Further discussion of a model of the REM sleep oscillator.

The limit cycle feature and the grounding of our model in physiology are endorsed by Daan and Beersma [Am. J. Physiol. 251 (Regulatory Integrative Comp. Physiol. 20): R1030-R1032, 1986.] as well as the fundamental postulate of the model that the latency, duration, and intensity of the first rapid-eye-movement (REM) period depends on whether the limit cycle is entered from an internal or external trajectory, and the fact that this trajectory is determined by circadian modulation of conditions at sleep onset. We describe our reasons for preferring a more explicit formulation of the sleep onset conditions than provided in our earlier "Karma" version of this model and provide additional details of how the control of the REM-off population decline is modeled. Additional empirical evidence is cited for the continuous circadian modulation of REM cycle parameters. We emphasize that, compared with the original simple model, the present version of the model adds only one additional "free" initial condition parameter (circadian phase) that is used to model normal sleep begun at different circadian phases and the resultant variations in REM latency, duration, intensity, and period length. We present specific predictions of the model and new supporting empirical data.