The field of cochlear prostheses is reviewed, emphasizing the neurophysiological principles necessary for the development of a successful cochlear prosthesis. Pertinent auditory physiology is reviewed, and four conditions are proposed which, if met, should result in speech recognition by the patient with an implant. These conditions are: (a) The surviving neural population must be adequate over the frequency distribution required to deliver the necessary information for speech discrimination. (b) The central neural processing pattern for decoding and recognizing speech must have been established and still persist. (c) The processing of the acoustic signal by the auditory system up to the level of the prosthesis interface must be understood and predictable from the acoustic stimulus. (d) The relationship between the response patterns of the neural elements and the electrical stimulus must be well documented and controllable. The degree to which these conditions can presently be met is discussed, with some suggestions for future development. The function of the cochlear prosthesis is separated into a signal processing section and an electrical stimulus section. Two signal processing strategies are analog processing of the acoustic signal and speech feature extraction from the acoustic signal. Four possible electrical stimulation strategies are direct analog stimulation from the signal processor, pulse amplitude modulation, pulse width modulation, and stimulation designed to optimize the neural responses to electrical stimulation. Some of the present generation cochlear prostheses, including five approved by the FDA as investigational devices, are discussed according to this classification. The clinical results of testing these devices as aids to lip reading and as stand-alone speech reception aids are compared. All are a potential aid to lip reading. Some prosthesis designs have been implanted in patients who then showed exceptional open set speech discrimination. These exceptional patients have used both single channel and multichannel devices and devices with both analog processing and speech feature extraction strategies. These results are encouraging. More consistent and better speech reception is anticipated for the future as this field develops.