Eye movement. Real fMRI studies don't produce images like those seen in advertisements, popular science, TV, and even peer-reviewed journals. They produce hundreds and hundreds of still, black and white scans similar to MRI scans that one can scroll through and that show nothing other than whether the participant was moving her or his head (which will ruin a run) and my favorite and first discovery: people who are asked to close their eyes will rapidly move them left and right without knowing it.
My first studies were as the lowliest of the low (beginning graduate research). While the experimental designs were decent enough if not good, the use of statistical analyses required for everything from discerning significant activity to coloring in the images from the scans were fundamentally flawed. So I devoted a lot more time to physics and mathematics than I was supposed to.
That neural correlates are correlates. We can state that, given some neural activity and a sufficiently well constructed experiment, there is a necessary relationship between particular neural activity and specific mental states or conceptualization, but not a sufficient relationship. Essentially, we can say a lot about brain activity that accompanies thoughts, beliefs, sensory processing, etc., but little about the causal direction of such relationships.