Abstract
This paper redescribes some recordings of human sleep and waking made in several laboratories during the past decade under conditions of temporal isolation. Since 1972 it has been noticed that sleep duration depends mainly on the timing of prior sleep onset relative to a rhythm of 24- to 25-h duration. The present paper emphasizes four additional points: 1) that the dependence sometimes includes a remarkable discontinuity, 2) that such dependence is characteristic of a rhythmically modulated threshold process; 3) that the rhythm's period gradually changes in some experiments; and 4) that no comparable regularity has been detected in the timing of sleep onset. This last impugns the reliability of models that treat sleep onset and wake onset as complementary but comparable processes.