Since its discovery in the early 1990s, functional MRI has been the basis for more than 19,000 studies of the living, working brain. The technique allows scientists an unprecedented glimpse of the brain regions that are most active during particular tasks or states of mind, but it does not do so directly: the scans measure blood flow, which seems to increase around neurons that are firing. Neurons are not directly connected to blood vessels, however, so until now the mechanism underlying fMRI's robust success has been a mystery.

Now a team from the Massachusetts Institute of Technology reveals that the support cells dubbed astrocytes (because of their star-shaped bodies) form the link between neurons and blood vessels. The neuroscientists used a technique called two-photon microscopy, which harnesses light particles to image very small structures, to observe cells in ferrets’ brains. As the animals were shown different animated graphics, neurons responded within milliseconds, and astrocytes became active seconds later—matching the time delay that neuroimagers have long known accompanies blood flow to active brain regions. When the M.I.T. team blocked astrocyte function, the ferrets’ neurons fired as usual, but blood flow did not increase.

When researchers use fMRI, co-author Mriganka Sur explains, “we are really measuring astrocyte activation. Thus, anything that influences astrocytes is likely to influence fMRI.” This finding could add a layer of complexity to the interpretation of fMRI scans, because astrocytes may be subject to a different set of genetic and environmental influences than neurons are. But the more researchers understand about what is really happening when the brain “lights up” in an fMRI scan, the better they will be able to use the technology to learn about human cognition. [For more on the science of fMRI, see page 66.]