Summary: We report a longitudinal study of diary- and laboratory-based sleep measures in 50 healthy elderly subjects followed prospectively over a 3- year interval. Our hypothesis was that "old old" subjects (aged 75 to 87; n = 27) would show decline over time in measures of sleep quality, continuity, and depth, whereas "young old" subjects (aged 61 to 74; n = 23) were expected to show stability of outcome measures. Using analysis of variance-based planned contrast procedures, we found that this hypothesis was strongly supported for subjective sleep quality and laboratory measures of sleep latency, sleep efficiency, wakefulness after sleep onset, and slow-wave sleep percent. These changes were accompanied by increased napping in the old old. However, there was no change of habitual time in bed (total time or temporal placement of nighttime sleep), daily social rhythms, or sleep apnea. Change in medical burden scores did not correlate significantly with change in sleep efficiency or other outcome variables in the old old. Intervention designed to slow age-dependent decreases in sleep quality, continuity, and depth is discussed. The current results are representative of healthy elderly; sleep would probably deteriorate earlier and more quickly in elderly with more serious health problems and heavier medication use.