Active neurons control their energy supply by dilating blood vessels to increase local blood supply to active regions. As well as balancing the brain’s energy supply with demand, this “neurovascular coupling” means that changes in blood supply, detected noninvasively in humans using fMRI/BOLD imaging, can be used to indicate regions of increased neuronal activity. Understanding the mechanisms that underlie neurovascular coupling is therefore critical for understanding how brain activity is fuelled and exactly how BOLD changes relate to neuronal activity. Classically, neurovascular coupling was thought to occur when signaling molecules, released from astrocytes and neurons, dilate arteriole smooth muscle cells. More recently, however1, contractile pericytes on capillaries have been found to constrict and dilate so could contribute to neurovascular coupling. Here (and see Ref. 2), I will show that capillaries in brain slices dilate in response to bath-applied glutamate or in response to neuronal activity. This dilation is mediated by prostaglandin E2 and is modulated by 20-HETE and nitric oxide. Capillaries also dilate in vivo, often before arterioles, suggesting that capillary pericytes are the first vascular cells to sense an increase in neuronal activity. The size of the capillary dilation, and the increased resistance of the capillary bed versus other components of the vascular tree3, suggests that capillary dilation produces most of the increase in blood flow that is observed in response to somatosensory stimulation. Finally, pericytes constrict and die following cerebral ischaemia, indicating that pericyte dysfunction may underlie the prolonged hypoperfusion observed following cerebral ischaemia. Protecting pericytes may therefore be a useful therapeutic strategy following stroke.