A Decentralized Architecture for Multisensory Neural Information Integration

Multisensory integration is important in our brain. However, understanding how the brain integrates multiple sensory cues in its neural circuitry remains a challenge. In this study, using biologically realistic neural network models, we propose a novel mechanism of how multisensory information might be integrated in a distributed fashion across interconnected brain areas without the need for a central integration unit. We show that this decentralized system can integrate information optimally in a biologically realistic setting, and is in good agreement with anatomical constraints and the experimental observations.