The relationship between the severity of obstructive sleep apnea (OSA) (measured by sleep study) and daytime sleepiness is poor. Variation in the degree of arousal accompanying obstructive respiratory events might help explain this poor correlation. Polysomnographic records from patients with OSA were reviewed in order to extract representative examples of apneas and hypopneas (in 10 patients), as well as events both supine and decubitus (in 12 patients). The EEG accompanying each obstructive event was processed with a neural network technique to describe sleep depth on a second-by-second basis. The lengths of any visually evident microarousals were also measured manually. There was considerable interindividual variation in the degree of sleep disturbance using the neural network technique (p < 0.005), but not using the lengths of the visually scored microarousals (p = 0.6). The arousals accompanying apneic events caused greater variability in sleep depth quantified using the neural network technique (p = 0.03), and also lasted longer based on the visual scoring (mean, 12.6; SD, 1.7 s) than the hypopneic events (mean, 9.9; SD, 2.4 s; p = 0.02). There were no significant differences between events occurring supine versus decubitus with either technique (p = 0.7). These differences in arousal magnitude may explain some of the poor correlations between conventional measures of sleep apnea severity and daytime sleepiness.