The effect of cochlear hearing loss and of probe tone level on slopes and sharpness of compound action potential tuning curves was investigated. Thirty-one simultaneously masked isoreduction (50%) tuning curves were determined in 26 adults with cochlear hearing losses up to 60 dB. Probe tone frequency was 2 or 3 kHz. Probe tone level was chosen as close as possible to the action potential threshold, usually within 30 dB. In 5 cases a second tuning curve was determined at a 20-30 dB higher probe tone level in order to differentiate between effects of hearing loss and of probe tone level itself on decrease of selectivity. Tuning was analysed in terms of high- and low-frequency slopes of the tuning curves, both in the steepest parts near the tip and overall, and in terms of Q10dB. Slopes and tuning quality diminished with increasing hearing loss up to 60 dB. Part of the decrease in Q10 could be attributed to increased probe tone level, implying that frequency selectivity is also a level-dependent property. In the same group of subjects so called 'narrow-band' (or 'derived response') compound action potential latencies were determined at 90 dB pe SPL and a derived frequency similar to the probe tone in the tuning curve experiments. Narrow band latencies did not change significantly out of the normal range (2 periods) with increasing hearing loss. This implies that narrow band latencies are not related to hearing loss, but reflect only the probe-level dependent impulse response delay. Analysis shows that it is possible to derived Q10dB from narrow band latencies with probe level as a parameter.