented in the contralateral visual field, a known characteristic of OPA, to understand the egocentric nature of its representations.
  • Dissociation from EVC: By comparing OPA responses to those in EVC (early visual cortex), the researchers aimed to distinguish between representations driven by low-level visual features (like motion energy) and higher-level, navigationally relevant information.

Datasets

The study did not introduce new datasets or benchmarks. Instead, it used custom-generated dynamic movie stimuli designed to specifically probe navigational affordances and ego-motion. These stimuli were created using Unity software and presented in an event-related design.

Key Insights and Ideas

  • OPA as a Navigation Hub: The most significant finding is the OPA's specialized role in processing both navigational affordances and ego-motion. This suggests it's not just a scene-processing area but a critical component for active navigation.
  • Beyond Static Scenes: The use of dynamic stimuli was crucial. It revealed that OPA's sensitivity to navigationally relevant information is most apparent when experiencing motion, highlighting the limitations of static image studies.
  • Conflict Detection: OPA not only processes affordances and ego-motion independently but also shows sensitivity to their conflict (e.g., turning towards an open doorway versus turning away from it). This implies a sophisticated integration of information for decision-making during navigation.
  • Egocentric Representation: The stronger responses in OPA to stimuli in the contralateral visual field support the idea that it encodes navigationally relevant information within an egocentric (self-centered) reference frame, essential for guiding one's own movements.
  • Functional Dissociation: The OPA's responses were distinct from those in PPA and MPA, which showed weaker or no sensitivity to affordances. Furthermore, OPA's representations were shown to be more abstract than those in EVC, indicating they are not simply driven by low-level visual features like motion energy.

Key Results

  • Navigational Affordance Decoding: OPA showed significant decoding of navigational affordances in multivariate analyses, with this effect being significantly stronger than in PPA and MPA.
  • Ego-Motion Decoding: OPA also demonstrated significant decoding of ego-motion direction, with similar strength to PPA and MPA in some analyses, but a stronger preference for forward motion.
  • Contralateral Bias: OPA exhibited a stronger response to doors in the contralateral visual field compared to paintings, and a preference for turns towards the contralateral visual field, supporting an egocentric representation.
  • Conflict Sensitivity: OPA showed a significant response to the conflict between navigational affordance and ego-motion information, particularly for turns in the contralateral visual field.
  • Dissociation from EVC: While EVC showed some sensitivity to basic ego-motion, OPA's representations of ego-motion were more robust, especially when generalizing across affordance information, suggesting higher-level processing. OPA also showed greater sensitivity to navigational affordances than EVC.

Limitations and Future Work

  • EVC vs. V3A Overlap: The paper notes the proximity of OPA to V3A, an area known for ego-motion sensitivity. It acknowledges that distinguishing between these regions is challenging and suggests future work is needed to clarify their relationship and potential overlap.
  • Generalizability of Conflict: While OPA showed sensitivity to the conflict between affordances and ego-motion, this effect was not entirely specific to OPA, also being present in PPA and MPA. Further research could explore the precise nature of this integration.
  • Hemispheric Integration: The study found contralateral biases, but future work is needed to understand how information from each hemisphere is integrated to form a coherent representation for navigation across the entire visual field.
  • Low-Level Feature Influence: While efforts were made to control for low-level visual confounds, the authors acknowledge that some influence might remain, particularly for ego-motion, which generates optic flow. Future studies could further refine stimuli to isolate higher-level representations.

Summary updated

Abstract

Humans effortlessly use vision to plan and guide navigation through the local environment, or “scene”. A network of three cortical regions respond selectively to visual scene information, including the occipital place area (OPA), parahippocampal place area (PPA), and retrosplenial complex (RSC) – but how this network supports visually-guided navigation is unclear. Recent evidence suggests that one region in particular, the OPA, supports visual representations for navigation, while PPA and RSC support other aspects of scene processing. However, most previous studies tested only static scene images which lack the dynamic experience of navigating through scenes. Accordingly, here we used dynamic movie stimuli to test whether OPA, PPA, and RSC represent two critical kinds of navigationally-relevant information: navigational affordances (e.g., can I walk to the left or right?) and ego-motion (e.g., am I turning left or right?). We found that OPA is sensitive to both affordances and ego-motion, as well as the conflict between these cues – e.g., turning toward versus away from an open doorway. These effects were significantly weaker or absent in PPA and RSC. Responses in OPA were also dissociable from those in early visual cortex, consistent with the idea that OPA responses are not merely explained by lower-level visual features. OPA responses to affordances and ego-motion were stronger in the contralateral than ipsilateral visual field, suggesting that OPA encodes navigationally relevant information within an egocentric reference frame. Taken together, these results support the hypothesis that OPA contains visual representations that are useful for planning and guiding navigation through scenes.
BibTeX
@article{kamps2024representation,
  author = {Frederik S. Kamps and Emily M. Chen and Nancy Kanwisher and Rebecca Saxe},
  title = {{Representation of navigational affordances and ego-motion in the occipital place area}},
  journal = {Journal of Vision},
  year = {2024},
  doi = {10.1167/jov.24.10.568},
}
Figure