1. Erasmus University Medical Center, Rotterdam, the Netherlands
2. Academic Medical Center, Amsterdam, the Netherlands

Most papers show that compression does not improve speech intelligibility while users generally like compression in hearing aids because it takes away the necessity to continually readjust the volume setting. The comfort aspect of compression should be realized by a slow system (automatic volume control). A short attack and a long release time make the system comfortable from the point of high-level protection and adjustment to overall level changes. The speech signal is not affected by compression except in terms of achieving a high speech score over a larger range of input signals. There is even some evidence that slow compression in many independent channels may not disturb speech intelligibility. Wearing comfort may only be hampered by sudden changes in speaker level resulting in some pumping. Shortening of the release time of the compressor can reduce the pumping effect (syllabic compressor) but may then interfere with the slowest modulations of the speech signal. The number of independent compression channels, the compression threshold, the compression ratio and the value of the time constants are now relevant factors. Speech intelligibility should be checked including supra-segmental information. Fast compression should be seen as a speech enhancement device: the compressor is used to minimize temporal masking avoiding the masking of a soft consonant by a loud vowel. The number of independent channels should be limited to avoid spectral blurring. We found in a two-channel system a positive effect of phonemic compression on speech intelligibility in quiet, while the same compressor in a 4-channel system did not show the improved intelligibility. The effects of compression in a noisy situation are to be considered separately. The slow system, particularly if it can be combined with an algorithm that can distinguish between speech and noise, can be used to reduce noise annoyance. In fast compression systems, we found a negative effect of our two-channel phonemic compressor on the speech intelligibility in noise contrary to the effect in quiet. This negative effect disappeared after about 4 weeks of wearing a prototype hearing aid.