Find in Library

A key question in cognitive neuroscience is how the brain combines low-level features processed in remote sensory cortices to represent meaningful multisensory objects in our everyday environment. Models of visual object processing typically assume a feedforward cascade through the hierarchically organized ventral stream. We contrasted this feedforward view with an alternate hypothesis in which object processing is viewed as an interactive, feedforward and feedback process. We found that higher-order regions in anterior temporal and inferior prefrontal cortex performed audio-visual integration 100ms earlier than a sensory-driven region in the posterior occipital cortex, and were modulated by semantic variables (congruency), from as early as 50-100ms. We propose that the brain represents familiar and complex multisensory objects through early interactivity between higher-order and sensory-driven regions. This interactivity may underpin the enhanced behavioural performance reported for semantically congruent audio-visual objects.