Contextual Role Object Representational Geometry

---
title: "Contextual Role Modulates Object Representational Geometry in the Human Brain"
source: "arXiv:2605.23111"
authors: "Julien Dirani, Shankar Chawla, Leila Wehbe, Bradford Z. Mahon"
category: "ai_collection"
tags:
  - neuroscience
  - fmri
  - brain-network
  - object-representation
  - representational-geometry
  - contextual-role
---

# Contextual Role Modulates Object Representational Geometry in the Human Brain

**arXiv:2605.23111** | Submitted: 22 May 2026 | q-bio.NC

## Summary

This fMRI study investigates how the human brain dynamically remaps object representations as the same object shifts across contextual roles — specifically, when objects are passive elements in a scene versus targets of goal-directed actions. Using naturalistic movie viewing, the authors demonstrate a **double dissociation** in representational geometry:

- **Target objects** (action targets) engage a **parietal action network** (supramarginal and postcentral gyri), organized by **action affordance** and **hand posture affordance** dimensions.
- **Passive objects** recruit a **distributed occipito-temporal network** (visual object recognition), aligned with **semantic dimensions**.

Outside context-specific networks, representational content retains **context-invariance**, showing flexibility and invariance operate at different levels of the same representational system.

## Key Findings

1. **Neural remapping**: Object representations are dynamically remapped based on moment-to-moment changes in contextual relevance within naturalistic scenes.
2. **Double dissociation**: Target vs. passive objects recruit distinct brain networks with different representational geometries.
3. **Context-invariant structure**: Visual representational structure remains stable across contexts outside specialized networks.
4. **Action affordance organization**: Target objects in parietal cortex are organized by what actions can be performed on them.

## Methodology

- **fMRI** combined with **naturalistic movie viewing** (rich ecological validity)
- Representational Similarity Analysis (RSA) to characterize representational geometry
- Comparison of object representations across contextual roles (target vs. passive)

## Implications

- Challenges purely invariant views of object representation in the ventral visual stream
- Supports **dynamic, context-dependent** models of neural representation
- Provides computational constraints for **brain-inspired AI** systems that need flexible object representations
- Demonstrates that **flexibility and invariance** are complementary, not competing, properties of neural systems

## Activation Keywords

contextual-role, object-representation, representational-geometry, fmri, brain-network, action-affordance, visual-cognition, neural-remapping