Because the two models make different predictions about the activity in a phoneme processing region, testing model predictions and identifying candidate regions will be a joint process.
Here I need to describe what are, in reality, pretty technical analytical methods. There's a lot of detail that I could present to make my analysis more specific, but you just don't have space for that. Note my very qualitative description of the two competing theories, and how I am reusing & extending the framework I set up in the introduction.