Ten users of multielectrode cochlear implants participated in an evaluation of the perceptual effects of input-signal compression. A syllabic compressor was introduced into the microphone circuit of Spectra-22 or SPrint sound processors. The post-compression gain was adjusted to provide similar loudness for speech at an average level of 65 dBA with compression either enabled or disabled. Sentence recognition was measured at three levels. Averaged across all listeners, statistically significant score increases were obtained at each level with compression enabled (45 dBA: 19.6 percentage points, p < .0001; 55 dBA: 16.6 percentage points, p < .0001; 70 dBA: 3.1 percentage points, p = .031). A test of speech intelligibility in noise showed no significant effect of compression. Generally, participants in the trial reported improved perception of low-level sounds with compression, although a few disliked the increased loudness of some background noises. Some participants suggested that the ability to enable or disable compression with a manual switch would be helpful. Overall, the results show that input compression can improve the performance of these sound processors for users of cochlear implants, especially when listening to speech at low levels.