[Physiological basis for a cochlear prosthesis (author's transl)].

For the attempt to develop a cochlear prosthesis, which allows some understanding of speech, it seems--at least for the first attempts--to be appropriate to mimic natural conditions as far as possible. The auditory nerve contains about 30,000 afferent fibres. Qualitatively, their behavior is similar but quantitative measures show considerably differences (2.3). Nothing certain can be said at present however about the spiral fibres originating from the outer hair cells. The quantitative differences between single afferents concern tuning, frequency selectivity, thresholds, intensity functions and--of particular interest for electrical stimulation--differences in timing of the activity pattern, brought about by differences in travelling time along the cochlear duct (2.3). The time differences seen in the activity pattern of different fibres are in the order of several ms (2.3.6;2.3.7). Actionpotentials elicited by natural acoustic stimuli show probabilistic behavior, that is they are not strictly determined. It is obvious that with artificial electrical stimulation not every surviving single fibre can be selectively stimulated. An electrode will always stimulate a group of fibres simultaneously. With any conceivable electrical stimulation all fibres in the suprathreshold region of the electrode will be synchronously activated (3.2); a fundamental difference to the natural situation. To estimate the number of channels, necessary to stimulate the auditory nerve with sufficient accuracy to allow speech perception we consider some psychoacoustic data. These have shown that the auditory system possesses the ability to differentiate a great number of different pitches, but on the other hand it is capable of integrating different frequency areas to a so called critical bandwidth. Sound energy falling into one critical bandwidth is integrated to a uniform auditory sensation. If one is to integrate various fibres of the auditory nerve to one channel of stimulation it seems to be adequate to use the critical bandwidth as a measure (3.1). According to this criterion 15 channels would have to be introduced into the speech region of the cochlea. This would allow 1.2 mm of cochlear length for each channel. Perfect electrical separation of the channels is required. Considering the severe distortions in neuronal activity pattern, introduced by electrical stimulation in comparison to the natural conditions it is not clear even whether the number given would be sufficient. On the other hand, current spreading would appear to prohibit any higher electrode density. As far as coding of sound parameters within one channel is concerned it is proposed that full use should be made of frequency analysis according to the place principle. In respect to coding of periodicity and loudness it is proposed to approach natural conditions as far as possible (3.3). Here delay times between the individual channels and a probabilistic character of the stimuli should be introduced to avoid dominance of periodicity pitch...