Functional magnetic resonance imaging (fMRI) measures brain activity indirectly through the blood-oxygen-level-dependent (BOLD) signal — changes in the ratio of oxygenated to deoxygenated hemoglobin that accompany neural activity. Since its introduction in the early 1990s, fMRI has become the most widely used neuroimaging method in cognitive neuroscience, generating thousands of studies mapping the neural correlates of perception, attention, memory, language, emotion, and decision-making.
Principles and Methods
When a brain region becomes active, local blood flow increases, delivering more oxygenated hemoglobin than the tissue consumes. This surplus creates a change in the magnetic properties of blood that can be detected by the MRI scanner. The BOLD response peaks approximately 5-6 seconds after neural activity, limiting temporal resolution. Spatial resolution is typically 2-3mm. Analysis methods range from simple subtraction (comparing activation between task and control conditions) to multivariate pattern analysis (MVPA, decoding cognitive states from patterns of activation across voxels).
fMRI has important limitations: it measures a hemodynamic proxy of neural activity (not neural activity directly), it has poor temporal resolution (~seconds), the BOLD signal reflects input and local processing more than output, and the spatial resolution (millimeters) averages over millions of neurons. Statistical challenges include multiple comparisons correction across thousands of voxels. Despite these limitations, fMRI has provided invaluable insights into the functional organization of the human brain.