maxcollard
commented
9 years ago Another interesting neurological question: is "functional connectivity" even "real" from a biological standpoint? Functional connectivity is a fuzzy term, but usually involves correlated metabolic activity (in the fMRI literature), correlated bandlimited electrical activity (coherence in EEG/ECoG), or predictive/generative modeling (point process models with interaction terms for spikes, AR "effective connectivity" in EEG/ECoG). The interpretability of each kind of result is ... pretty sketchy, at best. For example, what does it actually physiologically mean for two regions' blood flow to be correlated (fMRI resting state networks)?
A daunting mechanistic question arises for people who look at time-varying connectivity measures, which pervade EEG/ECoG. How do certain parts of the brain "pay attention" only to certain things at certain points during a task, if it's just a bucket of neurons? Functional connectivity seems largely correlative (in the sense of, "Look, activity at sites A and C predicted activity site B 200ms after a visual stimulus with p < .05") and somewhat difficult to interpret ("Connections between A/C and B are important for visual processing") at this point.
Then again ...
It does seem that the "predictiveness" of sites changes rather dramatically over the course of a particular task, and is vastly different between tasks [citation needed]. So, perhaps a better term would be that functional connectivity is largely phenomenological: it happens, but no one really know why, how, or why anyone cares.
Can you talk a little bit about the pros and cons of each and in which kinds of situations/applications one would be more useful than the other? Like why pick one over the other if you had to pick?