To determine whether polysomnography is necessary to assess the presence and severity of sleep-disordered breathing, bedside observations by physicians were compared with the results of polysomnography in 37 patients with clinically suspected obstructive sleep apnea. Physician observations correlated with objective findings from polysomnography in detecting the presence of obstructive apnea (p less than 0.01), and had a high specificity and positive predictive value. The 20 patients correctly identified by clinical observation had a longer duration of apneic episodes (p = 0.02), increased severity of snoring (p = 0.02), resuscitative snoring (p less than 0.02), and paradoxic thoracoabdominal movement (p less than 0.05). However, 11 other patients with sleep-disordered breathing were not identified clinically; therefore, the sensitivity (64.5 percent) and diagnostic accuracy (70.3 percent) of brief clinical observation were low. Furthermore, the physicians' determinations of the severity of the condition on the basis of bedside estimates of disordered breathing rate, duration of episodes, and the degree of associated hemoglobin oxygen desaturation did not correlate with objective measurements. These findings suggest that a single, brief clinical observation alone is an ineffective screening procedure for detecting obstructive sleep apnea.