We investigated the role of visual information in prehension movements, especially in the manipulation (grasp) component. In Experiment 1, we examined the fundamental properties of reaching and grasping movements. Subjects performed target-directed grasping under natural conditions, with no constraints imposed on the duration of movement. In Experiment 2, we investigated the stage at which visual feedback was effective, as well as the difference between prehension movements to a real object and prehension based on an “internal image”. A metronome was used to control the subject's movement duration. Subjects reached and grasped a cylinder (diameter of 6cm), wearing liquid crystal shutter goggles under four conditions. (1) The goggles opened (normal condition); (2) the goggles closed 0ms, 150ms, 500ms, or 700ms after the beginning of an arm movement (restricted condition); (3) subjects pretended to reach and grasp the cylinder (pantomime condition) and (4) subjects reached and grasped the cylinder without visual information (blind condition). In condition (3) and (4), their vision was occluded 3 seconds before initiating a movement. Kinematics of the wrist trajectory (transport component) and the distance between the thumb and middle finger (manipulation component) were analyzed. The results showed that the manipulation component was influenced by the shutter speed. Fast speeds (0ms, 150ms) caused a significantly large overshoot. Results also indicated that wrist velocity profiles in the acceleration phase are remarkably similar to the no visual and pantomime conditions, although the kinematics of the manipulation component in these two conditions was significantly different. These findings suggest that visual feedback is important during the early stage of movements and that the movements in pantomime and no-visual conditions probably involved an image-driven process, which is different from online visuomotor performance.