OBJECTIVE The purpose of the present experiment was to determine the relationship between most comfortable listening level and preferred listening levels for linear and slow-acting compression hearing aids as a function of variations in speech and noise level. METHODS A digital hearing aid test system was used to simulate six hearing aids having compression ratios of 1, 1.5, 2, 3, 5, and 10:1. Speech was presented in three different noises (vent, apartment, and cafeteria), with speech input level being varied (55, 70, 85 dB SPL). Subjects were 20 listeners with sensorineural hearing loss (half with a dynamic range < or = 30 dB and half with a dynamic range >30 dB). The boundaries of the most comfortable listening range were measured to estimate most comfortable listening level. Preferred listening level was measured by having subjects adjust the output of the hearing aid for satisfactory listening. RESULTS On average, the deviation of preferred listening level from most comfortable loudness (MCL) was less than 5 dB. Dynamic range, noise type, and input level were all found to have small, but significant, effects on the deviation of preferred listening level from MCL. On average, subjects with a small dynamic range listened slightly below MCL, and subjects with a larger dynamic range listened slightly above MCL. For favorable signal-to-noise ratios, preferred listening levels were highest for high input levels and for conditions that resulted in high output levels before level adjustment. Although the pattern of average performance differed slightly at poorer signal-to-noise ratios, all preferred listening levels were close to MCL. CONCLUSIONS The gain of a slow-acting compression hearing aid should place the output within 5 dB of MCL. The output for low and medium inputs should approximate MCL and the output for high input levels should be slightly above MCL. This pattern of gain may be obtained with mild compression ratios and a gain rule that places a speech input of 70 dB at MCL.