I tend to think of fMRI based neuroscience as analogous to trying to reverse-engineer how a computer works by measuring the electric fields it generates. A lot

I tend to think of fMRI based neuroscience as analogous to trying to reverse-engineer how a computer works by measuring the electric fields it generates. A lot could be learned, but only to a degree. Except that there was no evolutionary pressure for understandability on the brain, so it's quite probably a convoluted kludge. This complexity cuts across levels of abstraction, with interplay between the genetic, neural and environmental levels. I suspect the entire thing is too complex for us to understand -- without better mathematical and computational tools. Anti-depressants work for some cases for some reasons, but we might need some paradigm leap to eventually understand why. The current hypothesis of neurotransmitter imbalance always sounded childish to me. A bit like trying to fix a software bug by improving the conductivity of the circuits, or something like that.