Night-to-night variability of breathing and oxygenation during sleep was examined with portable monitoring equipment in 30 residents of a retirement village. Subjects had a variety of health problems as might be expected in the elderly, but all were living independently in self-contained units. None had clinical features to suggest obstructive sleep apnea. Two pairs of consecutive nights were studied, separated by 4-6 months. Satisfactory recordings on all four nights were obtained in 15 subjects, and in these subjects variability of measurements was examined across nights 1-4 using the kappa (K) statistic. There was low but significant agreement in estimated total sleep time (K = 0.23, p less than 0.01) and estimated wakefulness after sleep onset (K = 0.18, p less than 0.05) as assessed with a wrist actigraph. Good agreement was found among measures of disturbed breathing during sleep whether expressed in terms of numbers of events [respiratory disturbance index (RDI), K = 0.62, p less than 0.0001], their duration (event minutes, K = 0.53, p less than 0.0001), or associated disturbance of oxygenation (% cumulative time less than 90% SaO2, K = 0.50, p less than 0.001, n = 9). Twenty-eight subjects had at least two nights' satisfactory recordings. Although some of these individuals showed considerable variation in RDI, this had little overall effect on classification of them into normal (RDI less than or equal to 15) and abnormal groups. The accuracy of the first night's recording in predicting classification derived from recording on three or four nights was 83%.(ABSTRACT TRUNCATED AT 250 WORDS)