1. Transducer currents were recorded in turtle cochlear hair cells during mechanical stimulation of the hair bundle. The currents were measured under whole-cell voltage clamp in isolated cells that were firmly stuck to the floor of the recording chamber. 2. Stimuli were calibrated by projecting the image of the hair bundle onto a rapidly scanned 128 photodiode array. This technique showed that, while the cell body was immobilized, the tip of the bundle would follow faithfully the motion of an attached glass probe up to frequencies of more than 1 kHz. 3. The relationship between inward transducer current and bundle displacement was sigmoidal. Maximum currents of 200-400 pA were observed for deflections of the tip of the bundle of 0.5 microns, equivalent to rotating the bundle by about 5 deg. 4. In response to a step deflection of the bundle, the current developed with a time constant (about 0.4 ms for small stimuli) that decreased with the size of displacement. This suggests that the onset of the current was limited by the gating kinetics of the transduction channel. The onset time course was slowed about fourfold for a 20 degrees C drop in temperature. 5. For small maintained displacements, the current relaxed to about a quarter of the peak level with a time constant of 3-5 ms. This adaptation was associated with a shift of the current-displacement relationship in the direction of the stimulus. The rate and extent of adaptation were decreased by lowering external Ca2+. 6. Adaptation was strongly voltage sensitive, and was abolished at holding potentials positive to the reversal potential of the transducer current of about 0 mV. It was also diminished by loading cells with 10 mM of the Ca2+ chelator BAPTA. These observations suggest that adaptation may be partly controlled by influx of Ca2+ through the transducer channels. 7. Removal of adaptation produced asymmetric responses, with fast onsets but slow decays following return of the bundle to its resting position; the offset time course depended on both the magnitude and duration of the prior displacement. 8. In some experiments, hair bundles were deflected with a flexible glass fibre whose motion was monitored using a dual photodiode arrangement. Positive holding potentials abolished adaptation of the transducer currents, but had no influence on the time course of motion of the fibre. We have no evidence therefore that adaptation is caused by a mechanical reorganization within the bundle.