Resting state fMRI has rapidly emerged as a novel way to identify and investigate functional brain networks. This development raises a number of philosophical questions. In what ways is resting state work a "paradigm shift" in neuroimaging? To what extent does the technique improve upon, or inherit difficulties associated with older methods?
In this talk, I examine attempts to map cognitive functions onto resting state networks. I argue that resting state studies hold the promise of discovering functional brain networks in a "bottom-up"
fashion that avoids many of the problems associated with task-based fMRI. However, I propose that the same features that enable this exploratory science may also generate a novel confound. I argue that some features of resting state networks may be artifacts resulting from sampling a "mixture distribution" of heterogeneous functional states rather than genuine features of the brain's functional organization. This picture complicates the psychological inferences (e.g., network X performs function Y) drawn from resting state research.