The human visual system processes visual information hierarchically, that is, local visual features such as line orientations are extracted in early visual areas, and then complex patterns are recognized in higher cortical areas by integrating these local feature segments. On the other hand, it is known that, as well as the bottom-up hierarchical connections, there are top-down connections from the higher visual areas to the lower-order areas in the brain. There are many computational models that explain the computational role of the bi-directional connections, and many of them suggest that the visual system predicts an input stimulus via the top-down connections and corrects the recognition error. Here we examine the property of the human pattern recognition with masked visual search paradigm. The result showed that the reaction times (RT) were modulated by the mask stimuli, and the RT modulations depended on the types of mask pattern. This result suggests that the existence of the prediction-correction process in the pattern recognition system.
