Abstract: In virtual environments, virtual hand interactions play key roles in the human-computer interface. Specifically, the virtual grasping of 3D objects provides an intuitive way for users to interact with virtual objects. This paper demonstrates the creation of a sophisticated virtual hand model simulating natural anatomy in its appearance and motion. To achieve good visual realism, the virtual hand is modeled with metaball modeling, and visually enhanced by applying texture mapping. For realistic kinematics modeling, a three-layer model (skeleton, muscle and skin layers) is adopted to handle the motion as well as the deformation of the virtual hand. We also present an approach for virtual grasping of 3D objects with the realistic virtual hand driven by a CyberGlove dataglove. Grasping heuristics are proposed based on the classification with the shapes of objects, and simplified proxies for the virtual hand are used for the purpose of real-time collision detection between the virtual hand and 3D objects.