After exposure to a loud, non-traumatic low-frequency tone, auditory thresholds are elevated. Thresholds recover to normal in a non-monotonic manner, decreasing rapidly at first before increasing again, until they finally decrease monotonically towards normal. Although the transient elevation of thresholds after the initial improvement was originally called a 'bounce' by Hirsh and Ward (1952), Kemp (1986) suggests that the initial rapid recovery is the oddity: under some conditions a low-frequency tone can produce hypersensitivity in otoacoustic emissions, psychophysicaI thresholds, and perceived loudness (Kemp's 'bounce') without a later elevation of threshold (Hirsh and Ward's 'bounce'). Kemp also suggested that the transient hypersensitivity was caused by changes in the sensitivity of the active process within the cochlea. We have investigated the origin of this transient hypersensitivity (Kemp's bounce) in guinea pigs, recording cochlear potentials (CM, CAP, SP and EP) and otoacoustic emissions (DPOSEs at f(2)-f(1), 2f(1)-f(2), 2f(2)-2f(1) and 3f(1)-2f(2)). Our results indicate that the bounce does not require neural activity, but is probably produced by nonneural cochlear mechanisms, possibly a transient decrease in the permeability of the organ of Corti which produces a small but significant change in standing current through outer hair cells. At least part of these changes, which are reduced as the stimulation frequency increases, and absent above 2 kHz, seem due to a small and transient movement of the cochlear partition towards scala tympani, probably due to a transient osmotic imbalance.