For fMRI "typical sample sizes produce only modestly replicable results". What's interesting is that this is for tasks with large behavioral effect sizes. Perhaps common patterns of behavior are due to diverse neural responses. My brain != your brain. tinyurl.com/yaw2hjfc 
More detail: the high sample sizes needed to get consistent results for fMRI seem to be due to high between-subject variability. Yet this is observed on basic cognitive tasks that have consistent behavioral results (e.g. N-Back). So similar behavior belies neural diversity.
This is something physiologists have been documenting for a long time (especially Eve Marder and colleagues: nature.com/articles/nrn19…).
So maybe averaging over brains isn't such a great idea. Moreover, maybe the idea of a 'depressed brain' or a 'schizophrenic brain' is silly/foolish/Quixotic. When will neuroscientists start thinking critically about the fact that nervous systems are designed to become unique?
