Functional neuroimaging using positron emission tomography has recently yielded original data on the functional neuroanatomy of human sleep. This paper attempts to describe the possibilities and limitations of the technique and clarify its usefulness in sleep research. A short overview of the methods of acquisition and statistical analysis (statistical parametric mapping, SPM) is presented before the results of PET sleep studies are reviewed. The discussion attempts to integrate the functional neuroimaging data into the body of knowledge already acquired on sleep in animals and humans using various other techniques (intracellular recordings, in situ neurophysiology, lesional and pharmacological trials, scalp EEG recordings, behavioural or psychological description). The published PET data describe a very reproducible functional neuroanatomy in sleep. The core characteristics of this 'canonical' sleep may be summarized as follows. In slow-wave sleep, most deactivated areas are located in the dorsal pons and mesencephalon, cerebellum, thalami, basal ganglia, basal forebrain/hypothalamus, prefrontal cortex, anterior cingulate cortex, precuneus and in the mesial aspect of the temporal lobe. During rapid-eye movement sleep, significant activations were found in the pontine tegmentum, thalamic nuclei, limbic areas (amygdaloid complexes, hippocampal formation, anterior cingulate cortex) and in the posterior cortices (temporo-occipital areas). In contrast, the dorso-lateral prefrontal cortex, parietal cortex, as well as the posterior cingulate cortex and precuneus, were the least active brain regions. These preliminary studies open up a whole field in sleep research. More detailed explorations of sleep in humans are now accessible to experimental challenges using PET and other neuroimaging techniques. These new methods will contribute to a better understanding of sleep functions.