Spatiotemporal Representations of Contextual Associations for Real-World Objects

Area of Science:

• Neuroscience
• Cognitive Science
• Visual Perception

Background:

• Objects in the real world are rarely isolated, appearing within specific scene contexts and alongside co-occurring objects.
• Previous research indicates that contextual associations influence neural object representations, even without explicit scene cues.
• Key questions remain regarding the relationship, temporal emergence, and neural mechanisms of these contextual representations.

Purpose of the Study:

• To investigate how representations of contextual associations relate to perceptual and categorical information in the visual cortex.
• To determine the temporal dynamics of contextual association representation.
• To explore the neural mechanisms underlying contextual association representation for isolated objects.

Main Methods:

• Combined functional Magnetic Resonance Imaging (fMRI) and Electroencephalography (EEG) in human participants.
• Participants viewed isolated objects belonging to distinct scene contexts.
• Multivariate pattern analysis (MVPA) was applied to neural data to decode representations.

Main Results:

• Objects from the same context were represented more similarly than objects from different contexts in object-selective (LOC) and scene-selective (PPA) areas.
• These contextual representations emerged later in visual processing than perceptual and categorical representations.
• Contextual associations appeared to be implemented through object-to-object and object-to-scene associations, particularly in the anterior PPA.

Conclusions:

• Contextual association representations are automatically formed for isolated objects, even without specific tasks.
• These findings suggest that the brain actively uses contextual frames to support visual cognition in natural environments.
• The study demonstrates that contextual information shapes neural object representations, enhancing efficient visual processing.