Functional MRI (fMRI) studies have shown that activity in resting-state networks (RSNs) fluctuates over seconds. To assess functional connectivity with a temporal resolution higher than that of fMRI, Baker et al. used a novel, magnetoencephalography-based approach. This revealed several brain 'states' that were each characterized by a functional network resembling an fMRI-based RSN. Transitions between certain states were more likely to occur than transitions between other states. Importantly, each state remained stable for only 100–200 ms, indicating that functional networks are shorter-lived than suggested by fMRI data.