Positron emission tomography (PET) is currently the only technology affording three-dimensional measurement of the brain's energy metabolism which is closely coupled to brain function. Studies of glucose metabolism by PET of (18F)-2-fluoro-2-deoxy-D-glucose are therefore widely applied to show the contribution of various brain structures in the performance of a variety of tasks or their participation in functional deficits associated with various diseases. Although glucose metabolism decreases slightly with age to a regionally different degree, most types of dementia show severe changes in glucose metabolism. Alzheimer's disease (AD) is characterized by metabolic disturbances most prominent in the parietotemporal association cortex and later in the frontal lobe, whereas primary cortical areas, basal ganglia, thalamus, brainstem, and cerebellum are not affected. It is this typical pattern that distinguishes AD from other dementia syndromes. A ratio calculated from the metabolic rates of glucose of "affected" and "nonaffected" brain regions was able to separate patients with AD from age-matched controls and permitted the discrimination of patients with cognitive impairment of other origin in 85%. The discriminative power can be further improved by activation studies. A continuous visual recognition task increased the metabolic rate in normal subjects by 21% and in patients with AD by 6% on average, with significant regional differences. During activation the significant relation between severity of disease and temporoparietal metabolic rate became even stronger. In the assessment of effects of treatment on disturbed metabolism, PET studies demonstrated an equalization of metabolic heterogeneities in patients responding to a muscarinergic cholinagonist, whereas general increases in glucose utilization were observed with piracetam, pyritinol, and phosphatidyl-serine. The therapeutic relevance of such metabolic effects, however, must be proved in controlled clinical trials.