I am planning an experiment whereby rTMS would be administered to subjects while they perform a language task, with the hopes of improving performance on this task as compared to sham stimulation.
Due to the novel nature of this task, it is difficult to predict the most optimal region which to choose as a stimulation site. Someone in our group suggested to run this task in the scanner, and choose as the stimulation site the main cluster that the task activates, in other words the cluster that comes out of the fMRI contrast relevant for this task.
This however seems to me excessively data-driven, as well as somehow circular (though it would not be correct to quite call it double dipping).
Does this approach make sense at all?