Study of the epidemiology of disturbances of breathing during sleep was hampered until recently by the need to conduct studies in the laboratory, with attendant inconvenience and limited sample sizes. We assessed the accuracy of a microprocessor-based portable monitoring system (Vitalog PMS-8, Vitalog Corp., CA) to detect and classify episodes of disturbed breathing during sleep in 14 patients with sleep apnea by simultaneously recording oxygenation and thoracoabdominal motion on the portable system and a polygraph. Each patient slept in the laboratory for 1 night. In two subjects, the portable system failed to record thoracoabdominal signals. In the remaining subjects, the portable system detected 78% of 2,340 episodes of disturbed breathing, but the recorded information was not sufficient to allow confident classification into central or obstructive events. The positive predictive value of disturbed breathing detected by the portable system was 64%, Respiratory disturbance indices (RDI) computed from the polygraph and portable records were correlated (r = 0.70; p less than 0.01), and all patients with sleep apnea were correctly diagnosed by the portable system. The portable system overestimated arterial oxygen saturation (SaO2) recorded by an ear oximeter (Biox IIA, Ohmeda, CO) but the error was less than 10% of the true value at SaO2 greater than 60%. Seven normal subjects were studied while awake to examine the accuracy of volume measurements made by the portable system and the system's ability to detect paradoxical thoracoabdominal motion of various degrees. Absolute measurement of tidal volume was inaccurate, but detection rate of paradoxical thoracoabdominal motion was excellent (97%). We conclude that the portable system is sufficiently sensitive to allow detection of patients with breathing disorders during sleep, but further developments are necessary before the system can be relied on for accurate classification of apneas and hypoventilation.