An attempt is made to discover the physical processes in the cochlea which would yield results in agreement with observations. It is shown that the assumption of a 'passive' cochlea, where elements are brought into mechanical oscillation solely by means of the incident sound, is not tenable. The degree of resonance of the elementls of the cochlea can be measured, and the results are not compatible with the very heavy damping which must arise from the viscosity of the liquid. For this reason the 'regeneration hypothesis' is put forward, and it is suggested that an electromechanical action takes place whereby a supply of electrical energy is employed to counteract the damping. The circumstantial evidence for such a process is considered, and it appears that the cochlea microphonic potential, hitherto an unexplained by-product of the action, forms an important link in the chain of events. Some implications of the theory are discussed, and ways of testing it are suggested.