When interacting with our dynamic environment, fast and precise transformation of sensory motion information into motor behaviors is essential. However, when sensory information is missing or imprecise, relying on prior expectations and predictions based on past experiences can make our behavior more reliable. The Bayesian inference framework has effectively explained this type of reliability-based information integration. In this talk, I will present two studies that investigated the neural mechanisms underlying this phenomenon. Through human EEG and single-cell recording studies that employed a smooth pursuit eye movement task, we found that predictive information reaching the sensory cortical neural activity contributes to the enhancement of motor behavior. These results suggest that optimal integration of sensory information and prediction signals occurs in sensory cortical areas, which are sufficient for guiding enhanced motor actions.