The impetus to study sleep changes in a cluster population arose from a recent hypothesis that predicted the finding of sleep apnea in this disorder. It holds that cluster attacks may occur in response to oxygen desaturation. Proposed mechanisms involve impairment of carotid body activity secondary to hypothalamic-vasomotor regulatory dysfunction. Five chronic and five episodic cluster patients underwent nocturnal polysomnography, utilizing standard equipment for monitoring sleep status, cardiac activity, nasal and buccal air flow change, chest and abdominal breathing, muscle activity and oxygen saturation. All episodic patients and one of five chronic patients were found to have sleep apnea (60%). Mean apneas per hour during NREM sleep were similar to that of REM sleep; 26.7 and 28.2, respectively. Six patients with sleep apnea experienced 14 cluster headache attacks during the study period. Eight attacks (57%) followed episodes of oxygen desaturation ranging from 65% to 85%. In the sleep apnea group, 8 out of 14 attacks (57%) were associated with REM; three without, and five following oxygen desaturation. Of the non-apnea group, all of whom had chronic cluster headache, none of 5 attacks were associated with oxygen desaturation, and only 2/5 attacks occurred in relation to REM. Thus, our study showed that sleep apnea was a common finding in a randomly selected group of episodic cluster patients; and most nocturnal attacks were preceded by oxyhemoglobin desaturation and REM-related. These findings were uncommon in the chronic cluster group.