The major aim of this study was to examine laboratory- and diary-based measures of sleep in a group of healthy ("successfully aging") "old old" subjects (> or = 75 years of age), as contrasted with a group of "young old" subjects (60-74 years of age), who were followed longitudinally for 2 years. We hypothesized that sleep would deteriorate to a greater extent over time among the old old subjects than among the young old. The study group consisted of 50 elders (21 male, 29 female; 23 old old, 27 young old), each studied at baseline and then again at 1- and 2-year follow-up. Analysis of variance was used to determine main effects of age group, gender and time on key sleep measures. Most measures were found to be remarkably stable over time. However, some decay was detected in sleep efficiency among the old old, but not among the young old. In a multiple regression model for the cohort as a whole, age, cognitive status and medical burden at baseline predicted subsequent declines in sleep efficiency over the 2-year period. To our knowledge, this is the first longitudinal data set on sleep in the healthy old old.