The human hearing system shows both high sensitivity and fine frequency resolution. Its sensitivity to nonlinear distortion contrasts with its own large distortions. Although a receiver, the ear also produces sound, albeit faint. Peripheral preprocessing seems to be responsible for most of these effects. A mixture of linear and nonlinear preprocessing elements, however, prevents a complete description of the system's behavior. Nevertheless, the following simplifying comparison may be helpful: Former feedback radio receivers were used most effectively by turning the feedback knob so that ringing just failed to occur. At that point, the receiver was most sensitive and had its best frequency selectivity. Feedback, however, was a capricious tool; frequent ringing was annoying. To avoid ringing, our hearing system seems to use nonlinearities at very low levels, so that ringing (emissions) can occur only at inaudible amplitudes. Consequently, sensitivity and frequency selectivity are only high at low levels. Moreover, nonlinearity acts within one channel only, although the ear uses many channels along the basilar membrane. The relations among psychoacoustical data, data from emissions, and models are described and discussed.