This article describes the development and validation of a model for predicting multi-finger movements in grasping activities. The model builds upon a newly proposed approach that incorporates forward dynamics and a system identification procedure, and is amenable to empirical tests. A database of multi-fingered grasping movements performed by 28 subjects was established and divided into four sets, one for model development and three for model validation. In the development phase, model parameter values were estimated by the iterative system identification procedure through a physics-based heuristic algorithm. The estimated parameter values were then statistically synthesised and integrated into the prediction model. In the validation phase, the model was applied to three novel datasets containing different grasping movements involving objects of varied sizes and different subjects. The results demonstrated the model's ability to predict hand prehensile movements with error magnitudes comparable to the inter-person variability in performing such movements. New insights into the control of multi-fingered hand prehensile movements at the systems and joint levels emerged from the model development and validation process. The current study contributes to building a foundation for long-term development of realistic biodynamic simulation of multi-finger hand movements. Such simulation capabilities will aid in design of hand-operated tools, devices or hand-intensive work for proactive ergonomics and in evaluation as well as treatment of functional impairment of the hand.