Evaluating the usefulness of automatic continuous positive airway pressure (CPAP) in treating the sleep apnea-hypopnea syndrome (SAHS) is not easy because the algorithms for automatic CPAP implemented in the devices available are not well known and are probably dependent on the device. In addition, at present it is not possible to test the behavior of automatic CPAP devices in response to well-defined breathing patterns. Our aim was to implement a bench test to characterize the responses of automatic CPAP devices by subjecting them to breathing patterns of patients with SAHS. To this end, a variety of typical breathing patterns (normal, apneas, hypopneas, flow limitation, snoring) previously recorded in patients with SAHS during sleep were reproduced by a breathing waveform generator. Five commercially available automatic CPAP devices were tested. The responses of the devices to apneas, hypopneas, flow limitation, and snoring were considerably different. In some devices, the response was modified by air leaks similar to the ones found in patients. Consequently, the effectiveness of automatic CPAP assessed in clinical tests performed by using particular devices has no general validity. Testing automatic CPAP devices in a bench study is a useful first step in evaluating the performance of this new type of device in adjusting nasal pressure for each patient.