We have hypothesized that REM sleep is functionally and homeostatically related to NREM sleep rather than to waking. In other words, REM sleep rather than to waking. In other words, REM sleep occurs in response to NREM-sleep expression and compensates for some process that takes place during NREM sleep. Under normal conditions, the need for REM sleep does not accrue during waking. The primary basis for this hypothesis is the fact that REM-sleep expression is a function of prior NREM-sleep expression. That is, REM sleep follows NREM sleep within sleep periods, REM-sleep episodes occur at intervals determined by the amount of NREM-sleep time elapsed, and total time spent in REM sleep is consistently about 1/4 of prior NREM-sleep time, regardless of how much time is spent in NREM sleep. Our experimental tests of the hypothesis support it. (1) REM-sleep propensity accumulates quite rapidly during a 2-hr interval spent predominantly in NREM sleep. (2) The timing of individual REM-sleep episodes is controlled homeostatically, by accumulation within NREM sleep of a propensity for REM sleep. The NREM sleep-related model of REM-sleep regulation (Fig. 1) explains a number of phenomena of REM-sleep expression, including the frequent and periodic occurrence of REM-sleep episodes throughout sleep periods, that have been accommodated by the waking-related model but are not functionally accounted for by it. In our opinion, the NREM sleep-related model of REM-sleep regulation recommends itself partly by its simplicity. According to the waking-related model, two independent and competing sleep propensities accumulate during waking and are discharged in two distinct sleep states that perform different waking-related recovery processes. One behaviour, sleep, is thought to perform two independent and competing functions that alternate at regular intervals. In the NREM sleep-related model of REM-sleep regulation, sleep debt simply reflects a need for NREM sleep. That is, the cerebrally less activated state of NREM sleep enables some form of restoration made necessary by the cerebrally activated state of waking. Periodic occurrence of REM-sleep episodes is explained without postulating an oscillatory mechanism to gate expression of NREM sleep versus REM sleep. In assessing the comparative merits of the waking-related and NREM sleep-related models of REM-sleep regulation, one should consider the influence of time-worn habits of thought.(ABSTRACT TRUNCATED AT 400 WORDS)