Two- to threefold variations in sleep length were observed in 12 subjects living on self-selected schedules in an environment free of time cues. The duration of polygraphically recorded sleep episodes was highly correlated with the circadian phase of the body temperature rhythm at bedtime and not with the length of prior wakefulness. Furthermore, the rate of REM (rapid eye movement) sleep accumulation , REM latency, bedtime selection, and self-rated alertness assessments were also correlated with the body temperature rhythm.

No abstract

Since Cannon first formulated the concept of homeostasis 60 years ago, attention has been focused on the corrective responses initiated after the steady state of the organism is perturbed. In this lecture it is argued that the concept of homeostasis should be extended to include the precisely timed mechanisms of the circadian (and circannual) timing system which enables organisms to predict when environmental challenges are most likely to occur. A mature understanding of homeostasis should encompass both "reactive" responses to changes in physiological variables which have already occurred and the "predictive" responses initiated in anticipation of predictably timed challenges. Such predictive responses have particular value because they enable physiological mechanisms to be utilized immediately, even if they involve a delay of several hours, by activating them at a suitable time in advance of a probable challenge. However, conflicts may often occur between predictions and reality. Examples from sleep-wake behavior, thermoregulation, blood volume homeostasis, and the regulation of potassium balance show that predictive responses often compromise the effectiveness of reactive homeostatic mechanisms even to the point of risking the survival of the organism. It must be concluded that the day-night cycle of the natural environment has played a fundamental role in shaping the evolutionary development of homeostatic mechanisms because of the dominating predictability of diurnal changes in illumination, temperature, food availability, and predator activity.

Human subjects during extended isolation from environmental time cues show complex variations in timing and duration of sleep with a progressive pattern, which eventually results in rest-activity and body temperature rhythms having different average periods. We present a model where temperature and rest-activity are each governed by an oscillator of the van der Pol type, denoted x and y, respectively. The oscillators affect one another through "velocity" type coupling, the effect of x on y being about four times greater than y on x. Periodic zeitgeber, z, is modeled as forcing only on y. We find that the entire pattern sequence can be realistically reproduced by causing only the intrinsic period of the y oscillator to increase while that of x remains stable. Desynchronization between x and y is the result of the intrinsic periods of the two oscillators becoming so disparate that the coupling is no longer able to enforce synchrony. Prior to desynchronization both human subjects and our model exhibit "phase trapping" wherein the relative phase of x and y is slowly modulated although the average x and y periods match. The model phase relations between temperature and both the timing and duration of sleep are, throughout, in good agreement with entrained and free-running human data. Most importantly, the model shows that the dramatic change in the length of the rest-activity cycle when desynchronization occurs is actually due to a relatively small variation in the governing variable, y.

No abstract