This study investigated variation in respiratory disturbance during sleep. Sixty-six healthy elderly subjects (mean age = 67.2) underwent two consecutive nights of polysomnography. Respiratory disturbance was assessed by Respiratory Disturbance Index (RDI), the number of events per hour of sleep. Results indicated an increase in RDI from Night 1 to Night 2. Fourteen subjects increased their RDI by over 2.5, and eight by over 5.0, events per hour on Night 2. The increase was not a function of increased REM on Night 2. Age was positively related to RDI on individual lab nights but unrelated to nightly variation in RDI. Subjects without complaints of insomnia were more likely to show increases in RDI, perhaps reflecting the sounder sleep of this group on the second laboratory night. This study suggests that a single night of polysomnography is likely to underestimate the absolute level of respiratory disturbance seen in a subsequent recording night. Studies placing prevalence of such disturbance in the elderly at approximately 30% are thus likely to be underestimates. Whether this "error" is important will depend ultimately upon the meaning of various absolute levels of respiratory disturbance in healthy older persons.