The present study sought to investigate the meaning of subjectively good sleep, using a longitudinal and intraindividual design. Eight subjects slept in an isolation unit according to an irregular schedule of 6 h sleeps and 1 h naps, designed to give normal amounts of time in bed (1/3 of total), but variable sleep quality. Eight sleeps and eight naps were used for longitudinal simple and multiple regression analyses with standard polysomnographical sleep variables as predictors and subjective sleep quality as dependent variables. The results showed that subjective sleep quality (and related variables) was closely related to sleep efficiency, but not sleep stages. At least 87% efficiency was required for ratings of 'rather good' sleep. In addition, sleep quality ratings improved with closeness (of the awakening) to the circadian acrophase (17.00-21.00 hours) of the rectal temperature rhythm. The subjective ease of awakening differed from most other other variables in that it was related to low sleep efficiency. Objective and subjective homologues of sleep length and sleep latency showed high mean intraindividual correlations (r = 0.55 and 0.64, respectively). It was concluded that objective measures of sleep continuity were closely reflected in perceived sleep quality and that sleep quality essentially means sleep continuity.