FMRI · Grey Matter

fMRI (functional magnetic resonance imaging) maps brain activity across the whole head by detecting the changes in blood oxygen that follow it, and it is the method that made it possible to see, without surgery, which areas of a healthy brain work during a task.


FMRI. fMRI (functional magnetic resonance imaging) maps brain activity across the whole head by detecting the changes in blood oxygen that follow it, and it is the method that made it possible to see, without surgery, which areas of a healthy brain work during a task.

Active tissue receives more oxygenated blood than it uses, so in an active patch the share of deoxyhaemoglobin falls. Deoxyhaemoglobin is paramagnetic and disturbs the MRI signal around it, so less of it means a slightly stronger signal; this blood-oxygen-level-dependent (BOLD) contrast was described by Ogawa in 1990. A scanner takes a whole-brain image every second or two, and statistics compare images across conditions to find voxels whose signal follows the task.

It sees everywhere, slowly. Unlike EEG and MEG it reaches deep structures with millimetre precision, but the blood response rises and falls over several seconds, so fast sequences of events blur together.

It reads blood. A voxel holds hundreds of thousands to millions of neurons, and the signal reflects their inputs and local processing as much as their output.

It can decode meaning. Trained on many hours of a person's recordings, a decoder reconstructed the gist of stories they heard or imagined; it worked only with that person's cooperation, during training and use.

It is expensive and immobile. A scanner costs millions, needs a shielded room and a still subject, which keeps it a research and clinical tool.

fMRI trades time for space.

It is the only routine method that sees the whole brain in millimetres, and it pays by seeing it seconds late through the plumbing.

Questions: Can fMRI decode what a person is thinking without their cooperation? In 2023 a decoder trained on about 16 hours of a person's fMRI recordings reconstructed the gist of stories they listened to or imagined, and of silent films they watched. It worked only for the person it was trained on and only with their cooperation: decoders trained on other people failed, and participants could defeat it by thinking of something else. The authors presented those limits as a current safeguard and called for policies to protect mental privacy before decoders become less dependent on cooperation.