A fundamental challenge for the visual system is to integrate visual features into a coherent scene, known as the binding problem. The neural mechanisms of feature binding are hard to identify because of difficulties in separating active feature binding from feature co-occurrence. In previous studies on feature binding [1-5], visual features were superimposed and presented simultaneously. Neurons throughout the visual cortex are known to code multiple features [6]. Therefore, the observed binding effects could be due to the physical co-occurrence of features and the sensory representation of feature pairings. It is uncertain whether the mechanisms responsible for perceptual binding were actually recruited [7, 8]. To address this issue, we performed psychophysical and fMRI experiments to investigate the neural mechanisms of a steady-state misbinding of color and motion [9], because feature misbinding is probably the most striking evidence for the active existence of the binding mechanisms [10]. We found that adapting to the color-motion misbinding generated the color-contingent motion aftereffect, as well as the color-contingent motion adaptation effect in visual cortex. Notably, V2 exhibited the strongest adaptation effect, which significantly correlated with the aftereffect across subjects. Furthermore, effective connectivity analysis using dynamic causal modeling showed that the misbinding was closely associated with enhanced feedback from V4 and V5 to V2. These findings provide strong evidence for active feature binding in early visual cortex and suggest a critical role of reentrant connections from specialized intermediate areas to early visual cortex in this process.
展开▼
