Most techniques presently available to measure cerebral activity in humans and animals, i.e. positron emission tomography (PET), autoradiography, and functional magnetic resonance imaging, do not record the activity of neurons directly. Furthermore, they do not allow the investigator to discriminate which cell type is using glucose, the predominant fuel provided to the brain by the blood. Here, we review the experimental approaches aimed at determining the percentage of glucose that is taken up by neurons and by astrocytes. This review is integrated in an overview of the current concepts on compartmentation and substrate trafficking between astrocytes and neurons. In the brain in vivo, about half of the glucose leaving the capillaries crosses the extracellular space and directly enters neurons. The other half is taken up by astrocytes. Calculations suggest that neurons consume more energy than do astrocytes, implying that astrocytes transfer an intermediate substrate to neurons. Experimental approaches in vitro on the honeybee drone retina and on the isolated vagus nerve also point to a continuous transfer of intermediate metabolites from glial cells to neurons in these tissues. Solid direct evidence of such transfer in the mammalian brain in vivo is still lacking. PET using [(18)F]fluorodeoxyglucose reflects in part glucose uptake by astrocytes but does not indicate to which step the glucose taken up is metabolized within this cell type. Finally, the sequence of metabolic changes occurring during a transient increase of electrical activity in specific regions of the brain remains to be clarified.