The first sections of this paper survey the history and recent developments relevant to the major neurotransmitters and neuromodulators involved in REM sleep control. The last portion of this paper proposes a structural model of cellular interaction that produces the REM sleep cycle, and constitutes a further revision of the reciprocal interaction model This paper proposes seven criteria to define a causal role in REM sleep control for putative neuro-transmitters/modulators. The principal criteria are measurements during behavioral state changes of the extracellular concentrations of the putative substances, and electrophysiological recording of their neuronal source. A cautionary note is that, while pharmacological manipulations are suggestive, they alone do not provide definitive causal evidence. The extensive body of in vivo and in vitro evidence supporting cholinergic promotion of REM sleep via LDT/PPT neuronal activity is surveyed. An interesting question raised by some studies is whether cholinergic influences in rat are less puissant than in cat. At least some of the apparent lesser REM-inducing effect of carbachol in the rat may be due to incomplete control of circadian influences; almost all experiments have been run only in the daytime, inactive period, when REM sleep is more prominent, rather than in the REM-sparse nighttime inactive period. Monoaminergic inhibition of cholinergic neurons, once thought to be the most shaky proposal of the reciprocal interaction model, now enjoys considerable support from both in vivo and in vitro data. However, the observed time course of monoaminergic neurons, their "turning off" discharge activity as REM sleep is approached and entered would seem to be difficult to produce from feedback inhibition, as originally postulated by the reciprocal interaction model. New data suggest the possibility that GABAergic inhibition of Locus Coeruleus and Dorsal Raphe monoaminergic neurons may account for the "REM-off" neurons turning off. However, the source(s) of GABAergic influences suggested by anatomical studies has yet to be definitively identified by electrophysiological recordings of GABAergic neurons that show the requisite inverse time course of activity relative to monoaminergic neurons. New and still preliminary microdialysis data suggest that reticular formation neurons, the effector neurons for REM sleep phenomena, might be disinhibited during REM sleep by decreased GABAergic influence, perhaps stemming from REM-on cholinergic neuronal inhibition of reticular formation GABAergic neurons. Whether the postulated cholinergic inhibition of GABAergic neurons is present is testable with in vitro recordings and double labeling. Taking into account the observed data on neuro-modulators/transmitters, a structural model incorporating interaction of REM-on and REM-off neurons and GABAergic influences is proposed. Finally, with respect to orexin and REM sleep, it is hypothesized that orexinergic activity may be a principal factor controlling REM sleep's absence from the active period in strongly circadian animals such as rat and man.