Functional magnetic resonance imaging (fMRI) is a novel neuroimaging technique that has enjoyed explosive growth during the past 7 years. It can be implemented relatively easily on many already existing MRI systems, it is noninvasive, and functional images may be obtained within tens of seconds. However, it measures a secondary effect of neuronal activity, the blood oxygen concentration (the blood oxygen level-dependent, or BOLD, effect), which is a somewhat sluggish and blurred measure of the actual time course of neuronal activity. Here we discuss the present limitations to temporal resolution and the degree to which they can be overcome by using specific assumptions about the coupling between neuronal activity, blood flow, and oxygen metabolism, or by specific experimental designs, in particular time-resolved fMRI.