Tang et al. show that continuous language can be decoded from functional MRI recordings to recover the meaning of perceived and imagined speech stimuli and silent videos and that this language decoding requires subject cooperation.