Pitch is a complex hearing phenomenon that results from elicited and self-generated cochlear vibrations. Read-off vibrational information is relayed higher up the auditory pathway, where it is then condensed into pitch sensation. How this can adequately be described in terms of physics has largely remained an open question. We have developed a peripheral hearing system (in hardware and software) that reproduces with great accuracy all salient pitch features known from biophysical and psychoacoustic experiments. At the level of the auditory nerve, the system exploits stochastic resonance to achieve this performance, which may explain the large amount of noise observed in the working auditory nerve.

The perceived pitch of a complex harmonic sound changes if the partials of the sound are frequency-shifted by a fixed amount. Simple mathematical rules that the perceived pitch could be expected to follow ('first pitch-shift') are violated in psychoacoustic experiments ('second pitchshift'). For this, commonly cognitive cortical processes were held responsible. Here, we show that human pitch perception can be reproduced from a minimal, purely biophysical, model of the cochlea, by fully recovering the psychoacoustical pitch-shift data of G.F. Smoorenburg (1970) and related physiological measurements from the cat cochlear nucleus. For this to happen, the cochlear fluid plays a distinguished role.