We evaluated the performance of an event recorder system in a large, consecutive series of referred monitored patients to determine the relative incidence of true apnea and true bradycardia, false alarms, and alarms for movement or a loose lead. In addition, we developed an event classification system based on the reason for the event being recorded. The recorder stored transthoracic impedance and electrocardiogram signals on a floppy disk before, during, and after each monitor alarm. These events on 302 disks from 83 patients were analyzed and classified as true, false, or movement-loose lead. Of 14,131 events, only 8% were caused by apnea or bradycardia (true events). Of true events, 70% were triggered by apnea and 30% by bradycardia. These true events occurred in 48% of the patients. False alarms constituted 23%, and movement-loose lead 69%, of all events. Even when movement-loose lead events were excluded, nearly three of four events were found to be false. Event recording proved helpful clinically, allowing discontinuation of the monitor in 49% of patients, modification of monitor alarm settings, or reassurance and counseling for parents.