During an fMRI, if you do language testing consisting of naming pictures, the same level of brain activation is observed on the images whether the patient says the name of the object aloud or simply thinks of the name silently. Consequently, we perform the testing with the patients thinking the name silently, since speaking results in head movements which produces movement artifacts in our images (our resolution is 1.5mm x 1.5mm x 4.5mm, so we can even detect motion from breathing).
Another fun fact: my avatar is an image of my actual brain (structural spgr) overlayed onto the bone structure of my head/face using the images acquired via MRI and a program called BrainVoyager.