A discussion is presented on the control aspects of upper-limb prostheses, with emphasis on the areas of necessary improvements in current designs. Arguments are presented to indicate that it should be possible to obtain a substantial improvement in prostheses control by properly training the amputee, improving the dynamics responst of the prostheses, and improving the quality of the forward-path control signal. Augmentation of feedback information, although useful, may not be essential. The limitations of the myoelectric (muscle) signal as a forward-path control signal, especially for multiplc degrees-of-freedom prostheses, is discussed. Most of the limitations of the myoelectric signal can be overcome if the neuroelectric (nerve) signal is used as a forward-path control signal. Results of a series of experiments which demonstrate the feasibility of constructing an electrode capable of being implanted around severed nerves and of detecting neuroelectric signals for prolonged periods of time are presented. A possible scheme for employing neuroelectric control is also presented.