Temperature as a regulator of sleep is considered. Phenomenological data are presented indicating that there is a causal (not just a correlative) relationship between the changes in brain temperature and sleep phases in the wake-sleep cycle. An earlier suggested phase-transitional concept of the recovery function of sleep was shown to be the theoretical background for this relationship. According to the concept, sleep is an evolutionary developed mechanism of purification of the molecular composition of membranes in fast synapses, whose exocytosis depends on fluid-to-solid phase transition in the membrane. The concept suggests the answer to the question of why the recovery function is incompatible with the wake state and it states that the temperature changes (its decrease during the slow-wave sleep) is a necessary condition of the recovery process. Finally, some practically valuable issues from the concept are considered, including those that at first glance may seem paradoxical.