Is it possible to predict what a person is thinking of - or even what they are planning to do - based alone on their current brain activity? Recent advances have made it possible to decode and predict a person's thoughts from functional magnetic resonance imaging (fMRI) data. The key is that each thought is associated with a unique brain activation pattern that can be used as a signature or for that specific thought. It is possible to train pattern classifiers to recognize these characteristic signatures and thus read out a person's thoughts from their brain activity alone. This research can reveal how neural representations of mental contents are stored and transformed in the brain. It also gives rise to many potential applications, as for example in the control of computers and artificial prostheses by brain activity or in the detection of concealed mental states.