Recently Wilson et al. (1991) presented a new speech coding strategy for cochlear implants, the Continuous Interleaved Sampling (CIS) strategy. We implemented this strategy in real-time on a laboratory setup with a Digital Signal Processor, the Motorola DSP56001, and tested speech recognition on a group of 14 patients users of the lnerald multichannel cochlear Implant. This group included patients with performances ranging from optimal to poor and patients speaking different languages (French, German, Italian, Spanish, etc ... ). Results demonstrate (1) that recognition of non-sense speech sounds are significantly better with the new CIS strategy than with the Compressed Analogue (CA) strategy of the processor they dally use and (2) that speech reception for each patient is influenced by the speech coding strategy that is used and by some individual variables. These observations open new promising perspectives for all users of the Ineraid multichannel cochlear implant.